The transition toward circular economy principles in water treatment requires advanced process systems engineering tools to evaluate the trade-offs between environmental sustainability and economic viability, particularly for energy-intensive Zero Liquid Discharge (ZLD) systems. While classic ZLD systems treat concentrated brine as waste, circular ZLD (CZLD) systems incorporate salt recovery technologies that generate marketable salt product. This study presents a comprehensive technoeconomic assessment framework for CZLD systems integrated with renewable energy. The framework is developed to evaluate different CZLD configurations that generate saleable sodium chloride. The assessment methodology integrates solar photovoltaic systems with increasing capacities (100-1400 kW) to analyze renewable energy penetration and energy storage requirements. The renewable energy integration model incorporates hierarchical energy dispatch algorithms prioritizing direct solar utilization, battery storage, and grid backup systems. Solar PV integration demonstrates renewable fractions ranging from 15-100%, with corresponding emissions reductions of up to 72% compared to grid-powered baselines. The technoeconomic analysis reveals a critical trade-off: while renewable integration substantially reduces carbon emissions, it significantly increases operational costs due to higher solar costs compared to subsidized grid electricity. However, configurations with substantial salt revenue generation maintain economic viability even under complete renewable operation. Beyond configuration screening, this work makes three specific contributions to process systems engineering: (i) a hierarchical energy-dispatch model for intermittent solar integration in energy-intensive separations; (ii) a revenue-function formulation that embeds salt valorization into ZLD process economics, shifting the design objective from cost minimization to net revenue maximization; and (iii) systematic identification of the thermodynamic basis for membrane–thermal hybrid superiority through analysis of osmotic pressure versus latent heat energy barriers across 26 configurations. These contributions establish a reproducible framework applicable to brine management in resource-constrained industrial settings.

Plastic waste represents an abundant and underutilized resource that can be converted into valuable products through microwave-assisted pyrolysis. In this research, a novel microwave-assisted processing plant that converts mixed plastic waste to olefins and aromatics is developed and simulated on Aspen Plus (v.14) guided by laboratory-scale experimental data. The experimental results show that at a bulk temperature of 400°C and ambient pressure, approximately 95% of a solid waste plastic feed comprised of equal portions of polypropylene and polyethylene is converted to gases, with nearly two-thirds of the resulting effluent gas composed of olefins. Simulation results show that 2889.1 kg/h propylene, 2088.0 kg/h ethylene and 96.3 kg/h aromatics (benzene and toluene) are produced as main products from 8000 kg/h of mixed plastic feed. High-purity propane and ethane streams were also recovered and sold as byproducts. A technoeconomic analysis is subsequently conducted, revealing that the proposed novel plant achieves a positive net present value of USD 61.6 million at a processing capacity of 8000 kg/h of mixed plastic waste feed. The levelized cost of propylene production is also calculated to be USD 481.1 per ton, estimated while keeping the price of the other coproducts fixed at their average market selling price. Such results confirm the economic feasibility of the microwave-assisted novel plant at the investigated scale.

Conventional urea production is a centralized and fossilintensive process associated with significant greenhousegas (GHG) emissions and limited flexibility for deep decarbonization. As an alternative, the Integrated COnversion of NItrate and Carbonate steams (ICONIC) project is developing innovative electrochemical urea (eurea), via the co-electroreduction of nitrogen and carbon sources using renewable power. While recent research advances in electrocatalysis have demonstrated promising Faradaic efficiencies (FE) toward urea, the design of electrochemical systems involves inherent tradeoffs between key performance indicators (KPIs) such as current density, cell voltage, and FE. Crucially, the implications of electrolyzerlevel performance on plantlevel economics and environmental impacts remain poorly understood. To address this gap, we integrate process modelling with technoeconomic and lifecycle assessment (TEA–LCA) to evaluate the trade-offs of KPIs from a process systems perspective. Labscale electrolyzer KPIs are transformed into process-level environmental/economic metric, using Pythonbased electrolyzer modelling and Aspen Plus downstream simulations. Fossil urea is used as a benchmark, and production scenarios are evaluated for windpowered operation in the Netherlands and solarpowered operation in Spain. The results identify quantitative performance targets for electrolyzer design and demonstrate that urea recovery in the downstream are decisive bottlenecks. This work reveals the essentials for expanding the system boundary beyond the electrolyzer cell, as to identify R&D prioritization and deployment strategies during early-stage technology development and lays the groundwork for scalable and low-carbon urea production.

The utilization of low-quality vegetable oils as raw materials helps to reduce the production costs of biodiesel. Waste cooking oils are examples of this type of raw material, having a high content of free fatty acids. While biodiesel is a sustainable alternative to fossil fuels, conventional production methods face challenges due to low reaction rates and high energy demands. This study investigates pathways for biodiesel production from waste cooking oil using process intensification technologies in combination with biowaste-derived heterogeneous catalyst. Conventional and intensified (reactive distillation-based) processes are compared using Aspen Plus simulations as an analysis tool, focusing on the production of biodiesel from waste cooking oil (WCO) using CaO as a catalyst. According to the results, the conventional method at 60°C and 6:1 methanol-oil ratio achieves 95% conversion but suffers from high methanol use and long reaction times (65 min). The intensified process at 65–70°C reduces energy needs by more than 29%, requiring shorter residence time while enhancing efficiency. Overall, it demonstrates that intensified and sustainable biodiesel production can reduce environmental footprints and support the transition to cleaner energy systems. This work emphasizes the transition from traditional processing routes to innovative, intensified technologies for sustainable fuel production within the framework of the circular economy.

An integrated multiproduct facility for the valorization of apple pomace in the green production of high value-added chemicals such as phenolic compounds and pectin, bioethanol, as well as apple juice, was optimally designed. The process includes green extraction technologies relying on subcritical water extraction and ethanol produced on-site through fermentation of residues. Two scenarios were evaluated: one based on purchasing the ethanol and another with on-site ethanol production. The units of the process were modeled using first principles. The process superstructure was formulated as a mixed-integer nonlinear programming problem, solved by decomposition. Investment and production costs of both alternatives were similar, with unit production requirements ranging from 1.09 €/L to 1.13 €/L. The discounted payback periods were 6.7 years and 6.4 years for the on-site ethanol production and purchased ethanol scenarios, respectively, while the internal rates of return were 36.4 and 37.8% for both processes, respectively. Both scenarios showed economic profitability with higher returns under the purchased ethanol scenario, but the on-site production scenario allowed waste reduction, self-sufficiency and ethanol reuse. A green extraction process based on subcritical water for the pectin extraction and the solvent recycling avoided the use of conventional acids and promoted the circular economy. The findings of this project demonstrated the potential of apple pomace as a valuable feedstock for integrated biorefinery applications in the agri-food sector.

This work presents a hybrid modelling approach for a downdraft sewage sludge gasifier within the Shit2Power (S2P) process. The gasifier is represented in CHEMCAD NXT by a series of four standard reactors that combine stoichiometric and equilibrium models with a data-driven correction step to account for deviations from ideal Gibbs equilibrium. Reaction conversions in the correction reactor are fitted to experimental synthesis gas compositions reported by Werle (2014) [7] for 30 operating points with varying equivalence ratios and reactor inlet air temperatures. The calibrated hybrid reactor model is evaluated against these data and shows conservative agreement for the combustible gas components of the synthesis gas. To overcome the limitations of linear interpolation between fitted operating points, several machine learning approaches are evaluated to predict the reaction conversions, and boosted neural networks are selected as a compromise between prediction accuracy and smooth behaviour in the operating window. After implementation in the CHEMCAD flowsheet, sensitivity studies with respect to equivalence ratio and reactor inlet air temperature show good agreement between predicted and fitted gas compositions, considering the limited data available. This demonstrates that the combination of flowsheet simulation and machine learning provides a promising framework for modelling sewage sludge gasification in the S2P process, but that significantly more experimental and training data are needed for practical application.

The transition from a linear, fossil-based polymer economy to a circular bio-economy is critical for mitigating resource depletion and greenhouse gas emissions. This study provides a rigorous comparison of two biomass-to-plastic pathways: a biochemical route (PLA via enzymatic hydrolysis) and a thermochemical route (bio-PE via gasification and MTO). Based on Aspen Plus simulations and a “lifetime-adjusted” lifecycle assessment framework, we evaluate the environmental performance of these routes in the transition from linear to circular systems. Unlike standard “cut-off” methods, the lifetime-adjusted model accounts for virgin make-up and molecular retention across multiple recycling cycles. Results indicate that at current 15% recycling rates, PLA exhibits the lowest global warming potential due to significant biogenic carbon sequestration. However, as recycling rates reach 75%, process efficiency becomes the dominant factor; the precise biochemical recycling of PLA continues to outperform mechanical HDPE recycling, whereas the energy-intensive thermochemical bio-PE route loses its competitive advantage. Our sensitivity analysis reveals that LCA modeling choices significantly diverge at intermediate recycling rates, resulting in different values for environmental impact. Ultimately, while bioplastics serve as a vital agent for CO2 sequestration during the transition to circularity, long-term sustainability necessitates a shift toward “recyclable-by-design” materials.

The increasing popularity of electric vehicles (EVs) leads to an expected rise in the quantity of end-of-life lithium-ion batteries (LIBs) that require efficient management. This paper presents a State Task Network (STN) based optimization model to analyze and optimize the supply chain for LIBs, allowing for the selection of optimal processing routes, facility locations, capacities and reintegration of recovered materials, as well as analyzing the possible trade-offs between different end-of-life management strategies. Based on available data from the literature, the model is demonstrated with the LIB supply chain considering both primary production and different end-of-life strategies for spent LIBs (recycling and reuse). The case study reveals that mechanical pretreatment followed by hydrometallurgical recycling is the optimal pathway and it outperforms the linear supply chain in both costs and emissions. The cost optimal solution opts for more centralized collection and disassembly, whereas when minimizing the total greenhouse gas emissions, the optimal collection and disassembly center locations are decentralized. The model can aid in the design of circular supply chains to select and locate the optimal processes and to examine how the network should evolve over time.

The wine industry generates large volumes of organic effluents, whose inadequate management poses significant environmental challenges but also offers opportunities for resource recovery. In this work, an integrated biorefinery scheme for the valorization of winery effluents is proposed and evaluated through steady-state simulation in Aspen Plus®. The biorefinery converts winery wastewater into a portfolio of value-added chemicals and biofuels, including levulinic acid, propylene glycol, formic acid, light gases, naphtha, sustainable aviation fuel, green diesel, and bioethanol, while enabling water recovery and carbon dioxide management. Two alternative CO2 capture routes are analyzed and compared: a conventional CaO-based carbonation–calcination process and an innovative absorption system using deep eutectic solvents (DES), specifically choline chloride–urea. Technical performance is assessed through chemical oxygen demand (COD) removal, recovery, conversion, yield, and product mass ratios. Economic feasibility is evaluated using profit-based indicators, while environmental performance is quantified through CO2-equivalent emissions associated with utility consumption. Results show that the proposed biorefinery achieves a COD removal efficiency of 99.99%, producing treated water compliant with Mexican regulations for internal reuse. The DES-based configuration reduces raw material costs by 48.86%, enables hydrogen recovery as an additional valuable product, and increases overall profit by 2.86% compared to the CaO-based scheme. Although the relative reduction in total CO2 emissions is modest (˜0.5%), the DES configuration achieves an absolute annual reduction of 2, 198 t CO2. Overall, the results demonstrate that integrating DES-based CO2 capture into winery effluent biorefineries enhances economic performance and supports circular economy principles through waste valorization, water reuse, and emissions mitigation.

Gallium is a critical material with increasing demand driven by compound semiconductors such as gallium nitride (GaN) and gallium arsenide (GaAs) used in power electronics and optoelectronics and a highly concentrated supply chain, with 98 % of refined production occurring in China. While recycling remains limited, GaAs semiconductor manufacturing generates wastewater that can contain gallium concentrations ranging from 1–35 mg·L^-1, representing an underutilized secondary resource. This study evaluates the technical and economic feasibility of recovering gallium from GaAs semiconductor wastewater across an input range of 10–100 m³·d^-1 using process modelling and a techno-economic analysis comparing two alternative separation routes: ion exchange (IX) and solvent extraction (SX). Using a real-world industrial wastewater composition, IX achieves a higher overall recovery than single-stage SX (80.40 % vs. 62.54 %), which translates into consistently lower levelized costs of gallium. The resulting levelized cost decreases from 563 to 144 €·kg^-1 for IX and from 701 to 161 €·kg^-1 for SX. The results demonstrate that separation step efficiency is the dominant determinant of economic performance and show that gallium recovery from semiconductor wastewater can be economically viable at fab-relevant scale. The analysis provides a proof-of-concept for integrating gallium recovery into semiconductor wastewater management and highlights its potential contribution to resource security and circular material supply.

The accumulation of automotive plastic waste poses a growing environmental threat; while recycling has the potential to address this, its use remains limited by the complexity of the materials used in vehicle components. Specifically, the presence of mixtures of polypropylene (PP), polyethylene (PE), and polyoxymethylene (POM) in the materials makes mechanical recycling challenging due to the difficulty of separation. To address the inefficiency of current end-of-life management, we present a systematic computational framework integrating computer-aided molecular and process design (CAMPD) with technoeconomic assessment (TEA) and life cycle analysis (LCA) to design a solvent-based recycling process capable of producing near-virgin quality resins. This framework involves utilizing the SAFT-gamma Mie equation of state to predict thermodynamic properties and employing nonlinear programming (NLP) to perform process optimization. From an evaluation of 875 solvent candidates, we identify 72 feasible solvent combinations, amongst which cymene and cyclohexanone is the most promising solvent pair to separate polyolefins selectively and dissolve POM, enabling reduced dissolution temperatures and minimized unit capital costs. TEA and LCA indicate that the proposed process is highly competitive, as all feasible designs yield a minimum selling price (MSP) below the market price of virgin polymers and achieve global warming potentials (GWP) significantly lower than those of virgin polymer production. Sensitivity analysis confirms the robustness of the design, showing that both economic viability and environmental competitiveness relative to virgin production are maintained, even under conservative solvent loss scenarios of up to 5%. These findings suggest that solvent-based recycling offers a commercially viable pathway to meet emerging EU circularity mandates for the automotive sector.

The biopharmaceutical sector has traditionally focused on cost-efficient process design and capacity planning to meet rising demand. Recently, sustainability pressures have increased, driving efforts to reduce the environmental footprint of manufacturing and supply chains; however, strict quality and sterilization requirements can limit the implementation of fully circular resource-use strategies. In this space, adopting an industrial-cluster systems view could unlock opportunities to improve sustainability of industrial clusters through coordinated material and energy exchange, supporting resource efficiency at cluster level and still meet sector-specific quality/sterilization requirements. In this work, we present life cycle assessment (LCA)-based comparative analyses to investigate the potential of industrial symbiosis within monoclonal antibody (mAb) manufacturing, whereby LCA process models are based on comprehensive techno-economic analyses that quantify resource inputs and waste streams. The presented results discuss impact reduction pathways including integrating renewable wind-based electricity, low-grade waste heat use, heat recovery through incineration and mechanical recycling of single-use plastics. The compound reductions of industrial symbiosis and energy source decarbonization are estimated, corresponding to 75% reductions in GWP potential, 70% in fossil resource scarcity and 60% in water use, thus highlighting the benefit of adopting a large-scale system view rather than siloed approaches.

While the transition from traditional energy sources to renewable energy is necessary to reduce greenhouse gas (GHG) emissions, it introduces new challenges related to material use, both in quantity and type, potentially leading to resource scarcity, biodiversity loss, and waste accumulation. Therefore, incorporating circular economy (CE) principles into the design and planning of energy systems becomes essential. Despite the growing recognition of circularity, current assessments in energy systems focus on economic performance and GHG emissions. In this work, we propose a metric for quantifying circularity of energy systems based on the CE assessment framework MICRON, addressing the gap between CE metrics and energy systems planning. The framework is adapted to energy systems by accounting for the specific characteristics of energy technologies and by incorporating metrics associated with critical material use, scarcity, and durability. Its applicability is demonstrated through a case study of energy system planning at the University of Wisconsin-Madison, considering a grid-connected system with solar, wind, and lithium-ion battery technologies. Results show that wind-only portfolios achieve higher overall circularity scores than solar-only and hybrid systems, reflecting the higher efficiency and availability of wind energy. Hybrid systems exhibit higher durability and more efficient material use by avoiding system oversizing. Regarding decarbonization strategies, reducing grid reliance and associated emissions does not necessarily improve circularity, as energy storage is required to ensure reliability. Storage systems increase material demand, the share of critical materials, and replacement frequency. Finally, a sensitivity analysis was performed, highlighting that end-of-life recovery is a key factor influencing circularity.

Critical metals are indispensable in renewable, low-carbon, and hydrogen technologies due to their unique catalytic and electrochemical properties. They are primarily sourced through mining, which is associated with significant environmental impacts and geopolitical risks due to the uneven global distribution of ore deposits. As a result, efficient recovery of these metals from secondary sources such as electronic waste has become increasingly important. In this context, liquid-liquid extraction (LLE) has emerged as a promising separation technique due to its high selectivity and scalability. The development of intensified, continuous-flow LLE in small channels offers further advantages in terms of mass transfer efficiency, solvent utilization, and process sustainability, making it an attractive approach for the recovery of critical metals. A flow pattern known as segmented flow further enhances mass transfer in LLE in small channels. This work presents a hybrid modelling approach for developing a predictive model of a segmented flow LLE process, intended for digital twin implementation in critical metals recovery. Within the hybrid modelling framework, mass transfer is modelled using a lumped approach, which allows to treat mass transfer independently from flow hydrodynamics. Further, hydrodynamic and mass transfer models are developed in parallel using Gaussian process (GP)-based active learning (AL) and model-based design of experiments (MBDoE), respectively. Prior knowledge of flow regimes is used in developing the hydrodynamic model. The method was tested in an in silico case study and shown to efficiently develop reliable models for segmented flow extraction in small channels.

Regional deployment of bioenergy and bio-based products is often constrained by the seasonality, heterogeneity, and dispersed availability of lignocellulosic biomass. This work demonstrates a computer-aided process engineering (CAPE) workflow that integrates experimental characterization, process modeling, and life cycle assessment (LCA) to support regional symbiosis design using mixed feedstocks from agriculture and forestry. A case study is developed for Tanegashima, a remote Japanese island where unused woody residues and sugarcane bagasse are locally available but temporally mismatched. Torrefaction is modeled in an autothermal configuration: char is the main product, while torrefaction gas and condensables are recovered for internal heat supply and any excess is treated as an energy coproduct. Laboratory measurements (220–400°C, 20°C interval) provide temperature-dependent yields of char, tar, aqueous condensate, and gas, alongside ultimate analysis and heating values of solids and representative gas composition. These data are translated into a mass–energy inventory for LCA foreground modeling up to pelletization. The functional unit is 1 t of wet biomass (40 wt% moisture) processed on-island. Results show that increasing temperature decreases solid yield but increases charification and heating value. Autothermal operation becomes feasible above approximately 340°C under a mid-level drying-energy assumption, and the threshold is highly sensitive to drying performance. Across assessed feedstocks and temperatures, life-cycle GHG emissions are negative when credits are given for substituting fossil coal and kerosene-equivalent fuels. The proposed workflow enables composition- and seasonality-aware screening of regional biomass portfolios under consistent inventory assumptions.

This study conducted a life cycle assessment to assess and design regional symbiosis at plant-oil production. These industries face challenges including dependence on fossil fuels and the generation of underutilized by-products, while effective regional symbiosis requires the selection of diverse regional unused resources and assessment based on process models that consider future technological prospects. Mathematical models for plant-oil production were developed using industrial data from literature to calculate inventory data. The case study showed that introducing woody biomass combined heat and power reduced GHG emissions by 8% in the Cradle-to-Grave system boundary, while recycling technology for soap stock using Kolbe electrolysis achieved a 3% reduction. Regional analysis indicated that 33 prefectures in Japan could meet woody biomass demand through sustainable forestry management, potentially reducing GHG emissions in Japan by approximately 0.041%. These results suggest that regional symbiosis can enhance resource-use efficiency, reduce environmental impacts, and contribute to sustainable regions.

This paper presents OpenAD-lib, an open-source Python framework for anaerobic digestion (AD) digital twins, unifying mechanistic models, machine learning (ML) surrogates, and model predictive control (MPC) within a modular ecosystem. OpenAD-lib addresses the critical fragmentation in AD digitalisation by bridging mechanistic and data-driven paradigms under explicit uncertainty. By integrating uncertainty-aware feedstock characterisation with robust process control, the platform enables the transition from isolated research tools to fully integrated digital twins, delivering economic and environmental value in AD systems.

The integration of industrial symbiosis (IS) and carbon capture utilization (CCU) is recognized as a key strategy for achieving net-zero emissions in industrial parks (IPs). However, the optimization of these processes in combination remains an undeveloped research area. This work presents a multi-objective optimization framework implemented in Pyomo and linked with OpenLCA to simultaneously maximize material exchanges and minimize operational costs while evaluating CCU deployment. Applied to Portugal’s largest industrial park, the model identified 26 feasible symbiotic exchange routes involving 14 enterprises and 7 potential CCU technologies. Maximum material exchange optimization yields 3,042,107 ton/year across 26 symbiotic routes and 7 CCU units, achieving 89.8 % reduction in climate change impact (from 13.5 to 1.76 million CO2eq/year); Cost minimization achieves 2,223,298 ton/year with 27 % fewer exchanges, delivering 87.0 % environmental reduction and net revenue of €65.5M/year via carbon tax avoidance. The proposed trade-off configuration deploys 2 CCU units (both for methanol production), establishes 12 symbiotic routes, exchanges 2,925,556 ton/year, and generates €65.5M/year net benefit. Decision variables encompassed inter-enterprise flows, binary decisions to select exchanges, and CCU unit selection informed by technology readiness level (TRL). The epsilon-constraint method generated a Pareto front revealing the cost-benefit trade-offs of the symbiotic network. IS with CCU integration can reduce IP climate impact by 87-90 % while generating substantial economic returns. Methanol emerges as the critical CCU pathway for replacing fossil-imported feedstocks. This framework provides decision-makers with actionable optimization strategies for achieving near net-zero industrial parks by 2050, not only through CCU but also through IS.

Existing gas network infrastructure are important national energy assets, transporting mostly fossil-derived natural gas to end-users. Biomethane, methane derived from anaerobic digestion (AD) of organic matter, presents a potential route to replace fossil fuels with home-grown renewable gas. Combined with carbon capture and storage (CCS) of the CO2 in the biogas potentially results in carbon negative energy. This work seeks to understand the feasibility of operating a part of the gas network isolated from the main natural gas network fully on biomethane in Scotland. We present an integrated techno-economic optimisation framework for designing self-sufficient biomethane islands, applied to the Inverness network. The model, implemented as a nonlinear program (NLP), maximises annual net profit from biomethane sales and Green Gas Support Scheme (GGSS) tariffs subject to practical constraints such as GGSS-compliance of =50 % waste-derived biomethane, seasonal supply, land/scale, demand balancing with centralised liquefied natural gas (LNG) storage, and a life-cycle global warming potential (GWP) metric. Three archetypes are analysed: Type A (crop-dominated, manure co-digestion), Type B (food/industrial wastes, grass/manure support), and Type C (distillery residues + grass/manure). In Inverness, feasible solutions include: Type A (2 large ~92, 000 m³ digestion plants at ~23 ha/site producing 97.39 Mm³ y^-1 of gas; net revenue £13.4 M y^-1; GWP ~42.2 ktCO2e/y), Type B (1 ~97, 000 m³ plant at ~24 ha, producing 48.69 Mm³ y^-1 gas; net revenue £9.4 m y^-1; GWP ~48.0 ktCO2e/y), and Type C (2 large ~110, 000 m³ plants at ~27 ha, producing 97.4 Mm³ y^-1 of gas; net revenue £13.0 m y^-1; GWP ~47.2 ktCO2e/y). Type B is most profitable per unit capacity due to gate-fee feedstocks but carries higher GWP (mostly from grass-silage cultivation). The model balances a combination of dynamic feeding of different recipes with using a centralised LNG storage to buffers seasonal deficits and maximise asset utilisation; optional CO2 liquefaction (~87.7 kt y^-1 per large site at ~151 kWh t^-1) enables near/net-negative operation under low-carbon power. Our results find that the business model is feasible for Inverness and highlight the value of systems thinking and the need for policy reform (particularly lifting the 250 GWh y^-1 cap for GGSS and rewarding carbon intensity rather than just waste-derived methane) to unlock larger, efficient, low-emission regional systems.

Achieving and sustaining net-zero greenhouse-gas emissions will require the long-term deployment of green hydrogen. Proton exchange membrane water electrolysers (PEMWEs) are attractive for variable renewable electricity (VRE) because of their fast dynamic response; however, they rely on iridium (Ir) anode catalysts, and Ir supply is severely constrained. Here, a Japan-specific dynamic material flow analysis (DMFA) model is developed for 2025–2100 to quantify Ir circularity in PEMWE deployment under a backcasting-oriented hydrogen production pathway. The model tracks Ir in anode catalysts only and represents: (i) Ir demand for new capacity additions and replacements, (ii) end-of-life (EoL) outflows governed by a Weibull lifetime distribution, and (iii) closed-loop recycling characterised by an overall recycling rate across collection, separation/pre-processing, and refining. Sensitivity analyses show that long-term primary Ir requirements are governed by the coupled effects of catalyst durability and recycling performance. Although lowering Ir loading reduces initial Ir intensity, accelerated degradation can increase replacement frequency and cumulative primary Ir demand unless recycling rate improves concurrently. Importantly, conclusions depend on the analysis horizon: Ir loading dominates cumulative demand up to 2050, whereas recycling becomes increasingly decisive toward 2100. These results highlight that Ir-thrifting catalyst designs must be evaluated jointly with durability under dynamic operation and recyclability to ensure scalable and continuous PEMWE deployment.

Palm oil mill effluent (POME) represents a major environmental burden in the palm oil industry while offering opportunities for resource recovery. This study develops and applies a value-based assessment framework to examine how technological choice influences the integrated environmental–economic performance of POME valorization. Biomethane production and bio-hydrogen production are selected as representative mature and emerging technologies, respectively. Life-cycle environmental performance is quantified using greenhouse gas (GHG) emissions midpoint indicator and natural resources endpoint indicator, reflecting broader environmental damages. A techno-economic assessment is performed to show the economic performance. In addition to conventional return of investment (ROI), the benefits of mitigating environmental impacts are accounted using the return of value (ROV) methodology. The results indicate that the attractiveness of POME valorization pathways depends strongly on how environmental gains are represented. Midpoint-based valuation shows the climate mitigation potential of the biomethane and bio-hydrogen, whereas endpoint-based valuation reveals higher environmental impacts for the energy-intensive hydrogen route, favoring the mature biomethane option in terms of integrated sustainability performance. This contrast demonstrates that technological maturity is reflected not only in cost structure and process stability but also in the balance between avoided impacts and newly introduced environmental burdens. Overall, the proposed value-based assessment framework enables environmental impacts and economic performance to be evaluated on a common value scale, providing a consistent basis for sustainability-oriented decision-making in POME valorization and supporting strategic planning for wastewater management and bio-economy development in the palm oil sector.

Plastic pyrolysis is widely promoted as a techno-economic industrial scale recycling strategy. Nevertheless, the fate and reactivity of plastic chemical additives during pyrolysis are mostly overlooked in product quality and environmental release assessments. Here, we present an integrated modeling framework to elucidate the role of additives in plastic pyrolysis and evaluate the implications of their transformation products and environmental releases. Using high-density polyethylene (HDPE) as a case study, chemical additives of concern are selected based on occurrence, concentration data, and potential risk to human health and the environment. Bond dissociation energies are predicted using a machine learning model to identify dominant radical species formed under pyrolytic conditions. These additive-derived radicals are incorporated into an automatic chemical reaction mechanism generator that constructs kinetic models composed of elementary chemical reaction steps. These kinetic models are simulated using kinetic Monte Carlo (kMC) methods to predict product distributions and yields. The results show that common additives readily form stabilized alkyl and aryl radicals at energies accessible during pyrolysis, enabling their active participation in polymer degradation pathways. These interactions influence product formation and may contribute to the generation of environmentally relevant by-products. Overall, this study provides a mechanistic and risk-informed perspective on plastic pyrolysis, emphasizing the importance of explicitly accounting for additive chemistry in the development of safer and more sustainable chemical recycling technologies.Disclaimer: The views expressed in this work are those of the authors and do not necessarily represent the views or policies of the EPA.

The case of banana peel as a rich source of pharmaceutically important polyphenols is well-known. Our research focuses on techno-economic and environmental impact assessment of sourcing polyphenols from banana peel. Firstly, phenolic acids and flavonoids, the two major classes of polyphenols, are extracted from dried banana peel powder. The extraction process is carried out with different solvents with varying properties and different extraction techniques with differences in their working mechanism. These extraction protocols were compared on the basis of techno-economics. Distilled water was the most favored solvent in that regard. On the other hand, from the perspective of carbon dioxide emission potential, ethanol stood out as the best. Recovery of extraction solvent was considered and its impact, on techno-economics and emission potential of the protocols was analyzed. The exact nature of the impact was a strong function of the specific protocol and the target polyphenols.

China’s cement industry accounts for over half of global production and contributes 8% of global CO2 emissions, making its decarbonization critical for achieving climate targets. While carbon capture and storage (CCS) and carbon capture and utilization (CCU) are essential deep decarbonization technologies, existing research has not adequately addressed the regional and temporal variations needed for optimal pathway selection across China’s diverse provinces. This study develops a comprehensive whole-systems optimization model to design provincial-scale decarbonization pathways for China’s cement industry from 2025 to 2060. The model reveals significant spatial and temporal heterogeneity in optimal technology combinations. Before 2050, traditional cement processes integrated with CCS (TCP-CCS) represent the dominant bridging technology for low-carbon transition. However, reaching carbon neutrality by 2060 necessitates an eventual shift toward widespread deployment of novel chemical processes combined with green hydrogen (NCP-H2) as the ultimate decarbonization pathway. Notably, regional natural gas prices significantly affect technology feasibility and deployment timing. The optimized transition pathway reduces unit clinker costs by approximately 58% compared to 2025 levels while successfully meeting the 2060 carbon neutrality target. This research framework provides quantitative decision support for policymakers to design cost-effective, province-specific decarbonization strategies that align with local resource endowments and economic conditions, ultimately advancing deep decarbonization across China’s industrial sector.

Sustainable aviation fuel (SAF) is a key pathway for mitigating greenhouse gas emissions in aviation, yet its large-scale deployment is constrained by high energy demand and production costs. Among available conversion routes, the hydroprocessed esters and fatty acids (HEFA) pathway is the most commercially mature, but it requires substantial hydrogen input and high-temperature heat, affecting both economic and environmental performance. This study presents a decision-support framework for SAF production via integrated biorefineries, using rapeseed oil extraction coupled with HEFA conversion as a case study. Detailed process simulations are combined with energy integration and mixed-integer linear programming optimization to enable system-level analysis. Material integration strategies include internal hydrogen generation via glycerol steam reforming and valorization of rapeseed meal through gasification for syngas production. Heat integration considers cross-process heat recovery and the application of high-temperature heat pumps to reduce external energy demand. The results demonstrate that coordinated optimization of heat and material flows significantly improves energy efficiency and reduces reliance on fossil utilities. The framework provides a practical tool for identifying economically and environmentally robust SAF production pathways by integrating feedstock conversion, co-product utilization, and utility optimization. These findings support strategic decision-making for next-generation biorefineries and strengthen circular economy principles in sustainable aviation fuel production.

The ecological transition demands innovative frameworks to reduce industrial resource consumption and environmental impacts. Industrial ecology, particularly through Eco-Industrial Parks (EIPs), provides a promising pathway by enabling exchanges of materials, energy, and water between firms. However, the deployment of EIPs is limited by the lack of standardized industrial profiles and transferable modeling approaches. This study develops a generic framework for representing industrial actors as standardized input–output black-box models, consolidating data on resource consumption, energy demand, by-products, and waste streams. These profiles are structured into a harmonized database to support resource-exchange analysis and scalable optimization across diverse contexts. Complementary mappings of processes and resources, as well as energy and heat demand profiles, enhance the feasibility of identifying synergies such as heat cascading and material reuse. The framework is designed to integrate with optimization approaches to evaluate trade-offs between economic and environmental objectives. The framework highlights the benefits of generalizing industrial profiles for robust and transferable EIP modeling, while also identifying gaps in the integration of social indicators, uncertainty analysis, and governance aspects. This work contributes a methodological foundation for supporting circular, low-carbon industrial systems and provides practical tools for the design and deployment of sustainable industrial zones.

The increasing deployment of lithium-ion batteries (LIBs) requires effective recycling strategies to reduce environmental impacts and dependence on critical raw materials. In this study, a comparative life cycle assessment (LCA) of two LIB recycling routes, a pyrometallurgical process (Pyro) and a hydrometallurgical process with co-precipitation (Hydro), was performed using a Python-based process modeling framework. The LCA was carried out using an attributional approach, with impacts referred to 1 kg of spent LIBs treated at the recycling facility inlet, considering a representative mix of battery formats and cathode chemistries. Results showed that, when normalized per kilogram of treated batteries, the Hydro route is more impactful than the Pyro one, particularly in terms of global warming potential. The Pyro process does not enable direct cathode regeneration but allows the recovery of high-purity metal salts, whereas the Hydro route enables the production of re-formed NMC-111 cathode material that exhibits lower environmental impacts than virgin cathode production. Overall, the study highlights how rigorous process modeling integrated with LCA enables a detailed and non-trivial assessment of LIB recycling sustainability, revealing trade-offs that are not evident from simplified or purely qualitative analyses.

The reliable identification of feasible and optimal operating conditions is a key challenge in the design and optimization of thermochemical conversion processes, where kinetics, limited data availability, and strict physical constraints coexist. In this work, a novel data-driven strategy based on Physics-Informed Neural Networks (PINNs) is proposed to explore the operability space of a bubbling fluidized bed (BFB) plastic pyrolysis process. The approach integrates mechanistic knowledge through explicit mass balance constraints with data-driven learning, enabling accurate prediction of and feasibility boundaries. An adaptive sampling framework is employed to iteratively augment the training dataset. The trained PINN surrogate is then used to predict feasible regions and perform constrained optimization aimed at minimizing tar production, which is one of the most problematic byproducts in plastic pyrolysis processes. Beyond classical optimality, a robustness-oriented uncertainty quantification methodology is introduced, combining local perturbations with feasibility filtering to assess the reliability of the optimal solution. Quantitative robustness metrics, including relative objective uncertainty, feasibility retention ratio, and an overall robustness index, are introduced and used to track convergence and select the most reliable operating point. Results show that the proposed framework efficiently converges to accurate feasibility regions and identifies operating conditions that balance optimal performance with robustness against uncertainty. This work provides a methodology for physics-driven process optimization and operability analysis, with potential applicability to a wide range of complex energy and chemical systems.

Biodiesel production from sustainably cultivated plant sources holds extremely high promise globally [6]. Microalgae have been intensely explored as a next-generation source for transport fuel production [9], as they combine attractive characteristics: rapid growth, high lipid content, and environmental benefits. Nevertheless, technical challenges abound regarding the feedstock potential, cultivation process, and its fatty acid mono-alkyl ester (FAME) properties. Performance evaluations for specific microalgal strains [2, 4, 10, 15] are thus of particular interest. The case of Indonesia is particularly significant due to the country’s large size, population, biodiversity of terrestrial and marine plant species, and the variety of microalgae that can be harvested and used on an industrial scale for biodiesel production, especially in different media and cultivation methods. Many strains, such as Botryococcus braunii, Chlorella sp., Chlamydomonas sp., and Nannochloropsis sp., have a high identified potential [1, 5]. This paper presents the implementation of an established statistical methodology (Principal Component Analysis, PCA) to evaluate the potential of some nitrogen sources for sustainable biodiesel production from Indonesian microalgae strains.

Hydrogen from biomass gasification combined with carbon capture and storage (CCS) can lead to negative emissions and support Europe’s energy transition. This study presents a mixed-integer bilinear optimization model for the cost-optimal design of a Northern European hydrogen supply chain with integrated CCS, focusing on exports from Norway to Germany and CO2 sequestration in Norway. The model is formulated as a superstructure problem and implemented in Pyomo, considering multiple locations for infrastructure nodes and transport options for hydrogen, wood chips, and CO2. The results show that shipping wood chips and CO2 is generally more cost-effective than shipping compressed hydrogen. Supply chain costs range from 35–55 NOK/kg H2, and net-negative emissions (scope 1 and scope 2) are achieved at CO2 capture rates above approximately 30%.

In the urgent effort to reduce greenhouse gas (GHG) emissions in the industrial sector, biogas-derived green hydrogen and power co-generation represents a promising solution. Biogas, a renewable and carbon-neutral resource, provides a flexible feedstock for decentralized energy systems, particularly in regions with well-developed agricultural or waste biomass infrastructure. This approach allows the deployment of cost-efficient systems aligned with climate targets and industrial decarbonization roadmaps. Compared to steam methane reforming (SMR), sorbent-enhanced SMR (SE-SMR) with integrated calcium looping (CaL) CO2 capture reduces process emissions while enhancing hydrogen yield. This study investigates the economic and exergy-based implications of partially splitting hydrogen from a SE-SMR-CaL system producing 50, 000 Nm³/h of H2 from desulfurized biogas. Following heat integration using the PINCH methodology, an electrically self-sufficient base case was established. Economic and exergy analyses were conducted in Excel, with cash flow allocation based on exergy contributions. An automated optimization routine identified competitive levelized costs of electricity (LCOE) and hydrogen (LCOH). Results show that increasing the hydrogen split to power generation maintains nearly constant LCOE (<1% variation) while reducing overall exergy efficiency. Partial hydrogen splitting achieves LCOE below 30.5 €/MWh at the expense of increased LCOH, highlighting a trade-off between electricity and hydrogen economics in flexible, multi-vector systems. The workflow demonstrates that incorporating electricity generation into a sorbent-enhanced hydrogen production system using air combustion is technically and economically feasible.

When designing and planning Green Hydrogen Supply Chains (GHSCs), sustainability considerations are increasingly recognized as essential, particularly in light of decarbonization goals and climate policy targets. Existing research has largely focused on economic and environmental however, social sustainability aspects remain significantly underexplored. This work aims to develop a mathematical programming model to design a GHSC, considering simultaneously economic and social aspects. Solar PV, wind power, and PPA (wind) as energy sources are integrated, while transportation options include the construction of new pipelines, compared to the use of existing highways for trucks carrying liquefied or compressed hydrogen to deliver hydrogen to an oil refinery. The model is applied to a case study conducted in the Brazilian state of Bahia, where different social indicators will be explored, characterizing the case study context while allowing generalization to other contexts. Results allow us to establish a trade-off analysis between economic and social concerns, offering valuable insights for both policymakers and companies, supporting strategies that strike a balance between economic competitiveness and regional social development.

Green hydrogen constitutes a strategic energy vector for achieving the Sustainable Development Goals (SDGs 7, 9, 12, and 13) due to its high energy density, flexibility for renewable energy storage, and direct emission-free operation. However, its production critically depends on the supply of high-purity water, which is unsustainable in the context of a projected 40% global water deficit by 2030. Given that more than 97% of available water is saline, integrating desalination processes with electrolysis constitutes an essential strategy for transitioning toward circular economy models in water resource management. This work presents the conceptual design, detailed modeling, and optimization of an integrated process for the sustainable production of green hydrogen from saline water. The system couples a desalination technology (Solar Distillation) with two electrolysis technologies (AEL and SOEC), modeled through physicochemical, electrochemical, and thermodynamic principles. The objective is to determine technological configurations, materials, and operating conditions that maximize the energy efficiency and minimize the LCOH, contributing to the development of viable routes within the energy transition and circular economy. The results shows that the viability of a integrated Seawater Green Hydrogen system with a detail model is possible. In contrast to black-box models, this model yielded a detailed, geometrically accurate simulation that captures phenomenological effects. It enables the development of processes that contribute to the energy transition, reduce freshwater consumption, and assess their economic viability. The sensitivity analysis revealed promising solutions, with LCOH values ranging from 4.22 to 7.10 USD/kg and overall energy efficiencies between 67% and 82%.

The objective of this paper is to present a single equation format for quantifying the net carbon balance (NCB) in the evaluation of CO2 capture technologies, and to discuss the benefits of this approach. The equation must take into account indirect emissions, especially the contributions from utility generation systems (heating, cooling and electricity), making use of efficiency values and emission factors. The idea is to synthesize, in a single expression, the quantification of the environmental footprint of a technology, in a practical way so that it could be used as an efficient metric in technical evaluation studies, or as objective function/constraint in optimization problems. It also facilitates demonstrating the relationship between capture efficiency and environmental performance, as well as the contribution of each term to total emissions, and to compare different technologies in terms of time, location and available energy sources. To illustrate the application of the proposed NCB formulation, two processes were compared in two different scenarios: post-combustion and pre-combustion capture with MEA absorption, in a Base Case formulated with literature premises and an NCB Zero Case meaning carbon neutrality. A process simulator – Aspen HYSYS – was used to solve the mass and energy balances associated with each process, providing the necessary information for NCB calculation, whose parameters were obtained from the technical literature and were discussed in relation to their importance to the proposed analysis.

Flexible, electrified systems for chemical and energy production are promising alternatives to traditional, hydrocarbon-based processes. Flexible systems have the potential to reduce costs and emissions, but the interconnection between design and operation makes these systems challenging to implement. We use an operation-informed design framework to model a flexible, electrified ammonia synthesis system. We examine the levelized cost and carbon intensity of ammonia in response to different grid emissions (0-420 kg/MWh). We find levelized costs from 700-1200 $/ton-NH3 and observe non-monotonicity in carbon-intensity with respect to grid emissions. We rationalize this trend as a design transition from large, grid-reliant systems to smaller, flexible designs that are grid independent. We then study how synergies in demand and unit-operation flexibility can lower both the price and carbon-intensity of ammonia production. We find that for seasonal, or yearly demand (rather than hourly), a fully flexible Haber Bosch process can achieve 20% lower costs and reduce its carbon-intensity by ~70-100%. We highlight major challenge in decarbonization, but also the importance flexibility plays in reducing carbon-intensity. Together, these analyses demonstrate that flexibility helps electrified industrial systems achieve financial and environmental goals.

Behind-the-meter generation from variable renewable energy is a potential pathway for new data centers to obtain power more quickly and more sustainably than interconnecting to existing electrical grids. Energy storage is needed to accommodate the variability of wind and solar energy across multiple timescales. Hydrogen from electrolysis and ammonia made from this hydrogen can be used as fuel for dispatchable power generation while offering lower $/MWh storage costs than batteries. In this work, we analyze the economics of using hydrogen, and/or ammonia along with batteries in hybrid energy storage systems to enable data centers to be powered by 100% renewables. We perform this analysis using an optimization model for the selection, sizing, and coordinated hourly operation of constituent energy storage technologies toward minimizing the levelized cost of energy (LCOE). The model uses an hourly resolution scheduling horizon of five years to account for hourly, seasonal, and interannual renewable variability. We use this model to specifically analyze the economics of to 5 MW, 50 MW, and 500 MW sustainable data center energy systems in three relevant American locations. We find that including ammonia energy storage in hybrid systems with batteries and hydrogen enables lower LCOE across all scales and locations by reducing energy storage capital investments despite deploying an order-of-magnitude more storage capacity while also enabling less investment in renewable generation. Ammonia is used for seasonal and interannual energy storage and its inclusion in these hybrid storage systems is most advantageous for larger-scale data centers in locations with high wind potential.

Direct Air Capture (DAC) is a negative emissions technology whose performance is inherently linked to ambient conditions, which directly affect its primary feed stream (air). A common simplification in DAC model simulations is the use of fixed weather conditions, which can bias the predicted performance under weather variability. In response, this study quantifies the impact of local meteorological variability and temporal weather aggregation on the performance of DAC units. Building on a previously developed and validated 1D mechanistic model of a fixed-bed Steam-assisted Temperature Vacuum Swing Adsorption (S-TVSA) DAC process, we simulate its operation using weather data from the Met Office station at Buchan (UK), near the Saint Fergus terminal – a strategic hub for Carbon Capture and Storage (CCS) activities in Scotland. A two-branch methodological framework is developed combining optimization and forward simulations. Operating conditions are optimized using a multi-objective genetic algorithm (NSGA-II) to maximize productivity (Pr) and minimize specific equivalent work (Weq) at two temporal resolutions. Furthermore, daily weather inputs are aggregated on a monthly and yearly scale to assess the impact of data resolution on model predictions and real operational gains. Results show that temporal weather aggregation to yearly averages biases DAC key performance indicators, overestimating Pr by up to 5% while underestimating Weq by up to 31%, relative to performance based on daily weather variations. Moreover, optimization strategies that explicitly account for monthly weather variability present monthly gains, by increasing Pr by up to 10%. Yet, these monthly gains do not necessarily translate into significant operational performance benefits at the annual scale when daily weather data is propagated in the process model.

This study presents the modeling and operation optimization of an adsorption heat storage to improve the supply of renewable heat to a house. The system configuration is an open system with water being carried by an air flow and adsorbed on zeolite 13X beads in a packed bed. A numerical model is developed based on mass and energy balances, using a Langmuir adsorption isotherm and a Linear Driving Force (LDF) mass transfer equation. The model is implemented in Pyomo and solved with the NLP solver IPOPT. A sensitivity analysis on the discretization parameters is performed to choose a good compromise between accuracy and computational time. The chosen model is then validated against experimental data from the literature, with a mean absolute percentage error less than 5%. The dynamic optimization of the operation of the system to satisfy a heat demand is then performed. The trajectory for the inlet fluid velocity is optimized in several heat demand scenarios. The results show that this numerical framework is able to follow realistic heat demands for a house even in the case of large peaks corresponding to domestic hot water production, with quadratic errors between the demand and supplied power less than 1%.

The present work explores computationally the potential of hybrid operation of a direct reduction shaft furnace with a feed gas mixture of methane and hydrogen, considering the operation of the associated gas handling and conversion units (reformers, condensers, heat exchangers, compressors, heaters, etc.). The state of the system is dependent on both the performance of the shaft furnace, which is sensitive to the feed gas composition and quantity, and the reformer, which converts the methane and the recycled top gas from the shaft furnace into feed gas. This coupling gives rise to nonlinear behavior of the system with respect to changes in the boundary conditions, which justifies a model-based approach for holistic analysis. The system is modelled in Aspen Plus to simulate the operation of an industrial-scale plant using a detailed model of the process units. Three configurations of the system are evaluated: (i) co-feeding of hydrogen with methane, (ii) feeding hydrogen directly to the shaft furnace, and (iii) splitting the hydrogen feed between the two injection points. Further, the effect of H2 share in the feed was evaluated in terms of energy consumption, CO2 emissions and energy cost with low- and relatively high-carbon electricity. The results show that increasing the share of hydrogen in the fresh feed leads to lower CO2 emissions, considering low-carbon electricity.

The reduction of CO2-emissions in the chemical industry is essential to meet European climate targets. Particularly, the reliance on fossil fuels for process heat supply is a key factor for CO2-emissions. Electrically driven compression heat pumps are a promising option to reduce fossil fuel consumption by upgrading low-temperature waste heat to a higher temperature level, provided that low-carbon electricity is available. However, the integration of heat pumps into chemical utility systems remains a challenge due to economic constraints and the high complexity associated with site-wide heat integration and retrofit of existing structures. This work presents a mixed-integer linear programming (MILP) approach for the optimization of utility systems with integrated heat pumps. To address computational complexity, candidate utility temperature levels are pre-selected, and feasible heat pump coefficients of performance (COP) are precomputed. The framework is applied to both greenfield and retrofit scenarios for a synthetic case study consisting of 400 process streams. In the greenfield scenario, optimal utility temperature levels and heat pump integration configurations are identified. For the retrofit scenario, temperature levels of an existing utility system are modified to reduce total annual costs (TAC). Additionally, sensitivity analysis is conducted to assess the influence of key economic and environmental parameters. The presented case studies demonstrate short solution times, highlighting the suitability of the proposed framework for screening studies and systematic sensitivity analyses in early-stage design and retrofit applications.

Cryogenic fuel tanks used in ships are continuously subjected to heat ingress and motions which affect the thermal behavior of the fluid inside the tanks. In this study, the ullage space of a liquid hydrogen (LH2) tank subjected to heat ingress and periodic rolling motion is analyzed using Computational Fluid Dynamics (CFD). A two-dimensional transient model utilizing the dynamic mesh approach is created to represent the ullage space of a partially filled LH2 tank. Three cases are studied where the properties of the thermal insulation of the tank model are varied, resulting in a different heat ingress for each case. A low-frequency motion is applied to the model domain, which induces mixing of temperature layers and cooling due to vapor contact with wetted walls. After 60 s of tank motion, most mixing is observed in the case with the smallest heat ingress, whereas in the cases with larger heat ingress and, consequently, larger thermal and density gradients, separation into a warmer and colder vapor region can be observed. The model developed in this work can be used to gain insight into how liquid hydrogen tanks used on ships are affected by heat ingress and ship motions.

Heat pumps offer the possibility of reducing CO2-emissions in the chemical industry. However, the integration of heat pumps, especially in non-continuous processes, faces several challenges. Energy storage facilitates a way to enhance heat integration by providing a continuous supply of heat flows. By doing so, the question arises as to whether this implementation should be applied to the process or to the utility level. At the process level, there is usually more freedom, as one is not bound by the existing temperature levels of the utility system, which are mostly difficult to retrofit. Therefore, this study presents an approach that generates heat integration concepts at the process level based on two different criteria. These criteria influence which process streams are grouped for a storage implementation and therefore influence the heat integration. The aim is to maintain the heat flows as continuous as possible by integrated heat storages. Finally, the possible heat integration concept is evaluated in terms of energy efficiency by a know method for continuous process streams, here the pinch analysis.

Refineries are major sources of direct CO2 emissions, primarily from steam generation, fluid catalytic cracking, and hydrogen production. This study develops a superstructure optimization framework to evaluate the economic and environmental viability of retrofitting existing refineries with small modular nuclear reactors (SMRs) for cogeneration of heat and electricity. A multi-period mixed-integer quadratically constrained program is formulated, simultaneously minimizing the present cost of retrofitting and CO2 emissions over the time horizon. This problem is solved to generate a Pareto frontier via the e-constraint method. Two cases are analyzed for a medium-scale refinery, considering 1) inflexible operation under average annual electricity prices and 2) flexible operation under hourly prices with the possibility of installation of storage devices. Compared to a benchmark without SMRs in the superstructure, allowing their installation leads to reduced costs at lower or comparable emission levels. The results show that SMRs are primarily used for high-pressure steam generation. Flexible operation and the inclusion of thermal energy storage further reduce costs. Overall, SMRs appear in multiple non-dominated solutions, highlighting their potential as a cost-effective refinery decarbonization strategy.

The transition toward low-carbon ammonia production increasingly relies on highly efficient routes for renewable hydrogen generation, with Solid Oxide Electrolysis (SOE) representing a particularly promising solution. SOEs, operating with steam at elevated temperatures, offer intrinsic thermodynamic advantages and are attractive when integrated with processes that can effectively utilize or supply high-grade heat. This context opens the possibility for advanced energy coupling between hydrogen production and the exothermic ammonia synthesis process. This work investigates such an energy integration strategy by simulating a small-scale ammonia plant where hydrogen is produced through an SOE system thermally coupled to the ammonia synthesis loop. Specifically, high-grade heat available in the Haber-Bosch (HB) reactor outlet is recovered via a “Heat Recovery Steam Generator” (HRSG) to provide a substantial fraction (55 wt%) of the steam required by the electrolyzer. The assessment demonstrates that the integrated configuration significantly reduces the external electrical demand for steam generation by 84%, achieving a substantial enhancement in system-level electrical efficiency. Total plant energy efficiency increases from 57% to 62%, and the electricity-to-hydrogen ratio decreases from 45.9 to 42.1 kWh/kg. These improvements translate directly into a decrease in the Levelized Cost of Ammonia (LCOA) by 4-5%, underscoring the relevance of high-temperature electrolysis supported by effective plantwide heat recovery.

This study aims to compare different pathways for achieving CO2 emission reductions during the production of methanol and, subsequently, formaldehyde, i.e., its major derivative. An equation-oriented model of the formaldehyde sector is developed, incorporating a superstructure of various transition pathways including feedstock switching (biomass, biogas, waste), process electrification (Power-to-X), and CO2 capture. The OSMOSE tool is used to evaluate the superstructure and compare the alternative production pathways on the basis of thermodynamic, environmental, and economic key performance indicators for future scenarios (2025 and 2050). Furthermore, to cope with the limitations of predefined pricing scenarios, a parameter sweep is performed, exploring a broader set of economic conditions and seeking to identify the zones of economic optimality associated with each configuration through the solving of a Mixed-Integer Linear Programming cost minimization problem, while generalizing the analysis results beyond a specific year or geographic context. Finally, the potential of industrial symbiosis implementation, between clinker and formaldehyde production facilities, as a cost-reduction measure for Power-to-X processes is assessed.

“Green” hydrogen production via polymer electrolyte membrane (PEM) electrolyzers must overcome significant energy penalties and high costs to become competitive in renewables-based energy systems. Adaptive operating strategies for PEM electrolyzers—by dynamically adjusting current density, pressure, and temperature—have demonstrated efficiency improvements in simple energy systems. However, their effectiveness in the context of complex power-to-X energy systems featuring variable downstream synthesis processes remains unclear. This work shows that integrated optimization of PEM electrolyzer operating parameters in conjunction with downstream methanation processes (MP) delivers substantial system-wide efficiency and cost benefits under dynamic hydrogen demand and pressure conditions. To demonstrate this, an equation-oriented process model of a PEM electrolysis system is embedded within a higher-level energy system model to compare sequential optimization (where the electrolyzer adapts to predetermined MP operating decisions) against integrated optimization (where electrolyzer and MP operating decisions are determined simultaneously). Sequential optimization delivers 7.3% operating cost savings compared to conventional fixed-parameter operation. In comparison, integrated optimization achieves 9.9% cost reductions and 7.8% decreases in electrolysis system electricity consumption. Operating pressure emerges as the most critical parameter, with electrolyzer electricity consumption exhibiting markedly higher sensitivity than MP electricity consumption. The analysis reveals fundamental trade-offs between component-level and system-level efficiency, demonstrating that prioritizing electrolyzer efficiency optimization yields superior system-wide performance. These findings establish that system-wide coordination through integrated optimization approaches substantially enhances the economic viability of “green” hydrogen supply in complex power-to-X energy systems.

The aviation sector represents one of the most pressing challenges in the energy transition due to its strong reliance on energy-dense liquid fuels and established fuel infrastructure. Sustainable Aviation Fuel (SAF), particularly from agricultural residues, offers a near-term mitigation pathway; however, large-scale adoption is shaped by policy mandates, infrastructure expansion, market price formation, and passenger demand responses. These coupled dynamics are difficult to capture using aggregate or equilibrium-based models. This study develops an agent-based model to analyze SAF transition pathways and applies it to India’s civil aviation system. Results show that SAF adoption emerges from the coordination between infrastructure entry, cost learning, and market responses rather than mandate ambition alone. Even moderate mandates fall short of intended adoption levels without timely infrastructure expansion, while aggressive mandates become infeasible under binding supply and price constraints. Passenger demand feedbacks further influence outcomes by linking fuel cost increases to airline operations and route-level allocation decisions. The findings highlight infrastructure coordination and price formation as critical leverage points for aviation decarbonization and demonstrate the value of agent-based models for evaluating realistic SAF transition strategies under policy and market uncertainty.

Recent geopolitical disruptions and extreme weather events have underscored the importance of resilience in global energy supply chains, particularly for import-dependent economies pursuing ambitious energy transition targets. These events have exposed the limitations of supply chain designs focused solely on cost minimization that lack the flexibility and redundancy required for secure operation under stress. As energy systems evolve toward higher shares of variable renewable energy and increased demand uncertainty, episodic manual re-planning becomes inadequate, highlighting the need for modeling frameworks that integrate predictive modeling, optimization, and control to enable intelligent and adaptive supply-chain design and operations under uncertainty. This work presents a comprehensive data-driven modeling and optimization framework for adaptive energy supply-chain networks under evolving demand. The framework integrates three layers: (i) a machine-learning model for demand forecasting and scenario generation; (ii) a multi-period stochastic optimization model for strategic network design and operations; and a (iii) learning layer that monitors performance metrics and triggers strategic recourse when demand patterns shift significantly. The operational stage, which acts as a learning layer, is posed as a rolling horizon control problem determining necessary recourse decisions to adapt to changes in demand patterns. An illustrative case study, encompassing multiple energy generation hubs, energy carriers and transportation modes is shown to demonstrate the applicability of the framework.

The rapid expansion of AI data centers is straining electricity grids alarmingly, forcing data center planners to navigate two-pronged challenges: (1) lengthy interconnection queue delays undermining immediate grid access, and (2) volatile electricity prices that spike dramatically during high demand events. This convergence forces planners to reconsider traditional grid-only strategies. While behind-the-meter (BTM) generation offers a solution, existing research lacks comprehensive frameworks for identifying technology portfolios under combined uncertainties of grid access delays and market volatility. This study develops a two-stage stochastic optimization framework with binary capacity constraints co-optimizing data center location and BTM energy portfolios under these challenges. The model evaluates conventional (gas turbines), renewable (solar, wind, batteries), and emerging technologies (hydrogen fuel cells, small modular reactors) across four progressive scenarios spanning emission targets, demand flexibility, grid curtailment, land constraints, and queue delays, contrasting stochastic and deterministic solutions. Applied to a 5 GW data center expansion in ERCOT, three insights emerge: First, queue delays drive 2.7 GW bridging investments that transition to 92% grid reliance once interconnection is available, while stochastic optimization maintains higher BTM utilization against price volatility. Second, demand flexibility reduces required BTM capacity through load shifting during grid curtailment events, decreasing gas deployment from 2.1 to 1.7 GW. Third, land-constrained decarbonization shifts from wind-dominated portfolios to capital-intensive solar-hydrogen-SMR solutions, with stochastic optimization tripling SMR deployment to prioritize reliability under uncertainty.

Heavy-duty road transport remains a challenging sector to decarbonize, as full electrification of long-distance trucking is currently constrained by limitations in energy density and charging infrastructure. Alternative fuels such as hydrogen, biodiesel, and e-fuels are thus gaining increasing attention. In parallel, the cement industry is a major source of unavoidable, process-related CO2 emissions, offering an opportunity to use captured industrial CO2 as a feedstock for e-fuel production. This study evaluates the production of e-methanol and Fischer-Tropsch (FT) diesel from captured CO2 at an Austrian cement plant as a base case. Several system configurations are analyzed, including different electricity supply options across Europe and the use of biogenic versus fossil CO2. An integrated framework combining process simulation, techno-economic analysis, and life-cycle assessment is applied to compare both fuel pathways. Results show that the climate impact of e-fuels is highly dependent on the electricity mix. When non-renewable electricity is used, climate impacts are 100-440% higher than those of fossil diesel. In contrast, a wind-based Austrian scenario achieves the lowest impact, corresponding to a 44-50% reduction compared to fossil diesel. Overall, cement-plant-based power-to-liquid concepts offer limited near- to mid-term mitigation potential under current European energy conditions and deliver climate benefits primarily when based on biogenic CO2, high process efficiencies, and low-carbon electricity. The study further highlights key differences between the fuels, with FT-diesel being a certified drop-in fuel, while e-methanol still requires technical and regulatory validation, underscoring the need for technology-specific and system-level assessments.

Early-stage system-level assessment of emerging technologies is essential for achieving climate neutrality and a circular economy; however, such assessments are often constrained by the lack of representative life cycle inventory data. In thermal energy systems, performance strongly depends on scale, making direct application of laboratory- or bench-scale experimental data potentially misleading in life cycle assessment (LCA). This study investigates the influence of experimental scale on system-level evaluation using a zeolite-based thermal energy storage (TES) system as a case study.LCAs were conducted using performance data from laboratory-, bench-, and pilot-scale experiments and compared with predicted commercial-scale performance derived from numerical simulations. The TES system stores waste heat via water vapor desorption from zeolite and generates pressurized steam using a moving-bed with indirect heat exchanging system. Heat recovery ratios of 36%, 50%, and 61% were obtained at laboratory, bench, and pilot scales, respectively, while commercial-scale performance was predicted to reach approximately 80% due to reduced heat losses. Accordingly, LCAs based on small-scale data indicated higher greenhouse gas (GHG) emissions than conventional technologies, whereas assessments using commercial-scale data demonstrated clear GHG reduction potential.The results highlight the risk of erroneous conclusions when small-scale data are directly used for system evaluation. By integrating experimentally validated numerical modeling with computer-aided process engineering, scale-dependent data can be transformed into system-representative inputs, enabling fair and consistent assessment of emerging technologies.

European industry accounts for approximately 20% of total European CO2 emissions, with heat demand representing one of the largest energy consumers. Solar thermal collectors offer an efficient renewable alternative to fossil fuels to cover the heat demand. However, due to the temporal mismatch between the solar thermal generation and process heat generation a thermal storage is needed to maximize the renewable utilization. This article presents a novel optimization framework for integrating an ideal heat storage with solar thermal systems in multiperiod heat exchanger network synthesis. We derive an analytical approach to optimize the heat storage by using physical insights from Pinch Analysis: heat can only charge the storage below the lowest pinch point in a given period and discharge above the highest pinch point. We show both how to do it for a storage of infinite size and of finite size, and that the infinite size storage is much more efficient to solve. The approach is validated using real industrial data from a lubricant plant and optimizing the implementation of solar thermal at the plant. By increasing the size of the storage more of the solar thermal heat is utilized, however, the increase in utilization becomes smaller as the storage becomes larger. While 100% solar thermal heat utilisation requires a storage size of 1040 kWh, 90% utilisation can be reached with a storage of 250 kWh, compared to only 70% without a storage.

Human activities have already transgressed several planetary boundaries, yet energy system models remain largely focused on greenhouse gas mitigation, reflecting their original purpose of addressing climate change. Recent integrations of Planetary boundary-based Life Cycle Assessment into Energy System Optimisation Models show that even cost-optimal low-carbon pathways systematically violate multiple planetary boundaries, indicating that supply-side decarbonisation alone is insufficient for absolute environmental sustainability. At a 2050 horizon, where energy supply is largely decarbonised and technologies are assumed mature, further impact reductions through techno-economic optimisation become limited, positioning final energy demand as a key remaining lever for restoring feasibility under planetary constraints. To address this gap, we ex-tend an Energy System Optimisation framework coupled with a Planetary Boundary framework by explicitly treating final energy demand as a decision variable and exploring energy sufficiency configurations within a multi-objective formulation minimising system cost and environmental pressures. The national-scale EnergyScopeTD model is coupled with the RHEIA uncertainty analysis framework to propagate uncertainty in final energy demands and to identify demand-side drivers through global sensitivity analysis. Applied to Belgium, the results show that individual mobility demand is the dominant driver of environmental pressures across multiple planetary boundaries, with consistently higher influence than other end-use demands. While large-scale vehicle fleet electrification substantially reduces climate impacts, it simultaneously shifts pressures toward other environmental dimensions, indicating that technological substitution alone, even combined with moderate sufficiency, is insufficient to achieve absolute sustainability. Overall, the results confirm that energy sufficiency, particularly in mobility, is necessary for operating energy systems within planetary limits, and that anchoring system design in absolute sustainability thresholds provides a robust basis for prioritising demand-side mitigation strategies.

The increasing deployment of variable renewable energy (VRE) is essential for achieving a sustainable society; however, its inherent variability poses challenges for maintaining a stable electricity supply. Vehicle-to-grid (V2G) technology enables bidirectional electricity exchange between electric vehicles (EVs) and the power grid and can enhance the utilization of renewable electricity by charging EVs during periods of VRE output curtailment. This study developed a regional V2G system model and evaluated its potential through energy flow simulations and life cycle assessment (LCA). The model explicitly considered hourly operation schedules of individual EVs, the spatial distribution of V2G infrastructure, and minimum output constraints of thermal power generation. The number of EVs is assumed to increase to up to 10, 000 units. In the energy flow simulations, EV charging and discharging were calculated on an hourly basis over one year. LCA was conducted to assess greenhouse gas (GHG) emissions of regional V2G system. Fuel consumption, VRE utilization, and GHG emissions were quantitatively evaluated. The results demonstrated that regional V2G systems could reduce VRE output curtailment and GHG emissions and, under certain conditions, could also reduce thermal power generation. The developed model enabled a quantitative evaluation of the potential of regional V2G systems. For practical deployment, integrated regional platforms would be required to coordinate electricity flows and to manage financial transactions. By assessing the potential of regional V2G systems, this study would provide valuable insights to support the design and implementation of environmentally sustainable V2G systems.

Cost reduction of green ammonia production is critical to advancing the hydrogen-ammonia economy, as ammonia capable of cost-effective storage and transportation is a promising hydrogen carrier and energy carrier to alleviate the intermittency and geographical limitations of renewables. Optimisation and techno-economic assessment based on rigorous model are essential to accurately investigate techno-economic feasibility and fully explore optimisation potential. This work estimates the levelized cost of ammonia (LCOA) of an integrated system including a hydrogen generation process employing Proton Exchange Membrane (PEM) water electrolysis, a nitrogen generation process from flue gas recovery, and an ammonia synthesis process based on Haber-Bosch Process. To enhance the reliability of LCOA, detailed equipment sizing and costing is conducted according to stream data from rigorous modelling in Aspen Plus. A novel optimisation strategy is proposed to enhance the computational robustness by sequentially largening the group of optimised variables and improving the quality of initial points. Homotopy-continuation (HC) method is employed to ease the convergence difficulties for solving the flowsheet with multiple recycle streams and scaling up the plant capacities between 100 tNH3 d–1 and 600 tNH3 d–1 from an initially feasible point. Heat integration is also considered to realise the combined supply of heat and power by introducing a steam turbine power cycle and utilising the waste heat from exothermic ammonia synthesis.

The unprecedented expansion of Artificial Intelligence is adding increasing electricity demand to Europe’s power system. While incumbent plans pursue a net-zero future by 2050, they fail to consider the implications of large-scale AI-based data centres. In this study, a spatially explicit optimisation model is developed to assess how hyperscale data centres may reshape energy infrastructure investment, and emissions trajectories, across different AI demand growth scenarios. The results indicate that, after 2030, AI capacity deployment increasingly shifts toward regions with the ability to expand nuclear and gas-based generation, as firm and flexible power sources are essential for supporting the deployment of high-capacity AI data centres. By 2050, AI-driven electricity demand under high growth scenarios may reach up to 450 TWh, corresponding to 7% of total Europe’s demand, with installed AI capacity reaching approximately 85 GW. This additional load leads to an increase of nearly 25 MtCO2 in cumulative emissions between 2030 and 2050. Our analysis indicates that, depending on the AI growth scenario, meeting AI-related electricity demand by 2050 requires between 37 and 323 GW of additional capacity across Europe, ranging from the pessimistic to the lift-off scenario, including investments in nuclear (2-12 GW), gas (2-7 GW), wind (13-100 GW), solar (20-134 GW), and battery storage (0-70 GW).

Biogas offers a promising biogenic carbon source for renewable methanol, but differences in CH4/CO2 ratio across feedstocks and possible upstream CO2 handling can shift syngas stoichiometry away from the methanol synthesis target range. This work quantifies how biogas composition and reformer operation influence the stoichiometric number (SN) and the associated conditioning requirement needed to meet methanol synthesis targets. A steady-state Aspen Plus® model of an integrated biogas-to-methanol process is used as the analysis framework. A base-case operating point is defined, followed by parametric evaluation of biogas CH4/CO2 ratio, reformer temperature, reformer pressure and steam-to-methane (S/C) ratio. The studied CH4/CO2 ratio range covers CO2-rich to CH4-rich cases that may occur across sites and upgrading levels. The resulting SN shifts are tracked and converted into a quantitative correction requirement to maintain the methanol design target (SN = 2.01). Temperature determines the upper limits of conversion and SN, while pressure and S/C ratio have secondary effects once high-temperature operation is reached. By comparison, the CH4/CO2 ratio has the strongest influence on syngas composition, defining H2-deficiency and H2-excess regimes. Methanol production increases as SN approaches the required target range but shows diminishing gains beyond it, indicating limited benefit from excess hydrogen under fixed synthesis conditions.

The aviation sector is difficult to decarbonize due to limits on aircraft electrification, making sustainable aviation fuel (SAF) a critical near-term solution. This study integrates Aspen-based process modeling with game-theoretic optimization to design a multi-agent SAF production network comprising coal gasification and CO2-assisted natural gas reforming for syngas production, and Fischer–Tropsch (FT) synthesis for SAF production. Techno-economic parameters from Aspen simulations inform an agent-based model in which agents maximize their net present value subject to capacity and demand constraints. Three decision-making frameworks are compared: (i) social welfare optimization, (ii) cooperative bargaining – symmetric (equal bargaining power) and asymmetric (bargaining power weighted by agents’ competitiveness outside cooperation), and (iii) competitive equilibria modeled as generalized Nash equilibrium. The results show that social welfare maximization excludes coal and yields the highest total profit, strongly favoring the FT agent’s pay-off. Cooperative bargaining includes all agents in the system and promotes CO2 and water recycling with only a 3.5% profit reduction; symmetric bargaining shifts profit toward coal, while asymmetric bargaining partially restores FT agents’ share. Under bargaining, complete CO2 recycling is achieved, integrating coal without direct emissions. Under strict competition, upstream syngas producers do not participate in the network, and reliance on external syngas reduces overall profit by 12.4%. These results show that gametheoretic modeling for process synthesis reveals strategic incentives for process design that are typically obscured in more traditional superstructure-based optimization frameworks.

The development of energy technologies with low CO2 emissions is increasingly important for achieving the United Nations Sustainable Development Goals. In this context, power plants based on the Allam-Fetvedt Cycle appear promising because this cycle (introduced in 2012) features inherent CO2 capture. In this context, its association with biogas as fuel enables its application as a Bioenergy with Carbon Capture and Storage (BECCS) system. This study evaluates the technical, energetic, and economic feasibility of an Allam-Fetvedt Cycle power plant fueled by biogas. The methodology is based on detailed process simulations performed using Aspen Plus® v14. Two biogas production scenarios were assessed: Case 1 and Case 2, corresponding to the processing of 8 and 24 million tons of sugarcane per year, respectively. The economic analysis indicated a high capital investment, primarily in the Air Separation Unit (ASU) and the Balance of Plant (BOP). Nevertheless, a significant reduction in the Levelized Cost of Energy (LCOE) was observed, decreasing from 1443 R$/MWh (€0.227/kWh) in Case 1 to 848 R$/MWh (€0.134/kWh) in Case 2. Overall, the results suggest that integrating biogas into the Allam-Fetvedt Cycle represents a promising technological pathway for power generation with negative carbon emissions. Further research and optimization are required to improve economic viability and strengthen the role of this technology as a strategic option for decarbonizing the energy sector.

Trinidad and Tobago emits 11.3 metric tonnes of CO2-eq per capita per year, making it one of the highest per capita per year greenhouse gas (GHG) emitters globally. An estimated 87% of these emissions are linked to the industrial sector, including power generation. This study aims to reduce the national environmental impact of the power generation and waste disposal sectors, with the goal of reducing the country’s reliance on natural gas and promoting sustainable power generation via waste-to-energy (WTE) pathways. This work seeks to implement a mixed-integer linear programming (MILP) framework, along with techno-economic analysis (TEA) and life-cycle assessment (LCA) to determine potential sustainable objectives with respect to T&T’s power system. The study implements supply chain optimisation considering single objective optimisation (SOO) with life-cycle (LC) endpoint externalities included. The key constraint of the system was the production of sufficient electricity to sustain the national power grid. A WTE model was proposed in which incineration, gasification and anaerobic digestion (AD) were implemented to supplement the country’s natural gas based power generation system. When considering LC endpoint externalities, inclusion of WTE along with improved natural gas power generation efficiency resulted in a maximum reduction of 31% in total annualised life-cycle cost (TALC) and 38% in total annualised GHG emissions (TAGE). Thus, indicating the potential of WTE as a key factor in achieving sustainable development goals in SIDS. This work provides a roadmap for the implementation of sustainable power generation practices for policymakers.

Power-to-X (PtX) technologies play a central role in renewables-based energy systems by enabling the conversion of renewable electricity into multiple energy carriers. However, due to the multiple energy conversion stages inherent to such energy systems, they often suffer low system efficiencies and high operational costs. In this context, waste product utilization offers significant potential for improving system performance. Directly integrating waste product utilization into energy system operational problems, however, is computationally challenging, as it requires high model granularity to capture waste product characteristics and introduces additional complex constraints.This work proposes a method to integrate waste heat utilization into operational optimization problems, aiming to improve the overall performance of PtX energy systems. Detailed process models, together with pinch analysis, are used to generate surrogate models for the thermal (by-)products and their associated temperature levels. The resulting optimization problem is decomposed into multiple subproblems. To balance problem complexity and solution optimality, two alternative approaches, “ex-post” and “co-optimized” are developed, differing in whether waste heat utilization is addressed ex post or jointly optimized with the energy system operation.The proposed method is applied to optimize the daily operation of a PtX energy system. Compared with a reference case without waste heat utilization, the ex-post approach achieves a cost reduction of 7% within identical computational time (0.6h). The co-optimized approach yields cost savings of up to 22% with a computational time of 3.4h. These results demonstrate the effectiveness of the proposed method for integrating waste heat utilization into energy system operational optimization.

Achieving net climate neutrality will likely require negative-emission technologies such as Direct Air Capture (DAC). Potassium hydroxide (KOH) absorption is one of the most mature DAC approaches, but it can cause significant emissions due to natural-gas-based thermal regeneration. Electrochemical regeneration methods, such as electrolysis and electrodialysis, have recently been proposed as alternatives, yet their relative performance and environmental impacts remain unclear. We present a comparative cradle-to-gate life cycle assessment (LCA) of three KOH-based DAC configurations: (i) the established Ca-looping thermal regeneration, (ii) the electrolysis regeneration (DAC-ELY), which co-produces hydrogen, and (iii) the electrodialysis regeneration (DAC-ED). The results show that, expectedly, electricity demand dominates life cycle impacts across all configurations. With the current German electricity mix, the established DAC has the lowest overall impacts, while DAC-ELY and DAC-ED exhibit higher global warming impacts due to the high electricity requirements of electrochemical regeneration. Infrastructure impacts are more noticeable for DAC-ED due to its low current density and large membrane areas. This study highlights that electrification alone does not guarantee lower life cycle impacts and that the sustainability of electrochemical DAC depends on process efficiency, material use, and the expected use of renewable electricity.

CO2-to-methanol process is an attractive option to simultaneously reducing the anthropogenic CO2 while producing value-added chemicals. In this work, two distinct CO2-to-methanol process routes specifically, single step and dual step are evaluated based on their economic and environmental performance. First, a multiobjective optimization (MOO) framework is formulated to develop the optimal process configurations. Three conflicting objectives including methanol production rate, total annual cost (TAC) and carbon intensity of methanol are considered. For this MOO, the elitist non-dominated sorting genetic algorithm (NSGA-II) is employed to get the Pareto front. From the Pareto front, a balanced compromise solution is identified by the technique for order of preference by similarity to ideal solution (TOPSIS) with entropy information as weighting criteria. Then, the comparative performance analysis is conducted across the Pareto front. At the TOPSIS-selected configuration, the single step process offers economic advantages (1.97% lower levelised cost of methanol (LCoM)), but performs poorly on environmental perspective (3.03% higher carbon intensity of methanol). In contrast, at the cost-optimal point the dual step process is marginally favourable in both aspects (0.55% less LCoM and 1.26% reduction in carbon intensity of methanol). Overall, both processes exhibit comparable economic and environmental outcomes, with the dual step showing better environmental performance and reduced variability across the Pareto space.

Decarbonization of hydrogen-intensive industrial clusters is essential to meet the European Union’s net-zero targets. Although hydrogen can replace fossil-based feedstocks and fuels in refineries and chemical industries, its production remains largely dependent on natural gas. Therefore, cost-effective and low-emission supply routes require a system-level approach that integrates regional resources, technologies, and industrial demand. This study applies a multi-objective optimization framework to design a low-carbon hydrogen supply system for Galicia (northwestern Spain), addressing two gaps in regional energy system modeling: model transferability across regions and integration of social criteria beyond techno-economic assessment. The model quantifies trade-offs between total system cost and greenhouse gas emissions, and an employment indicator is integrated via post-processing using TOPSIS. The results show that meeting 100% of the projected 2030 demand (105 kt H2/a) yields a single feasibility-limited solution dominated by biomass gasification (69% of hydrogen production). When the demand was reduced to 60% coverage (63 kt H2/a), 28 non-dominated solutions were obtained. The top-ranked compromise combines 55% PEM electrolysis and 45% biomass gasification, supported by a wind-dominated electricity mix, under equal weighting of economic, environmental, and social criteria. When either emissions or employment is prioritized, this compromise remains top-ranked, while the cost prioritization shifts preference to a solution with hydrogen production from PEM (51%) and alkaline (3%) electrolysis and biomass gasification (46%).

Industrial decarbonization has heightened interest in electrifying major chemical processes, but existing planning methods typically assume fixed electricity prices and overlook how industrial power use affects the grid. This work introduces a grid-aware optimization framework that captures two-way interactions between industrial electricity usage and the power flows within the grid. We use the DC Optimal Power Flow (DC-OPF) model to generate Locational Marginal Prices across refinery demand levels and embed a surrogate reflecting the relationship between the power demand and the prices into an operational optimization problem for a partially electrified refinery. The surrogate model is embedded within the optimization problem using disjunctive reformulations and off-the-shelf packages such as OMLT (Optimization and Machine Learning Toolkit). In a case study considering an oil refinery with installed electric boilers, electrolyzers, H2 storage, and post-combustion carbon capture infrastructure, the grid-aware approach lowers operating costs by 7% relative to a price-taker model by anticipating how the refinery’s own demand shifts electricity prices. The method is also shown to incorporate the effect of demand uncertainty at other grid nodes by embedding a surrogate model trained using data generated by a chance-constrained DC Optimal Power Flow.

The manufacturing industry is the second largest emitter of CO2, with the cement industry being one of the main contributors (7-8 % of the global emissions). Carbon capture and utilization (CCU) technologies are promising decarbonization solutions for the cement industry, addressing both fossil fuel-related (40 %) and process-derived emissions (60 %). Within a cement plant, producing synthetic natural gas (SNG) from captured CO2 is particularly suitable, as it is sufficient to fully replace solid fuels in the rotary kiln. On the other hand, the use of zero-carbon fuels, such as green ammonia, is also recognized as a promising approach for decarbonization. In this work, a superstructure was developed to explore alternative routes for producing SNG and green ammonia from CO2 and N2 in cement plant flue gas, respectively. The routes were modelled in Aspen Plus® V14, and their economic viability was assessed. Currently, the most promising route, at a cost of 109 €/tonne of flue gas, involves CO2 capture (90% efficiency) and storage, with the N2-rich stream emitted. However, all the routes remain uncompetitive compared with the full emission of the flue gas (22 €/tonne). Routes for green ammonia production from N2 in flue gas were compared with those from atmospheric air but were found to be uncompetitive. Nevertheless, with future developments, the economic feasibility of integrating pathways for synthesizing high-value chemical products into the cement industry is expected to increase.

Hydrogen is increasingly recognized as a vital energy carrier for a sustainable future. Among the various methods for hydrogen production, alkaline water electrolysis (AWE) stands out as a well-established and commercially viable option. However, their more effective deployment requires more advanced, portable, and scalable designs. This study explores systematic model-based shape optimization of the next generation AWE based on computational fluid dynamic (CFD) aimed to enhance the hydrodynamics and electrochemical performance. Several design geometries and arrangements were proposed including flow baffles to enhance hydrodynamic and facilitate detachment of oxygen and hydrogen bubbles. The findings indicate that the optimal design and location of the baffles improve fluid mixing and enhance bubble detachment, resulting in a more uniform electrolyte distribution and decreased concentration polarization. Several key performance indicators were considered to analyse the performance of proposed designs including gas production rates, polarization curves, and fluid flow velocity profiles. The insights gained from this research offer valuable recommendations for optimizing flow field designs in alkaline water electrolyzers, aiming to enhance efficiency and operational robustness.

Thermal management and heat transfer optimization remain central challenges in next-generation concentrated solar power (CSP) systems employing solid particles for thermal energy storage and heat transfer. Conventional particle receiver concepts, such as fluidized beds and falling particle curtains are constrained by limited particle-wall contact, flow instabilities, and restricted operating temperature. This work presents a combined computational and experimental investigation of a gravity-driven dense-phase moving packed bed receiver featuring a flat conduit geometry and sub-millimeter particles. A multiphase modeling framework is developed and validated against pressure-drop measurements and particle velocity data obtained from dedicated experimental setups. The validated model is subsequently used to quantify dense-flow stability and thermal performance under indirect heating conditions. Results demonstrate stable dense-phase operation with particle volume fractions of approximately 0.6, leading to enhanced particle-wall transfer. Alumina particles achieve wall-to-particle heat transfer coefficients of up to 2.7 kW·m-2·K-1, more than twice those obtained with silica sand under comparable conditions. The findings indicate that flat-conduit dense-phase receivers provide a mechanically simple and thermally efficient alternative to conventional particle receiver concepts, supporting high-temperature and more efficient CSP systems.

This work develops a high-spatial resolution multi-stage stochastic programming (MS) model for power grid capacity expansion that co-optimizes generation, transmission, and energy storage system investments under uncertainty. Traditional two-stage stochastic programming (TS) models determine all investments in a single stage, limiting their ability to adapt to changing conditions such as evolving capital costs, policies, or supply chain disruptions. In contrast, the proposed MS formulation introduces sequential decision stages where partial information is revealed over time, allowing for adaptive, scenario-contingent investments. We compare TS and MS formulations using a modified IEEE 24-bus case study to quantify the Value of the Multistage Solution, which measures the economic benefit of allowing investment decisions to adapt over time as uncertainty is progressively resolved. Results show that while MS models are computationally more challenging, they achieve lower expected costs and yield flexible investment strategies that better respond to geopolitical and market uncertainties.

Computer-Aided Process Engineering (CAPE) has transformed how we analyze, design, and optimize energy processes. Yet, even advanced models rest on uncertain ground: their reliability depends on how well future operating environments are described—environments that are dynamic, complex, and deeply uncertain. In practice, uncertainty is often reduced to local parameter variations, driven by limited data, computational burden, and overconservative robust formulations. This narrow treatment creates a false sense of confidence: Designs that perform well in theory often fail in real-world operation. In a century marked by economic, climatic, and technological volatility, designing under uncertainty is no longer optional; it is essential.We have developed approaches that place uncertainty at the core of energy process modeling and design. This paper provides an overview of these methods and how uncertainty can be explicitly represented, quantified, and embedded into the design process.We present approaches to characterize uncertainties under data scarcity, including imprecise probabilities to distinguish between aleatory and epistemic uncertainty. To propagate uncertainty through computationally intensive models, we introduce surrogate-assisted techniques that exploit structural sparsity, enabling the analysis of systems with large numbers of uncertain inputs (100+) while mitigating the curse of dimensionality. These methods are integrated into optimization frameworks that target expected performance, robustness, and, for the first time, antifragility—systems that can benefit from variability rather than merely withstand it. We illustrate these approaches across applications ranging from detailed process models to system-level energy analyses, advancing a shift in CAPE toward designs suited for an unpredictable world.

Stochastic programming (SP) is a powerful framework for decision-making under uncertainty, but its practical adoption in industry is often hindered by the difficulty in understanding the causal relationships that drive optimal solutions. In the two-stage SP, strategic first-stage decisions are coupled with operational second-stage recourse decisions. When the number of scenarios under consideration is large, understanding the direct link between the uncertainty realization and optimal recourse strategy becomes computationally and cognitively demanding. Common approaches to improve interpretability include trained classification trees or scenario reduction, replacing the large scenario set with a representative subset. This is often achieved through post-hoc clustering (e.g., k-means) based on uncertainty realizations or optimal recourse decisions. While useful, these methods only provide a statistical approximation of the solution space and may fail to reveal the underlying structural properties of the recourse problem that drive optimal first-stage decisions. This work introduces a novel, deterministic approach to explainability using multiparametric programming (mp) within a Benders decomposition framework. We reformulate the recourse subproblem as a multiparametric linear program, generating an explicit map of Critical Regions (CRs), which are polyhedral partitions of the uncertainty space. This allows us to cluster scenarios analytically rather than statistically. We demonstrate this methodology on a supply chain planning under demand uncertainty. Our results show that 100 stochastic scenarios map to exactly six critical region clusters. This mapping allows us to explain optimal capacity planning decisions as a precise trade-off between specific operational modes, providing a fully transparent interpretation of the stochastic solution.

Although rare, unexpected events such as financial crises, geopolitical conflicts, and pandemics have reshaped reality in recent years. Despite their strong potential to affect the energy transition, such events are still largely overlooked in energy planning studies. Ignoring them can lead to poorly informed decisions that may jeopardize the transition. Identifying early-stage decisions that remain robust under unexpected events is therefore essential. To address this challenge, EnergyScope Pathway, a whole-energy system model with limited foresight, is applied to Belgium. To increase the likelihood of a successful transition, the Modeling to Generate Alternatives approach is used to diversify early-stage decisions in 2035. These alternatives are allowed to be up to 10% more expensive than the cost-optimal solution. However, the large number of alternative designs is difficult to navigate for decision makers. To address this, a decision-support framework based on the Multi-Armed Bandit framework is used to identify early-stage decisions that are most robust to future unexpected events. In this step, the remaining transition phases are optimized under unexpected events sampled within predefined impact ranges. The results show that, under normal conditions, there is a high degree of flexibility in the decision space for the 2030–2035 phase, with many technologies or resources that can be entirely omitted. However, robust early-stage decisions rely on a diverse energy generation portfolio, with a stronger emphasis on wind deployment, early mobility shifts toward battery electric vehicles, and the import of e-fuels. These insights can help decision makers steer the energy transition toward a robust path from the beginning. While Belgium is used as a case study, this framework is transferable to other contexts.

Resilience is becoming a top priority across industrial sectors, with increasing pressures to assess it systematically. In this work, we present an optimization-based framework for proactive design and planning under uncertainty of multi-product manufacturing networks, and testing of the reactive strategies available to withstand unforeseen disruptions. Specifically, the design problem is formulated as a two-stage stochastic optimization, integrating multi-period planning and scheduling, aimed towards mitigation against uncertainty. Designs are then fixed and tested through simulated outcomes from out-of-sample uncertainty distributions, with feasibility of operation monitored through the time-to-recover post disruption. Infeasibility triggers a scenario-update procedure via K-means clustering, whereby critical uncertainty information based on simulated outcomes is integrated in the proactive planning step, including low-probability high-impact scenarios. Modular and non-modular designs are compared to quantify the value of recourse actions with respect to responsiveness to disruptions. Results highlight that adaptable modular designs enable shorter recovery times combined with lower cost commitments. Overall, the framework underpins the development of tools for the anticipation and mitigation of disruptive events and systematic re-design based on continuously updated uncertainty information.

Wine consumers are increasingly health- and environmentally conscious. At the same time, white wine and rosé drinkers favour freshness and varietal aromas, which requires low-temperature regimes that extend fermentation time and increase energy demand. Additionally, global warming accelerates grape ripening which increases alcohol level in wine. To reduce cost and alcohol levels while maintaining quality, predictive tools that forecast how fermentation conditions impact fermentation time, and primary and secondary metabolite concentrations, can provide practical benefits to wineries by expediting oenological decisions-making and in turn reducing energy demand. Additionally, literature highlights static models in smart manufacturing suffer from performance degradation with data drift. In light of this, we successfully developed and evaluated pipelines for the automated design and training of three ML methods – support vector regression, random forest and artificial neural networks – to predict fermentation outcomes from ten initial features. The dataset employed captures the fermentation of nine commercially available Saccharomyces cerevisiae strains, each performed in triplicate, across two synthetic media and four temperature regimes, sampled at nine time points. The targets we focused on included time to complete fermentation, Ethanol, Acetate, given its undesired spoilage effects and ten secondary metabolite concentrations, given their synergistic impact on aroma. We were able to recommend effective pipelines for each task and identified our feature set contained adequate predictive signal for ethanol, acetate and fermentation duration predictions, but inadequate for the multi-output regression of secondary metabolite concentrations, which exhibited target heterogeneity. Ultimately, embedding these pipelines into optimisation frameworks, offers an actionable route to tailoring wine characteristics to evolving consumer preference, while reducing energy demand.

Ensuring reliable operations of modern power systems under uncertainty remains a key challenge, particularly due to the non-convex nature of Alternating Current (AC) power flow equations and the presence of high-impact disturbances from load and renewable generation fluctuations. In this work, we address the robust AC Optimal Power Flow (AC OPF) problem by developing a robust spatial branch-and-bound (RsBB) algorithm. Robustness is achieved by identifying worst-case uncertainty realizations and iteratively incorporating robust cuts to eliminate constraint violations. To accelerate convergence and tighten bounds, Optimization-Based Bound Tightening (OBBT) and Feasibility-Based Bound Tightening (FBBT) techniques are integrated into the framework. The proposed method yields global robust solutions with certified optimality gaps below 0.01% across standard PGLib test cases.

Mathematical models are increasingly used in pharmaceutical process development within quality-by-design (QbD) frameworks to reduce experimental effort and enable rational process design. However, model validation is still often based on deterministic performance indicators, which do not explicitly account for experimental variability, measurement noise, and model uncertainty. This work proposes an uncertainty-aware framework for model validation in pharmaceutical processes that quantifies predictive reliability in probabilistic terms, consistent with regulatory concepts. The framework explicitly integrates uncertainty in operating conditions, measurements, and model parameters, and evaluates model performance based on the probability that prediction satisfy predefined acceptance criteria rather than on single-point accuracy indicators. An in-silico case study of crystallization was performed to demonstrate the approach, where synthetic experimental data with controlled uncertainty were generated. This enables systematic assessment of the effect of data quality, data quantity, and methods of uncertainty analysis on model validity results. Overall, the proposed framework provides a structured and robust approach for uncertainty-aware, regulatory-aligned model validation in pharmaceutical process development.

Traditional centralised manufacturing offers efficient economies and broad market reach but faces increasing limitations with the rise of complex products requiring rapid localised delivery and greater supply chain resilience. The logistics demands of hospital-compounded therapies expose vulnerabilities in existing infrastructure, accentuating the need for rigorous evaluation of alternative paradigms. This study investigates the comparative performance of centralised and decentralised pharmaceutical manufacturing models, applying an agent-based simulation framework designed for specialised or time-sensitive drug product orders. The work implements an agent-based simulation to model both centralised and decentralised scenarios using key structural, resource, and demand parameters identified within the supply chain ecosystem. Comparison criteria include labour requirements, sustainability (as measured by environmental emissions and operational efficiency), and end-to-end supply chain lead times, informed by the geospatial distribution of manufacturers, hospitals, and clinics. The centralised case considers a single facility supplying major hospitals and several clinics from a designated hub, while the decentralised case models multiple smaller production sites supplying care centres more directly. Demand frequency, emergency inventory buffers, personnel allocations, and practical constraints are explicitly built into the simulation inputs. Preliminary simulation results reveal trade-offs between manufacturing paradigms across multiple performance dimensions. The decentralised model shows potential advantages in reducing supply chain lead times and improving responsiveness to localised demand surges, while the centralised model demonstrates efficiency gains in resource utilisation under steady-state conditions. The framework enables quantitative comparison of throughput, cost implications, delivery timeliness, and system resilience under varied operational scenarios. These findings will inform strategic design decisions for patient-centric and resilient pharmaceutical supply chains, facilitating adoption of flexible models capable of meeting modern healthcare delivery needs within the medicines manufacturing and supply ecosystem.

The objective of this work is to investigate strategies for the calibration of crystallization models aimed at predicting particle size distributions (PSDs) of active pharmaceutical ingredients (APIs) when using industrial datasets, which are limited in terms of number or information for the modeling exercise. In this work, the calibration task relies on two kinds of measurements, commonly performed in industrial crystallization practice: offline measurements of PSDs and API solute concentration carried out only at the beginning and at the end of experiments, and online measurements of chord length distributions (CLDs). Particularly, a strategy is proposed to use CLDs data from focused beam reflectance measurement (FBRM) probes as proxies of the PSD, which is the main key performance indicator for the model exercise. Industrial data concerning a seeded batch cooling recrystallization of an API in an organic solvent are used as a case study. The PharmaPy process simulator is used for parameter estimation and process simulation. Results demonstrate that, with proper data processing and feature extraction, all parameters can be estimated with sufficient precision. The model performance is satisfactory for most of the batch duration, even though some shortcomings highlight possible limitations in the data and/or in the model itself. From the industrial perspective, results pave the way for a quantitative usage of FBRM probes to enhance process understanding and to guide process development and scale-up.

Crystallisation is a promising and scalable alternative to chromatography for biologics purification. However biologics such as proteins and peptides often crystallise only in narrow operating windows, limiting process flexibility. Template-induced crystallisation can lower supersaturation requirements and expand feasible operating ranges, yet the template dependence of nucleation and growth kinetics remains difficult to parametrise mechanistically. To address this, we develop and experimentally validate uncertainty-aware hybrid models for lysozyme crystallisation on hydroxyl- and carboxyl-functionalised silica templates. A mechanistic population-balance model is coupled to a data-driven regressor that maps operating conditions and template variables to effective nucleation and growth rates. We compare a neural network baseline against a structured neural power-law surrogate, which embeds a supersaturation-dependent power-law form. Both hybrid models are trained in-the-loop via differentiable simulation, and variational inference is used to obtain posterior parameter distributions and calibrated predictive uncertainty. Across cross-validation and off-grid tests at previously unseen combinations of temperature and template loading, the hybrid models accurately reproduce solute concentration dynamics and capture key particle-size trends, while the neural power-law surrogate provides improved robustness and faster uncertainty quantification. These results support hybrid, uncertainty-aware PBMs as practical tools for prediction, design-space exploration, and comparison of template-enabled protein crystallisation processes.

Mathematical models play a central role in biopharmaceutical manufacturing, especially within the Quality by Design framework. For these models to be effectively used in optimization tasks, they must be both reliable and capable of delivering results in an affordable computational time. This work proposes a strategy to model aggregate formation during viral inactivation in the context of monoclonal antibody downstream processing. These units often display mixing-sensitive behavior because aggregation kinetics is controlled by local pH, whose spatial heterogeneities arise from titrant addition at a defined feed point. To address this challenge, compartment models (CMs) are employed. This modeling approach captures spatial inhomogeneities within the unit by leveraging flow-exchange information derived from a single steady-state Computational Fluid Dynamics (CFD) simulation involving only the solution of mass, momentum and turbulence equations. Results obtained by comparing compartment models with both perfectly mixed models and full CFD simulations including aggregation kinetics demonstrate that CMs can reproduce the CFD results with good approximation, while reducing computational time by orders of magnitude.

Continuous pharmaceutical manufacturing (CM) offers improved quality assurance, operational agility, and supply resilience, yet process development remains dominated by expensive trial-and-error experimentation and high-dimensional space exploration. Motivated by ICH Q13, we develop a generative inverse-design framework that maps target product quality to feasible process recipes for an integrated twin-screw wet granulation and segmented fluidized-bed drying line. The framework integrates three components: (i) a Conditional Variational Autoencoder (CVAE) generator that proposes process parameter sets conditioned on desired Critical Quality Attributes (CQAs), (ii) a Gaussian Process (GP) surrogate validator that screens candidates for manufacturing feasibility, and (iii) SHapley Additive exPlanations (SHAP) to interpret the generated designs. Training data were produced from a validated gPROMS digital twin of the Diamond Pilot Plant (DiPP) ConsiGma-25 line, covering liquid -to-solid ratio, drying temperature, drying time and air flowrate, with CQAs including granule moisture content, average particle size and porosity. The trained CVAE generated ~50, 000 candidate recipes and learned constrained feasible regions of the design space. Across seven operating scenarios, generated recipes achieved low deviations from targets. For a target-center case, deviations were 2.0% (moisture), 0.4% (particle size) and 0.5% (porosity). Edge cases remained acceptable, with the largest deviation observed for moisture in a high-moisture scenario (8.8%). SHAP analysis highlighted drying time and liquid-to-solid ratio as the dominant drivers of moisture, while liquid-to-solid ratio governed particle size and porosity. Overall, the approach enables rapid, explainable exploration of CM design spaces, reducing experimental burden and supporting QbD-aligned development for integrated continuous processes.

Accurate calibration of spectroscopic measurements is essential for reliable real-time monitoring and control of crystallization processes. In this work, calibration strategies for Attenuated Total Reflectance Fourier Transform Infrared (ATR-FTIR) spectroscopy were systematically evaluated for concentration monitoring in batch cooling crystallization of paracetamol in ethanol. Linear regression (LR), Partial Least Squares Regression (PLSR), Principal Component Regression (PCR), and symbolic regression (SR) were compared using both peak-based features and full spectral representations. Peak-based models provided a transparent baseline, with peak-area-based models consistently outperforming peak-height-based models. For LR, incorporating multiple absorption bands reduced the mean squared error (MSE) by nearly one order of magnitude compared to single-peak models. Using the same peak-based inputs, SR further improved performance, reducing prediction bias at high concentrations and yielding higher coefficients of determination (R² > 0.99) compared to LR. A substantial improvement was achieved when full spectral information was used. Among all evaluated approaches, SR with unprocessed spectra yielded the best overall performance, achieving an R² of 0.996 and an MSE of 1.4 × 10^-6 on the validation dataset. This model also demonstrated strong generalization on an independent solubility test dataset, closely reproducing the reference solubility curve over the full temperature range with minimal deviation. In contrast, PCR and PLSR models showed increased sensitivity to preprocessing choices and exhibited larger errors on the test dataset. SR provided an accurate, robust, and interpretable calibration framework for ATR-FTIR, with reduced reliance on spectral preprocessing and potential for real-time process analytical technology and control applications.

Oligonucleotides are short, sequence-defined nucleic acid chains with major therapeutic and diagnostic potential. Their industrial production is currently dominated by solid-phase oligonucleotide synthesis (SPOS), which suffers from mass-transfer limitations, limited scalability, lack of real-time process monitoring, and high process mass intensity. Membrane-enhanced liquid-phase oligonucleotide synthesis (LPOS) has emerged as a scalable alternative, in which oligonucleotide chains are grown on soluble anchors and organic solvent nanofiltration is used (OSN) to remove excess reagents and by-products between each reaction steps. However, diafiltration also introduces a risk of large cumulative product loss over multiple addition cycles, which requires fine-tuning of design and operational strategies in practice. This paper presents the results of a comparative assessment of two LPOS variants with either a single- or dual-stage diafiltration against a state-of-the-art SPOS, within a unified dynamic modelling framework. A recent mechanistic kinetic model of SPOS is transferred to LPOS, including a description of OSN diafiltration dynamics and internal recycles. A feasibility analysis is conducted to identify operating regimes under which LPOS outperforms SPOS in terms of selected yield and impurity performance indicators and to expose the underlying trade-offs. Overall, the results establish LPOS with dual-stage diafiltration as a promising configuration towards more flexible and sustainable oligonucleotide manufacturing.

Identifying cost-optimal yet environmentally friendly crystallisation processes in the production of small molecule pharmaceuticals is a highly complex task, since multiple solvent systems often exist which could be used to purify a given drug to a similar standard. It is, however, rarely possible to test each of these solvent systems within a laboratory setting, since this would be time-consuming and incur large material costs. It has, therefore, been suggested that process modelling tools should be used to screen different crystallisation processes available to produce a new drug prior to studying them experimentally – essentially allowing a shortlist of promising process candidates to be created. Indeed, it has been shown by several authors that this sort of work can be conducted without any need to establish the crystallisation kinetics associated with each drug-and-solvent combination considered: the reason being that crystallisation processes may be defined as simple solid-liquid equilibria (SLE) problems using thermodynamic models alone. Considering this, the present work has used SLE modelling to assess the cost and environmental impact of different solvent systems available for the crystallisation of pregabalin – an anticonvulsant and anxiolytic drug used to treat epilepsy, anxiety and neuropathic pain. Moreover, it has used entirely data-led approaches to achieve this: leveraging established costing methodologies and green metrics (i.e., Scope 1 and 2 CO2e emissions) to assess their economic viability and carbon footprint. This modelling framework can be used to identify viable crystallisation strategies for other small molecules.

Computer-aided molecular design (CAMD) is an efficient product design method that is gradually attracting attention at present. It mainly uses data mining technology to extract information from the existing chemical molecular data and use this information to generate potential excellent molecules. However, the key that CAMD can truly provide accurate and reliable results lies in the efficient utilization of chemical data. In this paper, a series of chemical data analysis methods based on molecular similarity are proposed to enhance the data utilization efficiency of CAMD, which mainly includes 3 applications: adaptive modeling, reliability assessment and advanced data preprocessing including molecular recommendation, data consistency test and data augmentation. We propose specific methodology for each application, and use multiple cases to verify the effect. The results show that molecular similarity can help to improve the accuracy of property prediction at the data level, provide quantification for property prediction reliability, recommend potential excellent molecules with similar structures, locate data mistakes and perform reliable data augmentation, finally enhancing the data utilization efficiency of CAMD.

This work presents Pareto Front Guided Sampling (PFGS), a model-guided Design of Experiments (DoE) strategy for bioprocess development that makes the exploration–exploitation trade-off explicit and integrates human expertise into experiment selection. Starting from an initial experimental design, PFGS fits a probabilistic surrogate and then proposes new experiments by solving a multi-objective design problem that simultaneously rewards (i) high predicted performance (posterior mean) and (ii) high information gain (posterior uncertainty). Rather than collapsing this trade-off into a single acquisition value, PFGS generates a Pareto set of candidate experiments, that reflect different balances between improvement-seeking and learning. To prevent wasted runs, an automated screening step is performed to remove candidates in (i) low predicted-mean regions unlikely to yield near-optimal performance and (ii) low-uncertainty regions already well explained by the surrogate, concentrating effort on promising yet under-explored areas and encouraging coverage of multiple near-optimal basins. For industrial application, PFGS was extended to batch selection by choosing a diverse subset of Pareto candidates to support parallel experimentation without substantial loss in sample efficiency, and a termination rule to stop experimentation once user-defined accuracy thresholds are met. PFGS is benchmarked against Latin Hypercube Sampling (LHS) and Bayesian Optimisation (BO) on two representative problems: (i) a bioprocess case study with shallow performance gradients, where naive exploitation is unreliable and non-adaptive designs are inefficient, and (ii) a multi-modal landscape, where methods that focus on a single basin can miss alternative optima. Across these settings, PFGS allocates experimental resources more effectively than non-adaptive baselines and, in the multi-modal case, uniquely identifies and characterises all stationary regions while maintaining strong overall performance. By combining Pareto-transparent decision support, automated filtering, batch practicality, and principled stopping, PFGS provides an interpretable and resource-efficient DoE methodology for biopharmaceutical process optimisation.

Managing cellular resources, especially transcription and translation machinery, constitutes a significant constraint in the effective synthesis of useful bio-compounds from synthetic gene circuits. Although light-based optogenetic control approaches provide precise temporal and spatial control that can balance growth and production. The increased complexity adds on to the competition for cellular resources such as ribosomes and RNA polymerases. The optimization of regulatory elements in bioengineering is a vital task, as the selection of promoters and ribosome binding sites (RBS) directly affects transcriptional initiation rates and translation efficiency, hence influencing resource allocation. The vast number of potential parameter combinations requires systematic approaches to narrow the design space and determine which bioparts most significantly influence system performance. Furthermore, it is crucial to ascertain the uncertainty regarding which bioparts exert the most significant influence on the performance of engineered gene expression. In this study, we conduct an exploratory analysis by combining two established global gene expression models with two optogenetic TCS regulation models within a 2×2 framework. The first resource model (McBride & Del Vecchio, 2021) illustrates resource competition via coupled transcription-translation dynamics, but the second model (Santos-Navarro, 2021) identifies translation as the primary resource-limiting factor. We illustrate the utility of Sobol Global Sensitivity Analysis (GSA) to evaluate which model parameters (bioparts) influence target gene expression in the CcaS/CcaR optogenetic system regulating GFP expression. The findings indicate that the formulation of resource competition, rather than regulatory complexity, fundamentally dictates the rankings of parameter sensitivity. The sensitivity ranking explicitly indicates which design parameters require tight control and which may be relaxed, resulting in a significant reduction in design space.

Bioprocesses are inherently multiscale, spanning intracellular metabolism to production-scale reactors. Simulation models that integrate these scales offer potential strategies to study the effect of changing metabolic states and enable efficient integration of biological knowledge gathered from lab-scale experiments. In this study, we demonstrate the potential of such simulation model towards the production of mevalonate, an important pharmaceutical drug compound produced through fermentation of a fungal species Aspergillus terreus. We integrate a genome-scale metabolic model of the organism with a plant-wide simulation model for the bioprocess that encompasses several upstream and downstream unit operations. Through this integration, we identify potential targets for metabolic engineering towards increased product flux and simultaneously estimate the associated oxygen requirements. This framework serves as a foundation for developing digital twins of bioprocesses that bridges strain engineering with process design and operations.

Recombinant protein production in Escherichia coli is a widely used system in industry for biopharmaceuticals, enzymes or other proteins. For protein expression, lactose poses as a more favorable and cost-effective induction agent over the common IPTG trigger. It imposes less stress on the cells and is fully metabolizable by the strain used. Therefore, lactose serves as an additional substrate source and adds a degree of freedom through tunable induction levels. To harness this opportunity, a physiological bioprocess model was created, describing the growth and production dynamics of this 2-feed system. Green fluorescent protein is expressed as a model protein in a fed-batch process using glucose as the main substrate and lactose as the digestible inducer. A suitable production kinetic is chosen by fitting a number of models to a collected dataset. The resulting model is used to highlight opportunities for improved process design and control of a 2-stage fed-batch process. It emphasizes the benefit of model-based methods to directly optimize for productivity over traditional design of experiment approaches.

The volumetric gas-liquid mass transfer coefficient (kLa) is a key performance parameter in stirred tank reactors and is commonly determined through extensive experiments across the operational space. This work presents an automatic, closed-loop framework for kLa determination based on model-based design of experiments (MBDoE), in which agitation and aeration inputs are adapted in real time.During each experiment, dissolved oxygen data is collected and used to estimate the parameters of a Van’t Riet kLa relation. The parameter uncertainty is quantified using the covariance matrix, and the experiments are iteratively selected based on D-optimality or E-optimality MBDoE, until a threshold of RSEi < 0.15 is reached for all parameters. The MBDoE approach is evaluated through repeated runs and compared against random designs, full factorial (FF) design, and a full grid design.The results demonstrate that the closed-loop MBDoE framework can significantly reduce the number of experiments required to characterize kLa while maintaining accurate predictions of gas-liquid mass transfer over the design space. D-optimal MBDoE converges in fewer than four experiments on average, while E-optimal MBDoE converges in less than six experiments on average, translating into approximately a 60% reduction of the number of experiments as compared to a full factorial DoE design.

The decarbonization of pharmaceutical manufacturing is critical for achieving the industry’s net-zero targets, and electrochemistry is emerging as a promising green technology that could play a key role in this transition. This work evaluates a continuous-flow electrochemical route for ibuprofen synthesis through electrochemical carboxylation of 1-chloro-(4-isobutylphenyl) ethane as a low-carbon alternative that can be directly coupled with renewable electricity. Experimental studies have demonstrated the selective formation of ibuprofen using a silver cathode in the ionic liquid N-methyl-N-propylpiperidinium bis(trifluoromethanesulfonyl)imide (PP13 TFSI). While the reaction mechanism is based on laboratory-scale, batch experiments, this study develops a conceptual design and scale-up methodology for the continuous route to provide an evaluation of the industrial feasibility of this electrochemical pathway through a rigorous plant-wide simulation in AVEVA® Process Simulation. Global sensitivity analysis is employed to identify key operating variables and evaluate their impact on reactor and process performance, energy consumption, and ionic liquid recovery. These insights provide a robust foundation for informed decision-making in process intensification, demonstrating the technical viability and scalability of continuous flow electrosynthesis as a sustainable alternative to conventional API manufacturing.

Real-time monitoring of bioprocesses is hindered by sparse, heterogeneous measurements of key biological states, such as biomass, substrate, and product concentrations. Extended Kalman Filter (EKF)–based soft sensors offer a physics-grounded solution but are sensitive to limited observability, sensor bias, and process-model mismatch—conditions common in industrial fermentations. This work proposes a Levelized Hybrid Estimation Architecture (LHEA) that systematically enhances physics-based state estimation through increasing robustness and adaptivity while preserving model transparency and regulatory interpretability. The approach is evaluated using the KTB1 benchmark simulation model for continuous lovastatin production under an industrially realistic, cost-constrained sensor configuration combining dissolved oxygen, biomass proxy, volume measurements, and sparse HPLC product assays. Three estimator levels are investigated: (L1) a baseline EKF, (L2) a bias-augmented EKF for sensor-drift robustness, and (L3) a hybrid EKF with physics-constrained residual learning driven by sparse assays. Results show that L1 provides stable state reconstruction under nominal conditions but is sensitive to bias and model mismatch. L2 effectively isolates measurement bias, while L3 adapts to evolving process kinetics and achieves the lowest estimation errors under mismatch. These findings demonstrate that a structured, levelized integration of machine learning can significantly enhance soft-sensor reliability without sacrificing interpretability, providing a practical pathway toward robust digital twins.

Roller compaction is widely used in pharmaceutical manufacturing to improve powder flowability and enable robust tablet production. Although often treated as producing a homogeneous granule population, ribbons may undergo splitting during compaction, generating structurally distinct granules that affect downstream tableting. This study investigates the impact of ribbon splitting on tablet critical quality attributes (CQAs) for 10% and 20% acetaminophen (APAP) formulations. Reduced-order models (ROMs) proposed by Bachawala et al. [1] were applied to predict tablet density, elastic recovery, tensile strength, and tablet weight under split and non-split conditions. Although ribbon splitting alters the granule size distribution (GSD) and ribbon density, tablet CQAs such as tensile strength, elastic recovery, and tablet density are accurately predicted by the existing ROM framework, provided that GSD and ribbon density are known. In contrast, tablet weight predictions deteriorate when split and non-split data are combined, reflecting the sensitivity of die filling to granule packing changes induced by splitting. Separating the datasets improves weight prediction for non-split granules, while split granules remain challenging to model. A multitask Gaussian Process regression model is further used to analyze splitting as a process disturbance, highlighting the need for extended or data-driven approaches to accurately predict tablet weight under split conditions.

In pharmaceutical process engineering, accurate prediction of tablet properties is crucial for ensuring product quality, optimizing manufacturing efficiency, and advancing sustainable production practices. This study presents a physics-informed neural network (PINN) framework for predicting the physical properties of pharmaceutical tablets from near-infrared (NIR) spectra. The PINN framework integrates revised Kubelka-Munk theory and physical constraints to ensure physically consistent predictions while requiring less training data than conventional artificial neural networks. Tablets were manufactured using acetaminophen and microcrystalline cellulose formulations with varying compositions and compression settings. The PINN framework successfully predicts critical quality attributes, including tensile strength, porosity, and density. It offers a data-efficient, interpretable solution for pharmaceutical tablet quality control.

Probabilistic design spaces enable pharmaceutical manufacturers to balance regulatory compliance and operational efficiency. In this context, the “edge-of-failure” separating compliant from non-compliant operation is not a fixed line, but an uncertain region driven by model parameter uncertainty. Traditional methods typically map this probability of failure across the whole design space, which is a computationally expensive task. We propose a more direct approach, reformulating the problem using quantile functions to search for a deterministic boundary at a desired confidence level. These functions are approximated via Gaussian process surrogates using a novel adaptive sampling strategy. An industrial peptide acylation case study identifies the probabilistic design space using 13 experiments versus 18 from a traditional DoE; a 28% reduction. The framework quantifies trade-offs between operational robustness and yield optima.

The formulation of oral solid dosage forms can have a significant impact on drug bioavailability, particularly for poorly soluble drugs. However, traditional formulation development relies heavily on extensive experimental testing, which limits its efficiency and effectiveness in oral drug product design. In this study, we present an integrated framework to support rational formulation design and exploration of optimal dosage regimens. This framework combines population balance-based tablet disintegration and dissolution modeling with physiologically based pharmacokinetic (PBPK) modeling to link critical material attributes (CMAs) with the pharmacokinetic response. The anticoagulant drug rivaroxaban is selected as a model compound for calibration and deployment of the framework, enabling systematic investigation of the effects of crystal size distribution (CSD) and tablet porosity on in vivo performance. The results demonstrate that CSD has a pronounced impact on in pharmacokinetics, whereas tablet porosity exhibits a smaller but non-negligible effect. Furthermore, optimization is implemented to identify the optimal dose amount under a given formulation for producing the desired pharmacokinetic profile for average patient group, demonstrating the potential of this framework for the digital design of both drug efficacy and treatment strategies.

Single-cell protein (SCP) fermentation efficiently converts carbohydrates into high-protein food products but typically relies on purified glucose. Better understanding of SCP growth on mixed sugar substrates could allow for use of lower-cost, less processed sugars and waste-derived feedstocks. Glucose–fructose mixtures are particularly relevant, as these are the main sugars in sucrose hydrolysates and are also common in many food and beverage waste streams. In this study, the growth of Fusarium venenatum on glucose–fructose mixtures was investigated experimentally and modelled using a lagged dual-substrate Monod framework incorporating inhibition of fructose growth by glucose. Batch fermentations were conducted at a fixed total sugar concentration (15 g/L) with four different initial substrate compositions. Model parameters were estimated using both single-experiment and multi-experiment fitting strategies using differential evolution and wild bootstrap uncertainty analysis. A shared parameter set reproduced biomass growth and substrate depletion across all conditions with only a modest increase in prediction error relative to individual fits. Bootstrap analysis revealed substantial correlation among uptake and inhibition parameters, while yield coefficients were well constrained. These results demonstrate that mixed-sugar growth of F. venenatum can be described using a unified kinetic model, supporting the feasibility of transitioning from glucose-only to mixed-substrate SCP fermentation.

Decisions taken throughout a bioprocess lifecycle are often guided by heuristic knowledge that is difficult to summarize and sort, scattered across heterogeneous tools and documents, and partly retained as tacit expert mental models alongside fragmented computational models. This fragmentation remains a central barrier to reproducibility, transparent provenance, and systematic reuse of prior learning across comparable development projects. In this paper, it is argued that a key missing link toward Bioprocessing 5.0 is the digitalization of FAIR knowledge through a Cognitive Digital Thread that couples semantic knowledge graphs with AI methods to connect experimental data, protocols, workflows, and decision rationale with mathematical models and digital twins in a machine-actionable and auditable manner. A digitalization roadmap is outlined as a sequence of capability stages—from local device and data integration, to reproducible workflow execution and metadata capture, to semantic knowledge digitalization and explainable reasoning, to secure cross-institutional exchange of FAIR data and knowledge (optionally anchored via permissioned blockchain and smart contracts for authorship, versioning, and IP provenance), and finally to human-in-the-loop collaborative environments for model-informed bioprocess design and control. The result is a layered conceptual architecture that clarifies required functions and dependencies and provides a practical framework for implementing cognitive decision support across the bioprocess lifecycle.

Recent advances in computer-aided retrosynthesis (CAR), flow chemistry, and continuous manufacturing collectively offer new opportunities to enable environmentally sustainable development and manufacturing practices across the pharmaceutical development and manufacturing value chain. However, the implementation of these methods and technologies remains scattered and fragmented, preventing full realization of their potential to address one of the most urgent needs in the pharmaceutical and related sectors. This work introduces a holistic digital framework for the design and optimization of an end-to-end manufacturing process for paracetamol (acetaminophen). The framework integrates Green-by-Design synthetic and purification routes of the active pharmaceutical ingredient (API) aims to deliver cost efficiency and robust quality, safety, and environmental sustainability assurance. The approach integrates AI-driven CAR with plant wide modelling, Techno-Economic Analysis (TEA), and prospective cradle-to-gate prospective Life Cycle Assessment (LCA) to evaluate designs options and greener, chemically feasible, and more selective synthetic pathways. An end-to-end mathematical model is implemented in gPROMS Formulated Products® to optimize operating and design parameters and generate inventory data for TEA/LCA. The framework also embeds Quality by Digital Design (QbDD) to identify CPPs and CMAs that influence CQAs, enabling definition of a robust design space. Finally, a multicriteria decision-aiding approach is implemented to determine the best trade-offs among yield, productivity, resource efficiency, cost, and environmental performance, providing a transferable Safe-and-Sustainable-by-Design (SSbD) methodology for developing greener and more resilient pharmaceutical manufacturing systems.

Magnetic Inductive Swing Adsorption (MISA) is a carbon dioxide capture process similar to Temperature Swing Adsorption that uses direct electromagnetic heating instead of classic heating systems for the regeneration step of the process. However, the lack of validated dynamic models hinders process optimization. This work introduces an open-source MISA model in the IDAES framework, incorporating Specific Absorption Rate (SAR) physics (SAR ? B²) to capture electromagnetic heating. Binary Sips isotherm parameters for Fe3O4@HKUST-1 were fitted to experimental data, achieving high statistical agreement (R2 > 0.996, RMSE < 0.022 mol/kg). Comprehensive validation was performed against adsorption isotherms, dynamic breakthrough curves, and desorption profiles. The model predicts breakthrough time with only 9% error and saturation time with 6% error. Crucially, the coupled thermal transport and SAR heating model capture temperature evolution during desorption within 5% error across all field strengths. Although the use of Linear Driving Force kinetics introduces minor systematic overprediction, the model successfully bridges the gap between laboratory feasibility and industrial design. This validated tool enables the first systematic investigation of cycle configurations, providing a platform for techno-economic comparisons and scale-up of energy-efficient magnetic carbon capture.

In the field of biochemical process design, the accurate modeling of microbial growth is essential for the development and optimization of biological reactors used in the production of high-value compounds. Achieving this objective requires a detailed understanding of how environmental factors—such as pH and nutrient availability—influence microbial dynamics across the four distinct growth phases: lag, exponential, stationary, and death. Traditionally, reactor design relies heavily on the Monod model, which provides a simplified representation of microbial growth, focusing primarily on the exponential phase under constant operating conditions (1). However, this model presents substantial limitations when applied to dynamic environments where key parameters vary over time. To overcome these constraints, the present study proposes a data-driven modeling approach using a multilayer perceptron (MLP) artificial neural network for the prediction of microbial growth trajectories under varying pH conditions and substrate compositions. The yeast strain Candida guilliermondii was selected as the model microorganism due to its industrial relevance. Experimental growth data were collected through optical density measurements using a Multiskan™ FC Microplate Photometer (Thermo Scientific), covering a pH range from 6.0 to 8.5 and two substrate scenarios: pure xylose, and a 1:2 glucose–xylose mixture. The experimental data were used to train the MLP neural network, which generated predictive models capable of estimating growth behavior under the specified input conditions. This modeling approach enables the simulation of microbial growth curves at any given time point within the defined parameter space, providing a more flexible and comprehensive tool compared to classical models. The results of this study demonstrate that the proposed MLP-based model is a powerful computational tool for both the design and real-time control of bioprocesses. The integration of these tools into predictive control strategies is a key aspect of the bioprocess research. The model’s versatility and its ability to predict microbial behavior make it a very important tool for bioreactor control.

This work presents the dynamic optimization of glucose feed in cell cultivation considering the balance between productivity and impurity generation. We first developed a mechanistic model considering cell growth promotion by glucose and cell growth inhibition by osmolarity for a newly developed, high-productivity CHO-MK cell line. For model development, fed-batch cultivation experiments were conducted at a 250 mL scale under three different glucose feeding profiles. Results from a single-objective dynamic optimization, using the glucose feed profile as a design variable, were compared to those from multi-objective problem settings with varying weights assigned to productivity and final impurity concentrations. Simulation results suggested different glucose feed profiles depending on the priority given to the mAb and impurities, where the main difference was in the generated viable cell density profiles. Productivity-focused profiles employed a low–high–intermediate feeding strategy, in which the total glucose feed over the cultivation period was set to a balanced level, with an initially low rate, followed by a gradual increase, and a later decrease. Profiles focusing on impurity reduction adopted a feeding strategy that maintained a high glucose feed throughout the cultivation period. This work demonstrated the power of mechanistic-model-based process design and is expected to encourage further model-based process design in biopharmaceutical manufacturing.

In this work, artificial neural networks are used to represent advanced oxidation processes, such as electrochemical oxidation, electro-Fenton, and photoelectro-Fenton, for the degradation of two dyes. The effect of treatment time, initial dye concentration, and current density on the degradation percentage is studied. Additionally, a network is developed to include discrete variables, such as treatment type and dye type, as input features, enabling it to predict the system’s performance across different technologies and pollutants. Operating conditions are optimized using the universal ANN as a surrogate model and the adaptive differential evolution algorithm to maximize dye removal efficiency. According to the results, after optimizing the architecture of the artificial neural networks using Bayesian optimization, deviations of 3.9% or less are obtained for purple RL removal predictions, while for Green A, deviations of 8% or less are obtained. These models may serve as a basis for more robust representations in the future, including the effects of electrochemical device size, among other relevant variables.

Reliable kinetic models are essential for the design, optimisation and operation of methanol synthesis reactors in power-to-X applications. However, parameter estimation is frequently performed without prior assessment of parametric identifiability or the information content of experimental conditions, often resulting in poorly constrained parameters and inefficient experimental campaigns. This study introduces a systematic, pre-calibration, information-driven framework for identifiability analysis and experimental design. The framework integrates local sensitivity analysis, structural and practical identifiability metrics and a Sequential Information-Driven Experimental Selection (SIDeS) strategy to guide experiment selection prior to parameter estimation. The methodology is applied to four literature kinetic models for methanol synthesis, spanning varying levels of mechanistic detail. A Sobol-sampled design space is first screened for feasibility and local information content, followed by cumulative Fisher information analysis. Results demonstrate that conventional screening campaigns provide limited gains in parameter precision, whereas SIDeS identify compact experimental sequences that significantly improve practical identifiability within a constrained experimental budget. Cross-model comparison reveals that increased mechanistic description does not necessarily translate to improved identifiability under fixed experimental resources. The proposed framework supports informed model selection, reduces unnecessary calibration effort, and provides actionable design-of-experiments recommendations for methanol synthesis systems.

Geologic hydrogen has emerged as a promising low-carbon energy vector, but its sustainable recovery requires effective stimulation and production strategies. This study presents an integrated process-modeling framework for evaluating hydrogen extraction from hydraulically stimulated reservoirs. The framework combines fracture propagation, damage evolution, Darcy-scale multiphase flow, permeability–aperture dynamics, and a dual-porosity dual-permeability (DPDP) representation to simulate hydrogen production in fractured source rock systems. In addition to production dynamics, the framework tracks operational water and energy inputs and incorporates a simplified reaction-extent formulation to represent hydrogen generation under data-limited conditions. The model was benchmarked against published shale gas production results to evaluate its ability to reproduce fracture-controlled production behavior and was subsequently applied to a representative multi-well hydrogen development scenario. Simulation results indicate that hydrogen production is strongly fracture-dominated, with fracture pathways accounting for more than 98% of total modeled output. Under the simulated conditions, the system produced approximately 1.13 million kg of hydrogen, with an estimated operational energy intensity of 3.4 MJ/kg H2 and water consumption of 0.02 m³/kg H2. These values represent model-based estimates within the defined operational system boundary and illustrate the potential performance indicators that the framework can generate. Per-well analysis reveals variability in water and energy efficiency across the multi-well system, highlighting the importance of fracture design and reservoir characterization. The simulations also indicate that matrix contributions to total recovery are limited under the assumed conditions, suggesting potential benefits from strategies that enhance matrix reactivity or transport processes. Overall, the framework provides an integrated approach for exploring hydrogen production dynamics together with associated resource requirements. As field and experimental data for geologic hydrogen systems become available, the framework can be further refined through improved parameterization, sensitivity analysis, and calibration to support future technical, environmental, and economic assessments.

The objective of the present work is to optimize the simulation of an electrodialysis reactor for the desalination and regeneration of multi-ionic wastewater with high salt contents and conductivities, within the framework in the Sustainable Development Goal 6 (clean water and sanitation) and remarking the Electrodialysis (ED) as a highly energy-efficient and sustainable technology. The mathematical modelling has been carried out by using a semiempirical model that involves an algebraic system of differential equations, including mass and charge balances (taking into account the ions present in the wastewater: Na+, Ca2+, Mg2+, Cl-, SO42-, and HCO3-), and the total electrodialysis stack voltage considering ohmic drops (in the dilute and concentrate compartments), the potential of membrane in each cell pair, and the electrode potentials. In the simulation process, different theoretical and experimental parameters are necessary such as number of cells, membrane working areas, efficiency, diffusion coefficients, molar conductivity at infinite dilution, etc.The experimental parameters have been obtained initially with a specially designed batch reactor printed with 3D technology, to imitate the hydrodynamics and the behavior of the industrial final reactor used. Additionally, to increase the accuracy of the simulation results, same parameters such as the resistance of the membranes (rm), the transport numbers and the mass transfer coefficient (km) can be optimized by using a metaheuristic derivative-free optimizer, specifically a proprietary version of Particle Swarm Optimization (PSO) implemented in MatLab®.The results obtained yield a satisfactory reproduction of the observed experimental behavior, analyzing the evolution over time of the intensity, conductivity, and concentration of the concentrated and diluted currents, respectively. In addition, the evolution of the different contributions of the stack voltage and the membrane resistances are shown, in order to confirm the coherence of the results obtained.

Bioprocess modelling, optimisation and scale-up are central components for improving sustainable manufacturing within pharmaceutical and chemical industries. However, developing accurate bioprocess digital twins remains a challenging process. Conventional mechanistic models are difficult to construct because of limited mechanistic understanding and large complexity of cellular metabolisms. While data-driven models have gained popularity, they often require large amounts of experimental data that is often time consuming to obtain and lack any quantitative description of the process. Hybrid modelling methods have emerged as promising alternatives however fail to provide physical insight to the root cause of model error. This work therefore presents a promising solution by developing a data-efficient symbolic regression (SR) based framework to enable the automated discovery of interpretable bioprocess models. A universal kinetic model backbone was used to capture overall process behaviour, while SR was applied to strategically uncover the structures of critical kinetic terms within the backbone. Two frameworks, embedding SR directly into the kinetic model backbone or identifying time-varying parameter profiles prior to SR, were benchmarked using an in-silico yeast fermentation case study. The results demonstrated that independently identifying individual kinetic terms was crucial for recovering the ground-truth model, while refining SR-generated candidates through a novel local iterative structural correction strategy significantly improved convergence to the true kinetic expressions, surpassing model-based design of experiments in data efficiency. This study therefore enables automated yet interpretable model construction for small-data bioprocess applications, paving the way towards augmented intelligence driven bioprocess modelling and accelerating digital twin development for process optimisation and control.

Brazed aluminum heat exchangers (BAHX) are key cryogenic equipment, but their simulation is sensitive to phase change modelling. This work benchmarks ProSec, CO-ProSec Reaction, and Aspen EDR on an ethylene-chiller BAHX using ethane in thermosiphon mode. ProSec’s interpolation scheme of tabulated data is validated against CO-ProSec Reaction’s approach that employs a thermodynamic model for the evaluation of thermodynamic and physical properties; an initial 25% pressure drop gap is traced to different void-fraction models (slip vs homogeneous). Against Aspen EDR, default duty/quality are ~15% higher; differences are mainly due to nucleate-boiling treatment and distributor pressure drop modelling. Harmonizing options reduces duty mismatch to ~5%.

Model adaptation to production environment is critical for reliable Machine Learning Operations (MLOps), less attention is paid to developing systematic framework for updating the ML models when they fail under drift. This paper compares the transfer learning enabled model update strategies including ensemble transfer learning (ETL), all-layers transfer learning (ALTL), and last-layer transfer learning (LLTL) for updating the failed feedforward artificial neural network (ANN) model. The flue gas differential pressure across the air pre-heater unit installed in a 660 MW thermal power plant is analyzed as a case study since it mimics the batch processes due to load cycling in the power plant. Updating the failed ANN model by three transfer learning techniques reveals that ETL provides relatively higher predictive accuracy for the batch size of 5 days than those of LLTL and ALTL. However, ALTL is found to be suitable for effective update of the model trained on large batch size (8 days). A mixed trend is observed for computational requirement (hyperparameter tuning and model training) of model update techniques for different batch sizes. These empiric insights obtained from the batch process-based industrial case study can assist the MLOps practitioners in adapting the failed models to drifts for the improved monitoring of industrial processes.

The conceptual design and performance of a novel Methylcyclohexane-Toluene-Hydrogen based chemical heat pump was studied using steady state simulations. The distillation operating parameters of the chemical heat pump were optimized to maximize the Coefficient of Performance based on heat quantity (COP) and its corresponding Coefficient of Performance based on electric work input (COPW) was calculated. The best operating temperature ranges of the endothermic and exothermic reactor are 200°C-225°C and 250°C-275°C respectively. An endothermic temperature of 200°C and an exothermic temperature of 250°C results in a COP of 0.1357 and a COPW of 13.3. By integrating this chemical heat pump with a vapor compression heat pump COP increased to 0.1445 while COPW reduced to 4.9.

As the demand of freshwater continuous to rise, the water desalination can act as prominent solution to address the global water scarcity. However, significant environmental concerns arise as the process is mainly powered by fossil fuel, which contributes to greenhouse gas emissions and thus leading to global warming. This research intends to explore the potential of renewable energy sources to effectively reduce freshwater production cost. Specifically, it intends to estimate the power generated and freshwater production cost of a multi-effect distillation (MED) desalination process, powered by solar and wind energy sources in addition to a comparison against the performance of an MED process powered by fossil fuel based conventional steam boiler. The comparison of energy efficiency and freshwater production cost is conducted in two different UK locations: Wick in the north of Scotland and Watchet in the south of England. MED process model with both wind power and solar energy models are developed to estimate the freshwater production cost while assuring 12kg/s of freshwater for small coastal communities. The results ascertain that both energy sources are able to produce thermal energy required for the MED process. Statistically, for a full day, 4029 kWh and 16383 kWh amounts of electricity can be generated in Watchet and Wick, respectively. Referring to wind powered MED, the freshwater production cost in Watchet is estimated at $20.49 per m³ of freshwater whereas it is reduced to $5.56/m³ at Wick as due to stronger and more consistent wind speeds, which is less than a third of the Watchet freshwater production cost. However, solar powered MED shows a much smaller difference between the locations: $1.87/m³ in Watchet and $3.11/m³ in Wick. Based on the results of freshwater production cost, the solar energy for MED process is more efficient in Watchet, whereas wind energy for MED process is more efficient in Wick.

A recurrent neural network (RNN) model for a CO2 methanation reactor was developed based on synthetic data generated from a validated mechanistic model of the same unit. The model was used to predict the main properties of the reactor – methane productivity and hotspot temperature – during a dynamic operation of the unit. The dynamic profile of feedstock availability was simulated taking into account the H2 flow that can be produced from PV-powered water electrolysis using solar irradiation profiles over a year in Milan, Italy. The dataset therefore consists of 366 data instances (one for each day), each composed of one datapoint per minute of sunlight. The best agreement between the predictions from the RNN and the target output values from the mechanistic model was found using a shallow RNN of 20 hidden-layer neurons, trained with a batch size of 10 and an 80/20 training-testing split. This showed that RNNs can constitute a reliable tool for dynamic surrogate modelling of energy conversion reactors. The surrogate model was embedded in an energy system model to provide dynamic predictions of the energy conversion efficiency in the reactor, providing a more realistic performance assessment compared to the average efficiency models commonly used in the literature.

Rapid, sustainable redesign of large functional molecules demands efficient exploration of vast chemical spaces. Chemical language models (CLMs), especially transformers, can learn long-range structure-property relationships and enable fast candidate generation after training. However, inverse molecular design is ill-posed – many structures can meet the same target – and conditioned generation often decodes to invalid or off-spec molecules. To address this challenge, we propose a CLM-based inverse design framework that optimises latent representations toward target properties and explicitly evaluates round-trip fidelity, i.e., whether decoded candidates remain on-target after decoding and forward re-evaluation. To improve reliability, we introduce post-decoding beam re-ranking using round-trip consistency and a predictor-guided minimal-edit repair step that corrects invalid near-misses while preserving closeness to the target property. We demonstrate the approach on surfactant critical micelle concentration (CMC) design, benchmarking existing large pretrained CLMs against our lightweight domain-trained CLM. The framework produces a high proportion of valid and diverse molecules (~90%) while maintaining target property error near 1%. Moreover, atom-level saliency analysis confirms that the generated structures follow established surfactant design rules, supporting interpretable structure–property control. Overall, the framework provides an efficient and broadly applicable solution to reliable inverse design of novel functional molecules.

The availability of temperature-dependent physicochemical property data forms the cornerstone of process simulation, optimization, and sustainable molecular and product design. However, a critical data gap persists, as experimental measurements are accessible for only a small subset of known chemicals. This renders experimental characterization resource-prohibitive, often compelling reliance on empirical estimation methods. Moreover, although many models offer single-point predictions at fixed temperatures, accurately modeling continuous temperature-dependent behavior remains challenging. Conventional methods frequently overlook intermediate variations, resulting in limited extrapolation capability. To overcome these limitations, we introduce a mechanism-guided hybrid modeling framework that integrates physical insights into data-driven models. This framework is built on two strategies. Strategy I targets trend correction by generating a continuous representation from discrete single-point predictions, incorporating descriptors and slopes. Strategy II addresses bias removal by anchoring a baseline to a high-accuracy point estimate and fitting the remaining deviations. The framework’s effectiveness is evidenced by evaluations across ten thermophysical properties: Strategy I achieves MSE reductions of 19.23% and 20.33% for the quantitative structure-property relationship and group contribution methods, respectively. Strategy II provides a more substantial improvement, attaining an 81.63% MSE reduction for the gradient boosting decision tree regression model. This work demonstrates that incorporating trend and slope constraints facilitates physically consistent, bias-corrected, and accurate predictions, offering a scalable approach to bridge the data gap and accelerate computer-aided engineering and design.

Accurately predicting chemical reaction yields can accelerate reaction optimization by prioritizing promising conditions; however, the complexity of multicomponent reactions and the limited availability of high-quality datasets remain significant challenges. While machine learning has achieved substantial progress in molecular property prediction, reaction-level modeling requires representations that capture three-dimensional structures, intercomponent interactions, and mechanistically critical features. In this study, we investigate the effective extension of molecular pretraining to reaction yield prediction from two complementary perspectives. First, we apply a reaction-aware architectural design based on molecular representation models pretrained via partial denoising to multicomponent reactions. Each reaction component is encoded as a three-dimensional stereoisomer embedding and concatenated into a reaction-level representation, with multi-head attention modeling intercomponent dependencies. We show that this approach achieves robust performance and generalizes well to reactions involving previously unseen components. Second, we design an architecture that extracts intermolecular relational embeddings through mechanistically meaningful pre-supervised learning using electron source-sink annotations derived from reaction mechanisms, followed by an active fine-tuning process. This mechanism-aware extension leverages large-scale reaction data to provide informative inductive bias and improves performance in data-constrained settings. We further confirm that this approach outperforms models that simply combine individual molecular embeddings, as it directly captures chemically meaningful information relevant to reaction outcomes. These results suggest promising directions for developing machine-learning models for reaction yield prediction that are more closely aligned with underlying chemical mechanisms.

The present work aims to demonstrate the synergy achieved through the integration of biomass gasification with a direct air capture (DAC) system to maximize overall CO2 removal capacity, while simultaneously converting waste into value-added products (hydrogen) and supplying the energy required for DAC operation (BG-H2P-DAC). The proposed configuration is modeled using Aspen Plus to investigate the synergistic interactions and key performance indicators of the BG-H2P-DAC system. Parametric analyses are conducted by varying gasification temperature, air inlet flow rate, and amine concentration and flow rate. The results indicate that increasing the monoethanolamine (MEA) concentration from 10% to 40% leads to a gradual decline in CO2 capture efficiency, accompanied by a reduction in CO2 slip. The system achieves a net specific electricity consumption of 0.0293 MWh/t CO2, confirming that the electricity generated from the integrated steam power cycle is sufficient to fully offset the electrical requirements of the DAC process. The regeneration heat requirement at a steam temperature of 150 °C is 1.32 MWh/t CO2, which represents the total net thermal demand of the DAC unit.

The rapid integration of renewable energy sources is increasing the volatility and non-stationarity of modern power systems, posing significant challenges for data-driven forecasting models. In particular, concept drift and uncertainty in exogenous inputs such as weather forecasts can severely degrade predictive performance over time. This work proposes a lightweight two-layer forecasting framework that decouples prediction from adaptation. A traditional offline regression model is augmented by an online meta-learner that continuously generates adaptive meta-features, enabling the system to respond to structural changes and noisy inputs without repeated retraining. The framework is evaluated on two real-world case studies. First, concept drift is addressed in nuclear power production forecasting, where abrupt and gradual capacity changes are inferred through an online meta-learner. Second, feature uncertainty is mitigated in day-ahead solar production forecasting by correcting noisy weather forecast inputs. Across both scenarios, the proposed approach consistently outperforms single-layer baselines, reducing root mean squared error by up to 10% and maintaining robust performance over multi-year horizons without retraining. The results demonstrate that meta-learning provides a practical and computationally efficient mechanism for improving forecast robustness in non-stationary energy systems, with applicability to a wide range of power-system forecasting problems.

Direct air capture is gaining significant interest due to its potential to achieve carbon-negative emissions. With the aim to reduce the energy consumption and in line with the electrification of chemical processes, the absorption direct air capture system is integrated into a bipolar membrane electrodialysis cell stack for solvent regeneration and carbon dioxide release. The scheme solution is characterized by carbon dioxide bubbles inside the cell reducing efficiency so that other regeneration schemes have been proposed in the literature. A direct comparison of those is missing in the existing state of the art: the present work wants to fill this gap. In addition to the above process, the reaction of the rich solution with a weak organic acid and the use of both nanofiltration and reverse osmosis membranes are considered in other process schemes. The three case studies are modelled in Aspen Plus software with the aim to compare the energy consumption and total cost while the environmental impact is evaluated with the life cycle assessment methodology. Results show that the best capture process from economic and environmental point of views is that using only a bipolar membrane electrodialysis for solvent regeneration and carbon dioxide release despite a higher energy consumption for the cell stack. Here total capture cost and environmental impact are respectively 480 $/tonCO2 and -0.9593 kgCO2eq/kgCO2 while the energy requirement in the BPMED stack is 408 kJ/molCO2.

Polystyrene (PS) recycling remains limited despite large waste volumes, largely because many routes struggle to recover high-value chemicals at scale. This work develops a full process concept that upgrades PS through low-pressure hydrothermal processing (HTP) and directs the resulting aromatic oil to separation and downstream conversion. The superstructure includes HTP at 30 bar and 350°C, three distillation columns for toluene/ethylbenzene/styrene splits, ethylbenzene dehydrogenation to boost styrene yield, steam reforming of the heavier C9+ fraction to syngas, and an ICI-type (Imperial Chemical Industries) methanol loop with inter-bed quench. The integrated flowsheet was simulated in Aspen Plus V14 using a consistent NRTL-RK (Non-Random Two-Liquid-Redlich–Kwong) property framework. Deep-vacuum operation (˜100–400 mbar) was applied where needed to limit styrene polymerization. Sensitivity and optimization focused on meeting a syngas stoichiometric number near 2; the best case occurred at ~998.8°C and 20.2 bar with ~37.8 t/h steam. Normalized to the polystyrene feed, the model predicts ~0.13 kg toluene, 0.16 kg styrene, and 1.27 kg methanol per kg PS, with only trace benzene/ethanol, at a water-to-PS ratio of ~3.77 kg/kg. When heat recovery is excluded, the overall heat demand corresponds to ~84 MJ per kg of PS processed. The flowsheet is technically consistent, but heat integration and a full TEA/LCA (Techno-economic Assessment/Life-cycle Assessment) will ultimately determine its viability.

The interdependence among energy, water, food, and land sectors has been addressed through the concept of Energy-Water-Food-Land nexus (EWFLN), where interconnections between different sectors generate complex feedback loops. In the field of EWFLN, transitioning from deterministic to stochastic approach is considered as the natural choice for policy makers. Working with a large number of uncertain parameters can make a stochastic system complex. Therefore, Identification of the significant uncertain parameters in the system would create a more acceptable model before transforming from a deterministic to a stochastic approach. This study incorporates uncertainty prioritisation in the mathematical model for the optimisation of EWFLN. The specific objectives of this study include creating a mathematical model, determining and prioritising uncertain parameters, and prescribing appropriate policy recommendations. This study points out clearly which specific parameters should be taken into consideration for risk hedging. The sensitivity analysis results present a quantitative hierarchical ranking of uncertain parameters based on their influence on the objective function. The results demonstrate that energy price, solar radiation flux and crop yield are the most sensitive among them all. Moreover, it provides information about prioritised uncertain parameters to the policy makers to avoid dangerous risk of food, water, and energy scarcity.

The global transition toward sustainable energy has intensified research into biofuels, with bioprocess optimization playing a central role in achieving decarbonization goals. Biobutanol, in particular, is a high-value molecule for sustainable fuel applications due to its superior energy density and compatibility with existing infrastructure. However, model-based optimization of its production is hindered by traditional semi-structured kinetic models that often suffer from limited predictive robustness. To address this challenge, within this study we developed a hybrid modeling framework for Clostridium saccharoperbutylacetonicum that integrates mechanistic mass-balance equations with Gaussian Processes (GPs) aiming to describe the biobutanol formation rate. Here, we investigate the effect of data normalization techniques on hybrid model’s prediction capabilities comparing min-max normalization, z-score normalization, and no transformation. For each data treatment strategy, 8, 000 hybrid models were trained using experimental fermentation datasets, and ensemble predictors were constructed to mitigate variability and enhance generalization. Results demonstrate that hybrid models significantly outperform parametric baselines, with the ensemble approach leveling the performance differences across all pretreatment techniques. Ensembles achieved validation RMSE reductions of approximately 10.1% to 10.4%, which are two times higher than the best individual models regardless of the scaling method used. Despite similar RMSE reductions, prediction envelopes varied along data transformations revealing tradeoffs between uncertainty and coverage. Notably, targeted hybridization of the butanol reaction rate propagated performance improvements to other states, such as glucose and biomass. These findings suggest that for hybrid modeling, ensemble design provides greater benefits for robustness than any data normalization technique.

Environmental concerns caused by CO2 emissions has attracted much attention by researchers worldwide. CO2 can be captured from large single sources such as power plants to reduce the CO2 emission. Solvent-based post-combustion carbon capture (PCC) process for large scale biomass-fired power plant could achieve negative carbon emission. However, capture level is commonly set at 90% in many studies. The small fraction of residual CO2 is still a large amount due to the high flue gas flowrate. In this study, a piperazine-based PCC process at 95% capture level for biomass-fired power plant was studied. The process was simulated in Aspen Plus® V11, validated and scaled up. The energy performance results showed that when the capture level is increased to 95%, the reboiler duty rises to 4.07 GJ/tCO2, corresponding to an increase of approximately 13.7% compared to the 90% case. This additional regeneration energy demand is offset by the reduction in residual CO2 emissions from flue gas or 0.23 million tons extra CO2 captured each year (from 3.86 million tons CO2/year to 4.09 million tons CO2/year). It is feasible for improving negative emission performance in BECCS systems. More analysis (e.g. different configurations and economic analysis) will be performed to further discuss the high capture level process.

The design of cyclic adsorption processes is computationally expensive as it involves screening many process designs, each of which involve a time-consuming simulation to reach cyclic steady state. In this work, we demonstrate how differentiable programming can be used to accelerate both the simulation and attainment of cyclic steady state for a four-step pressure vacuum swing adsorption (PVSA) process to concentrate carbon dioxide from flue gas. A mechanistic one-dimensional dynamic adsorption model was implemented in JAX, enabling automatic differentiation and just-in-time compilation for efficient solution and accurate sensitivity evaluation. The latter was exploited to implement a Newton-based direct determination method for accelerated convergence to cyclic steady state, avoiding repeated cycle simulations. Across 4096 designs sampled in a six-dimensional design space, the direct determination method converged in an average of 4.6 iterations, compared to 145 cycles required by successive substitution. When combined with the computational gains from the JAX framework, this resulted in an overall speed-up of over 20 times relative to the conventional MATLAB-based implementation.

Manufacturing kinetically stable products remains a challenge in the agrochemical industry. Current agrochemical formulation design relies on semi-empirical and trial-and-error methods. The inconsistency is caused by the lack of a mechanistic understanding of the formulation, making the design a black-box optimisation problem. In addition, validating the ground truth of the high-dimensional design space is expensive, driving chemists to explore possible solutions using data-driven methods. We proposed a Bayesian optimisation framework employing a Gaussian process as the surrogate model to intelligently guide the screening of the design space. The uniqueness of our framework is the application to the classification task to increase the number of hits of stable formulation recipes. The framework was tested on a provided industry dataset with a focus on emulsifiable concentrates. The performance reached a comparable accuracy with only ~25% of the data being sampled and hit more stable formulations than a Monte Carlo search. Our framework accelerates the discovery of stable recipes and guides formulation screening.

Biopharmaceutical manufacturing requires systematic identification, understanding and control of critical process parameters (CPPs) and critical quality attributes (CQAs) While deterministic machine learning models have proven valuable for describing complex systems and providing insights into process behaviour, the extension to probabilistic frameworks allows for the capture of intrinsic biological and process variability, improving robustness, understanding and safety. This work introduces a framework for autoregressive variational Bayesian neural networks (BNNs) that integrate uncertainty quantification into data-driven modelling. The approach is demonstrated on a multicolumn countercurrent solvent gradient purification (MCSGP) process for monoclonal antibody purification. The BNN is trained autoregressively on high fidelity data generated using a detailed mechanistic process model data. A heteroscedastic variance head is included to model signal-dependent observation noise, while the epistemic uncertainty is captured through stochastic network parameters. Model performance is assessed using both deterministic accuracy metrics -MSE and probabilistic scores namely coverage and continuous ranked probability scores (CRPS). Results show that the proposed method accurately recovers noiseless concentration profiles while maintaining well calibrated predictive distributions. Additionally, the framework is tested under multiple noise regimes to test robustness of the proposed approach.

The pharmaceutical industry is placing growing emphasis on sophisticated process modeling to enhance the efficiency of drug design and production pipelines. Optimal control over these models can significantly improve manufacturing performance by lowering costs, boosting productivity, and ensuring rigorous quality compliance. However, the intricate nature and heavy computational load of these models often require the adoption of more practical or simplified alternative strategies for optimisation such as simulation-based approaches. In this work, we introduce a simulation-based framework including a “top-level” gradient-based mathematical programming optimisation model coupled with a “low-level” simulation scheme, to optimise multi-scale drug substance manufacturing flowsheets. The proposed framework optimises critical quality attributes, such as yield and purity, including green metrics such as process mass intensity, aligning with digital platforms (e.g. gPROMS) used in the pharmaceutical industry. By constructing Pareto fronts, we capture the balance between conflicting objectives, providing valuable insight for informed decision-making. The results indicate that this simulation-based approach can effectively optimise complex process behaviors while improving computational efficiency as no preliminary data is needed. By applying the framework to a real-world case study, we demonstrate its potential in optimising complex pharmaceutical manufacturing flowsheets.

The need to reduce greenhouse gas emissions and diversify energy sources has driven the development of hydrogen (H2) as a leading carbon-free energy vector with high gravimetric energy density (33.3 kWh/kg) for stationary and transport applications. However, its storage and transportation remain challenging. Therefore, ammonia emerges as a promising hydrogen carrier due to its high energy density and ease of liquefaction and transport. This work presents the development of a phenomenological based mathematical model for an ammonia electrolysis cell operating in a zero-gap configuration with an anion exchange membrane. The model considers the contributions of different overpotentials (activation, ohmic, concentration) and incorporates empirical and semi-phenomenological expressions adjusted to experimental data from Zhang et al [1]. The model shows both qualitative and quantitative agreement with experimental measurements, achieving a determination coefficient (R2) of 0.9874. Beyond electrochemical modeling, the cell model is integrated into a nonlinear programming (NLP) optimization framework to minimize the total annualized cost (TAC) of the system. For hydrogen production of 1250 kg/h, the optimal solution corresponds to 353 K, 0.82 A/cm2, and 1302 cells, yielding a TAC of about 45.0 MUSD/year.

Simulation and modeling technologies play an increasingly critical role in enhancing feedstock utilization, improving product quality, and enhancing enterprise competitiveness in the refining industry. Molecular-level kinetic models for refining processes typically involve highly detailed and complex data, significantly increasing computational costs. Additionally, the data granularity within these models fails to align with the practical requirements of industrial production. To address these, this study proposes the Reaction Network Coarse-Graining Integration (RNCGI) algorithm to achieve flexible adjustments to model complexity and data granularity. Species with similar molecular structures and kinetic behaviors within the reaction network are identified and integrated. Subsequently, a kinetic parameter mapping approach is developed to preserve the critical reaction behaviors. Finally, the reaction coarse-graining process significantly reduces the number of reactions within the network. The RNCGI algorithm was validated using the naphtha catalytic reforming process. The original complete network with 206 species nodes was integrated into two versions with different data granularities: a medium-scale network with 42 nodes and a small-scale network with 5 nodes. Both coarse-grained networks calculated the product composition accurately and provided clearer insights into core species transformation mechanisms compared to the original network. Additionally, the RNCGI algorithm ensures the completeness of reaction routes. By flexibly handling detailed data in molecular-level models, this work preserves the necessary complexity of models tailored to industrial needs, supporting the development of refinery-wide production planning and scheduling strategies.

Chemical processes are characterized by high complexity and inherent hazards, necessitating systematic and comprehensive risk assessment methodologies. However, traditional risk assessment approaches are often limited to static analyses and small-scale models, resulting in insufficient coverage and a lack of detailed, unit-level insights. This paper proposes an automated Bayesian network–based dynamic risk assessment (DRA) framework for full-process risk analysis in chemical plants. The proposed method automatically extracts knowledge from established risk analysis documents, such as HAZOP reports, and integrates it with dynamic process monitoring data, expert knowledge, and reliability databases. By effectively leveraging multi-source heterogeneous data, an equipment-level causal inference network is constructed automatically, addressing the high labor demand and limited scalability of conventional DRA approaches. The proposed framework is validated through an industrial FCC case study at a large petrochemical enterprise in Zhejiang, China. The results demonstrate that the method can effectively construct Bayesian causal networks for chemical processes and quantitatively evaluate both the likelihood and severity of accident consequences under various failure scenarios, thereby enabling systematic and dynamic risk assessment of chemical processes.

Electrochemical reduction of CO2 (ERCO2) in gas diffusion electrode (GDE)-based electrolyzers represents a potential strategy for global decarbonization, achieving simultaneously the valorization of this abundant carbon resource. While significant progress has been achieved in enhancing CO2 conversion in these systems, further advances are required to enable their practical implementation at the industrial scale. Physics-based simulations offer a powerful tool to guide the optimization of design and operating parameters as well as for the efficient scale-up of CO2 electrolyzers. In this work, we have developed a three-dimensional multiphysics model of the cathodic compartment of a GDE electrolyzer for ERCO2 to formate. For that purpose, the software COMSOL Multiphysics has been used. The model is experimentally validated, confirming its accuracy at reproducing current density and Faradaic efficiency at cathode potentials in the range -1.2 V and -1.7 V. Moreover, kinetic parameters are fitted to experimental data performing several parametric sweeps to minimize discrepancies between simulated and measured current densities. We obtain charge transfer coefficients (alpha_c_k) of 0.07 and 0.34, together with exchange current densities (i0, k) of 30 mA·cm-2 and 10-4 mA·cm-2 for ERCO2 and the competing hydrogen evolution reaction, respectively. Finally, the model is used to predict formate concentration under varying applied potential conditions. Collectively, our three-dimensional multiphysics model reliably predicts CO2 conversion to formate, thus representing a useful tool for guiding system optimization and scale-up. Moreover, the systematic methodology followed for developing the model can be readily extended to the design and analysis of other electrochemical cells beyond CO2-to-formate conversion.

Multiscale models provide a powerful framework to link bioprocess operation conditions with polymer microstructure, yet their predictive capability for polymer attributes such as chain length distributions (CLDs) remains limited. In this work, an advanced multiscale modeling framework for the microbial production of poly(3-hydroxybutyrate-co-3-hydroxyvalerate) (PHBV) in Cupriavidus necator is presented, targeting the quantitative prediction of polymer microstructure. The model consistently integrates a structured macroscopic kinetic description of substrate uptake, biomass growth, and copolymer accumulation with an explicitly formulated microscopic polymerization model resolving initiation, propagation, termination, and depolymerization reactions of living and dead chains. A central contribution of this study is the quantitative calibration of the polymerization kinetics based on experimental size-exclusion chromatography (SEC) data. Polymerization rate constants were identified by fitting simulated CLDs to measurements obtained at multiple diagnostic time points during fed-batch cultivations using fructose and propionic acid as carbon sources. Parameter estimation was performed using a hybrid multi-start optimization strategy combining simulated annealing with Gauss–Newton refinement. Compared to literature-based parameter sets, the optimized model accurately reproduces both the temporal evolution of the weight-average molecular weight and the full CLD shape, including the emergence of high-molecular-weight tails. The results demonstrate that explicit polymerization modeling combined with parameter identification is essential for quantitatively linking process dynamics to PHBV microstructure. The proposed framework provides a sophisticated basis for model-based optimization of feed strategies and supports quality-by-design approaches for sustainable biopolymer production.

Hybrid process models are increasingly used for real-time decision support in dynamic operations, where models must remain reliable under changing operating conditions. In such settings, models are often required to extrapolate beyond previously observed batch trajectories, yet conventional validation strategies may fail to reveal weaknesses in extrapolative behavior. This work investigates the effect of hybrid model structure on extrapolative performance using a pilot-scale batch crystallizer as a test case. A structured set of mechanistic and hybrid models is evaluated using nested batchwise leave-one-out cross-validation (NBLOOCV), in which entire batches are withheld to assess extrapolation across operating regimes. For this specific application, the results show that hybrid models employing simple linear correction terms consistently outperform more flexible neural network–based formulations under extrapolation, despite comparable training performance. For the studied process, the findings highlight the importance of validation strategies that reflect intended model use and demonstrate that modest, physically consistent hybridization provides a robust and interpretable basis for operational decision support.

Near-infrared (NIR) sorting systems in waste sorting plants operate under multiple settings, creating distinct input–output relationships that challenge predictive modeling. Conventional neural networks, such as multilayer perceptron (MLP), often suffer from catastrophic forgetting under continual training, limiting reliability across settings. This study evaluates Kolmogorov–Arnold Networks (KAN) for continual regression modeling of multi-setting NIR systems. KAN assign nonlinear transformations to network edges using localized spline grids, enabling structural isolation between input regions. We introduce controlled input-space manipulations (shifting successive settings to adjacent or non-overlapping grid regions) and compare KAN performance with MLPs of comparable parameter count. We also examine single-input versus multi-input configurations to assess dimensionality effects. Results show that KANs with sufficient input-space separation maintain previously learned knowledge with perfect resistance to forgetting, whereas overlapping inputs induce interference. MLP forgetting depends on learning rate and training duration and cannot be fully avoided without additional methods. Single-input KANs achieve comparable accuracy to multi-input models in this system, suggesting limited benefit from additional inputs. These findings demonstrate that KAN’s structural locality, combined with controlled input-space alignment, provides a practical and robust approach for continual learning in industrial NIR systems.

Renewable-to-X products such as methanol (MeOH) and dimethyl ether (DME) offer scalable, carbon-neutral options for decentralized chemical production. Microreactors, with superior heat and mass transfer, provide more controllable reaction environments. This improved control enhances selectivity and conversion, making microreactors particularly well suited for intensifying CO2/CO hydrogenation within a Power to X framework for synthetic products. However, an assessment of MeOH and DME synthesis routes under microreactor operation is still lacking. To address this gap, a microreactor-scale model was developed where two reactor configurations were analyzed: i) parallel configuration, in which MeOH synthesis and subsequent dehydration to DME take place in the same reactor, and ii) series configuration, in which MeOH synthesis is carried out in the first reactor, followed by MeOH dehydration to DME in a second reactor. To capture realistic process behavior, the simulations incorporated non-isothermal, non-isobaric operation and fugacity-based reaction rates. The comparison reveals that reaction decoupling in the series configuration enables higher overall conversion of MeOH to DME, while the parallel route is constrained by strong kinetic and thermal coupling between hydrogenation, dehydration, and water formation. Despite requiring fewer unit operations, the parallel route yields mixed MeOH–DME product streams, whereas the series pathway favors higher DME productivity. The results highlight a fundamental trade-off between process intensification and conversion efficiency in microreactor-based Renewable-to-X systems.

This work proposes a parsimonious dynamic filtration model for superheated steam containing paper-derived dust, suitable for parameter identification, prediction, and future optimization under limited observability. The model is based on Darcy’s law, with the total resistance expressed as the sum of the intrinsic filter resistance and a time-dependent fouling contribution. Experimental data obtained from a dedicated superheated-steam filtration setup were used for parameter estimation and model validation under a single operating condition. Assuming a linear dust dosing rate, the model yields limited agreement with experimental data (R² = 0.24). By estimating the time-varying solid loading through minimization of the sum of squared errors between measured and predicted pressure drop, the agreement improves significantly (R² = 0.94). This demonstrates that uncertainties in the dust dosing rate strongly affect pressure drop predictions. The proposed model provides a foundation for extending the analysis to broader process conditions, enabling statistical characterization of fitted parameters, integration with a filter cleaning model, and the development of optimization strategies for long-term, energy-efficient operation of closed-loop superheated steam drying systems.

Catalyst deactivation significantly affects reactor performance, process efficiency, and economic viability in chemical processes. The precise estimation of kinetic and deactivation parameters in transient tubular reactors is essential but remains challenging due to strong parameter correlations, nonlinear dynamics, and limited prior knowledge of parameter values. Model-based Design of Experiments techniques for improving parameter precision (MBDoE-PP) has been shown to enhance parameter identifiability by optimally designing informative transient experiments even when catalyst deactivation occurs. However, MBDoE-PP is highly sensitive to parameter misspecification and can lead to suboptimal or infeasible solutions under model uncertainty, in which is the typical case in reaction systems exhibiting catalyst deactivation. In this work, a robust MBDoE framework for parameter precision (RMBDoE-PP) is proposed to explicitly account for processmodel parameter mismatch (PMPM) during the experimental design stage. In the framework, stochastic sampling is used to evaluate metrics of the Fisher Information Matrix (FIM) and the experimental design is carried out considering a worst-case scenario, where the globally least informative parameter corresponding to the minimum information content across the uncertainty domain is identified. The performance of MBDoE-PP and RMBDoE-PP is assessed in the sequential closed-loop experimental design framework under varying levels of PMPM. Results demonstrate that RMBDoE-PP consistently yields tighter confidence intervals and more accurate parameter estimates with fewer experimental runs, confirming its superior robustness for deactivation kinetics.

Direct Air Capture (DAC) is recognised as a critical climate mitigation technology necessary for achieving global net-zero emissions by balancing difficult-to-avoid emissions. Despite its importance, the commercial deployment of DAC technology is currently challenged by the substantial energy demands and resultant extremely high operational costs associated with handling the low atmospheric CO2 concentration. This study aims to address these two challenges through dynamic process modelling, simulation and rigorous multi-objective optimisation of a solid sorbent-based temperature vacuum swing adsorption (S-TVSA) cycle. The system utilises an advanced amine-functionalized sorbent, selected for its favourable low-temperature regeneration kinetics.A technical performance assessment was conducted using a first-principle mathematical model to accurately simulate mass and heat transfer in the adsorbent beds. To improve system viability, a multi-objective optimisation with NSGA-II in MATLAB was performed, maximising CO2 productivity and minimising energy demand. The optimisation targeted key parameters like adsorption/desorption time, temperature, and pressure. The Pareto frontier shows trade-offs, identifying optimal points that minimise Weq and energy use. These results provide a pathway to reduce energy costs and enhance the economic viability of solid-sorbent DAC technology.

Stochastic gene regulatory networks exhibit complex dynamics that require efficient methods for parameter inference and uncertainty quantification. In this work, we propose a surrogate modelling framework that combines a partial integro-differential equation (PIDE) formulation with polynomial chaos expansions (PCE) to efficiently approximate the stochastic dynamics of gene expression models under parametric uncertainty. The approach represents the time evolution of low-order statistical moments as polynomial functions of uncertain kinetic parameters, enabling fast evaluations and tractable inference. The method is demonstrated on a self-regulating gene network, achieving accurate parameter estimation and a reduction of approximately two orders of magnitude in computational cost compared to direct PIDE-based optimisation.

Developing predictive, high-resolution microbial models that retain mechanistic insight remains a central challenge in biochemical engineering. This paper addresses the challenge of preserving metabolic information while ensuring model accuracy and proper statistical definition. It employs the Hybrid Cybernetic Modelling (HCM) approach to integrate metabolic regulation with genome-scale information, and dynamically predict E.Coli phenotypes. The main aim of this work is to explore HCM’s parameter identifiability to advance its accuracy and robustness for a limited set of data. To bypass the computational burden of computing the elementary flux modes, the opt-yield Flux Balance Analysis (opt-yield FBA) is employed to identify a physiologically relevant set of yield-maximising metabolic pathways. Metabolic Yield Analysis (MYA) then reduces this to four key pathways which capture 99% of the original metabolic yield space. The cybernetic model is formulated where “artificial enzymes” are allocated among these key competing pathways, based on a return-on-investment criterion. Initial parameter estimates are derived through non-linear least squares regression for an in silico data set of biomass, glucose and acetate concentration. Practical identifiability analysis is employed to explore parameter estimates. Specifically, profile likelihood & sensitivity analysis results lead to model reduction by eliminating the combined acetate-biomass producing pathway. To further improve parameter uncertainty, Optimal Experiment Design (OED) is used to obtain an additional informative experiment. Finally, the calibrated model is validated against a new artificial dataset.

Predictive modeling of surfactant properties is constrained by limited experimental datasets, a common challenge in specialty chemical development where property measurements require specialized equipment and significant time investment. In this study, we address data scarcity through a semi-supervised generative augmentation framework that leverages both labeled and unlabeled molecular data for surface tension prediction. We implemented a two-stage variational autoencoder (VAE) training strategy using a curated database of 600 non-ionic surfactants. First, 461 unlabeled surfactant structures were used for VAE pre-training to learn latent representations capturing molecular connectivity patterns and amphiphilic relationships. Second, 125 molecules with surface tension measurements were used for fine-tuning to embed property-structure relationships. Our stratified generation framework produces surfactants matching target property distributions (Wasserstein distance = 0.030, KS statistic = 0.152) through stratified sampling and multi-criteria filtering, achieving 88.5% structural validity. We systematically evaluated augmentation effectiveness across seven levels (×1.25 to ×5) using graph convolutional networks with five random seeds per level. Results show optimal performance at ×3 augmentation (375 training molecules), where test R² improved from 0.60 (baseline, 125 molecules) to 0.78 (+28%), with RMSE decreasing to 0.056 mN/m. Multi-seed analysis demonstrates stable and reproducible training (test R² = 0.776 ± 0.030). This framework achieves performance equivalent to tripling the experimental dataset, providing a practical and scalable approach for specialty chemical property prediction where structural data is abundant but property labels are scarce.

The rapid electrification of the energy and transport sectors has increased the demand for efficient thermal management of Lithium-Ion Batteries (LIBs). LIBs operate optimally within a narrow temperature range (15–40 °C), effective Battery Thermal Management Systems (BTMS) are essential to prevent accelerated aging and performance degradation. Immersion cooling using dielectric fluids has emerged as a promising solution; however, a systematic comparison of their thermal and hydraulic performance remains limited. This study evaluates six dielectric fluids—Silicone Oil, Mineral Oil, Sunflower Oil, AmpCoolAC-100, E5-TM410, and Deionized Water—using Computational Fluid Dynamics (CFD) simulations of a lithium-ion battery pack composed of 32 fully immersed 18650 cells. The analysis focuses on key thermophysical properties, particularly the Prandtl number and thermal diffusivity, as criteria for coolant selection. The results demonstrate a strong correlation between cooling performance and the Prandtl number. Fluids with lower Prandtl numbers provide superior heat dissipation and lower hydraulic losses. Deionized water exhibited the best overall performance, maintaining maximum cell temperatures at 31.04 °C under a 2 LPM flow rate and achieving up to 31.33% better cooling than silicone oil. In contrast, high-viscosity oils showed poor thermal performance and significantly higher pressure drops. This work establishes the Prandtl number as a critical and reliable parameter for selecting dielectric fluids in immersion-based BTMS, highlighting deionized water as the most effective coolant among those evaluated.

As greenhouse gas emissions accelerate global warming, new capture and storage technologies are essential for reducing the industrial CO2 concentration in the atmosphere. This study addresses the urgent need for greenhouse gas capture technologies by developing a detailed dynamic mathematical model for Chemical Looping Process with Integrated Combustion and CO2 capture (CL-ICCC). In the CL-ICCC process configuration, CO2 capture is integrated into the chemical looping combustion system, resulting in a higher-purity, more efficient process. In this work Cu/CuO oxygen carrier material and CaO/CaCO3 sorbent materials were considered in a fixed bed reactor as solid phase to investigate Oxidation and Reduction/Calcination processes under different operating conditions. The simulation results were compared with experimental results from the literature. In case of the oxidation process, a sensitivity study was performed to investigate the behavior of the process for variation of different operating parameters under dynamic conditions: gas composition, initial gas temperature and solid phase mass composition. In the case of reduction/calcination, multiple fuel type and solid composition were investigated.

Styrene is mainly produced by dehydrogenating ethylbenzene over an iron oxide-based catalyst. The reaction is endothermic and thermodynamically limited when operated in conventional catalytic fixed-bed reactors. This makes the styrene production process a conversion-selectivity trade-off, where different objectives must be compromised. In this work, a one-dimensional reactor model accounting for changes in molar flowrate, temperature, and pressure is used to predict the performance of a membrane reactor. Three main hydrogen flux profiles were assumed along the reactor axial direction: constant, linearly increasing, and linearly decreasing. It is found that the styrene yield and selectivity in a membrane reactor operated with a linearly decreasing hydrogen flux profile are higher than those with constant or linearly increasing hydrogen flux profiles in both isothermal and nonisothermal cases. It is also observed that the styrene yield and selectivity of the membrane reactor operated with a linearly decreasing hydrogen flux can be enhanced by improving the membrane’s average hydrogen permeability. The influence of nonlinearity in the hydrogen flux equation is marginal and observed mainly at the reactor inlet. The improvement is due to enhanced hydrogen permeation, which is observed to be better for the linearly decreasing hydrogen profile than for the constant, linearly increasing hydrogen flux profile. When a permeation section is included in the reactor at the point where maximum hydrogen production is observed, the styrene yield is better than that of the completely permeated catalytic reactor.

This work presents a data-driven surrogate modeling framework for predicting distillation behavior assuming an infinite number of stages and distillation limits informed by residue-curve topology and pinch-point feasibility analysis. The framework provides a direct mapping from feed composition and distillate-to-feed ratio (D/F) to distillate and bottom product compositions, making it suitable for flowsheet synthesis and optimization applications. The approach combines three components: a classifier that identifies feasible singular-point splits, a boundary regression model that predicts D/F limits separating pure- and mixed-product operating regimes, and a neural network that interpolates product compositions in the intermediate regime. The method is demonstrated for the ternary system ethanol, benzene, and water at 1 atm using data generated from rigorous vapor–liquid–liquid equilibrium analysis. Results show that the framework provides reliable predictions for pure splits while retaining smooth interpolation behavior in the mixed regime.

Activity coefficients are key thermodynamic quantities for describing phase equilibria, but their experimental determination entails laborious and costly phase-equilibrium measurements, making predictive approaches highly desirable. The potential of machine learning for such predictions has received growing attention as an alternative to physics-based models that require experimental data or expensive calculations for parameterization. We propose a physics-informed edge-enhanced graph attention network (PEGAT) to predict activity coefficients in multicomponent mixtures, where each molecule is encoded as a graph in which the nodes correspond to atoms and the edges to chemical bonds. The excess Gibbs free energy of the mixture is predicted using the proposed model, including a nonlinear transformation in the final layer to ensure that the excess Gibbs free energy vanishes for pure components. To further enforce thermodynamic consistency, the relevant activity coefficients are obtained via the Gibbs–Duhem relation. Unlike machine-learning models developed primarily for binary systems, the proposed framework is directly applicable to arbitrary multicomponent mixtures. The PEGAT model is evaluated using a mixed dataset comprising both binary and ternary mixture data and demonstrates high predictive accuracy. Further validation on representative mixtures shows close agreement between predicted and reference activity coefficients. The results confirm that improved thermodynamic consistency can be achieved by embedding hard physical constraints into the graph neural network architecture. However, they also highlight that unphysical behaviors may still be predicted despite these constraints.

Despite the growing economic importance of the fragrance industry, perfume formulation remains largely guided by empirical knowledge and iterative trial-and-error approaches. The structured design of fragrances, typically organised into top, middle, and base notes through the blending of perfume raw materials, is therefore time-consuming, costly, and difficult to generalise. Existing computational approaches have begun to address this challenge, but are commonly limited to small ingredient sets or require extensive sensory data that are not always available. This work proposes a computer-aided optimisation framework for Eau de Parfum formulation that simultaneously maximises perceived olfactory intensity and minimises formulation cost. The resulting optimisation problem is formulated to preserve the structural balance of top, middle, and base notes inherent to the perfume pyramid. Application of the framework to a fruity–floral Eau de Parfum formulation demonstrates a substantial increase in perceived intensity alongside a reduction in total formulation cost of approximately 64%. The proposed approach is intended as a decision-support tool for perfumers, enabling rapid exploration of cost–intensity trade-offs once a candidate ingredient set has been defined. The framework offers a step towards more systematic, data-driven perfume engineering and provides a foundation for future extensions incorporating ingredient selection and sensory constraints.

This work presents a preliminary application of a developing methodology for assessing the operational flexibility of ammonia synthesis loops. Part of this methodology involves systematic dynamic testing in the synthesis loop. In the present case, a synthesis loop equipped with a two-bed, quench-cooled converter operating under variable feed conditions was considered. A high-fidelity model of the converter was developed in gPROMS Process using two-dimensional reactor models from its fixed-bed catalytic reactor library, and a synthesis loop configuration was modeled and designed in the same environment. A series of dynamic tests varied the make-up flow rate across four disturbance amplitudes (±25% and ±50%) and four disturbance durations (10 s, 600 s, 1800 s, 3600 s). The results showed that disturbances of ±50% magnitude led to the violation of one process operational constraint. These findings enable the construction of a preliminary operational map of this system, providing an initial determination of its process boundaries, and therefore its current degree of flexibility. Such knowledge is fundamental to understand how the synthesis loop can be made more flexible so that it suits Power-to-Ammonia applications. Overall, this study constitutes an initial step toward the development of a framework for enhancing the flexibility of ammonia synthesis loops.

Heat exchangers frequently suffer from fouling, which is the accumulation of unwanted deposits on heat-transfer surfaces. This issue reduces thermal performance, increases pressure drop, and raises energy use and operating costs. Predicting fouling resistance remains challenging in process engineering, yet it is important for monitoring, maintenance planning, and mitigation actions that reduce economic losses and environmental impacts. Symbolic regression (SR) is a machine learning approach that searches for an explicit mathematical expression that best represents the relationship between process inputs and a target output. Unlike many black-box models, SR can capture nonlinear behavior while producing compact, interpretable equations that are easier to deploy and analyze in industrial settings. In this work, a methodology to rapidly obtain algebraic models for fouling resistance in industrial heat exchangers using SR was proposed. Plant measurements of hot- and cold-side flow rates and inlet/outlet temperatures were transformed into dimensionless numbers, which were used as model inputs to predict fouling resistance. SR was applied to operational data from industrial heat exchangers, achieving coefficients of determination (R²) above 0.95 and 0.87 for the training and validation datasets, respectively, with low prediction errors. A selected equation was then transferred to a second industrial heat exchanger by re-estimating a reduced set of parameters using plant data. Validation on data not used during parameter estimation showed that the adapted model maintained high predictive accuracy, with R² values of 0.96 and 0.91 for training and validation, respectively. Overall, the proposed approach provided an efficient and interpretable framework for fouling resistance modeling and supported practical deployment across similar units.

We present a steady-state, optimization-based techno-economic study of a continuous fluidized temperature-swing adsorption (TSA) system for post-combustion CO2 capture from biomass-derived flue gas, using two adsorption stages and one desorption stage with integrated heat-pump thermal management. The GAMS/CONOPT4 model couples molar and energy balances, Toth adsorption equilibrium, fluidized-bed hydrodynamics and literature cost correlations. Optimization yields CO2 purity of 96% v/v and 95.5% recovery at low, safe pressures with particle Reynolds numbers of 2-11, indicating near-minimum-fluidization operation. The nominal capture cost is 87 USD/tonCO2 with an internal rate of return of 42%; utilities comprise 49% of annualized costs and the adsorption compressor dominates equipment capital. Disabling the heat pump increases modeled capture cost to 124 USD/tonCO2, highlighting the heat pump’s decisive role in reducing energy demand and costs. Adding adsorption stages lowers modeled cost further but produces impractically shallow beds, revealing a trade-off between mass-transfer performance and feasible bed geometry. A sensitivity analysis demonstrated a linear positive correlation between electricity cost and capture cost, with capture costs potentially as low as 50 USD/tonCO2 and an internal rate of return as high as 64%. While results demonstrate the technical and economic plausibility of a heat-pump-assisted fluidized TSA under the stated assumptions, the study recommends including flue contaminants, refined heat-exchanger and pressure-drop models, and experimental validation for scale-up.

Polylactic acid (PLA) is the most produced bioplastic, however, for it to compete with fossil-based plastics, maximum production efficiency is crucial. To achieve this, all scales, from catalyst scale to process scale, must be simultaneously considered. This challenge is particularly relevant for PLA production, where complex interactions of multiple phenomena occur in several unit operations and the development of active non-toxic catalysts is of major importance. For these reasons, developing a framework that allows a comprehensive understanding of the influence of design choices at different levels is of paramount importance. To address this, a multi-scale model was developed for the PLA production, coupling a custom devolatilizer reactor model to a process simulator that is subsequently linked to an environmental assessment tool via Python-based interfaces. With this model, sensitivity analysis was performed to assess the influence of operational variables on the most relevant KPI’s. Results showed that changes of only 5.5 torr in the devolatilizer pressure can lead to total utility consumption and climate change impact variations of 13% and 2%, respectively. This demonstrates the critical role of the devolatilizers in the process, highlighting the importance of developing custom unit operation models coupled to the process simulator. Additionally, the results show that, with this framework, the influence of reactor scale variables on industrial environmental performance can be seamlessly assessed. Overall, this work demonstrates the feasibility and value of integrated multi-scale modeling for sustainable process design and provides a solid foundation for the future development of a multi-scale environmental and economic optimization framework.

Ethanol derived from biomass is a promising renewable fuel; however, its long-term use as a gasoline additive is becoming increasingly uncertain due to the rise of electric vehicles and alternative propulsion technologies. This trend motivates the exploration of higher-value applications for ethanol, particularly in the food and pharmaceutical sectors, where product safety is critical. A key challenge in ethanol purification is breaking the ethanol–water azeotrope, as conventional entrainers such as ethylene glycol or glycerol can leave residual traces that limit ethanol’s use in sensitive markets. Magnesium chloride (MgCl2) offers an effective alternative, enabling high-purity ethanol without introducing hazardous organic residues, while exhibiting favorable hygroscopic properties and operational reliability. Simulating this system is challenging due to strong non-ideal and electrolyte interactions in phase equilibrium. Conducting a rigorous controllability analysis is also difficult; therefore, within an Integrated Design and Control (IDC) framework, accurate approximations are essential for successful implementation. This work presents a multi-objective optimization of a MgCl2-based ethanol dehydration column, simultaneously minimizing the Total Annual Cost (TAC) and the open-loop controllability criterion (A_gamma + gamma_sm). The optimization integrates Aspen Plus® with Python and employs validated surrogate cost models for the preconcentrator and salt concentrator processes. Dynamic boundaries ensure product purity by maintaining the ethanol/water feed below the salt feed. The ASF (Achievement Scalarizing Function) solution achieved a 64.33% improvement in controllability with a 13.37% increase in TAC relative to the reference case reported in the literature, illustrating the trade-off between economic and control objectives. This study demonstrates that incorporating simplified controllability metrics into multi-objective optimization enables the efficient and practical design of complex systems, providing a viable approach for managing cost–control trade-offs in processes with challenging dynamics.

Building a robust calibration curve is essential for accurate quantification of multicomponent mixtures. Matrix effects can distort the proportional relationship between the analyte concentration and instrumental response. In addition, classical machine learning models do not inherently incorporate simple physical constraints, such as the requirement that a zero response must correspond to a zero concentration, which can result in non-zero predictions. To address this limitations, prior-induced calibration models were proposed that inherently embed this physical constraint into the model architecture. A dataset was generated for ethanol electrooxidation products using headspace gas-chromatography-mass spectrometry (HS-GC-MS). Multiple linear regression (MLR), polynomial regression and artificial neural network (ANN) models were trained to investigate the effects of model complexity and the incorporation of physical information on predictive performance. Model selection and complexity were assessed using Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC). The prior-induced polynomial regression model achieved improved predictive accuracy while maintaining low model complexity, whereas the ANN provided comparable accuracy but was strongly penalized due to its substantially higher complexity.

This study develops a multi-objective computer-aided process engineering (CAPE) framework to evaluate integrated sugarcane-based agro-industrial systems combining a high-yield cultivar, Haru-no-Ougi, and the “Inversion Process, ” which reverses the conventional order of sugar crystallization and ethanol fermentation through selective fermentation of reducing sugars. The in-house CAPE tool SugaNol integrates agricultural, industrial, and environmental (life cycle assessment) models to simulate productivity, energy balance, greenhouse-gas (GHG) emissions, and relative economic performance on a per-hectare basis. Four scenarios were analyzed: NiF8–Conventional, KY01-2044–Conventional, Haru-no-Ougi–Conventional, and Haru-no-Ougi–Inversion Process. Simulation results showed that the combined Haru-no-Ougi and Inversion Process system increased total energy-equivalent productivity by approximately 40-45% compared with the baseline NiF8 system. Life cycle GHG emissions were reduced by 4–11%, while relative economic performance improved by approximately 35-40% due to enhanced ethanol and electricity co-production. These results demonstrate that CAPE-based multi-objective simulation enables holistic assessment and design of agro-industrial systems by explicitly linking crop traits and process design, providing a quantitative basis for sustainable sugar–ethanol–energy system design.

As economies advance towards decarbonization, industry follows suit. The ammonia (NH3) sector heavily relies on the energy- and carbon-intensive Haber-Bosch (HB) process, which accounts for nearly 2% of global CO2 emissions due to its reliance on fossil fuels. Emerging technologies are paving the way for fully renewable NH3 production, although the most mature green process still relies on the HB process, entirely replacing fossil fuels with electrolytic green hydrogen (H2). This work introduces the first developments towards the gradual incorporation of green H2 and air-separated nitrogen (N2) into a conventional NH3 production plant. Using Aspen Plus® V14 for modelling and simulation of Steam Methane Reforming (SMR), different scenarios incorporating 0 to 40 % of these alternative feedstocks were analyzed. An economic objective function is used in each scenario’s optimization. To improve green H2 incorporation and ensure operational constraints were met, the simulations used an adapted SMR section proposed in previous work. Preliminary results in the development of this methodology indicate that at 40 % green H2 and 40 % air-separated N2 incorporations, a 29.4 % reduction in carbon emissions was achieved, along with 30.1 % natural gas savings and a 35 % decreased in conventional operational costs (without green H2 and air-separated N2 costs) compared to the base case, where no alternative feedstocks were used. At current market prices, in the same scenario, incorporating alternative feedstocks can increase total operational costs by 52 %.

The zero-pollution ambition of the European Union requires improvements in wastewater treatment to meet increasingly stringent regulations at achievable cost. One promising approach consists in model-based optimal control. However, wastewater treatment plants involve highly nonlinear and time-varying processes, making existing mechanistic models such as the Benchmark Simulation Model no.1 (BSM1) challenging for direct use in online control. Therefore, this study explores a hybrid modeling approach using the Hammerstein-Wiener (HW) structure. The proposed model combines a mechanistic steady-state model, derived from BSM1, with a data-driven approximation of the system dynamics, incorporating low-order linear dynamic models. In this work, the HW model was used as a surrogate for BSM1. The HW surrogate model attained coefficients of determination (R2) often exceeding 0.95 across key water quality indicators, such as total nitrogen and ammonium concentration. This accuracy was found to be substantially better than that of a First-Order-Plus Dead-Time model as conventional low-order linear model, demonstrating the usefulness of the HW structure in capturing key nonlinearities.

Pyrolysis is an emerging green pathway to produce bulk chemicals and sustainable fuels. However, pyrolysis oil requires stabilisation via hydrotreatment, and this process generates an aqueous waste containing alcohols (mainly methanol and ethanol), carboxylic acids, and some ketones. To increase the economic sustainability of biofuels production, there is increasing interest in recovering these valuable chemicals from water. Reliable thermodynamics are necessary to address the separation and design of equipment to fractionate such complex mixtures, with multiple azeotropes and non-idealities. The Non-Random Two-Liquids (NRTL) models in both Aspen Plus V12.0 and COFE V3.7, a license-free software released by AmsterChem, do not accurately reproduce the equilibrium measurements of most of the binary and ternary mixtures involving water, a C1-C4 alcohol, and a light carboxylic acid. This work aims to retune the activity-based model to improve the NRTL model predictivity, using experimental data retrieved from NIST ThermoData Engine V10.1. The regression is performed in Aspen Plus V12 using the Maximum Likelihood algorithm. The resulting re-parametrised NRTL shows improved accuracy and reliability compared to the existing models. For all investigated binary and ternary mixtures, the Average Relative Deviation (ARD) is lower than 20%, and for 75% of them, ARD is lower than 10%. The ARD in the prediction of vapour phase composition drops from 16.0% and 12.1% of the original COFE and Aspen Plus models, respectively, down to 7.0%.

The growing integration of renewable energy sources such as solar and wind power has intensified the demand for advanced energy storage technologies. Lithium–air (Li–O2) batteries are particularly attractive due to their exceptionally high theoretical specific energy, which surpasses that of the conventional lithium-ion system. However, their practical application is hindered by poor reversibility during discharge, primarily due to the formation and decomposition of lithium peroxide (Li2O2), which causes cathode passivation and capacity fading. Since the electrochemical performance of Li–O2 batteries is strongly influenced by the morphology, size, and spatial distribution of Li2O2 crystals, understanding the mechanisms governing their nucleation and growth is critical. To address this challenge, this work proposes a computational framework based on population balance modeling (PBM) to describe Li2O2 crystallization dynamics during battery discharge. The framework integrates population, mass, and energy balances, allowing the coupled analysis of electrochemical kinetics, supersaturation effects, and the evolution of crystal size distributions. Compared with continuum-scale and phase-field models, the PBM approach offers reduced computational cost while naturally accounting for particle size distributions and linking microscopic crystallization phenomena to macroscopic battery performance and degradation. Although simplified assumptions were adopted in this initial formulation, the framework successfully captures the essential discharge behavior of Li–O2 systems. As such, it provides a robust foundation for future model refinements incorporating additional physicochemical mechanisms and more detailed electrochemical and transport phenomena.

The growing global energy demand and the need to reduce dependence on fossil fuels have intensified efforts toward developing renewable alternatives. Among these, biodiesel and ethanol emerged as viable and sustainable fuel sources. In this context, the use of palm oil and ethanol as raw materials represents a promising production route, due to their availability, high productivity per unit of cultivated area, and renewable characteristics. However, ethanolic transesterification still faces challenges, such as lower productivity compared to methanolic processes and higher energy consumption due to reaction characteristics that impact the whole process. Bearing this in mind, this work aims to develop process simulations in Aspen Plus to optimize biodiesel production from palm oil and ethanol, coupled with an energy integration analysis. The process was divided into three main stages: (i) feed preparation, (ii) transesterification reaction, and (iii) separation and purification of biodiesel and glycerin. The key operational parameters evaluated were temperature, pressure, and ethanol:palm oil molar ratio, important factors to ensure reaction efficiency and biodiesel quality. Additionally, energy integration was performed using the Pinch Analysis method, which enabled the identification of major energy demands and potential opportunities for process improvement. Overall, the proposed simulation framework contributes to enhancing the competitiveness of biodiesel production through ethanolysis, reinforcing its role as a sustainable and viable alternative within the global energy mix.

The pharmaceutical sector is essential for ensuring universal access to medicines, demanding ef-ficient supply chains that deliver drugs at optimal prices with minimal delays and shortages. Pharmaceutical supply chains (PSCs) face significant challenges, including strict quality controls, government regulations, drug perishability, high R&D costs, and complex transportation require-ments. The sector is undergoing a shift, driven by the rise of pharmaceutical components in emerging markets, unpredictable demand, and reduced R&D investments by major companies, which struggle to compete with generic pharmaceutical brands. Post-pandemic challenges and geopolitical risks have further exposed vulnerabilities in PSCs, leading to frequent supply disrup-tions, product shortages, and unreliable transportation. The increasing focus on regionalization highlights the need for more resilient supply chains to manage disruptions effectively. PSCs must incorporate robustness to address uncertainties and disruptions, which can be internal, such as operational, financial, and quality risks, or external, including supply, demand, and environmental uncertainties. While recent research emphasizes the importance of integrating these uncertain-ties into decision-making tools, the field still lacks comprehensive solutions for PSC optimization. Previous studies have used heuristic and metaheuristic algorithms to address individual uncer-tainties and disruptions, providing feasible solutions with lower computational effort. However, a more holistic approach is required. This work introduces a new tool based on a Meta-heuristic al-gorithm, which evaluates the robustness of PSCs under multiple uncertainties and disruptions. The tool enables the development of what-if scenarios, helping decision-makers improve supply chain robustness and adaptability in the face of several risks

Process modelling is essential to improve carbon capture unit design. However, depending on the modelling decisions made, such as the thermodynamic model and calculation method, the results obtained may vary significantly, hindering reliable process design. Nevertheless, studies that decouple the effects of thermodynamic packages and model approximations on simulation results are not well established. This contribution focuses on clarifying these effects and providing guidelines for the simulation of a CO2 capture process with monoethanolamine (MEA) at different liquid-to-gas (L/G) ratios and CO2 partial pressures in Aspen Plus. The calculations are validated with eight pilot campaign runs. The analysis reveals that the use of rate-based with kinetic reactions significantly improves the accuracy of the simulations. This approach, combined with the ENRTL-RK thermodynamic model, using Peng-Robinson for the vapor phase, provides the best performance, with average deviations below 3% in terms of CO2 capture efficiency and energy requirements. On the other hand, strategies to improve convergence success in closed-loop simulations include replacing a single heat exchanger with two heater blocks, and a calculator block to automate mass-balance calculations and account for losses.

The use of green hydrogen to reduce iron ore is a promising approach to drastically decrease CO2 emissions in the steel industry. To enable the rapid adoption of this technology, it is essential to start from the fundamentals, namely understanding the intrinsic kinetics of iron oxide reduction. In this work, a kinetic investigation was conducted in a PBRlike system using both pure and commercial iron oxide powders under a wide range of operating conditions. The thermal conductivity of the outlet gas was measured and innovatively correlated with the extent of solid reduction through a rigorous mathematical procedure. To simulate the reduction process and determine the kinetic parameters, a deterministic axial dispersion model was developed in conjunction with a dynamic multistep shrinking core model. The model incorporates the particle size distribution of the solid into the mass balance and includes a reactionfront control mechanism to ensure physical consistency during kinetic parameter estimation. The model successfully reproduced all experimental data, confirming that the controlling mechanisms were predominantly chemical, internal diffusion, or mixed, depending on the type of solid material used. Furthermore, the model not only demonstrates excellent predictive capabilities through the validation of kinetic parameters but also provides new insights into the key phenomena governing iron oxide reduction.

The identification of predictive mathematical model structures (i.e. set of model equations) is essential for the development of digital twin models of chemical reactor systems. Recent work demonstrated the use of artificial neural networks (ANNs) for kinetic model recognition in a conceptual batch reaction experimental system. In practical chemical processes, however, system behaviour is governed not only by reaction kinetics but also by reactor hydrodynamics and system thermodynamics. While a very recent study incorporated hydrodynamic effects, this work integrates the three aspects: reaction kinetics, reactor hydrodynamics, and system thermodynamics, to develop a general reactor modelling recognition framework. The framework, which comprises three modules: 1) model generator module; 2) data generation module; and 3) ANN classifier module, was applied to a case study of benzoic acid esterification in a Taylor vortex flow reactor system. Analysing the framework’s sensitivity, results showed that ANN performance in classification deteriorates under increasing measurement noise but can be improved by increasing the number of simulated experiments. Further results show that extensive hyperparameter optimisation of ANN architectures provides no benefit over a fixed ANN architecture. This study highlights the potential of ANN-based frameworks for reactor model recognition while underscoring the dominant role of experimental design and data quality over network hyperparameter tuning.

To produce complex injectable formulations, it is imperative that the contents of the formulation are not damaged during processing. Some formulations require the removal of excess solvent often through evaporation. Wiped Film Evaporators can be applied for these scenarios, since they can operate at relatively low temperatures with mass transfer being promoted through a large surface area. This is created by wiping the liquid against the inner heated wall of the equipment, with the proper operation requiring, as much as possible, a stable and continuous liquid film. In some cases, however, depending on operating conditions or physical properties of the materials, film ruptures and wetting/de-wetting dynamics are observed, and in more severe situations the liquid is dispersed in isolated globules and drops. These are undesirable situations of operation, where it is not possible to guarantee an optimal and homogenous heat and mass transfer. In this work, Computational Fluid Dynamic (CFD) simulations are used to predict a map relating operating conditions to these different flow regimes. From this fundamental knowledge, it is possible to infer operatory windows for a given task and ultimately develop a reliable model for process scale-up.

Fouling is a persistent issue in industrial heat-transfer equipment, increasing energy demand and reducing efficiency. This is also true in the dairy industry where falling-film evaporators are central to powder production. Most dynamic models, however, neglect gradual fouling, limiting predictive accuracy during extended operation. As a result, model-based control can become unreliable when fouling becomes significant. The dynamic models by Bojnourd et al. [1] are widely used but assume clean-surface operation. While this captures short-term thermal behavior, it cannot represent the progressive decline in heat-transfer performance caused by fouling. Díaz-Ovalle et al. [2] introduced a fouling-layer model that explicitly describes the growth of a fouling deposit over time. Building on this concept, the present work incorporates a simple dynamic fouling model for falling-film evaporators and validates it using industrial data from a four-effect evaporator using thermal vapor recompression. Two fouling-modelling strategies are evaluated: a mechanistic fouling layer model and an empirical model describing the observed loss in heat-transfer efficiency. The fouling layer model substantially reduces prediction error, demonstrating that it captures the dominant fouling behavior. Although the empirical model fits all observed discrepancies, it lacks predictive capability and therefore cannot be applied for model-based prediction or control.

Slug flow in multiphase production systems is a critical challenge in the oil and gas industry, characterized by complex and oscillatory dynamics, e.g., limit cycles. First-principle (FP) models often employ physics simplification, such as a virtual valve for the slug formation. To capture complex physics poorly modeled by FP models, hybrid models combine data-driven techniques and physical knowledge, such as the architecture known as universal differential equation (UDE). This work aims to employ a hybrid model based on neural networks to enhance the modeling of multiphase oil production systems. In the UDE model, a neural network is embedded within the structure of the FP differential equations. To demonstrate the feasibility of the methodology, the model was trained based on synthetic data, employing parameters estimated from OLGA simulations. Since the system faces oscillatory behavior, we trained the UDE in two stages: the first one employs smooth collocation on data to obtain an estimation of the derivatives. The results show that the two-stage strategy was successful in addressing the resolution of an oscillatory system. In the second stage, the algorithm reached a scaled MSE of 0.009 in 502 iterations. For the testing phase, using 20%, 40%, 60%, and 80% valve openings, the scaled MSE reported was 0.088. For the 100% valve opening, outside of the training part, the MSE reported was 0.039, showing a good extrapolation capacity.

Reliable data for the standard enthalpies and Gibbs free energies of formation, DHf° and DGf° are essential for process synthesis, energy integration, and lost-work analysis. However, many biomass-derived compounds lack reliable thermodynamic property data, limiting optimization of energy and carbon utilization in biomass conversion processes. This study proposes a composition-based method to estimate DHf° and DGf° for compounds containing carbon, hydrogen, and oxygen. The method exploits widely available heats of combustion data and establishes a linear correlation between the enthalpy and Gibbs free energy of combustion, DHC and DGC using tabulated organic compounds. The applicability of this relationship to biomass-derived compounds is tested using published data for cellulose, starch, and glucose. Thornton’s correlation between heat of combustion and oxygen demand is then incorporated to derive simple expressions for estimating formation properties directly from elemental composition. The results show that the proposed framework provides sufficiently accurate estimates for conceptual thermodynamic analysis, enabling rapid evaluation of reaction enthalpies, minimum work requirements, equilibrium constants, and reaction driving forces. The method is therefore well suited for early-stage process synthesis and screening of biomass conversion pathways.

Waste incineration with combined heat and power (CHP) supplies electricity and heat to cities and industrial clusters but remains a significant source of greenhouse gas emissions. This work develops an optimization-based, system-level decarbonization framework for an integrated waste-to-energy and chemical production cluster under operational, societal, and economic constraints. The framework is applied to a real-world case study including a municipal waste incineration plant, an energy utility system, and multiple chemical production units. A layered decarbonization strategy is implemented. First, energy efficiency is enhanced through waste heat valorization using heat pumps. Second, coordination between industrial actors is improved through solid waste storage management and operational alignment of heat and power supply with demand. Third, alternative feedstocks are introduced to reduce fossil-based inputs. Within the work material and heat-stream inventories are collected, and the required chemical synthesis pathways are modeled using Aspen Plus. The integrated system is optimized within the OSMOSE framework using a multi-period mixed-integer linear programming formulation that minimizes the annualized total cost while accounting for seasonal heat demand and electricity price variability. The results show that waste-heat upgrading alone enables fossil emission reductions of up to 22% while improving overall system profitability. More broadly, up to 80% of fossil-based Scope 1 and Scope 2 CO2 emissions can be reduced through energy efficiency improvements, system integration, and feedstock substitution prior to carbon capture deployment. To address the remaining hard-to-avoid emissions and enable net-negative operation, post-combustion carbon capture options are evaluated in combination with solid oxide electrolysis cells (SOEC).

Energy Baselines (EnBs) provide a reference for evaluating Energy Key Performance Indicators (eKPIs), and their establishment is mandated under ISO 50001. Since eKPIs are typically defined per functional unit, such as product, recipe or recipe phase, EnBs should not be averaged across heterogeneous operating conditions but instead be defined in a context-specific manner. This requires detailed mechanistic models or sufficiently rich operational data for statistical approaches, both of which are often unavailable in highly flexible, semi-continuous production systems.This paper proposes a four-stage framework for the automated generation of surrogate EnBs to address this gap. In the first stage, the relevant training data space is defined, including non-influenceable variables (e.g., equipment deviations), design parameters (e.g., material properties), and adaptable recipe parameters (e.g., operating conditions and control actions). In the second stage, these parameters are systematically varied to automatically generate a labelled dataset representing process dynamics and energy behavior. The third stage trains and evaluates candidate surrogate models using the PRESTO recommendation framework, followed by a global sensitivity analysis based on Sobol indices to identify the dominant drivers of energy performance.The framework is demonstrated using a stirred tank reactor at laboratory (2 L) scale. The results show that Random Forest surrogates provide reliable performance across a wide operating range and that recipe parameters, particularly mixer rotational speed, dominate energy efficiency. Future work will address scalability to other equipment types and reactive processes.

Liquefied natural gas (LNG) plays a key role in the energy transition, but its production is often limited by the high CO2 content of some reservoirs, which increases operating costs and solidification risks. This study evaluates the dual-pressure low-temperature cryogenic distillation process applied to a recently discovered gas field with a high CO2 content (25%) for LNG production. The critical properties of the gas streams and the process operating conditions were analyzed using Aspen Plus v14. The results indicate that reducing the CO2 concentration throughout the column is essential to prevent solid formation by maintaining a fluid composition with a freezing temperature below the operating temperature of the stages. It was also observed that the reflux affects LNG purity and freezing temperature in all stages. Furthermore, the adoption of a low-pressure separator upstream of the distillation proved crucial to producing condensates within commercial specifications. The work extends the understanding of the dual-pressure low-temperature cryogenic distillation to the 25% CO2 range and highlights the potential of cryogenic distillation to enable LNG production in fields with high CO2 content.

Biological systems are governed by coupled interactions between intracellular metabolism and bioreactor operation that span multiple time scales. Constraint-based metabolic models are widely used to describe intracellular metabolism, but repeatedly solving the optimization problem at each time step in dynamic models introduces numerical challenges related to infeasibility and computational efficiency. This work presents a multi-scale modeling framework that integrates genome-scale, constraint-based metabolic models with dynamic bioreactor simulations. Intracellular metabolism is described using positive flux variables in a parsimonious flux balance analysis, and the resulting embedded optimization problem is replaced by a neural network surrogate. The surrogate provides a smooth approximation of the embedded optimization mapping and eliminates repeated linear program solves during simulation. The approach is demonstrated for fed-batch fermentation of Escherichia coli, in which the surrogate model yields intracellular fluxes under substrate-limited conditions, whereas the underlying linear program would otherwise be infeasible. The framework provides a continuous representation of intracellular metabolism suitable for dynamic simulation of genome-scale models in bioreactor configurations.

Capacitive Deionization (CDI) is an emerging electrochemical technology for energy-efficient brackish water desalination. However, the rigorous design and scale-up of CDI systems are frequently hindered by the complexity of validating predictive models. The coupling of electrochemical double-layer kinetics with macroscopic mass transport often leads to structural parameter correlations, where multiple combinations of kinetic rates yield indistinguishable effluent trajectories. This paper addresses these challenges by proposing a simulation-driven Model-Based Design of Experiments (MBDoE) framework. We develop and implement a reduced-order Dynamic Langmuir (DL) model within the gPROMS platform, designed to capture cyclic adsorption-desorption dynamics with high computational efficiency. Sensitivity analysis reveals that information content is highly transient, concentrated primarily in the short time windows following voltage switching, and that the effluent concentration is significantly more sensitive to desorption kinetics than adsorption. Fisher Information Matrix (FIM) analysis of baseline experimental data confirms a strong negative correlation between kinetic parameters, resulting in a poorly conditioned estimation problem. To resolve this, a large-scale in-silico screening of the experimental design space—spanning inlet concentration, flow rate, and cell volume—is conducted using a D-optimality criterion. The simulation results demonstrate that operating at low flow rates and large effective cell volumes maximizes parameter identifiability by enhancing the separation of dynamic signatures. This work illustrates the critical role of dynamic simulation in guiding experimental strategy, minimizing trial-and-error effort, and improving the robustness of process models.

Liquefied Natural Gas (LNG) plays a strategic role in the global energy transition, as it represents a less carbon-intensive alternative to coal. Separation of CO2 from raw natural gas is a critical step for meeting LNG specifications and enabling Enhanced Oil Recovery (EOR) in offshore fields. However, high CO2 concentrations and formation of a CO2 ethane azeotrope increase the process complexity, often requiring extractive distillation with heavier hydrocarbons. Severe limitations are faced in offshore environments due to their weight, volume and high energy consumption. Due to that, Process Intensification (PI) seeks to enhance heat and mass transfer efficiency, potentially reducing equipment volume and weight. Rotating Packed Beds (RPB) have demonstrated significant potential for intensifying LNG purification by using centrifugal forces to drive liquid through a porous medium in contact with a gas stream. Experimental measurements of total pressure drop, and local liquid holdup are feasible in pilot-scale units, although typically requiring specialized equipment and instrumentation. Consequently, modeling approaches based on mass, momentum, and energy balances have been widely proposed in the literature to estimate the hydrodynamic behavior of RPB systems. While Computational Fluid Dynamics (CFD) models can provide detailed insights into equipment behavior, they are often computationally expensive for parametric studies and design optimization. In this context, this work proposes a one-dimensional model in cylindrical coordinates to analyze fluid flow in an RPB for natural gas/CO2 separation process. The governing equations are formulated and validated against experimental data for pressure drop, interfacial velocity, and liquid holdup along the RPB radius.

The valorization of lignocellulosic biomass represents a key pathway toward sustainable chemical production, as it enables the development of circular economy products with reduced dependence on fossil resources. Among the platform molecules derived from biomass, furfural stands out as a versatile intermediate that can be transformed into several high-value chemicals, such as furfuryl alcohol, 2-methylfuran, tetrahydrofurfuryl alcohol, furan, tetrahydrofuran, and maleic anhydride. In this work, an integrated biorefinery scheme for producing the main furfural derivatives is proposed and evaluated through process simulation and sustainability metrics. The process was modeled in Aspen Plus V14, using furfural obtained from lignocellulosic biomass (corn stover) as raw material, following hydrogenation and oxidation routes. The process is multi-product, meaning that the main furfural derivatives are produced simultaneously within the same plant. This was achieved by implementing splitters that work at 50% capacity in the division of raw materials. The mass and energy balances from the simulation were used to evaluate environmental, economic, energy, and efficiency performance using a set of indicators. Economic data were obtained from Aspen Economics, yielding a total capital cost of 7.65 MUSD and a total annual cost of 1.93 MUSD/year. The environmental assessment resulted in an EI99 of approximately 2.0×108 mPt/year, mainly influenced by hydrogen consumption and carbon monoxide emissions. From an efficiency perspective, the process shows favorable performance, with an MLI of 0.07 and an MCI of 69%, indicating high material efficiency despite the inherent complexity of multi-stage hydrogenation and oxidation pathways. Overall, the results demonstrate the technical and sustainability potential of an integrated furfural-based biorefinery, while highlighting opportunities for further improvement through process integration and hydrogen recovery.

Copper refining via electrolysis is a core metallurgical process that takes place in tankhouses, subject to strict temporal, spatial, and operational constraints. The efficiency and stability of this process depend critically on the coordinated scheduling of crane operations responsible for handling anodes, cathodes, and auxiliary tasks. In industrial practice, crane scheduling must simultaneously satisfy long-term production targets and short-term operational feasibility, while respecting process-dependent timing constraints imposed by electrochemical parameters. Inefficient or inconsistent schedules can lead to process delays, suboptimal resource utilization, and degraded electrolysis performance, ultimately affecting product quality and operational stability. This paper presents a modeling approach for optimizing tankhouse operations. The uniqueness of this case lies in the broad range of constraints, including human capacity, energy restrictions, metallurgical rules, and logistical specifications. The model operates on three levels, starting with generalized schedule optimization and culminating in the detailed optimization of crane movements and cathode stripping machine operations. Although the focus is on the specific use case of tankhouse operations within metallurgical production, the model can be adapted and fine-tuned for a wide range of scheduling applications. Additionally, we discuss the challenges encountered when integrating the model into existing production processes and outline directions for further work.

Carbon dioxide hydrogenation to produce methane, as an energy carrier or raw material, has great potential for the chemical industry. Since methanation reaction is strongly exothermic and sensitive to diffusion, radial flow reactors represent a clear solution thanks to their low pressure drop and effective heat removal. A two-dimensional mixing cell network (MCN) approach to model the carbon dioxide methanation in a radial flow reactor is proposed. The reaction is catalyzed by a bi-functional Ni–Ce zeolite 13X supported catalyst, combining catalytic and adsorption functions. This contribution outlines the ongoing work, starting from a straightforward MCN pseudo-homogeneous approach comparing it with a tubular packed bed reactor. Both methanation kinetics and water adsorption have been successfully implemented in both models, setting feasibility for further improvements. Future developments will be necessary aiming to aid the design of units employing Ni–Ce/13X catalysts.

The climate crisis and dependence on fossil fuels make the transition to renewable energy sources imperative, with biomass standing out for promoting decarbonization and circular economy. In this context, fluidized bed gasification emerges as an efficient route for converting waste into syngas, applicable to power and hydrogen generation. Given the variability of real operating conditions, dynamic models are essential to represent coupled fluid dynamic, thermal, and kinetic phenomena over time. In this work, a dynamic phenomenological model was developed using a lumped 0D approach, in which the reactor is divided into two interacting zones represented as continuous stirred-tank reactors (CSTRs): a dense bed, where drying, devolatilization, and heterogeneous reactions occur, and a freeboard, dominated by homogeneous gas-phase reactions. The model was validated against experimental data from a bubbling fluidized bed gasifier, showing good agreement for major syngas species (CO and H2, mean absolute error below 2%), with minor deviations for CH4 due to simplified devolatilization and tar conversion kinetics. Transient simulations revealed strong sensitivity of process performance to particle size and biomass composition. Compared with the reference case (6 mm), reducing particle size to 1 mm enhanced hydrogen gas efficiency, cold gas efficiency, and carbon conversion due to intensified heterogeneous reactions and improved solid–gas heat transfer. Additionally, biomass switching simulations demonstrated that real-time air flow control effectively stabilizes syngas quality under feedstock variability. Overall, the model provides a computationally efficient tool for process optimization, control strategy development, and integration of biomass gasification into flexible and hybrid low-carbon energy systems.

Traditional approaches for mechanistic modelling require in-depth understanding of the underlying chemical and physical phenomena to construct reliable and predictive models. However, at early stages of development, limited experimental data, incomplete expert knowledge, and non-observable states often hinder a full understanding of the underlying mechanisms. Symbolic regression (SR) enables systematic model discovery and offers a practical route to addressing these challenges by automating the identification of interpretable model structures and the estimation of associated parameters from available data. However, structural identifiability and observability (SIO), a critical property of such models, is often overlooked in SR, thereby limiting its broader adoption and effective deployment. To address these limitations, this study proposes a comprehensive framework, which leverages scarce prior knowledge in SR and incorporates SIO analysis, offering a potential solution to capture the effects of all state variables and ensure rigorous structure of the resulting models. The proposed strategy is demonstrated on a partially observed sulfide oxidation system, demonstrating how an SIO-assured model can be systematically identified from limited data and incomplete system knowledge. Overall, this work presents a potential pathway for extending model discovery to partially observed systems and enhancing model robustness, thereby positioning SR as an alternative tool with respect to traditional kinetic modeling strategies.

This work presents the scale-up and detailed analysis, including comparison of relevant key performance indicators (KPIs) and energy analysis, of three innovative technologies for producing biomethane and its subsequent upgrading. The SEMPRE-BIO project tested and validated three different innovative technologies and pilots within the Horizon Europe framework. The three pilots are in Spain (biomethanation of biogas from wastewater treatment fermentation), France (biomethanation of syngas from biomass gasification), and Belgium (purification of biogas from manure anaerobic digestion). The three case studies will be presented and discussed. The analysis will dive into the layout of the optimised process layout of the scale-up plants as well as an exhaustive comparison, presenting advantages, shortcomings and bottlenecks of each technology, accounting for the main KPIs, i.e., electricity and steam demand, consumables including cooling water and others specific to the technologies.

Power-to-ammonia (P2A) provides a carbon-free alternative to conventional ammonia production by replacing fossil-based feedstocks with electrolytic hydrogen and nitrogen from air separation. For decentralized P2A systems, pressure swing adsorption (PSA) offers a flexible alternative to cryogenic air separation. However, its industrial implementations are largely proprietary, and open, first-principles models capable of simulating its cyclic, nonlinear transport are scarce in literature. This work presents a first-principles, dynamic, one-dimensional model of a PSA superstructure for nitrogen generation, formulated with thermodynamically consistent equations of state, coupling multicomponent mass, energy, and momentum balances with kinetically limited adsorption on carbon molecular sieves. The resulting system of partial differential-algebraic equations is semi-discretized using the finite volume method, integrated using diagonally implicit Runge-Kutta methods, and cyclic steady states (CSS) are computed via shooting-based solution methods. The framework is implemented in Julia, combining analytical derivatives with automatic differentiation and utilizing sparse linear algebra for efficient solution of the arising large nonlinear systems. The framework is demonstrated on a two-bed PSA cycle for air separation, comparing spatial and temporal discretization strategies, CSS solution methods, and the effects of ideal versus real-gas thermodynamics on predicted nitrogen purity and recovery. The proposed framework establishes an extensible basis for PSA simulation and optimization.

The SEMPRE-BIO project tested and validated three different innovative technologies and pilots within the Horizon Europe framework. One of the pilots is commissioned in Belgium. The proposed technology purifies biogas from manure anaerobic digestion and delivers simultaneously pure biomethane and food-grade CO2, conversely to conventional purification technologies such as absorption and adsorption. Due to the severe cryogenic conditions, energy recovery for purge and waste streams becomes relevant to improve the energy demand of the process. The present work will show an effective solution to reduce the electricity demand of the process. Biomethane slip and other purge stream are valorised in a steam boiler and a two-pressure-level steam turbine to deliver both middle pressure steam as utility in distillation reboilers and produce electricity. The analysis will propose a simple, but rigorous methodology to maximise the steam turbine loop and the net power. The present work is based on solid and concrete assumptions for the design of the steam turbine. The analysis shows that the process allows for recovering up to 12% of the total electricity and around 55% of the total energy demand for the refrigeration loop to preserve the cold box of the bio-CH4/CO2 distillation.

This paper presents ProcessSimulator.jl, an open-source process simulation framework built in Julia that combines acausal, equation-oriented modelling with seamless integration of procedural code. The framework leverages ModelingToolkit.jl to enable modular construction of unit operations using symbolic–numeric representations, facilitating the extension of models with advanced thermodynamics, kinetics, and data-driven components. Inspired by the ModuSim concept [3], ProcessSimulator.jl introduces an extensible control-volume abstraction and connector-based composition at the unit-operation level. A steady-state CSTR case study is presented and compared against Aspen Plus, showing good agreement in key variables. The results demonstrate the feasibility of a flexible, open, and composable process simulation paradigm for research and education.

Nanoparticle (NP) synthesis has been extensively studied since the mid-1800s and are utilized across numerous fields due to their unique microscopic properties that collectively yield macroscopic benefits. Of particular interest are silver (Ag) NPs, whose controllable size and morphology impart distinct catalytic, electronic, and optical properties advantageous for environmental and energy-related applications. The theoretical understanding of NP nucleation and growth has advanced considerably starting with classical nucleation theory, evolving into the LaMer model centering on burst nucleation and diffusion-limited growth and resulted in near monodispersed hydrosols. Finke and Watzky later introduced the autocatalytic model considering a slow and continuous nucleation and autocatalytic surface growth not limited by monomer diffusion. However, the precise mechanisms remain the subject of active debate for the different homogeneous and heterogenous nucleation systems. In this study, simulation models for NP population balances are developed using both material balanced and constant number Monte Carlo methods to describe NP formation and growth under diffusion-limited and autocatalytic conditions. Perturbative analysis of the model provides insight into whether NP formation is diffusion-limited or autocatalytic and how addition(s) or subtraction(s) of precursor influence the NP growth mechanism. Perturbative analysis reveals the mean particle size and variance can be tuned by adding or removing precursor molecules during the reaction, although the overall size distribution trend remains consistent. The present study together with new experimental capabilities lays the foundation for a model-based design of experiments to explore the nucleation and growth mechanism of metal nanoparticles.

This work presents the conceptual design, rigorous simulation, and multi-objective optimization of an integrated multiproduct biorefinery focused on the sustainable valorization of lignin into catechol and a portfolio of high-value aromatic derivatives, including guaiacol, vanillin, and vanillic acid. The proposed methodology combines a critical review of synthesis pathways and kinetic data with detailed process modeling in Aspen Plus® V14. The process framework begins with lignocellulosic biomass pretreatment via acid hydrolysis, followed by catalytic lignin depolymerization to obtain catechol as a platform intermediate. Subsequently, dedicated reaction and separation sections are designed for the selective conversion of catechol into value-added derivatives, incorporating energy-efficient and thermally integrated separation schemes. A stochastic multi-objective optimization strategy based on Differential Evolution with a Tabu List (DETL) is implemented to simultaneously assess economic and environmental performance indicators. The results demonstrate that lignin can be effectively upgraded into commercially relevant fine chemicals, confirming its strategic role within next-generation biorefineries and reinforcing the transition toward circular, low-emission chemical processes.

In today’s competitive LNG market, reducing energy consumption is critical for enhancing both profitability and sustainability. The hot section of the LNG processing, which includes inlet receivers, acid gas removal, and dehydration units, is the most thermally demanding. Previous optimisation methods targeted each unit separately. On the other hand, this work details the development of a data-driven optimisation framework to minimise energy across these interdependent units. Preliminary application of the framework has yielded encouraging results. Utilising HYSYS process simulation data, the study successfully identifies critical operating variables—such as reboiler duty, amine circulation rate, and air-to-furnace stoichiometry—that drive production efficiency and energy consumption. Results indicate that a baseline condensate mass flow of 2, 048.71 kg/h is achieved at a stripper bottom temperature of 137.74 °C, while the AGRU produces sweet gas with 0.18 ppm H2S. Optimisation using Pareto frontier analysis reveals a “knee” point in the SRU at 96.9% recovery efficiency, balancing elemental sulfur production (7, 622.24 kg/h) against heating loads. The findings provide mathematical coefficients for plant control to maximise yield while maintaining strict environmental stack limits.

Sonochemical reactors are a promising technology in process intensification, offering a sustainable and energy-efficient means of enhancing chemical reactions. By harnessing acoustic cavitation – the formation, oscillation and violent collapse of bubbles in a liquid medium – these systems generate local hotspots that can accelerate reaction kinetics. Despite its potential, efficient design and scale-up of sonochemical reactors remain major challenges, mostly because the cavitation phenomena take place close to the ultrasonic transducer. This work presents a simulation-based framework for the optimization of sonochemical batch reactors by coupling microscopic-level bubble behavior with macroscopic-level reactor performance, focusing on the placement of transducers to maximize reaction activity.

The purification of therapeutic peptides represents a major bottleneck in biopharmaceutical downstream processing due to the structural similarity between target products and closely related impurities. In this study, a shortcut dynamic simulation model of a two-column Multi-Column Countercurrent Solvent Gradient Purification (MCSGP) process is implemented in Aspen Chromatography for peptide separation. Each column is described using a one-dimensional axial dispersion model coupled with mass transfer kinetics and a modulated Langmuir adsorption equilibrium, while time-dependent boundary conditions are applied to represent solvent gradient elution. The model explicitly incorporates internal recycle streams between columns using the cycle organizer approach, capturing the defining operational features of MCSGP. This enables a unified representation of chromatographic transport phenomena, gradient operation, and discrete recycle logic within a single flowsheet-based framework. The novelty lies in treating MCSGP as a hybrid dynamic process rather than a purely chromatographic unit, providing a transferable modelling framework suitable for operability analysis, process comparison, and future optimization of continuous peptide purification systems. Dynamic simulation results for a model mixture of closely eluting peptides demonstrate stable cyclic operation and a substantial improvement in mass recovery compared with conventional batch chromatography under identical transport assumptions. In particular, systematic recycling of intermediate fractions leads to an order-of-magnitude increase in recovered target product.

Low-temperature industrial processes release substantial amounts of waste heat, representing a largely untapped renewable energy resource. This study focuses on the brewery sector, encompassing both beer and whiskey production, along with its integration with manufacturing and city. The brewery industry generates approximately 0.061 kWh of waste heat per liter of beer, while whiskey production releases around 2.2 kWh per liter, with most of this waste heat available at temperatures close to 95 °C. Such low-grade heat is well suited to meet heating demands in manufacturing industries and urban district heating networks, where temperature requirements typically remain below 80 °C. Multiple technological options for meeting process heat requirements and recovering waste heat are evaluated using the OSMOSE energy integration framework. The study assesses the technical performance and economic viability of these options under varying assumptions for electricity prices, natural gas prices, and carbon pricing. Depending on market conditions, overall external utilities consumption reductions ranging from 22% to 63% are achieved. Price of utilities also has an effect on reduction of emission, as it is ranging from 25% to 90%.The results highlight the significant potential for cross-sectoral heat integration between breweries, manufacturing industries, and urban energy systems. A comparison is drawn between a case in which 5 th generation system (case 1) is activated and a case in which 5 th generation is not active (case 2). It was observed in most of the market conditions, case 1 is energetically, environmentally and economically more sustainable compared to case 2. In one of the market condition it was possible for both the cases to be energetically a carbon negative solution as it was economically feasible to capture biogenic CO2. The work helps in validating that strategic waste heat recovery and utilization can substantially enhance energy efficiency, support industrial decarbonization pathways, and improve the resilience of urban energy infrastructures.

This work presents guidelines for controlling and intensifying mixing in oscillatory opposed-jet mixers, focusing on Confined Impinging Jets (CIJs) as a model system where flow behavior is primarily governed by oscillatory parameters, decoupled from geometric complexity. Computational Fluid Dynamics (CFD) simulations were used to investigate the effects of oscillation amplitude and frequency on mixing. The results show that at high amplitudes, mixing is robust across a broad frequency range, as energy injection is sufficient to promote vortex formation and their propagation to the reactor’s outlet. At low amplitudes, mixing is highly sensitive to the oscillation frequency and occurs only near the resonance frequency, the specific frequency at which the flow’s response to the applied oscillation is maximized. At low amplitude, lower frequencies fail to inject sufficient energy, while higher frequencies promote flow segregation. Remarkably, effective vortex propagation and mixing were achieved even at very low net Reynolds numbers (Re_net=25), increasing the residence time from 0.042 s at Re_net=300 to 0.5 s at Re_net=25. Two quantitative mixing metrics corroborate the qualitative flow observations and were instrumental in defining practical guidelines for process intensification. These findings provide new insights into how oscillatory forcing can be tuned to control mixing scales, enabling efficient mixing under a wide range of operating conditions, including shear-sensitive processes. The study offers a foundation for the model-based design and optimization of intensified static mixers, linking oscillation parameters, Reynolds number, and residence time to application-specific mixing performance.

Laboratory safety is procedure-dependent: required personal protective equipment (PPE) and permissible actions vary across experiments and across experimental steps, yet most vision-based monitoring remains appearance-driven and often produces generic warnings without reliable procedural context. We propose a multimodal framework for step-aware safety monitoring in laboratory videos. The framework first localizes procedural context through clip-level step prediction and protocol alignment to identify the experiment and current step. Given this context, it retrieves step-specific safety constraints, extracts evidence of step-relevant equipment and interactions using an equipment database, and prompts a video-capable vision-language model (VLM) to generate structured (JSON) monitoring reports supported by retrieved constraints and visual evidence. Experiments on protocol-annotated molecular biology lab videos show that our approach improves the mean score from 0.4352 to 0.6430 and reduces the missing rate from 65.00% to 33.75% relative to a video-only baseline, demonstrating more faithful and step-specific safety judgments.

Mathematical models in process systems engineering (PSE) are widely used to support decision-making in design and operation, but they are mostly limited to deterministic models. For biochemical systems, the biological variability can give rise to stochastic dynamics. This work addresses the question of when to act in such processes, as the stochastic dynamics affect the timing of important events. We consider the case of batch production of malic acid using Ustilago trichophora. The goal is to predict when the substrate concentration falls below a predefined threshold. We extend an existing deterministic model of the process to a stochastic differential equation (SDE) formulation by introducing a Monod-like noise term. Simulations of the SDE model reveal a distribution of substrate depletion times and a deviation between the mean of the stochastic trajectory and the deterministic solution due to nonlinear effects. To determine optimal intervention times under uncertainty, we formulate a finite-horizon optimal stopping problem and solve it using the Longstaff–Schwartz algorithm, also known as the Least-Squares Monte Carlo (LSMC). The resulting distribution of optimal stopping times from the LSMC algorithm is shown to closely match the actual first threshold hitting times obtained from a posteriori analysis of the stochastic simulations, as confirmed by a two-sample Kolmogorov–Smirnov test. The results demonstrate that optimal stopping provides a framework for decision-making in stochastic biochemical processes, enabling risk-aware operational strategies beyond deterministic optimisation.

Digitalization, artificial intelligence, and autonomous experimentation are reshaping chemical process development by enabling data-driven system identification and model-based optimization. Despite these advances, mechanistic models remain a cornerstone for predicting chemical reaction behavior and supporting optimization. However, purely mechanistic models often exhibit limited predictive accuracy when key phenomena affecting kinetics, mass and energy transfer are not fully captured. To address limitations on kinetic modelling, a hybrid modelling framework is proposed in this work that integrates a lumped power-law kinetic model with a Gaussian Process (GP) residual model to predict the reaction rate across the experimental design space while quantifying the uncertainty of the predicted rate. The hybrid model is then coupled with multi-objective Bayesian optimization (MOBO) by employing a weighted-sum approach and an upper confidence bound acquisition function to guide experimental design by simultaneously maximizing reaction rate and minimizing the associated uncertainty on prediction. The approach is tested on a case study related to catalytic methane oxidation in an autonomous microreactor system, where the proposed hybrid model performs comparably to the ground truth model, identifying the same optimal operating conditions when validated against the original experimental data, with an average prediction uncertainty of approximately 2.7%. The results highlight the potential of hybrid modelling and uncertainty-aware optimization for guiding autonomous catalytic experiments in flow systems.

Mathematical models are paramount to the field of reaction engineering, facilitating reaction mechanism discovery, process optimisation, and informed decision making in academic and industrial settings. Nevertheless, the development of precise mechanistic reaction rate models remains experimentally intensive, requires expert knowledge, and is susceptible to the introduction of structural bias. Similarly, the identification of a suitable reaction network that depicts all chemical transformations remains a non-trivial task, with existing techniques often being ill-suited for large and complex systems, hence limiting their scalability and implementation within chemical and biochemical applications. This work develops a two-stage autonomous framework that exploits non-linear sparse optimisation to identify the minimum size global reaction network representative of the system under study, and subsequently proposes and discriminates between interpretable rate equations developed through symbolic regression (SR). The generated SR expressions are constrained to a mechanistically meaningful form through the use of a novel substructure decomposition strategy, largely reducing the search space that must be explored and increasing model interpretability. The framework is evaluated on two case studies; the first, depicting a catalytic methanol synthesis reaction network, and the second, an enzymatic kinetic resolution network. The methodology exhibited high levels of accuracy, scalability, data efficiency, and robust network identification consistent with known physics. Lastly, the potential of augmented intelligence, is discussed as a method to enhance fidelity. Therefore, this work represents a key step toward autonomous process modelling and digitalisation in reaction engineering, providing a foundation for accelerated design and development of chemical and biochemical processes.

Understanding the relationships between operating conditions and product formation pathways in biomass pyrolysis remains challenging due to the complex interactions among temperature, catalytic effects, and feedstock composition. In this work, principal component analysis (PCA) was applied to investigate the combined influence of temperature and catalyst-to-biomass ratio on the pyrolysis of sugarcane bagasse and Salicornia. To preserve mechanistic interpretability, two complementary analyses were performed: one considering only catalytic experiments and a second integrating both thermal and catalytic conditions. Separate PCA were conducted for product yields, gas and liquid compositions, and solid-phase FTIR features. The results reveal that thermal conditions promote severe cracking and solid carbonization, whereas catalytic operation favors secondary pathways associated with controlled dehydration and partial stabilization of liquid products. Distinct patterns between the two feedstocks were also identified, reflecting their intrinsic compositional differences. Based on the multivariate trends identified, a conceptual reaction scheme is proposed to describe the interplay between thermal and catalytic pathways.

In this study, we propose a novel trust-region funnel (TRF) optimisation framework for process systems that integrate external black-box models, such as rigorous models, within equation-oriented (EO) formulations. The framework is applied to optimise a synthetic natural gas production process combining direct air capture and catalytic CO2 conversion using dual-function material (DFM) technology, with the objective of minimising the total annualised cost. The problem is formulated in Pyomo and solved using IPOPT, treating the DFM reactor as an external black-box model. The TRF method achieves substantial improvements compared to published mixed-integer nonlinear programming and direct nonlinear programming approaches, reducing capture cost from 460 USD to 426 USD per tonne of CO2. Key design improvements include reducing the number of DFM units per train by one-third and achieving a 22% reduction in DFM capital costs. These results highlight the TRF framework’s ability to overcome numerical challenges and unlock economically superior designs for next-generation carbon capture and utilisation systems.

Addressing the urgent transition to low-carbon energy systems requires microgrids capable of locally coordinating electricity generation, storage, and flexible consumption. Their efficient integration calls for control strategies that are scalable, privacy-preserving, and robust to uncertainty. To address such a challenging control problem, this work proposes a decentralised Multi-Agent Reinforcement Learning (MARL) approach based on the Cross-Entropy Method (CEM) for the coordination of prosumers, equipped with renewable generation and vehicle-to-grid capabilities. To improve sample efficiency and robustness, the policy is warm-started using Behaviour Cloning (BC) from a classical Proportional-Integral-Derivative (PID) controller, resulting in a hybrid BC–CEM framework. The proposed method is evaluated in a realistic microgrid simulation with stochastic demand and real weather and generation profiles. Results show that BC–CEM accelerates convergence and achieves lower energy costs compared to both PID control and randomly initialized CEM, without sacrificing comfort or mobility requirements. The findings highlight the effectiveness of combining derivative-free optimization with imitation learning in complex MARL tasks, such as energy flexibility coordination.

This paper examines the mutually beneficial relationship between biomimetics and modelling and simulation tools, showing how each can enhance the other. Through a literature review and a detailed use case on anaerobic digestion, the study highlights how the complexity, multiscale organisation, and functional richness of biological systems challenge current modelling capabilities. By analysing the contributions of modelling and simulation to product development, such as early performance validation, rapid and lowcost iteration, and multicriteria evaluation, the paper questions whether integrating modelling and simulation tools to biomimetics would bring similar benefits to the design process. Several hypotheses are formulated regarding the potential contributions of modelling and simulation to biomimetics, particularly the improvement of biological system understanding through advanced visualisation and the assessment of functional viability using parametric modelling. Integrating such tools into biomimetics is envisioned as a means to reinforce the existing methodology, support more reliable principle transfer, and facilitate the wider adoption of biomimetics as a problemsolving design approach, ultimately promoting the emergence of biomimetic innovations across diverse applications.

As chemical processes become increasingly complex and costs of experimentation increase, understanding the practical effectiveness of Active Learning methodologies is essential. In this regard, an ongoing debate is occurring within the research community about the use of Design of Experiments (DOE) and Bayesian Optimisation (BO). However, this debate is limited by the scarcity of systematic comparative studies. Therefore, this work provides a comparative analysis of two widely adopted data-driven optimisation approaches: DOE and BO. The comparison is conducted across two distinct case studies reflecting different levels of complexity, regarding the quantity and variety of input variables involved. The first case study represents a realistic in silico experimental scenario, with multiple decision variables of different types (continuous, categorical and mixture), and two distinct single-objective optimisation goals, while the second one considers a simpler, well-known benchmark model with just two input continuous variables. Both studies were designed and analysed, while acknowledging the inherent conceptual and operational differences between DOE and BO. The results showed that a sequential DOE strategy, adapted here for optimisation purposes, consistently outperformed classical BO under the tested conditions. Particularly in terms of convergence towards the target response, robustness in identifying the optimal region, and overall experimental budget efficiency. Rather than asserting the superiority of one methodology over the other, this work highlights the need for case-specific adaptation when applying data-driven optimisation strategies in Chemical Engineering. The findings emphasise that theoretical guarantees alone under relatively strict assumptions are insufficient and must be complemented by problem-driven evaluation to support informed decision-making.

The bioprocess industry is actively exploring technologies associated with the fourth industrial revolution, with modeling offering considerable potential for process optimization. Nevertheless, model adoption in industry remains limited. This is partly because model development continues to depend heavily on offline sampling, and because relatively few industrial applications convincingly demonstrate their practical value. This study therefore first examines the benefits of online rheology and online biomass measurements for model development and demonstrates, among other aspects, that online biomass significantly improves model fidelity. The second part examines how electricity prices affect process conditions, a key factor in production, and finds that, contrary to common practice, maximizing all operating parameters is not the most cost-effective strategy. Finally, an insilico framework for model predictive control, applied to a reactor endfill scenario, demonstrates that oxygencontrolled processes can be dynamically optimized, highlighting the strong potential of bioprocess models for industrial usage.

Accurate life cycle assessment (LCA) depends on robust characterization factors (CFs), which quantify impacts such as ecotoxicity through the integration of fate (FF), exposure (XF), and effect (EF) factors. While databases such as USEtox and Ecoinvent provide essential CFs, significant data gaps remain, particularly in ecotoxicity endpoints like hazardous concentration 50% (HC_50), which directly inform effect factor calculations. Existing machine learning models can predict such values, but they often lack interpretability, which limits trust and transparency in environmental modeling. To address this, a machine learning framework is applied that utilizes symbolic regression (SR) and genetic programming (GP) to predict missing HC_50 values from physicochemical descriptors. A dataset with 14 descriptors was used to train SR models capable of generating interpretable mathematical expressions that link chemical properties to HC_50 values. SR models were benchmarked against prominent black-box models such as random forest (RF) and neural network (NN) models. SR performance was consistent across multiple parameter configurations, while revealing recurring patterns in variable selection. These results demonstrate that symbolic regression can both predict ecotoxicity values with comparative accuracy and provide transparent functional relationships, thereby filling existing data gaps in characterization factors needed for a complete LCA.

Cement is regarded as the most widely used construction material worldwide; however, its production is also recognized as a major contributor to global CO2 emissions. Strict control of cement quality is therefore required to prevent excessive consumption of raw materials and energy, which would otherwise increase the process environmental footprint. Cement quality is largely governed by clinker quality, which is primarily characterized by two quality control parameters: free-lime content and alite fraction. At present, these are characterized by costly and time-consuming laboratory analyses that are not optimal for real time process control and optimization. Hence, in this work, a soft sensor for the real-time estimation of the clinker alite fraction is proposed. The developed soft sensor is designed to adapt to process drifts and operating condition changes, capture nonlinear and dynamic behavior, and retain interpretability through a Partial Least Squares (PLS) modelling framework. To this end, a novel recursively adaptive local dynamic soft-sensing strategy (ALD-PLS) is introduced and implemented within a multi-model ensemble structure referred to as Quasi-Ensemble PLS (QE-PLS). Unlike conventional ensemble approaches, where model diversity is generated through data resampling or training–testing partitioning, the proposed framework constructs multiple sub-models using combinations of model hyperparameters, evaluated on the same evolving dataset. As a result, improved predictive accuracy and robustness are achieved, while estimation uncertainty is quantified. The proposed QE-PLS soft sensor is shown to outperform, in terms of R2 and RMSE, PLS-based and single-instance implementations ALD-PLS for a similar task. The novel methodology is validated against data collected from an industrial cement production plant.

Modular plants emerged as the key solution for reducing time-to-market and increasing flexibility in the process industry by combining different modules known as Process Equipment Assemblies (PEAs). While PEA automation is standardised through the Module Type Package (MTP), comparable tools for their documentation remain absent. This work presents the Semantic PEA Datasheet (SPEAD) ontology, which represents PEA documentation as a machine-readable knowledge graph that adheres to the FAIR principles. SPEAD integrates established standards such as DEXPI and the VDI 2776 guidelines and ensures data quality through comprehensive annotations and constraint-based validation. The ontology was evaluated against twelve competency questions derived from a representative use case as well as competency questions from the literature using a continuous stirred-tank reactor PEA as well as a dosing PEA as example systems. SPEAD successfully covers operational and design parameters as well as interface definitions, and provides provenance tracking through rich metadata. Comparison with existing information models OntoCAPE and DEXPI shows that SPEAD provides enhanced provenance tracking essential for trust in engineering documentation. Future work includes enriching capability and safety descriptions and exploring integration with the Asset Administration Shell (AAS).

Hierarchical optimization problems such as (generalized) semi-infinite optimization problems and bilevel problems appear in various disciplines of process systems engineering, such as flexibility analysis or parameter estimation. Adaptive discretization-based algorithms are a family of methods to solve these problems. In these methods, the original problem is decomposed into subproblems, which are solved with a standard optimization solver and then refined iteratively. Several related algorithms have been published. Until recently, computational studies have typically been performed using publication-specific implementations and benchmark problems. We recently published a software package – libDIPS – comprising existing adaptive discretization-based algorithms and a library of test problems for comparison. Several of the algorithms implemented in libDIPS exhibit strong parallelization potential in their algorithmic steps: In the algorithms of Mitsos [Optimization 60:1291-1308 (2011)] and Mitsos and Tsoukalas [J Glob Optim 61:1-17 (2015)], the upper- and lower-bounding procedures are independent. In the algorithm of Tsoukalas and Rustem [Optim Lett 5:705-716 (2011)], several objective function values can be probed independently for feasibility. However, algorithm statements in the literature typically only consider serial execution of the algorithm steps. While parallelization of the used subsolver is already possible in libDIPS, leveraging parallelization in the algorithmic steps has not been investigated until now. In this contribution, we present an overview of libDIPS and introduce our parallelized implementations of several existing algorithms. We investigate the effect of parallelization at the algorithm-level compared to parallelization within the subsolver. Furthermore, we propose two improvements to speed up the solution process of the algorithms compared to their original publications. Results are presented for the existing library of test problems collected from literature.

This work aims at verifying the CO2 absorption capture model using monoethanolamine (MEA) solvent developed by the U.S. DOE’s Carbon Capture Simulation Initiative (CCSI) and performing uncertainty propagation of mass transfer, liquid hold-up and reaction kinetics properties in the complete model, which includes absorber and stripper columns. The verification of the Aspen Plus CCSI model, based on pilot plant data from the National Carbon Capture Center (NCCC) for a CO2 flue gas concentration between 7 and 11% (mol) allowed uncertainty quantification (UQ) analysis for four different selected operational points using Monte Carlo Simulation (MCS), where low liquid mass transfer parameters exhibited an impact on calculation convergence. Gaussian Processes (GP) surrogate model was implemented, followed by a sensitivity analysis in order to correlate the most sensitive parameters with studied outputs.

The accurate prediction of pure component properties has consistently been a critical issue in fields such as chemical engineering, biomedicine, and environmental science. In recent years, end-to-end deep learning methods have shown significant improvement over traditional machine learning approaches. This is due to their ability to automatically learn task-relevant representations from raw molecular data. In addition to accurate property prediction, researchers have increasingly focused on how specific fragment structures influence molecular properties. However, existing fragmentation methods based on predefined rules and group libraries struggle to capture novel molecular structures, which hampers the development of new materials and drugs. To address these challenges, this work proposes a hierarchical molecular fragmentation method. This method can automatically segment molecules into multiple fragments containing key functional groups. Then a three-branch graph attention network was constructed to achieve multi-level representation. Finally, a multi-layer perceptron is employed to establish the mapping relationship between molecular features and physical property values. Twenty datasets were used for validation, which can be grouped into four categories: Thermodynamic Properties, Pharmacokinetics, Toxicological Properties, and Industrial Safety. The results show that the best performance is achieved, with the average error reduced by 6.8% compared to existing research.

Global optimization of industrial-scale chemical process flowsheets remains challenging due to nonlinearity, nonconvexity, and large problem scale. While equation-oriented modeling frameworks enable high-fidelity representation of industrial processes, obtaining globally optimal solutions is often computationally intractable for off-the-shelf solvers. In this work, we present a decomposition-based global optimization strategy that solves a high-fidelity flowsheet model from the IDAES framework with the Structured Nonlinear Global Decomposition (SNoGloDe) framework. The proposed approach exploits spatial decomposability by partitioning the flowsheet into coupled subproblems linked through a small set of complicating variables and solving them within a prioritized spatial branch-and-bound framework. The methodology is demonstrated on a hydrodealkylation (HDA) process for benzene production, a nonconvex and industrially relevant case study. The flowsheet is decomposed into reactor and separation subproblems, enabling efficient computation of valid global bounds while preserving convergence guarantees. SNoGloDe successfully identifies a (verifiably) globally optimal solution. These results illustrate the potential of decomposition-based global optimization for complex process systems and highlight the advantages of integrating rigorous process modeling with flexible, customizable global optimization algorithms.

Modular plant design requires assessing whether independently characterized process requirements and module capabilities are compatible—a challenge that established methods address incompletely. Feasibility and flexibility analysis, as well as Quality by Design, typically assume integrated single-domain models where all variables belong to one coherent description, yet modular design involves domains that originate from different sources, evolve independently, and connect through interface variables. This work proposes Quantified Constraint Satisfaction Problems (QCSPs) as a formulation for interface-level suitability assessment: universal quantification encodes properties that must hold across their entire admissible range (e.g., physical properties, uncertain or environment-dependent characteristics requiring robustness), while existential quantification encodes variables where at least one feasible value must exist (e.g., critical process parameters, control inputs, configuration options). By operating over domain interfaces rather than requiring integrated system models, the formulation accommodates heterogeneous information sources within a common problem specification, with uncertainty handled through probabilistic intervals governed by a global confidence threshold. The framework supports both finite-domain solving on pre-evaluated datasets and numeric solving through on-demand model evaluation. A bioreactor suitability case study provides proof-of-concept, illustrating how the framework accommodates realistic problem complexity and offering initial observations on method selection trade-offs.

In this work, we present quantum reinforcement learning algorithms for process flowsheet synthesis. Particularly, we discuss the implementation of encoding strategies to improve the algorithmic scalability. Reinforcement learning (RL)-driven flowsheet synthesis techniques provide a promising approach for conceptual process design, in addition to traditional optimization-based methods. These RL-based strategies identify the optimal flowsheet configurations from a maximum set of available processing units, without requiring to pre-postulate an interconnected superstructure. However, the resulting combinatorial design space for RL can scale extensively with the increased number of available processing units, which can render the algorithms to be computationally intensive or even intractable. To address this challenge, our prior work has introduced a quantum-enhanced approach to RL-driven process synthesis. However, this algorithm was limited in its capacity to solve larger flowsheeting problems as a large number of qubits were needed. To reduce qubit requirements and improve scalability, this work presents two different encoding strategies to embed information into the gates of quantum circuits instead of embedding it into qubits alone. We demonstrate the efficacy of these algorithms on three different scenarios of a flowsheet synthesis problem and interpret the results obtained from the IonQ quantum computing platform.

Batch manufacturing is central to fine chemicals, pharmaceuticals, and bioprocessing. Its operation evolves across phases and recipes, which yields high-dimensional trajectories and strong batch-to-batch variability. Meanwhile, key quality-indicative variables are often measured offline and cannot be used as online model inputs. This work presents an integrated deep learning framework that unifies soft sensing and process monitoring in a single module using only process variables as inputs. A multi-scale Temporal Convolutional Network with multiple kernel sizes extracts complementary dynamic features from sliding windows. These features are concatenated and pooled into a compact representation that feeds two task branches. A variational autoencoder branch reconstructs the input window and provides fault monitoring signals via reconstruction deviation while regularizing the latent space through KL divergence. In parallel, a prediction branch estimates the quality-indicative variable directly from the pooled temporal features without using the variational latent sample. This separation preserves a stable quality mapping while retaining a probabilistic reconstruction model for monitoring. During inference, reconstruction error and prediction error are fused into a joint state score that more comprehensively reflects system state changes than either deviation alone. Diagnostic heatmaps are produced from residual maps and optional SHAP attributions to highlight contributing variables and time segments. The framework is validated on the Industrial Penicillin Simulation, an industrial-scale penicillin fermentation benchmark. Results show stable convergence of reconstruction, prediction, and KL terms, clear fault-set monitoring trajectories, and interpretable heatmaps that support actionable diagnosis.

Accurate prediction of solubility limits of organic compounds in mixed solvents is critical for the design and optimization of chemical and pharmaceutical processes. Recent advances in machine learning have enabled fast and reliable prediction of physicochemical properties of molecules, including solubility. In this work, we present a Bayesian Optimization framework to identify optimal solvent combinations, compositions, and temperatures that maximize the solubility of active pharmaceutical ingredients. The optimization strategy leverages a multicomponent directed-edge message passing neural network trained on solvent mixtures to predict solubility in ternary systems consisting of a solute and two solvents. To enable efficient Bayesian optimization, we represented the solvents in a continuous space and compare three different strategies: integer enumeration, numerical descriptors, and deep embeddings. The proposed approach was tested on a dataset comprising 14 299 points in solvent mixtures. The results indicate that integer enumeration and embedding-based representations achieve the best performance in identifying solvent mixtures that maximize solubility for each compound at a fixed number of model evaluations. While demonstrated on ternary systems, the framework is readily extensible to mixtures with a larger number of components. With this work we aim to identify which chemical representation is most suitable for similar optimization problems, providing a solid foundation for further and broader exploration, for example in the optimization of crystallization processes.

Modular process simulators are widely used in industry due to their robust and detailed unit operation models. However, their application to gradient-based process optimization remains challenging, as these simulators are typically treated as black boxes, limiting access to internal equations and derivatives. As a result, finite difference methods are commonly employed for gradient estimation, despite their sensitivity to numerical noise and poor scalability. While previous studies have demonstrated the benefits of analytical derivatives in modular simulators, these approaches have largely relied on tangent differentiation modes. This work proposes a non-intrusive methodology that couples analytical derivatives with the adjoint mode of automatic differentiation to efficiently compute gradients for process optimization in modular simulators. The approach preserves the robustness of existing simulation tools by performing simulations normally to convergence, followed by external adjoint-based derivative evaluation using analytically available sensitivities. Theoretical advantages of adjoint differentiation are thus exploited, particularly for optimization problems where the number of decision variables exceeds the number of objective outputs. The proposed framework is evaluated through a case study based on a combined heat and power cycle, using three benchmark objective functions. Gradient computation cost, robustness, and overall optimization performance are compared against forward and centered finite difference methods. Results demonstrate that the analytical adjoint approach significantly reduces gradient computation time while maintaining high robustness, achieving up to 56% reduction in computational cost compared to finite differences. These findings highlight the potential of analytical adjoint differentiation as an efficient and reliable alternative for optimization in modular process simulators.

High-throughput analysis of microfluidic droplets and bubbles is essential for chemical engineering but remains challenging due to the inherent loss of high-frequency details in standard deep learning models. This study proposes a novel Hand-crafted Feature Fusion framework that explicitly integrates physical priors, specifically Local Binary Patterns and Discrete Wavelet Transform, into a two-stage instance segmentation network. We design an adaptive attention-based fusion module embedded within both the Feature Pyramid Network and Region Proposal Network to synergize explicit texture cues with implicit semantic features. Validated on a large-scale dataset comprising over 64000 instances, our method achieves a test mAP of 0.808, significantly outperforming state-of-the-art architectures. Crucially, the framework effectively resolves the detection bottleneck for minute targets and elevates the small-object accuracy to 0.764, representing an improvement of nearly 20% over the baseline. This work demonstrates that incorporating physical priors offers a superior strategy for precise scientific image analysis compared to generic data-driven models.

Type your abstract text here. This work presents an essential methodological framework oriented to the implementation of sustainability promotion measures in process industries. It makes use of a previously developed paradigm, designated as Water and Energy Integration Systems (WEIS), which are fundamentally conceptual systems based on the implementation of several technologies implemented with the end to minimize water use, energy use and related environmental burdens. The primarily conceptual nature of these systems is significant that these have not been significantly implemented in real-life, and that these have been essentially implemented in the virtual basis of digital twin-based computational models. This work extensively presents a methodology developed for the assessment of the integrity of WEIS models, which have been developed using the capacities of a customised computational tool designated as ThermWatt. Two previously approached case-studies have been considered to perform a proof-of-concept in respect to the applicability of the delineated methodology. The delineated methodology proved to be adequate for the whole process of the elaboration of the WEIS Engineering projects, paving way to the implementation of these in real-life in a future perspective.

This work explores Generative Flow Networks (GFlowNets) as a computational approach for sustainable fragrance design, focusing on generating novel molecules that reproduce the scent profile of vanillin while reducing reliance on resource-intensive synthesis and environmentally vulnerable natural sources. An integrated pipeline couples a GFlowNet generator with a fragrance note predictor, which guides learning toward a target odor by rewarding molecules predicted to be aromatically similar to vanillin. Chemical validity and realism are enforced through chemistry filters that penalize unstable or implausible structures and through an odorless-vs-odorant classifier, so only chemically and olfactorily plausible candidates are selected. The agent is trained in a hybrid offline–online regime, implementing reinforcement-based exploration, with hyperparameters tuned via Bayesian optimization. As an independent validation layer, an olfactory receptor docking model estimates binding affinities to receptors associated with vanillin, and several generated candidates show comparable or higher predicted affinities than vanillin while retaining structural novelty. Overall, the results suggest that GFlowNets can capture multidimensional olfactory patterns and support CAPE-style exploration of vast chemical spaces under uncertainty, enabling more efficient exploration of olfactory ingredients.

Nonconvex generalized disjunctive programs (GDPs) frequently arise in chemical engineering applications and are commonly reformulated as mixed-integer nonlinear programs (MINLPs). However, nonconvexities in these reformulations often lead to numerical difficulties, sensitivity to initialization, and degraded solution quality when solved with general-purpose MINLP solvers. This work proposes a two-phase strategy to mitigate these effects by generating improved initial points through the solution of a sequence of relaxed MINLPs, which are subsequently used to initialize the original formulation. The approach is evaluated on a family of purely disjunctive benchmark problems, referred to as the Crescent problems, with sizes ranging from 60 to 1000 binary variables. Numerical experiments using the DICOPT and SBB solvers assess performance in terms of objective value distributions, the percentage of feasible initial points, and average constraint violation. The results indicate that the proposed strategy improves solution quality, increases the likelihood of feasibility, and reduces the magnitude of constraint violations across all problem sizes.

Anomaly detection is essential for ensuring the safe and efficient operation of chemical plants. Although many deep-learning-based methods have been proposed in recent years, their evaluation remains largely limited to synthetic benchmarks such as the Tennessee Eastman Process (TEP) [1]. While these simulators enable controlled and reproducible comparisons, they fail to capture the noise characteristics, operational complexity, and irregular fault dynamics of real industrial plants, leaving the practical generalizability of many methods unclear. In this work, we extend our earlier ESCAPE study [2] beyond water-based systems to industrially relevant chemical processes. We analyze multivariate time-series data from two continuously operated pilot-plant scenarios at the Technical University of Munich, namely n-butanol/water heteroazeotropic distillation and poly(oxymethylene) ether purification, whose datasets were recently published at NeurIPS 2025 [3]. Using the open-source TimeSeAD library [4], we benchmark 30 anomaly detection methods, including 26 deep-learning-based and 4 classical approaches, under a unified preprocessing, model-selection, and evaluation pipeline. Performance is assessed using the F1-score and the area under the precision–recall curve (AUPRC). Our results show a substantial performance drop when moving from synthetic to real process data, with average scores far below those commonly reported for TEP. No single method performs consistently best across all datasets, and the ranking depends strongly on the chosen metric and process scenario. These findings highlight the limitations of synthetic benchmarks and underscore the need for more realistic industrial datasets, process-aware methods, and evaluation practices that better reflect real operating conditions.

Piping and instrumentation diagrams (P&IDs) encode the functional structure of process plants and are a critical yet underutilised source of engineering knowledge for digital twins and intelligent decision support. However, digitising legacy P&IDs remains challenging due to heterogeneous drawing standards and the reliance of existing methods on brittle symbol recognition and rule-based connectivity reconstruction. This work reframes P&ID digitization as the extraction of equipment tags and inference of process topology, rather than graphical reproduction. We propose a two-stage workflow based on multimodal large language models, in which visual extraction and topology reconstruction are treated as distinct reasoning stages guided by chemical engineering process knowledge. The approach is evaluated on two ANSI-standard P&ID case studies of increasing complexity. Results show that decomposing visual extraction and topology reasoning yields more accurate and structurally consistent process representations than end-to-end digitization, highlighting the potential of language-model-based, knowledge-guided workflows for scalable and semantically reliable P&ID digitization.

Machine learning (ML) has attracted growing interest in process systems engineering for its potential in process design, synthesis, and optimization. By learning complex patterns from data, ML methods complement traditional first-principles modelling and heuristic approaches, particularly for conceptual process design and the exploration of alternatives. Although current text-based representations capture unit-level connectivity, they lack a holistic view of process intent, equipment hierarchy, and contextual information to guide learning and inference. Consequently, models trained on such linear token sequences tend to reproduce syntactic structure rather than underlying process reasoning, thus limiting interpretability and user control. In this work, we introduce a contextual framework for representing process flowsheet information in ML models that embeds process engineering logic directly into the model inputs. The approach combines a structured, text-based representation of process topology with context descriptors that define the process type and synthesis task. The task context, such as simplified block flow design, enables engineers to steer learning and generation towards intended designs. This shifts control from post-training filtering to context-driven learning, allowing models to act as flexible, engineer-guided synthesis tools rather than passive pattern generators. Transformer models were trained on 100, 000 synthetic ethylene glycol process flowsheets, with and without context features. To emulate user-guided synthesis, the models were asked to generate different design configurations involving critical equipment (reactors and distillation columns). Models trained without contextual input achieved an average success rate of 24.6% on average (59.6% in the best case), whereas context-aware models achieved 98.8% on average.

Porous materials such as Metal-Organic Frameworks and Covalent Organic Frameworks are emerging adsorbent materials with tunable structures and chemistry, making them useful for applications such as carbon capture, drug delivery, gas separations, and storage. This work aims to design and develop a systematic approach to build a data-driven recommendation system that leverages the historical experimental data or simulation data to assist process engineers in identifying the most suitable adsorbents from a large candidate space. In general, only some of the adsorption properties are available for porous materials owing to limited experimental data. In this scenario, this problem can be formulated as a matrix completion problem, which aims to impute the missing data by exploiting the underlying pattern in the available data. To this end, we propose a parameterization of the kernelized probabilistic matrix factorization framework, which aims to determine the nonlinear latent factors that are parameterized. The resulting bi-convex Maximum a Posteriori objective with reproducing kernel Hilbert space penalties can be solved using an alternating minimization approach. We demonstrate this approach on a publicly-available COF dataset. Of the 16 adsorption properties considered in this study, 15 of them can be predicted very accurately for a missing percentage of up to 60%. Further, the proposed approach preserves the ranking of COF candidates, which helps in accurate screening of best COFs for a given task.

Large-scale mixed-integer programs (MIPs) typically cannot be solved by standard solvers with reasonable computational cost. MIP-heuristics decompose large-scale monolithic mixed-integer programs into polylithic programs such that they can be solved with reasonable computational cost at the price of loosing their optimality certificate. The decomposition is steered by hyperparameters that impact the solution quality and the computational cost diametrically. The proper selection of the hyperparameter values is a black-box optimization problem which is mostly solved by grid search or random search. In previous publications the authors proposed a novel hyperparameter optimization method based on Bayesian optimization and studied a use case from the PSE domain. Computational studies showed that the BO-based algorithm is superior for objective functions with few optimal solutions.This contribution generalizes the description of the MIP-Heuristic Optimization Problem (MIP-HOP) and the computer implementation of the solution-method. It supports the transferability of the method to other use cases. The general mathematical description of the MIP-HOP comprises eight expressions covering the monolithic and the polylithic MIP-model, the heuristic decomposition and composition operator, and the MIP-HOP. The formulation is explained and linked to the computer implementation of the solution method. The Python-code integrates state-of-the-art tools for mathematical programming (Pyomo and CPLEX) and Bayesian optimization (SMAC3) and will be published on GitHub.

Fault diagnosis is essential for ensuring safety and efficiency in chemical process industries. Conventional diagnostic systems often generate raw numerical outputs that require extensive human interpretation, increasing the operator’s workload and slowing decision-making during abnormal events. To overcome these limitations, this work introduces a model context protocol (MCP)-integrated fault diagnosis framework, where a Large Language Model (LLM) functions as the MCP client, coordinating multiple diagnostic tools through a unified protocol. Within the proposed framework, the LLM interacts with specialized diagnostic tools, including a convolutional neural network-based fault diagnosis model and an ensemble-based variant for uncertainty-aware analysis. The LLM synthesizes the outputs of these tools and generates operator-oriented natural-language reports that summarize diagnostic results and explicitly communicate uncertainty, thereby supporting more transparent and efficient decision-making.A benchmarking study based on the Tennessee Eastman (TE) process is conducted to evaluate the framework using multiple LLMs under identical settings. The evaluation assesses diagnostic accuracy, tool adherence, and operator report quality using an LLM-as-a-judge methodology. Experimental results show that LLM performance varies significantly across models, and that reliable tool calling and uncertainty-aware reasoning are more critical than model size for effective fault diagnosis assistance.Overall, the proposed framework demonstrates strong potential for enhancing fault diagnosis workflows in chemical processes by improving usability, interpretability, and decision-making efficiency without modifying existing diagnostic algorithms.

Dynamic simulation of distillation columns and chemical reactors remains essential for plant design, controllability analysis, and economic optimization. High-purity separations of close-boiling mixtures present significant computational challenges due to nonlinear thermodynamic behavior and stiff differential-algebraic equation (DAE) systems. This work presents a unified Python/JAX framework integrating four computational modules: (1) Peng-Robinson thermodynamics with complex-step differentiation, (2) nonlinear solvers (Newton, Broyden, Newton-Krylov) with automatic Curtis-Reid scaling, (3) DAE solver with Radau IIA collocation and intelligent auto-selection, and (4) constrained optimization using the Augmented Lagrangian Method with JAX automatic differentiation. The framework leverages JAX’s just-in-time compilation (JIT), vectorization (vmap), and automatic differentiation (AD) to achieve near-compiled-language performance. Validation includes: nonlinear solver benchmarks with Newton-Krylov achieving 100% success across seven problems (n=2 to 5000), Williams-Otto reactor optimization with 0.06% deviation from published literature and 1325× real-time speedup, and a 180-stage propylene-propane splitter with 18× real-time performance under three concurrent disturbances. The framework shows that is possible to have an open-source alternative suitable for real-time optimization, operator training, and advanced process control applications.

The use of equipment digital twins of standardized, multi-purpose units can accelerate process development and reduce experimental effort. Experimental data are essential not only for identifying critical process parameters and enabling model-based methods within a Quality by Design framework, but also for constructing and validating the simulation models that describe digital twin behavior. To achieve high-fidelity and robust predictive models, structured concepts are required to manage metadata and process-, product-, and resource-specific information exchanged between physical and digital twins. Electronic lab notebooks (ELNs), which contextualize experimental data, must therefore be structured and standardized to ensure interoperability and seamless data exchange. For integration into digital twin workflows and process transfer between equipment instances of the same category, ELNs must comply with FAIR (Findable, Accessible, Interoperable, Reusable) data principles. This work proposes a Semantic Integration Pattern for FAIR-compliant ELNs to document simulated experiments using domain-specific, vendor-independent standards such as DEXPI 2.0 and ISA 88/IEC 61512-1. DEXPI 2.0 supports standardized descriptions of equipment and process steps, while ISA 88/IEC 61512-1 structures experimental recipes. The resulting FAIR ELN Semantic Integration Pattern enables standardized data exchange between ELNs and digital twin tools. Its applicability is demonstrated through an illustrative example documenting simulated saponification experiments in gPROMS® using simple batch reactor model. The results will support subsequent real experiments conducted in a 2L stirred tank unit.

Artificial neural networks (ANNs) have emerged as powerful surrogate models in process design and optimisation, capable of capturing complex nonlinear process behaviour while significantly reducing computational cost compared to detailed first-principles simulations. However, ANN prediction errors in safety-critical applications can lead to suboptimal or vulnerable designs, necessitating rigorous treatment of approximation uncertainties. While probabilistic approaches exist for surrogate-based decision making, risk-averse contexts that require formal robustness guarantees face a fundamental challenge: the nonconvex nature of ANN-embedded models hinders the employment of standard robust optimisation methods. To this end, in this work we explore the global robust optimisation of process design problems with embedded ANNs. A robust spatial branch-and-bound (RsBB) algorithm to achieve global optimality is proposed while enforcing constraint satisfaction across all uncertainty realisations. This approach integrates spatial branch-and-bound with adversarial cuts generation, concurrently searching for the global optimum while ensuring robust feasibility despite ANN prediction uncertainties. We test the proposed algorithm on the ANN-embedded optimisation of a cumene manufacturing process, across a range of conservatism levels. The algorithm computes solutions within the prescribed optimality tolerance while quantifying the robustness-performance trade-off.

Real-time monitoring of metabolite concentrations is critical for optimising bioprocess performance. While Raman spectroscopy offers a non-invasive solution, translating spectra into metabolite concentration estimates requires robust machine learning models. Foundation models such as TabPFN demonstrate exceptional predictive performance but suffer from high inference complexity when trained on large calibration datasets, hindering their use in real-time laboratory settings. This study proposes a batch Active Learning (AL) strategy to efficiently calibrate TabPFN using a minimal subset of data. We employ a weighted K-means clustering strategy that balances model uncertainty and dataset diversity to select the most informative calibration samples. We evaluated this method on a dataset of nearly 7, 000 Raman spectra covering eight substances. Our AL strategy achieved a mean R² score greater than 0.95 with approximately 1, 000 samples, significantly outperforming random sampling. Notably, the method matched the accuracy of a model trained on the full dataset using only 20% of the data. This reduction lowers computational complexity by a factor of 25, enabling millisecond-scale inference times suitable for high-throughput bioprocess monitoring.

The increasing demand for energy, water, and chemical products signals the need for more sustainable and efficient process design methodologies. Traditional methods for conceptual process design constrains the exploration of novel and intensified process alternatives, as they rely on prior knowledge in defining the design space. Previous studies employing bottom-up approaches, such as phenomena building blocks (PBBs), suggest that the synthesis of complex bottom-up flowsheets remains computationally challenging and is thus limited to the synthesis of individual units of operation. This work proposes a bottom-up, data-driven framework for process synthesis and intensification based on phenomena building blocks (PBBs), in which process flowsheets are constructed from their underlying physical and chemical phenomena rather than conventional units of operation. The proposed framework introduces a phenomena-based text representation and data collection module. Furthermore, a sequence training and generation module is developed, which learns patterns governing PBB interactions and placement within flowsheets. A case study on ethylene glycol production is presented, in which nine fundamental phenomena yield 49 distinct PBB tokens. Results show that the trained model successfully reproduces established flowsheet structures and generates flowsheets using novel PBB combinations, highlighting its potential to support sustainable and intensified process designs.

Pharmaceutical and fine chemical industries face strong pressure to shorten time-to-market while maintaining compliance with complex regulatory frameworks. These conflicting demands require rapid process design, validation, and scale-up. Modular production plants standardized in VDI 2776 and VDI/VDE/NAMUR 2658 have emerged as a promising strategy to shorten engineering and validation efforts. The Product–Process–Resource (PPR) philosophy represents a key approach to efficient data management in modular plant engineering. It enables the separation of different flexibility dimensions into distinct, relevant aspects that can ideally be exchanged or modified independently. To realize this principle in practical applications, formalized information models and ontologies serve as a key enabler for structuring and managing semantic data. This work investigates several information models and ontologies for the process engineering domain regarding their suitability to support separation according to the dimensions defined by the PPR approach. As a metric for separability, coupling measures from Design Mapping Matrix (DMM) research are applied. This work provides the foundation for improving computer-aided engineering tools to efficiently handle modular processes while considering the reshaped information distribution.

A flexible start-stop operating policy that involves full shut-down and start-up may be beneficial for electrified plants under certain grid conditions, such as dispatchable demand response. This paper introduces a multi-period Hamilton-Jacobi reachability framework to explore the space of state trajectories for plant shut-down and start-up. Shut-down is defined in terms of operations leading to a stand-by state with no material flows or energy inputs, and variables within safety constraints. Candidate stand-by states are identified by constructing backwards reachability tubes from the desired steady-state operating point. The candidate shut-down/stand-by state is partitioned in fast and slow regions. Admissible control input trajectories are determined for the fast region, from which the minimum time trajectory is selected as optimal for fast start-up. A proof-of-concept simulation using a reaction/separation/recycle plant is presented.

In this paper, PhoSim V.0, a physics-based digital twin core of a wet-process phosphoric acid (WPPA) digestion reactor, is presented as an initial step toward a high-fidelity virtual representation of industrial WPPA plants. The simulator integrates dissolution, reaction, and crystallization phenomena under strongly non-ideal electrolyte conditions. The reactor is modeled as a batch reactor, where the dissolution of tricalcium phosphate is described using a shrinking-core approach, while gypsum precipitation is represented by a one-dimensional population balance equation capturing particle size distribution evolution. Supersaturation ratios governing nucleation and crystal growth are computed from non-ideal activities using a Pitzer-based thermodynamic model, ensuring consistent coupling between dissolution, reaction stoichiometry, and crystallization kinetics. The resulting stiff system of equations is solved to predict key process indicators such as speciation, supersaturation ratio, and characteristic crystal size. Implemented within a user-friendly simulation environment, PhoSim V.0 provides a modular and extensible foundation that will be progressively refined to mimic industrial WPPA digestion units and supports future monitoring, optimization, and control-oriented applications.

Microstructure segmentation is essential for quantitative materials analysis; however, supervised deep learning demands substantial annotation, whereas general-purpose foundation models such as the Segment Anything Model (SAM) offer limited domain-specific semantic control. This paper presents MatStudio, a human-in-the-loop framework for microstructure segmentation that is proposed and implemented end to end in this work. MatStudio couples an interactive workflow for batchwise micrograph annotation and model adaptation with a dual-head convolutional architecture and SAM-guided boundary refinement. The loop combines sparse supervision with SAM-assisted labeling, task-specific training, and iterative batch-level correction, typically converging within two to three cycles.. The network comprises a shared encoder initialized from a pretrained backbone and two decoders: a UNet-style segmentation head that jointly predicts class labels and pixelwise uncertainty, and a prototype branch that measures texture similarity against a memory bank of learnable prototypes. An uncertainty-aware training objective couples label estimation with confidence; uncertainty weighting emphasizes prototype learning in ambiguous regions, and purity filtering restricts prototype updates to phase-interior features to mitigate boundary contamination. At inference, decoder outputs are fused in a spatially adaptive manner according to local confidence. We further introduce SAMFusion, which treats SAM as a morphology-aware proposal source and accepts proposals that are consistent with the fused prediction under a hierarchical small-to-large activation criterion, yielding improved boundary fidelity without case-specific tuning of SAM prompts. Experimental evaluation on Ni-alloy micrographs and public benchmarks shows consistent gains relative to interactive scribble-based segmentation and SAM-centric reference methods.

Environmental sustainability is increasingly recognized as a critical consideration in pharmaceutical development, yet it is rarely incorporated at the scale of molecular-level design. This study introduces a strategy to predict cradle-to-gate indicators that can be flexibly incorporated into multiple early-stage molecular prioritization scenarios. A dataset of 150 pharmaceutical-relevant molecules was compiled, with each molecule described by structural descriptors, thermophysical properties, and ReCiPe endpoint indicators representing human health, ecosystem quality, and resource scarcity. A dual-branch multi-task model combining graph-based and descriptor-based representations was trained to predict these three endpoint indicators. Model performance was evaluated through validation metrics, local sensitivity analysis, and SHAP-based interpretability. A case study with solubility-based feasibility constraints was then used to illustrate how different sustainability weighting schemes affect molecular ranking and to demonstrate the potential for incorporating sustainability assessment into early-stage molecular prioritization. The results indicate that sustainability preferences can lead to distinct prioritization patterns, while some candidates remain comparatively favourable across scenarios.

Reliable multimodal monitoring in crystallization processes remains challenging due to heterogeneous PAT signal quality, sensor drift, asynchronous sampling and nonstationary noise. This work presents a machine-learning-assisted fusion framework that integrates multimodal PAT alignment, estimation and physics-guided regularisation to generate coherent concentration and particle-size trajectories. A mechanistically informed simulation platform is developed to produce synthetic Raman, FTIR, FBRM and image-based crystal size data with realistically simulated drift, heteroscedastic noise, dropouts and distortion patterns. Sensor reliability is inferred through a Random Forest model trained on variance-normalised discrepancies and quality metrics, which allows the dynamic adjustment of channel contributions. Across modalities, the Random Forest achieves MAE values of 0.03-0.20 for probability-type indicators and shows stable explanatory power for variance-inflation factors on particle-size channels (R2 = 0.81-0.93). Two representative PAT scenarios illustrate the performance of the proposed framework under different cross-sensor discrepancy structures, demonstrating improved robustness, reduced local variance and enhanced physical coherence compared with individual signals. Overall, the results highlight the potential of multimodal information fusion as a foundation for trustworthy online monitoring and modelling in crystallization.

We investigate ultrasound-assisted transesterification for biodiesel production. We use inline Raman spectroscopy to quantify its impact on reaction kinetics, catalyst reduction, and temperature sensitivity. We perform a systematic experimental study at different temperatures (50, 55, and 60 °C), different catalyst loadings, with and without ultrasound. The results show that ultrasound significantly accelerates early reaction kinetics at all temperatures, with the strongest effect observed at 55 °C, where both Fatty Acid Methyl Ester (FAME) formation rate and final conversion increase by up to 7 wt%. Under reduced catalyst conditions, ultrasound restores high conversion levels, leading to up to 20 wt% higher final FAME compared to operation without ultrasound and achieving performances comparable to, or exceeding, those obtained with (normal) catalyst without ultrasound. This is mainly because ultrasound primarily enhances mass transfer and phase contact, thereby reducing the system’s sensitivity to catalyst loading. These findings demonstrate that ultrasound enables catalyst reduction while maintaining high biodiesel yields, enabling more sustainable and intensified transesterification processes.

Predicting solvent effects on reaction activation barriers is central to understanding chemical reactivity and reaction kinetics, and guiding solvent selection. The solvent-induced change in activation free energy (DDG_solv‡) provides a quantitative descriptor of this effect, but remains costly to evaluate across vast reaction-solvent spaces, using quantum mechanical methods. Recent data-driven models have enabled prediction of solvent effects. However, most typically rely on two-dimensional representation of reactions and do not explicitly encode sufficient reaction context, such as transition-state information, or three-dimensional structural changes along the reaction, resulting in limited generalizability and predictive accuracy. In this study, systematic evaluation is presented of modelling strategies for predicting DDG_solv‡, with a focus on the role of reaction-state representation, input-geometry fidelity, and input modality. Using a large reaction-solvent dataset, models based on two-dimensional condensed reaction graphs are compared with models incorporating three-dimensional geometries of reactants, transition states, and products. The sensitivity of geometry-based models to structural accuracy is assessed by replacing quantum-chemically optimized transition states with structures predicted by a generative model. In addition, a dual-modality architecture combining two-dimensional graph-based and three-dimensional geometry-based representations is examined. The results show that explicit inclusion of both reactant and transition-state geometries leads to improved prediction accuracy relative to representations based on reaction endpoints or transition states alone. However, model performance depends strongly on the fidelity of the input geometries, with substantial degradation observed when low-quality structures are used. The dual-modality approach partially mitigates this sensitivity by adaptively reweighting two-dimensional and three-dimensional information, leading to performance recovery under low-fidelity conditions.

Nonlinear model predictive control (NMPC) can substantially improve performance and constraint handling for (bio)chemical processes, but its adoption is still limited by the effort required to build maintainable first-principles models and to implement efficient dynamic optimization-based controllers. This paper presents GlycoPy, an open-source, equation-oriented and object-oriented Python framework that supports hierarchical model construction and integrated workflows for simulation, parameter estimation, dynamic optimization, and NMPC. The case study of the monoclonal antibody glycosylation process based on a multiscale model demonstrates the capability of GlycoPy.

The increasing adoption of modular plant concepts in process engineering requires new strategies for design and safety evaluation. Conventional hazard analysis methods, such as HAZOP are time-consuming and must be repeated whenever a configuration changes, which contradicts the flexibility that modularization aims to achieve. This work combines concepts from the modular HAZOP method and the preHAZOP method to enable an automated, early-stage safety assessment for modular plants based on P&IDs. A workflow is developed that generates a combined P&ID for a modular plant from individual module (PEA) P&IDs provided in DEXPI format and performs an adapted preHAZOP analysis on the resulting plant representation.A key motivation is that P&IDs are always created during plant or module design and already contain relevant piping information, which can be reused for automated module interconnection and plausibility checks. In the proposed workflow, standardized interfaces enable automatic connection generation and validation. The safety assessment is performed by rule-based deviation propagation on a DEXPI-based P&ID graph and can operate with, without, or alongside process data. The workflow outputs structured scenario tables and selected design consistency checks and is implemented as a Streamlit web application, validated on a modular example system.

Fourier Transform Infrared Spectroscopy (FTIR) is a powerful technique for polymer analysis. Conventional FTIR analysis struggles with complex patterns, and expertise is required, limiting scalability in high-throughput environments. Machine Learning (ML) offers a promising route to accelerate data processing by objectively identifying and classifying spectral patterns. This work aims to identify polymers from their FTIR spectra with ML. A novel methodology inspired by chemical intuition is proposed, combining unsupervised dimensionality reduction of the FTIR spectra, bond fraction prediction via deep learning, and polymer identification through matching bond fraction predictions with a reference file. Several architectures are explored, with direct polymer classification used as a benchmark. Additionally, a Neural Network (NN) is designed to predict the oxidative degradation state of poly(ethylene) and poly(propylene) samples. For the polymer bond fraction prediction, the best results are obtained using a NN with latent values of the autoencoded FTIR spectra as input, achieving a square root of the mean squared error of 0.023, and correspond to the bond fractions from the polymer’s repeating unit. The highest classification accuracy (75%) is obtained after augmenting both the reference file and spectral training data, using Euclidean distance as a matching method to the reference file. The oxidation detection algorithm reached 100% accuracy by restricting the spectral range input to the carbonyl region (1, 800–1, 550 cm-1), where oxidation features appear. The proposed methodology enables the identification of both seen and unseen polymers during training, outperforming direct classification methods and establishing a scalable framework for automated polymer identification.

Dimethyl ether (DME) is a promising clean fuel and chemical intermediate, yet its synthesis from synthesis gas remains highly sensitive to both catalyst formulation and operating conditions. In this work, a data-driven framework is developed that combines machine learning surrogate modeling with multi-objective optimization to support systematic decision-making in DME synthesis. The novelty lies in the systematic comparison of different optimization approaches applied to an identical machine learning surrogate model for DME synthesis, thereby highlighting their respective strengths and limitations as decision-support tools under limited-data conditions. A dataset compiled from published literature includes catalyst composition, preparation methods, physicochemical descriptors, and operating conditions, with CO conversion and DME selectivity as performance indicators. After data preprocessing, feature analysis using correlation analysis and principal component analysis (PCA) is applied to explore dominant trends and interactions. Several supervised machine learning models are trained and benchmarked, and the best-performing model is selected as a surrogate for optimization. Two complementary multi-objective optimization strategies are then applied to the same surrogate model: a fuzzy-enhanced NSGA-II algorithm and a constrained multi-objective Bayesian optimization approach. Gradient boosting is found to provide the most reliable predictive performance among the tested models. Both optimization strategies identify similar catalyst formulations as optimal, while differences emerge in the recommended operating conditions. The comparative analysis highlights how different optimization paradigms influence compromise solutions when balancing CO conversion and DME selectivity.

Distributed deployment of process systems can benefit from modularity and shared components across multiple variants, reducing both manufacturing costs and engineering effort. Process family design formalizes this idea by simultaneously optimizing a family of process variants while determining a shared platform of common components. This results in a large-scale mixed-integer nonlinear program (MINLP) that couples nonlinear process models with discrete platform-allocation decisions. In this work, we solve the process family design MINLP using a progressive hedging (PH)-based decomposition strategy that exploits its block-angular structure. To improve convergence for this nonconvex problem, we introduce dynamic gradient-based penalty updates, a decoupled primal-dual strategy via separate PH runs, and parallel optimization-based bounds tightening of first-stage variables. Computational results on a water desalination case study demonstrate that the proposed approach improves solution quality and reduces computational time compared to baseline PH, while achieving a lower total cost of the process family than the previous discretization-based MILP formulation [1]. These results highlight the effectiveness of tailored decomposition strategies for large-scale process family design problems.

Machine learning (ML) reduces reliance on computationally expensive first-principles simulation while capturing complex nonlinear behaviors. However, poor extrapolation, overfitting, limited interpretability, and lack of strict consistency with governing laws limit the use of ML models in process applications. Current methods for learning optimization policies also struggle with constraint satisfaction and optimality guarantees. Approaches such as physics-informed neural networks (PINNs) incorporate constraints “softly” and do not ensure strict constraint enforcement—an issue that can be particularly detrimental in safety-critical applications, where even minor violations may lead to unsafe or infeasible decisions. To resolve these issues, we develop an ML framework with a differential projection layer that allows computationally efficient process modeling, parameter estimation, and nonlinear optimization with feasibility and optimality guarantees. The framework is general in a sense that, depending on the loss function and the projection layer, it provides different functionalities. For instance, an ANN followed by a distance minimization-based projection leads to a hybrid mechanistic/data-driven process simulator, KKT-HardNet, that ensures inference with strict satisfaction of linear/nonlinear equality/inequality constraints. An ANN with a gradient-based loss function and a Taylor approximation-based pointwise projection leads to DAE-HardNet, which solves and/or estimates parameters of ODE/PDE-based governing equations. Lastly, an ANN with a distance-minimization based projection followed by objective-based loss function leads to KKT-OptNet, which learns to solve nonlinear optimization problems for different parameter instances. We illustrate the use of these tools for the modeling, optimization, and parameter estimation of complex processes systems.

In this work, the synthesis, design and optimization of a quaternary double dividing wall distillation column (QDDWC) is presented. The effect of the feed composition over the performance of this intensified configuration is studied. The synthesis and design of the QDDWC takes place using as basis a conventional direct sequence for the separation of a n-butane/n-pentane/n-hexane/n-heptane mixture. The column is tested for three molar feed compositions: 40/10/10/40, 25/25/25/25, and 10/40/40/10. The configurations are optimized through a multiobjective genetic algorithm to simultaneously minimize the total heat duty and the number of stages. According to the results, the proposed structure allows savings in heat duty up to 59% but requiring up to 28% more stages than the conventional sequences.

Efficient deployment of carbon capture, utilisation, and storage (CCUS) within industrial clusters requires coordinated CO2 allocation under economic, technical, and environmental constraints, particularly when CO2 availability is limited. This paper presents a centralised optimisation framework for allocating captured CO2 from nine industrial sources to six utilisation and storage sinks within an industrial park in Qatar. A multi-objective mixed-integer linear programming (MILP) model is developed to minimise total system cost while accounting for capture, purification, transport, and utilisation processes, and enforcing an environmental feasibility constraint to ensure net CO2 reduction. The model is evaluated under four scenarios: a baseline case with sufficient CO2 to satisfy all sink demands, and three scarcity scenarios in which 15%, 25%, and 35% of total source emissions are available. Results show that under scarcity, allocations prioritise large EOR sinks supplied by high-volume, low-purity sources, while utilisation pathways are progressively reintroduced as availability increases. The findings highlight the critical influence of CO2 purity, sink requirements, and supply constraints on allocation outcomes and underscore the importance of centralised planning for robust CCUS deployment.

Simulated Moving Bed (SMB) chromatography is widely used for a variety of separations, yet, when applicable, these systems are typically operated using offline optimization strategies. Over time, process degradation and unforeseen disturbances may cause SMB units to deviate from the calculated optimal conditions, reducing overall performance. Real-Time Optimization (RTO) offers a promising solution by continuously monitoring and adjusting operating conditions to maintain optimal performance, despite such perturbations. However, experimental implementation of RTO in industrial SMB processes is costly and requires significant interdisciplinary coordination.To address this challenge, a virtual framework is proposed for the preliminary development of a model-based RTO system. The methodology employs a virtual plant–model pair, in which a representative plant model generates in silico experimental data, while a structurally distinct predictive model reproduces these results. Structural mismatch was intentionally introduced to mimic real-world differences between a plant and its mathematical model, and measurement noise was added to enhance realism. Within this methodology, the in silico experiments were successfully generated and the parameters of the predictive model were then estimated using a Particle Swarm Optimization algorithm that sought to minimize the residuals between the in silico experimental data and the predictive model outputs. The parameters were successfully estimated, allowing the predictive model to closely reproduce the behavior of a structurally distinct plant model without introducing additional complications, which is expected to be analogous to a real-scenario RTO system. Hence, this work establishes a critical step towards the foundation of a virtual RTO framework.

Cryogenic air separation units (ASUs) are the primary industrial technology for producing high purity oxygen, nitrogen, and argon gases at commercial scale. Cryogenic ASUs are large consumers of electricity, making them ideal candidates for market-driven operation research in today’s volatile and uncertain manufacturing environments. To maximize profitability, ASU operation must dynamically adapt to changing market conditions as they evolve. This work explores the implementation of a rolling-horizon scheduling (RHS) strategy for the real-time market-driven operation of a high-dimensional ASU model with inventory, responding to uncertainty in future plant demand and electricity price forecasts by periodically rescheduling in response to updated market information. A dynamic latent variable-based surrogate model (LV-SM) is used within the scheduling framework as a computationally efficient substitute for an existing first-principles-based ASU model. Results show that RHS and plant inventory are effective strategies for handling uncertainties in the future market forecasts, while the LV-SM shows computational performance suitable for real-time implementation. With this scheduling framework in place, extensions involving expanded case studies and uncertainty-aware optimization are planned for future work.

Deep Kernel Learning (DKL) has emerged as a powerful framework for Bayesian Optimization (BO), via combining expressive representation learning models with typical Gaussian Processes (GPs) surrogate models. However, conventional DKL typically relies on weight-based feature extractors (e.g., multilayer perceptrons (MLPs)), which often lack interpretability and may suffer from overfitting under data scarcity or training instability, potentially leading to a degraded uncertainty quantification in GP models. Grounded in the Kolmogorov-Arnold representation theorem, this paper proposes a novel DKL-KAN framework that employs Kolmogorov-Arnold Networks (KANs) as adaptive feature extractors, formulating a DKL-KAN surrogate model. Unlike MLPs, the KANs learn data-driven univariate functions, yielding more sample-efficient and stable representations for regression under limited data regimes. Followed by the GP, the DKL-KAN facilitates end-to-end learning of expressive latent representations while maintaining robust uncertainty quantification. The proposed DKL-KAN framework is further validated with the Williams-Otto (W-O) process benchmarks. Compared to classical GPs and DKL-MLPs, the DKL-KANs consistently exhibit superior predictive accuracy and uncertainty calibration even under high-dimensional and scarce dataset. The results show that the KANs generate more structured and informative latent embeddings, thus enhancing the ability to capture complex physical nonlinearities in sparsely sampled edge regions of the design space where traditional models often falter. Moreover, the DKL-KAN is embedded into the BO and yields the efficient optimization performance on the W–O process benchmark, achieving faster early-stage improvement and converging to higher objective values. These results demonstrate the potential of DKL-KAN surrogates for accelerating BO in chemical engineering processes.

Black-box optimization (BBO) deals with problems involving functions that are either unknown, imprecise, or costly to evaluate. Current BBO methods encounter multiple challenges, such as high computational demands from excessive function evaluations, difficulties in handling complex constraints, lack of theoretical convergence guarantees, and unstable performance due to significant variations in solution quality. This work presents a machine learning-powered feasible path (MLFP) framework for general BBO problems involving complex constraints. An adaptive sampling strategy is first proposed to explore optimal regions and pre-filter potentially infeasible points, thereby reducing the number of evaluations. Machine learning algorithms are utilized to build surrogates for black-box functions. The feasible path algorithm is integrated to accelerate theoretical convergence by updating only independent variables instead of all variables. Computational experiments demonstrate that MLFP can rapidly and robustly converge near the KKT point, even when training surrogates with small datasets. Compared to state-of-the-art BBO algorithms, MLFP stably delivers equivalent or superior solutions with fewer evaluations across benchmark examples.

The integrated recovery of heat and mechanical work has gained increasing importance in process integration due to the strong thermodynamic coupling between temperature and pressure changes in many industrial systems. This work presents a rigorous framework for the simultaneous synthesis of Work and Heat Exchanger Networks (WHEN), in which heating, cooling, compression, expansion, throttling, and pumping are optimized in a coordinated manner. The problem is formulated using Generalized Disjunctive Programming (GDP), allowing the explicit representation of alternative thermodynamic paths, phase-dependent behavior, and logical equipment choices. Process streams are defined by supply and target states, while only bounds are imposed on intermediate pressures, temperatures, and flow rates. Streams may change classification between hot and cold multiple times and may undergo several phase transitions.Rigorous thermodynamic correlations obtained from Aspen HYSYS are embedded in the optimization model, enabling a physically consistent treatment of condensation, vaporization, and isenthalpic valve expansion without restrictive assumptions. The heat exchanger network is represented implicitly through the Pinch Location Method, avoiding the combinatorial complexity of explicit exchanger matching while accommodating multiple utilities and unclassified streams.The approach is demonstrated through a self-refrigerated alkylation process case study. The optimal solution reveals a non-intuitive thermodynamic path involving alternating vapor and liquid compression stages and multiple phase changes, which enhances internal heat recovery. Compared to a conventional design strategy, the proposed framework reduces total annual energy integration costs by 12.5%. The resulting non-convex MINLP is solved to global optimality within seconds, demonstrating both robustness and computational tractability.

Decarbonizing ammonia by replacing grey with green hydrogen directly affects the operation of the Haber-Bosch (HB) process. When directly coupled to green hydrogen production from renewable energy, the HB process should be able operate flexibly to match variable hydrogen supply. This study presents a structured approach for designing a load-flexible HB plant, supported by a rigorous process model. First, we screen 2, 000 designs at high (100%) and low (10%) hydrogen loads to assess operability. Only 1, 100 designs are feasible for both loads, underscoring the need to account for multivariable interactions during design. Next, we assess the economic feasibility of a base design, comparing HB operation under constant and flexible loads. Flexible operation reduces the levelized cost of ammonia (LCOA) by about 5.8%, primarily by lowering green hydrogen production costs. This cost reduction results from downregulating hydrogen production during periods of high electricity prices. By contrast, HB design improvements yield only small LCOA reductions (about 0.4%), though lower reactor pressure and a larger reactor volume remain the best HB design options to further reduce renewable ammonia costs.

Manufacturing disruptions can be a major driving factor in the wastage of resources and delays which result in spiralling costs and cancelled orders. Operational decision making should therefore consider the potential for disruptions from as many sources as possible, encouraging improvements to operational resilience and agility. Our work presents a scheduling and rescheduling framework formulated as a rolling horizon problem for the emulation of real time decision making within a dynamically changing scenario. The framework is applied to a complex multistage problem with parallel lines susceptible to disruptions as a result of process or equipment failures, or ineffective inventory management that results in material shortages. The framework is demonstrated for a simple example case which highlights the impact of disruptions on the time taken to complete orders and the associated costs. It is observed that the inclusion of disruptions can alter equipment congestion, shifting focus for future process improvements. A scenario with intermittent raw material availability is explored, with greater mean and range for a large number of simulations performed, compared to the case with constant availability. A compounding effect is observed, whereby disruptions lead to a greater likelihood of further disruptions as machine runtimes increase and tasks are repeated. The presented framework presents a strong basis from which future works could be performed in a range of scenarios and with different operating policies.

For process simulation, Python and Aspen Plus can be combined to leverage the advantages of both. This work utilizes the integration of Python and Aspen Plus for the design and optimization of a lycopene production process from tomato waste. Three production pathways are studied: acetone and hexane as solvents, enzymes with ethyl acetate, and supercritical CO2 with ethanol. This allows for the scaling of laboratory-scale studies into industrial-scale analyses. Genetic algorithms are used for optimization, enabling the determination of the optimal process design, costs, and operating conditions, while minimizing the total annual cost. The process with acetone and hexane yields a final production of 0.21 kg/h, the process with enzymes and ethyl acetate, 5.13 kg/h, and the process with supercritical CO2 and ethanol, 0.13 kg/h. It is shown that the process with ethyl acetate has a higher production and the process with supercritical CO2 and ethanol results in lower production and higher costs.

In the pursuit of decarbonisation, process industries are turning to electrification as a solution to avoid fossil fuels for heating and processing raw material. Transitioning to renewable electricity couples the processes to varying electricity availability and requires more consideration for production timing and scheduling to support grid stability and avoid high electricity prices. However, practical challenges limit the capability for unforeseen rescheduling for large processes. This paper explores the idea of auxiliary flexibility in an electrified steel production process, where only the auxiliary systems can react to changing conditions. We model an H2-DRI-EAF inspired process with controllable Air-Separation unit, water electrolysis, pressurized hydrogen storage, gas liquefaction units, and a battery energy storage system to react to a production related demand delay. First, we compare hourly and 15-minute DA pricing and observe that without fast flexibility the cost difference is marginal, while fast-reacting flexibility (electrolyser ramping and batteries) enables a small additional benefit from 15-minute pricing, around 0.3% on average and up to 1% at maximum. Second, without a demand delay, allowing DA+ID decisions provides significant additional cost reduction via intraday arbitrage, yielding up to about 3% lower total costs compared to DA-only operation under perfect price knowledge. Third, when introducing 15–45-minute demand delays close to delivery, the model shows little to no cost increase on average compared to no delay. This highlights that price-taker models yield optimistic results in small markets, where open positions over 100 MW would realistically influence the market price significantly.

Today, beverage production plants are planned and designed from the material-handling context as a packaged-goods production facility, not as a process plant. Therefore, a lot of potential for optimization exists. This paper presents a new approach to the design of beverage production plants according to the design of process plants. A component library for the simple creation of beverage production plant process models is developed. All steps in the plant design process can be accelerated and automated to be used for the high number of existing and new installations around the world. As first use case an energy optimization upgrade for existing Carbonated-Soft-Drink production lines is described to save cooling and heating energy in warm climates.

To effectively distribute power to a system of multiple electrolyzer stack units, control strategies have been developed that now need to be applied to heterogeneous electrolysis systems. These are the ‘segment principle’, the ‘slow start principle’ and the ‘start-stop principle’. As there are many possible combinations to the system composition of a modular heterogeneous electrolysis system together with the most suitable control strategy, a discrete multi-criteria optimisation problem can be formulated. To solve this discrete multi-criteria optimisation problem, two discrete decision variables are introduced. One is the electrolysis system composition, represented by the power ratio/configuration (C). A total of 17 different configurations were used for this, consisting of different proportions of alkaline electrolysis (AEL) and proton exchange membrane electrolysis (PEMEL). The other one are the control strategies (R). For the control strategies, the conventional strategies, mentioned above, have been adjusted to deal with the heterogeneity of the electrolysis system. Variations are made in the priorization of certain technologies for power distribution and the availability of a base load operation. The objective function was designed to be the weighted sum of the three highest ranked decision criteria in the categories ‘flexibility’, ‘cost’, and ‘process engineering’ from an empirical expert survey, which led to the respective decision criteria LGC, LCOH and the reciprocal of eta_sys. The weights were derived from the ranking, a global grid search identified the combination of (R) and (C) which will be discussed in the paper together with a detailed analysis of the optimization problem.

Grid-connected electrolyzers with intermediate hydrogen storage offer significant potential for reducing electricity costs through flexible operation under dynamic pricing. A threshold-based scheduling optimization approach is developed that derives interpretable on/off production rules from electricity price signals. The method identifies local price thresholds separating high-price from low-price periods, yielding binary production schedules. Adaptive horizon partitioning—subdividing the scheduling horizon when constant thresholds become infeasible—is combined with a receding horizon strategy that implements only a portion of each optimized schedule before re-optimization. This procedure enables systematic investigation of how characteristics of Integrated Electrolyzer–Storage Systems (IESS) influence cost reduction potential while maintaining computational tractability for both offline analysis and online implementation. A case study applying the approach to historical German electricity prices across 2, 688 scenarios reveals that storage capacity is the primary determinant of cost-saving potential, while start-up and shut-down times within the tested range (up to 120 minutes) show negligible impact. Systems with storage capacities allowing for hydrogen production stops of up to 1.4 days achieved electricity cost reductions of up to 80% compared to steady-state production. Production cycles in optimal schedules predominantly follow 24-hour rhythms, with approximately 30 cycles per month observed for high-performing configurations. These findings indicate that conventional flexibility metrics—focused on ramp rates and start-up times—are inadequate for assessing the cost-saving potential of flexibly operated, grid-connected electrolyzers under dynamic pricing. Instead, equipment wear from the frequent production cycling required by optimal schedules emerges as a key consideration for economic operation.

Industrial ammonia synthesis remains a high-impact decarbonization target due to the combined effect of large production volumes and reliance on CO2-intensive fossil-based hydrogen. Replacing it partially with renewable-based (green) electrolytic hydrogen can minimize direct emissions; however, the intermittency of solar and wind complicates retrofit strategies for existing continuous plants. This paper addresses a techno-economic study centered on a retrofit-oriented hybrid concept: a conventional natural-gas ammonia plant is operated within a pre-defined “hybridization envelope”, where part of the process hydrogen is supplied by a renewable Power-to-Hydrogen subsystem. A steady-state process model (gray/blue baselines) is coupled with an hourly energy dispatch and a sizing optimization of solar/wind park, electrolyzer, battery, and hydrogen storage. A case study based on 2024 California data for renewables and carbon pricing illustrates how hybridization can reduce carbon intensity while ensuring all process constraints are met, and how the optimal renewable mix trades off between overbuild and storage. Results highlight the role of solar-wind complementarity in lowering the levelized cost of hydrogen and, consequently, the levelized cost of ammonia, while leaving a clear pathway for combining hybrid-green operation with carbon capture (hybrid-blue-green) as a transitional strategy.

Industrial decarbonization is crucial to reducing global emissions. Efficient processes lower energy use and reduce the environmental impacts, such as material use and waste, decreasing the overall industrial footprint. In this context, the present study explores the impact of reducing steam consumption (thermal energy) during the ammonia regeneration process in the production of sodium carbonate. A key feature of the Solvay process is ammonia recycling, which significantly reduces raw material consumption and ensures both economic and environmental sustainability. However, this stage is highly energy-intensive. To enhance energy efficiency in soda ash production, a study was conducted to analyze variation in temperature, pressure, and steam flow introduced into the ammonia regeneration system. The objective is to understand its impact on both ammonia recovery and the process’s energy consumption. Variations in steam pressure do not impact on energy consumption of the process. By reducing the temperature and steam flow introduced into the system, in order to reduce the energy consumption, it was found that there is no decrease in ammonia recovery. The reduction in both the total injected steam flow and operating temperature led to lower heat exchange requirements and a decreased the cooling energy demand, highlighting the potential for optimizing steam flow as an effective strategy to improve process efficiency. Instead, it underscores a potential for energy optimization, promoting a more sustainable and cost-effective operation.

An optimized flowsheet can be generated by numerous approaches. Process optimization via superstructure is one of the methods used to provide solutions that consider the interactions between different decision layers. A process simulator-based optimization is considered in this work, as it offers a reliable and rigorous modeling environment. It is then coupled with a P-graph-based framework to reduce the tedious mathematical writing of the logical constraints to guarantee the structural coherence of a sequence of unit operations.The developed framework consists of three algorithms. The first algorithm transforms the superstructure flowsheet into a P-graph. The second algorithm gets process sub-structures from the superstructure by searching for active units corresponding to a set of decisions made, for example, by an optimizer. The third one checks structural feasibility by verifying that the resulting structure satisfies the five axioms of the original P-graph framework and two additional connectivity tests proposed in this work.The proposed methodology is then applied to an ammonia synthesis process. The results obtained in this work enable a reduction in structural logical constraints while achieving outcomes equivalent to those of a previous study, proving the methodology’s efficacy.

Heat Pumps (HPs) can play a vital role in the decarbonization of heating in industry. The performance of a HP strongly depends on the refrigerant, the working fluid within the HP. In order to maximize HP performance, systematic selection of the refrigerant is key. Refrigerant choice affects the very feasibility of employing a HP to deliver heating to a process. A flexible and robust method is required to select refrigerants that are the best fit for a given heating application. A computer-aided molecular & process design (CAMPD) method is developed to design the optimal refrigerant that is tailored to process needs. The method is applied to three case studies across which the HP performance objectives and constraints, and heat source and heat sink temperatures are varied. In addition, the design of refrigerants with low (<150) global warming potentials and zero ozone depletion potentials is investigated. For all applications across all case studies, the CAMPD approach successfully identifies high-performance refrigerants including those that are known refrigerants, other known fluids and novel molecules.

In recent years, research for systematic process design approaches has gained traction, especially with the rise in popularity of generative machine learning models and reinforcement learning. However, works from the literature will often focus on proof-of-concept studies, limited to a specific process synthesis problem. Despite showing promising results, it is not clear how easily these methodologies could be transposed to new applications, and whether they would be successful. In this context, this work evaluates the possibility of using a Natural Language Processing model, which has already proven itself for thermodynamic cycle generation, for another different case: seawater desalination. The processes generated by this model will initially be those using reverse osmosis processes aimed at desalinating a seawater solution containing 25000 ppm of NaCl. Results show that the model has been successful in designing structural reverse osmosis desalination processes without defining assembly rules or a superstructure. Given the large number of these generated processes, a clustering method based on their similarity has been developed. This made it possible to identify different known heuristics (like multi-stage) in process engineering. This adaptation was made possible by modifying aspects external to the original model: a dedicated vocabulary, design rules and objective function.

Although solvent-based crystallization is widely adopted for separation and purification of crystalline pharmaceutical products, solvent choice and utilisation critically influence product quality, manufacturing cost, and the environmental performance of the pharmaceutical process. Escalating demands to reduce process mass intensity (PMI), together with increasing vulnerabilities in the supply chains, necessitate the development of more efficient and resilient process designs, incorporating solvent and active pharmaceutical ingredient (API) recycling. The conceptual design of crystallization processes offers a viable route to identify flowsheets with substantially reduced solvent consumption. In this paper we present a computer-aided molecular and process design (CAMPD) formulation to explore the benefits of solvent/API recycle for two processes/APIs: (i) a continuous cooling crystallization process for mefenamic acid (MA) employing a binary solvent mixture and (ii) a batch cooling crystallization for paracetamol using a single solvent. At the conceptual design stage, the process is assumed to operate at thermodynamic equilibrium and the SAFT-gamma Mie group contribution approach is used as a rigorous thermodynamic model. The optimal design for each process is determined based on the trade-off between two key performance indicators (KPIs), crystallization yield and solvent consumption. The resulting bi-objective optimization (BOO) problem is implemented in gPROMS. In addition to the solvent identities the proposed methodology also enables the simultaneous optimization of process temperatures, and the feed composition. The results demonstrate that solvents and API recycling can deliver significant benefits, leading to reduced process mass intensity (PMI) while maximizing the crystallization yield.

This work aims to systematically generate intensified process designs by integrating reinforcement learning (RL)-driven process synthesis and phenomena-based modeling via Generalized Modular Framework (GMF). Rather than considering flowsheet synthesis with conventional unit-operations, GMF utilizes fundamental building blocks, also known as mass and heat exchange modules, to describe the physiochemical phenomena and to enhance novel process discovery. At its core are driving forces which characterize the mass transfer feasibility based on the total change in Gibbs free energy of the system. RL is integrated with this phenomena-based modeling strategy to drive flowsheet generation by exploring much of the total action space and minimizing pre-postulation of stream connections. All possible inlets, outlets, and interconnections between modules are contained in a stream matrix. Deep Q-Network is used as the RL agent, which contains a multi-layer convolution neural network followed by a multi-layer feedforward neural network. This network computes q values for each possible stream connection contained in the stream matrix and determines an optimal action from the maximum q-value (e.g., varying the stream interconnections). New designs can thus be generated and iteratively improved. This approach is demonstrated using two case studies: (1) binary separation of benzene and toluene, and (2) membrane-assisted reaction for methanol production.

Batch production scheduling of resources to meet fluctuating product demand is a critical topic in the process industry. Existing optimisation approaches, based on heuristic and exact methods, trade off solution optimality and scalability to large problems. In this work, we investigate deep reinforcement learning as a powerful alternative in order to learn heuristics for batch scheduling. We formulate the batch scheduling problem as a Markov decision process operating on a state-task network representation encoded using graph neural networks, capturing relevant structural inductive biases. We propose a centralised training with decentralised execution architecture, in which agents placed on machines individually choose which tasks to complete using a global view of the network, cooperating towards task schedules that optimise the final production quantity. Preliminary results demonstrate that the proposed end-to-end framework learns to construct task schedules comparable to the optimal solution on small instances unseen during training, exhibiting strong potential for extension to more general graph structures and better scalability.

The pipeline transportation and distribution process of natural gas, from production sources to end consumers, can be divided into three stages: upstream production and supply, midstream storage and pipeline transportation, and downstream distribution and end-use. In the optimization of natural gas pipeline networks, considering the full life cycle, multiple uncertainties in planning, design, operation, and maintenance often affect the efficiency and quality of model optimization. This study addresses the uncertainty in end-user demand during the operation of natural gas pipeline net-works and investigates a scheduling method that simultaneously meets user demand while achieving coordinated optimization of operational costs and carbon emissions. Based on one year of historical demand data, a normal distribution is fitted to characterize the demand, and several representative scenarios with corresponding probabilities are extracted through clustering to capture demand uncertainty. A two-stage stochastic programming model is developed. In the first stage, the basic operational strategy of the pipeline network is determined, while in the second stage, flow, pressure, and compressor power are adjusted according to each demand scenario. The model prioritizes minimizing operational costs and employs the e-constraint method to transform the carbon emission objective into a carbon emission cap. By systematically varying the value of e, the Pareto frontier between cost and carbon emissions is delineated. Furthermore, non-linear constraints are addressed using a combination of piecewise linearization and McCormick re-laxation. The proposed model is implemented in a real-world case study in China using Python and the Gurobi solver. Results demonstrate that, compared to conventional methods, the proposed approach effectively reduce carbon emissions by 34.96%, 42.35%, and 36.84% in high, medium, and low demand scenarios. The method provides clear optimization solutions and supports in-depth analysis of results, thereby enhancing data management capabilities of pipeline systems. The model systematically accounts for uncertainties in natural gas demand, offering valuable research support and practical references for scheduling decisions in the natural gas pipeline industry.

To improve the energy efficiency of distillation processes, various process intensification concepts have been proposed, including direct heat integration and thermal coupling. Identifying the most suitable alternative for a given separation task requires a rigorous and consistent techno-economic optimization. Superstructure models typically rely on isobaric operation and fixed HETP values, in order to avoid treating column hydraulics when solving the already challenging mixed-integer nonlinear optimization problems. In order to overcome this limitation and evaluate the effect of the simplification, the current work extends a rigorous equilibrium-stage superstructure model to account for tray-specific pressure drop and HETP values. A polylithic solution approach is implemented to improve the convergence for the resulting optimization problems. The proposed approach is demonstrated for the optimization of heat-integrated distillation sequences operated at close to atmospheric and vacuum conditions, enabling a closer investigation of the impact of the classically applied simplifications. As the results illustrate that the overall energy demand and total annualized costs are only marginally affected for the considered wide-boiling mixtures the evaluation of competing process configurations will not be affected by the simplifications. However, considerable changes in column height and heat exchanger areas are observed, particularly under vacuum operation, such that column hydraulics should be considered in design optimization, in case column size restrictions or other considerations require a more accurate equipment sizing.

Aluminium cold rolling is an energy-intensive process that has a substantial impact on CO2 emissions and production cost, yet plant-level optimization remains challenging due to strong process nonlinearities and various operational constraints. This study develops a physics-consistent hybrid model that combines a Stone–Hitchcock–Ludwik analytical rolling-energy formulation with a residual deep neural network to predict the daily electricity consumption of three single-stand cold rolling mills. Using plant raw data, the hybrid model achieves lower prediction errors than conventional data driven model and yields line-specific physical parameters that agree well with the observed behaviour of each mill. On this basis, an NSGA-II-based tri-objective optimization is carried out to minimise daily energy use, CO2 emissions, and specific production cost (SPC) by adjusting pass-wise reduction and tension schedules and line-wise production allocation. Case studies on a representative operating day and additional plant data show that the optimised operating strategy shifts production load from less efficient to more efficient lines and smooths pass-wise operating conditions, thereby consistently reducing daily energy consumption and unit cost while moderately decreasing CO2 emissions without any hardware modifications. The proposed hybrid prediction–optimization framework thus provides a practical decision-support tool for integrated energy–environment–economic optimization in multi-line aluminium cold rolling operations.

Anaerobic digestion (AD) across Europe and the UK faces increasing economic and operational pressure from volatile feedstock supply under climate extremes. Existing stochastic programming (SP) approaches for feedstock planning often rely on limited historical observations and/or simplify yield uncertainty in ways that miss the joint, non-linear response of crops to weather variability, thereby understating downside supply risk. We develop an integrated decision-support framework that links climate uncertainty to AD procurement planning by coupling mechanistic crop simulation, generative surrogate modelling, and stochastic optimization. First, APSIM is used offline to generate a mechanistic yield knowledge base across weather trajectories and discrete planting-density choices. Then, a conditional GAN (CGAN) is trained to produce non-parametric joint yield samples for multi crops conditioned on scenario features and management, enabling fast Monte Carlo evaluation. At last, these samples are embedded in a two-stage SP that optimizes first-stage land allocation and planting densities, with second-stage recourse represented by spot-market purchases to cover shortfall. The architecture is designed for stage-based rolling updates as forecasts are progressively replaced by observations. We demonstrate the framework on a 15-ha unit under different contracts of biogas-equivalent output. Results reveal a target-induced regime shift in optimal procurement. Under moderate production targets, monocropping solutions minimize cost with negligible loss in reliability, reflecting broad operational indifference across several land allocation patterns. As contract levels approach the biophysical limits of a 15-ha system (~110×10³ m³), the optimizer transitions into a risk-reducing regime where maize–rye double cropping becomes increasingly necessary. At high targets, the feasible set collapses toward near-complete double cropping. At a representative contract of 140×10³ m³, the least-cost rye-only plan achieves 34.3% supply confidence, while full double cropping increases confidence to 85.7% but with higher cultivation cost. Even under full intensification, an irreducible tail risk remains (14.3% shortfall frequency), implying unavoidable reliance on spot-market procurement under extreme seasons. The APSIM–CGAN–SP integration translates climate-driven biophysical uncertainty into actionable procurement strategies. It reveals threshold behavior, tail-risk exposure, and the limits of intensification under fixed land constraints. The framework supports both pre-seasons planning and rolling in-season updates, providing a quantitative basis for contract feasibility assessment, hedging design, and resilient feedstock procurement.

In this study, we propose a method to uniquely determine robust operating conditions for simulated moving bed (SMB) chromatography, an essential continuous liquid-phase separation technique in the pharmaceutical industry, in the form of explicit algebraic equations. The proposed method incorporates process robustness—defined as the probability of meeting the target purities under flow-rate uncertainty due to pump errors—without requiring computationally expensive dynamic simulations. In a computational demonstration, the method achieved a joint probability of 0.960 for simultaneously attaining 99.9% purity in both extract and raffinate products.

A topology optimization approach is applied to the design of a compact fluid distribution device that achieves uniform flow distribution among parallel channels while minimizing the mean residence time under a pressure loss constraint. Such devices are widely used in applications including parallel microreactors, thermal management systems, and energy devices, where both compactness and precise flow control are required. The design problem is formulated using a density-based method in which the design variable represents the local void fraction of a fictitious porous medium. The governing equations are based on the incompressible Navier–Stokes equations with an additional Darcy-type resistance term. The objective function is defined as the spatially averaged design variable, corresponding to the minimization of the effective flow passage volume and thus the mean residence time. Constraints are imposed on both flow rate uniformity among parallel channels and the total pressure loss. A two-stage topology optimization strategy is proposed to obtain physically realizable geometries. The results show that the proposed strategy effectively removes non-functional regions while preserving flow uniformity, leading to compact designs. Furthermore, numerical investigations reveal that designs with fewer early-stage bifurcations tend to achieve smaller flow passage areas under identical pressure loss conditions. These findings demonstrate that topology optimization provides a rational and flexible framework for the systematic design of compact fluid distribution devices.

Flowsheet synthesis is an integral step in process design, entailing the selection of a set of unit operations and their connectivity to convert raw materials to products. Superstructure optimisation represents a promising class of synthesis approaches, allowing for the systematic exploration of the flowsheet design space. Despite this, many superstructure formulations suffer from numerical instabilities, combinatorial explosion, and/or rely on restrictive assumptions on the types of flowsheet alternatives that can be considered. The modified state-operator network (MSON) formalism has recently been proposed to address some of these issues for isobaric flowsheets. The constant-pressure assumption restricts the applicability of the MSON to real process applications as pressure is a key process variable in many unit operations, such as distillation, reaction, and extrusion, and is necessary to elicit flow. In this work, we present the extended MSON (E-MSON) which inherits the numerical stability of MSON, whilst removing the isobaric assumption. This is achieved through the introduction of new constraints that further improve the numerical behaviour of the MSON. The E-MSON is then applied to a simple non-isobaric superstructure optimisation problem, the results of which demonstrate that the E-MSON can serve as a framework for non-isobaric flowsheet synthesis, enabling a broader range of flowsheet alternatives to be considered.

Sohar Port and Freezone is well-positioned to drive Oman`s economic future through continued expansion, increased reliance on sustainable energy, and the integration of advanced digital tools. Sohar contributes 25% of the current GHG emissions in Oman. Decarbonizing Sohar industrial clusters is a strategic challenge for emerging economies striving to attain net-zero objectives. This research develops a superstructure optimization framework for designing an integrated Carbon Capture, Utilization and Storage (CCUS) value chain, specifically applied to the Sohar Port and Freezone in Oman. A mixed-integer linear programming (MILP) formulation that simultaneously selects capture technologies, identifies optimal transport modes (pipeline, trucking, and shipping), and allocates CO2 flows among utilization and geological storage alternatives. The novelty of this work lies in the multi-scale integration of techno-economic, spatial, and policy dimensions, coupled with GIS-based route generation into the optimization model. Currently, the total CO2 emission from Sohar Freezone is around 20.3 MtCO2/yr, after applying the CCUS value chain optimization, 7 – 14 MtCO2 annual net reduction could be achieved, corresponding to 30-70% cluster-level decarbonization at 50 – 85 USD/tCO2 levelized cost of CO2 capture, transport, and storage. Sensitivity analysis reveals that port-based CO2 hubs, combined with short-haul shipping to offshore storage, reduce the system cost by 23–31% compared to single-mode configurations. The framework offers decision–support tools for policymakers and industrial stakeholders, integrating techno-economic optimization with regional carbon control policies. The Sohar case study demonstrates the capabilities of computational superstructure design to expedite CCUS implementation in the industrial clusters in the Gulf region.

Shell-and-tube heat exchangers are critical assets in the process industries, yet their design optimization typically relies on deterministic formulations that ignore operational variability. This study presents a systematic comparison of four optimization approaches, spanning three paradigms for decision-making under uncertainty, applied to shell-and-tube heat exchanger design. The objective minimizes total annualized cost (comprising capital and pumping costs) subject to thermal duty, pressure drop, velocity, and geometric constraints. Six uncertain parameters are modeled across three categories: mass flowrates (coefficient of variation = 10%), inlet temperatures (coefficient of variation = 2%), and fouling resistances (uniform distribution). Shell-side heat transfer is computed via the Bell–Delaware method, while tube-side correlations follow Sieder–Tate. The four approaches benchmarked are: (1) deterministic optimization as a baseline, (2) sample-average approximation, a classical stochastic programming method, (3) box-constrained robust optimization, which guards against worst-case realizations without probabilistic assumptions, and (4) Wasserstein distributionally robust optimization, which hedges against distributional ambiguity. Results reveal a clear cost–robustness hierarchy: deterministic designs yield the lowest expected cost but lack formal guarantees, sample-average approximation balances cost and reliability with modest computational overhead, the distributionally robust approach achieves the tightest cost dispersion at a moderate premium, and robust optimization delivers the highest feasibility guarantees at the greatest cost. These findings provide practitioners with quantitative guidance for selecting an appropriate optimization paradigm based on reliability requirements, data availability, and computational budget.

While classical heuristics can be applied to decide on the preferred reactor concept for simple reaction schemes, more complex reaction networks require more sophisticated methods, such as the multilevel reactor design approach or superstructure optimization. Based on an analysis of the existing methods a nonlinear programming framework for a superstructure-based reactor network synthesis is presented, emphasizing numerical robustness and flexible network representation without relying on integer decisions. The approach, which is implemented in GAMS, allows for the combination of continuous stirred-tank and cross-flow reactor models. An exemplary application for the classical Van de Vusse reaction is first shown for validation, prior to the application to an enzymatic cascade based on the Weimberg pathway. Assuming fast co-factor regeneration, the performance of the resulting PFR cascade, which can also be interpreted as a sequence of batch reactions, is compared with a commonly applied single batch reactor. The results show that the two-reactor configuration consistently achieves higher product formation by mitigating inhibition effects, demonstrating the potential of reactor network synthesis for complex reaction systems like enzymatic cascades.

Pharmaceutical supply chains face increasing pressure to deliver high service levels while meeting environmental and social expectations. Vaccine supply chains amplify these challenges due to strict cold-chain requirements, demand uncertainty driven by acceptance and preferences, and the urgency of public-health objectives. This paper develops a multi-objective mixed-integer linear programming (MILP) framework for national-scale vaccine distribution that explicitly integrates economic cost, service level, greenhouse-gas emissions, and population-level vaccine effectiveness. Behavioral realism is incorporated by modeling vaccine acceptance and brand preferences as operational constraints rather than ex-post indicators. Trade-offs are explored using an e-constraint method that preserves the MILP structure and enables systematic recovery of Pareto-optimal solutions. The framework is applied to a 52-week national case study for metropolitan France during the 2021 COVID-19 vaccination campaign, focusing on Pfizer-BioNTech and Moderna mRNA vaccines. Results show that population-level effectiveness saturates early once acceptance-driven coverage is achieved, making cost and emissions the primary decision levers under service guarantees. Balanced solutions maintain near-maximal protection while achieving substantial emission reductions at moderate additional cost, whereas deeper decarbonization requires explicit service concessions. The proposed framework provides transparent, policy-relevant insights and supports informed decision-making in large-scale vaccine logistics.

This paper presents a Europe-wide optimisation framework for designing and operating a multi-period Carbon Capture and Storage (CCS) supply chain across Europe. A MATLAB preprocessing pipeline constructs an auditable techno-economic dataset (emission nodes, ports, aquifers, candidate pipeline/shipping arcs and costs) and exports it to a GAMS optimisation model. The planning problem is formulated as a two-stage stochastic MILP, where scenario-independent first-stage decisions select discrete pipeline and shipping capacity bands and port operating modes, while scenario-dependent second-stage decisions allocate capture, transport and sequestration flows. Uncertainty is represented through correlated scenarios of capture unit costs for four capture technologies (CV=0.35, rho=0.8, Ns=20). To address the computational burden induced by inter-temporal binary investments and scenario replication, we apply a two-phase arc-screening heuristic: an LP relaxation on the full network identifies promising corridors, and a reduced MILP is solved on the resulting candidate arc sets. The resulting stochastic model contains 1, 278, 661 continuous variables and 48, 860 binaries and is solved in 1, 347.9 s with a 4.85% optimality gap.

Renewable powered water electrolysis offers a promising strategy to decarbonize industrial sectors with high demand for hydrogen. Operational optimization of industrial electrolyzer systems is often formulated as mixed-integer linear programming (MILP) problems, where a constant hydrogen production to electrical power consumption ratio is assumed instead of the nonlinear relationship. Incorporating a nonlinear electrolyzer model into a linear optimal hydrogen dispatch framework remains a significant challenge. This study addresses this challenge by formulating the optimization problem in two ways. First, the model is solved as a nonlinear programming (NLP) problem by incorporating a nonlinear model of an alkaline electrolyzer (AEL) into the optimization framework. The binaries and integers are relaxed to continuous variables and associated penalty terms are added to the objective function to enforce integrality. The NLP is solved using the local nonlinear solver Interior Point OPTimizer (IPOPT) using a multi-start approach, where the solver is executed from random initial points and the minimum objective value is taken as the best guess for the global minimum. Second, a linearized model of the AEL is used in an MILP formulation. The univariate nonlinear terms within the model are approximated using three piecewise linearization techniques, i.e., convex combination, convex combination with SOS2 variables and a slope based big-M formulation. The bilinear terms are relaxed using the McCormick envelope. Results show that the MILP formulation achieves faster solution times, with varying accuracy depending on the linearization method, whereas the NLP approach more accurately captures the hydrogen production curve, albeit having longer convergence times.

The state task network (STN) representation is a widely used modeling approach for optimal multipurpose batch production scheduling. In practice, STNs have been traditionally formulated as mixed-integer programming (MIP) problems and solved using general-purpose MIP solvers relying on branch-and-bound and branch-and-cut. In the meantime, alternative modeling and solution paradigms for optimization have been developed, enabling the incorporation of alternative variable types and optimization algorithms. Specifically, this work relies on the Hexaly software, which introduced set-based models and their solution through general-purpose hybrid algorithms, i.e., methods that combine traditional MIP with constraint programming, local search, large neighborhood search, among other tools. So far, Hexaly has shown promising results when tackling optimal scheduling problems, however, set-based models and solution approaches for STN optimization have not been studied in the literature. Aiming to fill this gap, this work introduces the first set-based models for the STN problem and performs preliminary benchmarking tests with Hexaly’s general-purpose hybrid algorithms. Results from two case studies suggest that set-based models may perform better than the traditional MIP STN formulation when dealing with a simple STN with sequential connectivity and a long-term scheduling instance of a STN with fixed integer batching variables. Overall, this work establishes the foundations for advancing research on set-based approaches for network scheduling, opening new directions beyond traditional MIP optimization.

Ammonia has received in recent years significant attention as potential carbon free fuel. However, its combustion properties limit its direct application for both providing heat and in power generation through gas turbines. Ammonia cracking is one potential solution to circumvent the problem by producing hydrogen. When using the ammonia in gas turbines, it is possible to heat integrate the endothermic decomposition reaction with the exhaust gas from the gas turbine. Thermodynamic and kinetic limitations have however a major impact on the achievable ammonia conversion. Based on the consideration of these limitations, this paper presents a detailed investigation of key design parameters affecting the overall process efficiency utilizing both an equilibrium reactor model and a reactor model based on detailed kinetics and heat transfer. Ammonia decomposition should occur at sufficiently high pressure to avoid a) the com-pression energy demand for achieving the pressure of the combustion chamber and b) to reduce equipment size although the increased pressure results in a reduced conversion when considering an equilibrium reactor. It is crucial for the energy efficient integration of ammonia decomposition with gas turbines to avoid a partial combustion of the decomposed ammonia even as an increased temperature of the reactor results in an increased ammonia conversion. Furthermore, it is beneficial to operate the compression stages without intercooling to reduce the required fuel flow to the gas turbine.

The scheduling of multi-line polyolefin production is a complex decision-making process characterized by sequence-dependent changeovers, strict physicochemical constraints, and dynamic market environments. Traditional optimization methods often suffer from high computational costs and a lack of flexibility in online adjustments. To address these challenges, this paper proposes a Deep Reinforcement Learning (DRL) framework for dynamic scheduling tasks. We first construct a high-fidelity simulation environment that meticulously models realistic industrial constraints, including transition materials, shutdowns, and inventory limits. A Soft Actor-Critic (SAC) agent with a tuple-based action space is employed to mitigate the combinatorial explosion associated with multi-line decisions. Furthermore, a dynamic action masking mechanism embedded with domain knowledge is introduced to strictly enforce hard constraints and significantly improve sample efficiency. Case studies based on real-world industrial data demonstrate that the proposed method can autonomously generate valid schedules that satisfy complex production requirements. Comparative experiments further reveal that the action masking mechanism accelerates training convergence, and the DRL agent exhibits superior adaptability to dynamic price fluctuations.

This work explores symbolic regression (SR) as an interpretable surrogate modelling approach for the multi-objective optimisation of pressure swing adsorption (PSA) systems for CO2 capture. A first-principle model was used as a virtual plant to generate synthetic datasets covering the operating space defined by cycle step durations. Two surrogate frameworks were developed and compared: SR models derived through evolutionary search and deep neural networks (DNNs) trained via Hyperband-based tuning. Both surrogates were used as simulation models within an optimisation procedure based on a particle swarm optimisation (PSO) algorithm to maximise CO2 purity and recovery. While DNNs achieved the lowest prediction errors (MSE ˜ 10-6), the SR surrogates provided compact analytical representations and significantly faster optimisation. The SR framework yielded a denser and more diverse Pareto front (4345 vs 508 points). It was about 34 times faster (38.6 s vs 1331 s), confirming its efficiency for surrogate-based optimisation of cyclic adsorption processes.

The growing penetration of renewable energy sources and power-to-hydrogen (P2H) systems demands high-fidelity, large-scale optimization frameworks that capture the nonlinear physics of both AC power flow and hydrogen thermodynamics. However, existing approaches rely on DC approximations and simplified electrolyzer models, neglecting critical operational constraints. As a result, accurately modeling such systems leads to large-scale nonlinear programs that are computationally intractable for conventional CPU-based solvers. This motivates the need for scalable optimization frameworks capable of handling both physical fidelity and computational complexity. This paper proposes a fully GPU-native framework for solving large-scale multi-period AC optimal power flow (AC-OPF) problems with integrated power-to-hydrogen systems. High-fidelity thermodynamic models of hydrogen production, compression, cooling, and storage are coupled with AC power flow constraints, resulting in large-scale nonlinear programs (NLPs) with up to 14.4 million variables and 22.4 million constraints in the largest benchmark case (9, 591-bus network with a 168-hour horizon). To enable scalable solutions, condensed-space KKT reformulations and GPU-accelerated sparse Cholesky factorization are employed within an end-to-end GPU optimization pipeline. Numerical results on benchmark networks up to 9, 591 buses demonstrate 10–600× speedups over CPU solvers, whereas CPU-based solvers fail to converge on the largest instances. Operational studies further highlight the importance of thermodynamic constraints in realistic hydrogen system scheduling.

The optimization of chemical reactions requires navigating a high-dimensional design space composed of both discrete and continuous variables. Although one-hot encoding has been widely adopted, it lacks chemically meaningful information and suffers from sparsity and poor generalization. To address these limitations, we explored the use of pretrained molecular foundation models to generate latent representations as input variables for optimization. However, rigorously comparing different combinations of reaction representations and optimization algorithms remains a time- and resource-intensive challenge. In this work, we developed an end-to-end benchmarking platform that systematically evaluates diverse encoding schemes and optimization strategies under identical conditions. The platform automates the entire workflow from data preprocessing to result analysis, supporting fair comparison across multiple representation–optimizer combinations. Furthermore, we designed a custom reaction representation that integrates a 3D equivariant encoder with a bidirectional cross-attention module to explicitly capture interactions between reaction components. The proposed platform provides a scalable foundation for reaction optimization and advances the feasibility of autonomous experimental systems.

The electrification of regional aviation requires coordinated planning of airport energy systems that integrate renewable generation, energy storage, and hydrogen technologies in a cost-efficient and resilient manner. This paper presents a scalable techno-economic optimization framework that models multiple airports as collaborative energy hubs. An object-oriented mixed-integer linear programming (MILP) formulation is combined with a genetic algorithm (GA) to optimize infrastructure sizing and energy dispatch. The framework is applied to three Swedish regional airports—Västerås, Jönköping, and Visby. A set of scenarios, including parties operating under shared wind-energy contracts using power purchase agreements (PPAs) and dynamic pricing (DP), was studied. Detailed representations of battery energy storage, hydrogen production and storage, and market interactions are included. Results show that coordinated operation and airport collaboration under a smart energy management system can reduce total annual cost by up to 5% while enhancing operational flexibility and resilience to demand peaks and grid disturbances.

This study presents different approaches for introducing mixed integer problems into a meta-heuristic algorithm. The algorithms are developed to address the simultaneous design and optimization of a heat pump unit. A distinction is made between integer variables such as nominal tube diameters and the adsorbent employed in the process. The choice of adsorbent is named as a “key variable” due to its high impact on the process. To optimize the selection of these “key variables”, a branched version of Particle Swarm Optimization (PSO) is presented and compared with the non-Branched version and a deterministic solver (IPOPT). Advanced Convergence Criterion is also implemented to mitigate the computational effort of these approaches. In the studied cases, Branch_PSO presents a higher degree of consistency and can even outperform the traditional PSO in simultaneous process optimization and material screening. However, its computational effort in cases with a large number of branches might be an obstacle.

This study presents a techno-economic modeling framework integrating a modular blue H2N2 production unit with a hydrogen refueling station (HRS) across capacities ranging from 0.1 to 4.0 tpd. A model-based approach is used to size key process and refueling components and to estimate the resulting hydrogen retail cost. The analysis indicates that hydrogen retail costs range from 4.6 to 10.8 USD kgH2-1 over the considered capacity range. Relative to alternative on-site hydrogen production pathways, the proposed system demonstrates better cost-effectiveness while meeting clean hydrogen production standards. The approach is particularly suitable for regions with established natural gas infrastructure, as it leverages existing supply chains. Overall, the results provide actionable insights for policymakers and industry stakeholders in planning future hydrogen refueling infrastructure.

This work presents an integrated geospatial-technoeconomic optimization framework for siting modular blue and green hydrogen production units co-located with hydrogen refueling stations (HRS) along U.S highways, with a case study focused on New England. The workflow identifies geospatial highway networks and natural gas infrastructure intersections, estimates hydrogen demand based on heavy-duty truck flows from U.S. Freight Analysis Framework, and formulates a mixed-integer linear program (MILP) that selects technology candidates and their capacities to minimize total cost, subject to corridor coverage and supply-demand constraints. Two onsite hydrogen production scenarios are evaluated: a green hydrogen-only production case and a mixed configuration combining modular green and blue hydrogen. Results indicate that, under a 5% hydrogen adoption scenario in truck traffic, 29 HRS with onsite hydrogen production are needed in the New England region. These findings highlight the benefits of integrating local hydrogen production with HRS planning to reduce the reliance on centralized hydrogen delivery and storage infrastructure.

Membranes provide an efficient method for treating flue gases to capture CO2 from various point sources, achieving high recovery and purity rates. However, the lack of systematic process-level design tools has limited the translation of advanced membrane materials into large-scale technical and economic metrics. Thus, in this study, we present a superstructure model for the design of membrane-based carbon capture, both from highly energy-intensive industries and from power plants. The superstructure model enables the flexible design and global optimization of multi-stage membrane systems. Multiple membranes are compared under technical performance indicators (specific energy and specific area), while the already commercialized polymeric membranes Polaris and PolyActive are taken into consideration for estimating their economic performance. The presented framework establishes a robust link between material innovation and optimal process design, providing a key tool for the large-scale deployment of membrane-based carbon capture.

Water supply in the countryside of São Paulo state, Brazil, is based on groundwater resources that can be contaminated with substances such as heavy metals or fluoride, requiring the usage of water treatment technologies such as Reverse Osmosis (RO); however, RO systems create a stream of high-salinity brine, with negative environmental consequences. Besides, regulatory constraints demand that well operations must be interrupted for a daily contiguous period. In this work, a Mixed-Integer Linear program (MILP) was implemented to define water network topologies and well exploitation schedules, under these downtime constraints, aiming the minimization of RO plant capacity (and, therefore, of brine discharges). This model was then applied to the water supply of a small city in the São Paulo countryside, with around 8000 inhabitants, where high fluoride concentrations warranted the implementation of an RO system. Demand variations between weekdays and weekends (with demands 52.7% higher) were modeled using a multi-scenario formulation. This model was optimized under different conditions of tank capacity and membrane technologies. Different membrane systems (with salt retention rates varying from 85% to 99.5%) associated with different storage tanks (with volumes between 50 and 1000 m³) led to different system topologies, and a trade-off was observed between plant capacity, tank capacity, and operating pressure: while a low pressure system with a small tank capacity required larger plant capacities (up to 44.02 m³/h), high pressure systems associated with large storage tanks required smaller plants for treatment (with a capacity as small as 11.15 m³/h).

The design of Power-to-Ammonia systems (P2A) is a challenging task. While the technology for all its components, including renewable energy harnessing, electrolysis, Haber-Bosch synthesis, and auxiliary buffers, is mature, assembling such a system to meet the challenges of varying power profiles is not trivial. To ensure resilient, cost-effective designs, careful selection of unit capacities and coordination of all system operations are required. Specifically, this requires modeling system behaviour and enforcing operational constraints to capture system flexibility over a representative time frame. The first steps towards a novel P2A design framework are presented, with a focus on enhanced process operations and exploring new options for process flexibility. A general methodology is proposed, where the full system can be customized, namely by enabling or disabling: (i) multiple renewable energy harnessing sources, (ii) grid operations, and (iii) buffers, including battery and/or an H2 tank. The model is defined as a linear function of a set of design and operation variables, expressed through a set of matrices and vectors, which are then used to formulate and solve a Linear Programming problem. The design methodology is presented and illustrated via 3 distinct P2A configurations, with key results and analyses of system flexibility and performance. A discussion of future steps towards developing this design framework is also presented.

Self-driving laboratories (SDLs) increasingly use Bayesian optimisation (BO) to navigate expensive design spaces, yet high-fidelity simulations and experiments remain too costly to query at scale. Multi-fidelity Bayesian optimisation (MFBO) alleviates this by combining abundant low-fidelity evaluations with scarce high-fidelity observations. However, Gaussian process (GP) surrogates can become computational bottlenecks as data volume and dimensionality increase, motivating scalable alternatives. Here, we assess transfer learning based deep neural network (DNN) surrogates that pretrain on low-fidelity data and fine-tune on high-fidelity observations. We construct a chemical process benchmark for glacial acetic acid separation and purification with paired low- and high-fidelity flowsheets. The optimisation considers eight decision variables and minimises the minimum selling price (MSP), while enforcing a product purity threshold via a quadratic penalty. To reflect realistic resource constraints, we define seven training scenarios with approximately constant total simulation time, given the large cost gap between fidelities. Across all scenarios, the transfer learning DNN outperforms a GP regression baseline trained only on high-fidelity data. Test-set R² increases from 0.41 to 0.71 for the DNN, compared with 0.32 to 0.51 for GP, with gains already visible in the most data-scarce setting. These results suggest that transfer learning–enhanced DNN surrogates can offer accurate, scalable multi-fidelity surrogate models for process optimisation, and could be a practical component in BO-based automated discovery settings where low- and high-fidelity data coexist.

This work proposes a two-stage optimization framework for the optimal utilization of biogas from pig manure, integrating process-level design with short-term operational planning under dynamic electricity tariff schemes in Mexico. In the first stage, a multi-objective optimization based on 3E (Exergy, Environment, and Energy) analysis was performed. The results demonstrate that increasing the biogas split fraction for upgrading significantly reduces the environmental and exergy indices, enhancing thermodynamic and environmental performance without compromising the energy index. High upgrading flows (split > 0.7) emerged as the most favorable compromise across the evaluated metrics. In the second stage, support vector regression (SVR) surrogate models were developed to approximate nonlinear relationships between the operational split and process outputs. These surrogates were embedded in a Mixed-Integer Linear Programming (MILP) formulation to optimize weekly scheduling under the Mexican Net Billing scheme, incorporating an economic objective (+1E index). The SVR-MILP framework successfully captured market-driven decisions, prioritizing biomethane upgrading during base and intermediate tariff periods and switching to electricity generation during peak hours to maximize economic profit. Biomethane upgrading accounts for approximately 90-100% of the total positive economic benefits, while electricity generation contributes the remaining 10%, depending on the exposure to peak electricity tariffs. Thus, the proposed methodology offers a robust, computationally efficient decision-support tool for flexible biogas systems, bridging the gap between steady-state design and dynamic market responsiveness based on a 4E approach.

Pyrolysis, a key waste-to-X technology, enables converting a wide portfolio of biomass waste into valuable chemicals and fuels. However, raw pyrolysis oils are chemically and physically unstable. A multi-step stabilisation is necessary to reduce acidity and the content of reactive components, mainly carbonyls and carboxylic acids. During stabilisation, which involves deoxygenation and hydrogenation as the main steps, an aqueous phase is generated as a by-product. This stream contains mainly water, but relevant amounts of methanol and ethanol (2-8 wt%) are also present, together with minor concentrations of higher alcohols, C1-C4 carboxylic acids, and light esters. The aim of this work is to design and optimise a process to isolate the methanol and ethanol embodied in the aqueous phase and exploit them as intermediates to generate biofuels, biochemicals, and pharmaceutical products. The process consists of a train of four distillation columns to maximise the recovery rate and purity of the two targeted products. COCO-COFE, a CAPE-OPEN process simulator, is used for flowsheet development. The optimal number of theoretical stages and the optimal positioning of the feed to each unit have been optimised, considering a trade-off CAPEX-OPEX as the criterion. Results show that up to 95% and 98% yield can be achieved for methanol and ethanol, respectively. Methanol purity is limited to a maximum of 92 wt% due to the presence of volatile impurities, while ethanol contains residual propanol (6 wt%). The overall heat demand is in the order of 1 MJ/kg feed.

Critical metals are essential for clean energy technologies but, due to being mainly sourced through mining, the critical metal supply chain is susceptible to geopolitical risks. Electronic waste (e-waste), however, can serve as an alternative “urban mine”, but the recovery at high purities requires complex and resource-intensive processing. This work explores the modeling and optimization-based design for the intensification of liquid-liquid extraction in small channels as a means to recover critical metals from e-waste. Small channels can achieve high mass transfer rates while mitigating the environmental impact. A superstructure-based approach is employed to represent the alternative system configurations, while a plant propagation algorithm is used to optimize the multi-objective problem to recover Neodymium (Nd) and Samarium (Sm). The multi-objective problem aimed to tackle product quality, process economics, and environmental impact. The results demonstrated that optimally designed extraction can support efficient recovery of critical metals; however, decision making support may be essential for the selection of one process design from those identified as optimal in a multi-objective design approach.

The Heterogeneous Capacitated Vehicle Routing Problem (HCVRP) is a fundamental extension of the classical Vehicle Routing Problem in which customer demands must be satisfied using a fleet of vehicles with varying capacities and costs. In this paper, a novel and intuitive decomposition-based formulation for HCVRP is presented that decomposes the problem into two tractable subproblems: (i) a route generation and an optimal customer sequencing problem and (ii) a vehicle route assignment problem. In the first stage, all feasible customer combinations are constructed as routes, and for each route an optimization problem is solved to identify the optimal customer sequence that results in the minimum distance travelled. In the second stage, the optimal routes are selected, and vehicles are assigned using a mixed integer linear programming (MILP) formulation that minimizes the fixed cost of vehicle utilisation and total transportation costs, ensuring demand satisfaction for all customers while meeting vehicle capacity limitations. Computational results demonstrate that the decomposition strategy preserves optimality while significantly reducing solution time.

Production scheduling in the process industry is often treated as a static optimization problem, although real plants require frequent rescheduling due to disturbances such as rush orders, equipment breakdowns, and changes in processing times. This paper compares a simulation-optimization approach that couples a discrete-event simulator with an evolutionary algorithm (EA) with a sequence-based mixed-integer linear programming (MILP) formulation for real-time scheduling of multistage batch systems. Both methods are embedded in an event-driven rolling-horizon framework under strict computation time limits.In static experiments for a 3-stage, 2-machine flow-shop setting (10 products, 20 orders, random processing times), the EA achieved lower makespans across all tested time budgets, improving results by about 7–13% on average compared to the MILP approach. In real-time experiments (40 initial orders, maintenance, three rush orders, 10 s and 60 s periodic updates), the solution quality of the MILP approach was lower after disturbances under restricted computation times. Each experiment was repeated 25 times with identical randomly generated processing times across methods per run; in 4% of all runs no solution was obtained within the available response time. Including explicit allocation decisions for the EA increases flexibility but can reduce short-term responsiveness within the rolling-horizon setting.Overall, the results indicate that MILP can be competitive in stable scenarios when sufficient solver time is available, whereas simulation-optimization is better suited to reactive scheduling in real-time settings where rapid schedule adaptation is critical.

Brazil is the world’s largest producer of orange juice, generating significant peel residues that are currently underutilized. This study proposes a mixed-integer linear programming (MILP) framework for optimal supply chain design, utilizing Special Ordered Sets of type 2 (SOS2) to accurately represent non-linear investment costs. The model maximizes Net Present Value (NPV) by integrating production costs with multi-echelon logistics, including inland transport, port handling, and international maritime freight. Applied to a case study in São Paulo, the framework evaluates pathways for the co-production of D-limonene, pectin, and bioenergy. Results indicate a positive NPV of BRL 1.27 billion, with pectin contributing over 65% of total revenue. The optimization favors centralized configurations in Araraquara or Matão to exploit economies of scale while minimizing the transport of high-volume, wet biomass. Notably, total transportation costs represent only 1.13% of expenditures, as the high value-density of the bioproducts effectively absorbs global distribution costs. This work demonstrates that in regions with high biomass abundance, strategic success is primarily governed by process efficiency and industrial scale rather than logistical constraints, providing a robust tool for decision-making in the citrus bioeconomy.

The energy recovery of agro-industrial residual biomass offers a pathway to reduce fossil fuel emissions in thermal processes while valorizing waste. In practice, however, the primary bottleneck is logistical: feedstocks are geographically dispersed, with low bulk density and high moisture content, driving up collection, pretreatment, and transport costs. This work combines geospatial processing with mathematical optimization to design a multi-stage logistics network. The model incorporates intermediate densification options and technology selection (chipping, pelletizing, or briquetting) to supply one or more final waste-to-energy plants. The case study focuses on Northeastern Uruguay, considering forestry residues, meat-processing waste, and rice husks. We formulate a multi-period Mixed-Integer Linear Programming (MILP) model aimed at minimizing the total annualized cost, encompassing transportation, logistical operations, capital investment, and plant O&M, subject to supply constraints, capacities, and maximum transport distances. Results show that the optimal configuration corresponds to the briquetting scenario, which includes a fixed densification facility and processes 200, 160 t/y of biomass from 18 sawmills. Under this configuration, the system generates 276, 954 MWh, with a net margin of USD 5.77/MWh and a net profit of USD 1, 596, 662 over the planning horizon. Sensitivity analysis indicates that the electricity selling price is a critical parameter, with a tolerance margin of only 8%, while mobile densification is not economically viable under current market conditions. Beyond the specific case study, the main contribution of this work is the optimization model itself, whose general and configurable formulation allows its application to other regions, residue types, and market conditions without reformulating its mathematical structure.

This study investigates control strategies for the flexible operation of heat pump-assisted distillation processes, focusing on the heat integrated distillation column configuration. The methanol/water separation system was selected as a case study and modelled to achieve 99.9 wt% AA-grade methanol purity. A limiting piece of equipment for flexible operation of heat pump assisted distillation is the compressor. To assess its impact on flexible operation, dynamic simulations in Aspen Dynamics were conducted for two heat integrated distillation column control strategies: one using fixed compressor duty and one using variable compressor duty. The control performance for a 20% throughput disturbance, as well as for a 50% turndown ratio scenario was investigated. Results show that fixed-duty operation maintains robust stability and rapid disturbance recovery even at 50% turndown, while variable-duty operation delivers higher efficiency for moderate load changes but cannot sustain low-load stability. This work supports the electrification of distillation by enhancing the operational flexibility of heat pump-assisted distillation, enabling better integration with intermittent renewable electricity grids.

For the long-cycle process industry, operational cycles can be severely affected by equipment aging, catalyst deactivation, and safety limitations. As illustrated by the residue hydrotreating process, metal impurities gradually deposit on the catalyst during residue purification, leading to catalyst poisoning and eventual process shutdown. Such long-cycle processes require dynamic adjustments of operating conditions to balance immediate economics with long-term sustainability. While current practice relies on empirical tuning based on historical data, this work focuses on studying how to obtain an optimal operating trajectory to guide the monthly adjustments of operating variables. The long-cycle simulation of the residue hydrotreating process can be performed using the commercial software, PetroSIM. After adjusting the feed conditions, its embedded mechanistic model can calculate the deviation of average bed temperature from the set point and output the remaining operating time. Since Bayesian optimization (BO) is well-suited to address complex processes with unknown mechanisms and high computational costs, and can effectively seek optimal solutions, a BO framework incorporating constraints evaluated by PetroSIM is proposed in this work to optimize the monthly feed composition and maximize the total profit over the entire operational cycle. The results indicate that the total profit achieved through BO-optimized operation is 17.8% higher than that from empirical operation. The optimized strategy demonstrates practical rationality: in the early stage of high catalyst activity, more residue can be processed; in the later stage, increasing the light oil ratio helps extend the processing cycle. This strategy provides theoretical support for advancing the research toward industrial applications.

Centralized monitoring methods experience reduced fault detection sensitivity in large-scale industrial systems due to the masking effect arising from the aggregation of many interconnected variables. Decentralized monitoring, where variables are grouped into subsystems, has been shown to effectively address these limitations. However, the performance of this class of methods critically depends on how the network is partitioned, and the role of its structural factors on fault detection remains poorly understood. This work studies how network topology and causal structure affect decentralized monitoring in high-dimensional systems. Using SimCaNet, a DAG-based data simulator, where large-scale systems with 100-1000 variables were generated, we rigorously compared the performance of centralized and decentralized causal log-likelihood monitoring methods under process perturbations and sensor bias faults. Network partitioning is performed using the Leiden community detection algorithm and characterized at the network, community, and node levels. Fault Detection Sensitivity (AUCNORM) is quantified per variable, using the normalized area under the curve, which is computed from the true positive rate profiles across fault magnitudes. Results show that AUCNORM is primarily driven by the number and the size of communities across fault types. Smaller communities consistently achieve higher AUCNORM than larger communities, while excessive fragmentation or high modularity degrade performance. Community density and inter-community coupling further enhance fault detection sensitivity. The strength of node-level causal relationships explains most of the variability observed at root nodes for sensor bias faults and their robustness to information loss. Furthermore, ablation studies confirm the robustness of distributed systems to major disruptions in the system: more than 32% of community information can be removed without significantly impacting the fault detection sensitivity. These topology-driven partitioning findings constitute novel and valuable contributions to the theoretical foundations for designing better distributed monitoring systems in complex industrial plants.

This work demonstrates the application of a pulsed microwave system for single-step co-production of ethylene and ammonia from methane. To mitigate inherent production fluctuations from pulsed microwave reactors, a staggered manifold configuration was utilized to stabilize effluent flow for industrial-scale compatibility. Dynamic validation of the ammonia and ethylene purification columns confirmed that a rigorously tuned control strategy effectively rejects ±10% feed disturbances while maintaining process stability and product purity. Ultimately, this systematic approach establishes a robust foundation for the sustainable, electrified production of foundational chemicals by bridging the gap between laboratory-scale pulsing phenomena and industrial-scale operational reliability.

A safety-intelligent framework is presented for developing a multiple-input multiple-output (MIMO) risk-based explicit model predictive control (R-eMPC) for metal hydride storage systems (MHSS). These systems are susceptible to thermal runaway during the charging process as a result of the exothermic adsorption reaction within the metal alloy. To address this issue, deterministic and stochastic safety-intelligent control algorithms are designed and implemented by explicitly embedding a dynamic risk index (RI) or risk tolerance (() into the control law and decision-making. In closed-loop analysis, the deterministic R-eMPC regulates both core temperature and hydrogen storage capacity by forecasting fault occurrence, triggering alarms, and reducing the risk index by adjusting the optimal control actions, supply pressure and water flowrate. Meanwhile, the stochastic R-eMPC accounts for uncertainties in core temperature variation by incorporating risk tolerance through chance-constraints. When the temperature exceeds the strict operational limit, stochastic R-eMPC enables continued operation beyond conservative safety boundaries via dynamic updated probabilistic limits. Closed-loop responses demonstrate that the safety-intelligent controllers effectively minimize evolving thermal risks while maximizing storage capacity during charging operation of the MHSS. Overall, the risk-based controllers improve both safety and charging performance, demonstrating the potential of the proposed framework for real-time, safety-intelligent operation of metal hydride systems.

The Dividing Wall Column (DWC) offers significant potential in saving both energy- and capital cost compared to conventional distillation sequences. However, there are some issues regarding flexibility and control that require attention in reducing the risks or uncertainties in achieving the potential benefits in practical operation. This calls for control and optimization methods that rely on the available measurement data and less on simulation models. The “Perturb and Observe” method is a simple algorithm that seems suitable for this on-line optimisation task. A series of experiments have been carried out at the Kaibel-column pilot at NTNU and some key results are presented. The method is combined with a conventional control structure at the regulatory layer.

The extensive use of renewable energy to supply hydrogen production for chemical processes is hindered by the uncertainty in power generation and by strict operational limits.These challenges are addressed through a real-time dynamic optimization approach based on a receding-horizon strategy that provides optimal decision variables. The framework explicitly relies on imperfect weather forecasts and dynamically adapts the hydrogen reference production to guarantee the final productivity target. The optimization methodology focuses on minimizing grid electricity imports, limiting excessive equipment stress, and preventing constraint violations, while ensuring that the hydrogen production target is satisfied.The proposed approach yields competitive economic and environmental performance, with a levelized cost of hydrogen of 3.31 USD/kgH2, well within literature values (1.50–7.50 USD/kgH2) and below typical industrial costs (4–12 USD/kgH2). At the same time, carbon dioxide emissions are reduced by more than 90% compared to steam methane reforming, resulting in specific emissions below 1 kgCO2/kgH2. These results are supported by a year-long simulation based on CAISO renewable energy data. The optimized system achieves the annual hydrogen production target while respecting all operational constraints, whereas the absence of operational strategy and buffering systems leads to systematic constraint violations. The strategy can be implemented in existing industrial facilities with limited modifications.

Early and reliable prediction of convergence to cyclic steady state (CSS) is increasingly important in periodic downstream bioprocessing, where switching and cut decisions are tuned for a repeatable cyclic regime. This work addresses time-to-CSS (TCSS) forecasting and CSS-existence classification for multicolumn countercurrent solvent gradient purification (MCSGP) systems under run-to-run feed variability. We propose a Multi-Feature k-Nearest Neighbours (MF-kNN) framework that performs long-horizon one-shot trajectory forecasting from an early run segment. CSS outcomes are inferred by reapplying a peak-based convergence rule to the predicted trajectory, while CSS existence is predicted via neighbour-label voting. The approach uses multivariate, standardised features, run-level splits, and a windowed neighbour search to reduce computation. Hyperparameters are tuned with a CSS-oriented objective function that balances trajectory fidelity, TCSS error, and misclassification penalties. On an in-silico MCSGP dataset (98 runs; delta-t = 0.2 s; 6800 steps/run) with varying initial modifier concentration, MF-kNN produces accurate full-run forecasts from early data and enables operationally useful early go/no-go decisions. Across outlets, results support accurate CSS timing inference and high CSS-existence classification accuracy (up to 100% on selected outlets), indicating MF-kNN as a transparent and deployment-ready complement to cycle-to-cycle CSS monitoring and control.

Reinforcement learning (RL) presents a promising alternative to model-based advanced control schemes, such as model predictive control (MPC), whose application can be limited by highly complex system models. However, incorporating constraints in RL remains challenging and formulating a suitable optimization objective is not straightforward. Reinforcement learning from human feedback (RLHF) offers an approach to derive the RL reward function from human expert preferences, enabling the incorporation of process knowledge. In this work, we present the application of RLHF to fine-tune an approximate MPC controller with suboptimal performance. We demonstrate that combining conventional reward formulations with RLHF, along with varying trajectory segment lengths for collecting human feedback, improves the control methodology for a batch bioreactor by enhancing safety and accounting for long-term effects. Furthermore, direct-preference based policy optimization (DPPO) represents a promising alternative for directly fine-tuning learning-based controllers while circumventing explicit reward model design.

In modern industrial systems, increasing performance requirements and sustainability constraints have intensified the need for advanced optimization methodologies capable of efficiently handling complex process models. The Simulated Moving Bed (SMB) process is a well-established technology for continuous chromatographic separations, offering high productivity and reduced solvent consumption compared to batch operations. However, its optimization is challenging due to the underlying distributed-parameter nature of the process.This work presents the development of a dynamic simulation and parameter optimization framework for the SMB process, implemented in Python using the open-source CasADi framework. The SMB model accounts for axial dispersion and mass transfer using a linear driving force formulation and is discretized in space using the method of lines, resulting in a state-space representation compatible with CasADi’s numerical tools. Model accuracy was validated by reproducing a benchmark case from literature and comparing concentration profiles and product purities against results obtained using MATLAB’s PDEPE solver.The proposed framework was further applied to a parameter optimization problem involving the separation of fructose and glucose. The objective was to maximize the feed velocity subject to purity and recovery constraints for fructose, with the superficial velocities in each SMB section and the switching time treated as decision variables. The resulting nonlinear programming problem was solved using a direct single-shooting approach and the interior-point optimizer IPOPT. The optimized operating conditions and performance metrics closely match published results, while achieving the solution in less than 90% of the computational time reported for a gPROMS-based implementation.

The convergence of artificial intelligence (AI) and chemical process systems engineering is creating unprecedented opportunities to transform current refineries from conventionally operated plants into intelligent, automated, and resilient systems. However, the practical deployment of AI in these complex industrial environments faces several critical challenges. First, most existing process datasets contain minimal labeled data, making it difficult to apply supervised learning techniques that require extensive annotations to generate meaningful insights. Furthermore, refinery data typically consists of high-dimensional, multivariate time series, which pose additional complexities in capturing temporal dynamics and system interactions. Traditional Fault Detection and Diagnosis (FDD) frameworks often struggle to address these complexities, lacking adaptability to evolving process conditions, scalability to large plant networks, and explainability in their diagnostic reasoning. To address these gaps, we propose an integrated framework for unsupervised knowledge discovery, adaptive process monitoring, and AI-assisted fault diagnosis. The approach combines Dynamic Time Warping (DTW) with graph-based layout algorithms to extract and visualize temporal structures and similarities within complex process datasets. The resulting graphs function as an unsupervised model, enabling the identification of operational regimes through process clustering. These models can be deployed online to detect anomalies using adaptive fault thresholding, ensuring robustness to changing process dynamics. When faults are detected, the framework leverages large language models (LLMs) as AI assistants to support fault diagnosis. By bridging domain knowledge—such as piping and instrumentation diagrams (P&IDs) and signal flow diagrams—with data-driven insights, including machine learning metrics and feature contribution scores, the system delivers explainable and accurate diagnostic outputs. This hybrid methodology enhances interpretability and decision-making, offering a scalable and adaptive solution for next-generation AI-enabled refinery operations.

We develop a multi-fidelity soft-sensing framework to reconcile online (low-fidelity) industrial measurements with sparse (high-fidelity) laboratory samples from an alkylation unit in a refinery. A first-principles model is used to generate an additional low-fidelity dataset and train a surrogate that predicts the output variable. We investigate whether incorporating sparse high-fidelity laboratory data with the low-fidelity data improves prediction accuracy. A multi-fidelity predictor forms a corrected output by learning the residual between the high-fidelity observations and the low-fidelity surrogate using Gaussian process regression. The simplest model structure performs best, reducing test prediction error (computed against laboratory samples) by 31.1% relative to the currently deployed industrial analyzer and outperforming a standard high-fidelity-only model trained on laboratory data. Overall, the simplified surrogate model captures the main industrial trends well enough to serve as a reliable low-fidelity input for the multi-fidelity soft sensor.

Energy management plays a crucial role in achieving efficient and sustainable operation of industrial energy systems. With the increasing integration of renewable electricity and the growing complexity of hydrogen production networks, effective control strategies are required to minimize operational costs and carbon footprint. However, the uncertain nature of renewable energy sources, such as photovoltaic (PV) power, complicates their accurate forecasting and challenges the optimal energy management of system components. To deal with uncertainties, the rolling horizon approach (RHA) provides a practical framework for adaptive decision-making by repeatedly solving optimization problems over moving time windows while updating system data in real time. In RHA-based energy management, linear or linearized system models are often employed and optimized by linear methods to reduce computational complexity; however, these simplifications can compromise physical realism and lead to suboptimal decisions. Although RHA can also incorporate local, or global deterministic and stochastic algorithms for nonlinear problems, such approaches frequently suffer from high computational effort, slow convergence, local optima, and difficulty in ensuring constraint satisfaction in large-scale nonlinear systems. To overcome these limitations, this work employs the novel hybrid optimization method “BO-IPOPT”—a combination of Bayesian Optimization (BO) for global exploration and the Interior Point OPTimizer (IPOPT) for rapid local refinement. Applied to an industrial hydrogen production system, BO-IPOPT outperforms state-of-the-art approaches in accuracy and robustness by achieving lower operational costs at the same CPU time while satisfying all constraints. Finally, the influence of the uncertainties in PV generation on the performance of the energy management system is analyzed.

Maintaining mother liquor supersaturation at a setpoint within the metastable range is vital for achieving the best production yield in industrial sugar production. However, precise online measurement and control is challenging. In this work, we develop a model-based soft sensor for supersaturation monitoring, and we propose a new feedforward-feedback control structure for batch sugar crystallization. Supersaturation is estimated using standard process measurements, enabling a soft sensor that can be readily adapted to different production units. The soft sensor continuously estimates supersaturation from standard process signals, and the control strategy ensures it remains within the desired operating range, enabling simple and straightforward application to other sugar production units.

We study a data reconciliation problem in a petrol refinery. The problem is to reconcile inventory and flow measurements to estimate true values of measured and unmeasured flows respecting the mass conservation. The problem is formulated as a mixed-integer quadratic program (MIQP). Upon successful problem resolution, a neural network (NN) is trained to mimic the MIQP solver to study potential improvements in CPU time without compromising the solution quality. The results show a significant improvement in refinery monitoring and feasibility of NN-based reconciliation.

This research presents a novel framework for the surrogate-based dynamic optimization of control schemes within chemical separation and purification processes such as the biorefinery downstream processing. The current study investigated the downstream of an enzymatic bioreactor responsible for the synthesis of 5-hydroxymethylfurfural value-added derivatives, focusing on the critical balance between operational costs and productivity. Two high-fidelity long short-term memory neural network-based surrogate models were developed to predict energy consumption and economic gain, both achieving a coefficient of determination (R2) exceeding 0.97. These models were subsequently integrated into a multi-objective optimization architecture to address an operating efficiency testing scenario characterized by stepwise inflow parameter changes. By exploring the resulting Pareto front, an optimal set of operational (control) settings was identified and validated. The results demonstrate that while energy consumption remained nearly constant, the total economic benefit increased by approximately 20% over the whole studied timeframe. Critically, during the transient period between steady-states, the economic benefit surged by nearly 60%, highlighting the potential of surrogate-based dynamic optimization. The high computational efficiency of the developed models (optimal solution obtained in less than three minutes) suggests significant real-time process control application possibilities. Unlike computationally expensive first-principles models, the developed neural architectures allow for virtually instantaneous setpoint adjustments. This paves the way for their integration into model predictive control or real-time process control layers, enabling industrial plants to respond adaptively to feed fluctuations, minimize off-spec product during transitions, and maximize profitability.

In this work, we propose a methodology for monitoring the performance of model predictive controllers (MPCs). A sequence of binary classification machine learning models, organized in cascade, called Cascade Machine Learning Models (CMLM), is evaluated to give a diagnosis of the control conditions. The proposed methodology was assessed using two case studies: a benchmark problem (the van de Vusse reactor under nonlinear MPC, NMPC) and a simulated industrial debutanizer column under commercial MPC. The ML models evaluated were the Random Forest and the Multilayer Perceptron. The results show that the proposed approach outperforms both a single multiclass model and traditional MPC performance monitoring methodologies, while remaining adaptable and scalable to larger applications.

Predictive maintenance is a promising approach to increase safety and productivity in chemical plants. One notoriously difficult problem in predictive maintenance are hard to predetermine, non-deterministic changes such as drifts. The term “drift” can be found with different definitions in this context. Therefore, it is defined here as changes in variables and parameters that occur orders of magnitude slower than the nominal process dynamics and are not directly measurable. Previous research resulted in a hybrid method that detects and diagnoses drifts from two sources: process and equipment. This method combines model-based and statistical approaches and additional information from the equipment, such as measurement gain or power consumption, is envisioned to reduce uncertainty about the drift cause [1]. First case studies revealed significant problems regarding economically viable integration of additional information. These problems arise due to the amount of information in scenarios with multiple devices, making analysis costly and time-consuming. For solving these problems, an automated evaluation of the information is introduced. It analyzes which additional equipment information is most relevant to distinguish between different drifts, based on sensitivity analyses. Additionally, a drift index – a unique mathematical label for drifts – is defined. The extended method is then successfully applied to a scenario closer to the reality in chemical plants with multiple adjacent devices. It is shown that the additional information is evaluated automatically and drifts are diagnosed more efficiently than in the previous method.

Offshore natural gas compression systems are characterized by strong hydraulic coupling, nonlinear behaviour, and strict safety constraints, particularly in high-CO2 production environments. Conventional decentralized PID control with anti-surge protection ensures reliable local regulation but often leads to poor plant-wide coordination and persistent offsets when multiple compression trains, recycle loops, and separation units interact dynamically. Although multivariable control strategies such as model predictive control can address these issues, their industrial application remains limited by modeling effort, computational demand, and robustness concerns. This work presents a hybrid supervisory control framework in which reinforcement learning (RL) augments an existing PI-based architecture for an offshore gas compression system with membrane-based CO2 separation. A Proximal Policy Optimization (PPO) agent is trained on a dynamic digital-twin model of export, CO2, injection, and bypass compression trains with shared headers, recycle flows, and thermal constraints. The RL agent acts exclusively at the supervisory level, providing bounded incremental adjustments to pressure and bypass setpoints, while all regulatory PI loops, anti-surge protections, and safety logic remain unchanged. Closed-loop simulations under representative disturbance scenarios show that the RL-supervised architecture significantly improves plant-wide coordination. Compared to fixed-setpoint PI operation, total tracking error (IAE) and time-weighted error (ITAE) are reduced by approximately 54% and 56%, respectively, using smooth, low-magnitude supervisory actions. In addition, RL achieves performance close to that of a nonlinear MPC benchmark based on a random-shooting, MPPI (Model-Predictive Path Integral) formulation, while requiring substantially lower online computational effort. The results demonstrate that RL can be effectively integrated as a supervisory layer in offshore gas compression systems without replacing established industrial control infrastructure, although additional safety mechanisms are still required to ensure stricter constraint satisfaction.

Recurrent neural nets (RNNs) are now commonly used for the surrogate modeling of process systems, leading to better control and faster real-time optimization. However, when trained with small training data sets, most experiments show that RNNs exhibit poor generalization abilities outside the range of the training data space. Nonetheless, recent advances in deep learning research have shown that certain characteristics of the loss landscape of trained models, such as the flatness around the local minimum, tend to relate to generalization ability. This paper investigates this phenomenon for the case of RNN surrogates of the well-known Three Tanks case study, which is representative of many continuous processes. We trained a total of 200 LSTMs (long short-term memory networks) differing in initialization, architecture, and training dynamics on the same data of 500 samples. The number of model parameters ranges from 238 to 11, 353. We estimated the loss curvature of each trained model using the Hessian-vector products method and then evaluated their test accuracy on 100 pre-generated data sets with 50% larger input amplitude than the training data to force extrapolation. We found that the estimate of the top eigenvalue of the Hessian matrix is 75% rank-correlated to the one-step-ahead prediction accuracy averaged on all test data sets. We report other insights on the impact of other Hessian-based metrics to generalization ability. Our study can potentially inform how to control the training dynamics of RNN surrogates to improve their extrapolation ability in the absence of first-principles knowledge.

Multi-stack proton exchange membrane (PEM) water electrolysis systems are increasingly deployed to improve the scalability and flexibility of green hydrogen production. However, sharing balance-of-plant equipment introduces coupling between stacks, and differences in stack performance increase the complexity of plantwide operation. In particular, non-identical efficiencies and safety constraints, such as hydrogen-to-oxygen (HTO) ratio limits, can render single-stack or equal-power-sharing control strategies suboptimal. In this work, the steady-state optimal operating regime of a two-stack PEM electrolysis system is characterized using a plantwide optimization approach and active constraint mapping over a range of system power loads. Performance differences between the stacks are represented through variations in Faraday efficiency to emulate simplified degradation. For identical stacks, the system behaves similarly to a single large electrolyzer, where equal power distribution is optimal, and the active constraint regions closely resemble those of a single-stack system. As the stack performance differences increase, the optimal power distribution becomes asymmetric, with the more efficient stack preferentially loaded. However, HTO safety constraints in the degraded stack may limit the utilization of the more efficient stack and introduce additional active constraint regions, resulting in more complex operating regimes.

Carbon dioxide transport is a critical component of the carbon capture and sequestration (CCS) supply chain. Given the substantial energy requirements and dispersed locations of CCS facilities, optimizing pipeline operations is critical to minimize costs. Although CO2 in dense phase is typically favored for long-distance transport, gaseous phase transport is also a possibility for shorter distances and volumes. This study models a gaseous CO2 pipeline system. Since CO2 gas pipelines provide the benefit of line packing, owing to gas compressibility, this work leverages it to maximize throughput in the presence of disturbances. Pipeline pressures within each segment are perceived as an inventory (i.e. form of storage) and a model predictive control (MPC) formulation for optimal inventory management is implemented to maximize throughput. This study applies the formulation to pipelines arranged in series and parallel. It effectively maximizes throughput and optimally drains pipeline pressures in the presence of an inlet flow rate disturbance.

Plastic injection molding is a widely used polymer-processing method. As the requirements for processing accuracy have become increasingly stringent, defect analysis in plastic injection molding is necessary to improve the product yield. Causal discovery has recently gained attention for defect analysis in many processes. Because injection molding is a spatial process involving the distribution of physical quantities, spatial autocorrelation should be considered. Although the linear non-Gaussian acyclic model (LiNGAM) is a well-known causal discovery method, it cannot properly model spatial autocorrelation. In this study, a new causal discovery method for a spatially autocorrelated dependent variable, referred to as the Causal Structure Search for the Spatial Autoregressive Model (CASSPAR), is proposed. It models the causal relationships among the observed points without prior knowledge of the spatial structure. The proposed method represents the causal relationships among the observed points as a causal graph and estimates the adjacency matrix of the graph. The adjacency matrix is estimated using LiNGAM, and the model parameters are estimated using two-stage least squares (2SLS). The usefulness of the proposed CASSPAR was demonstrated using simulation data of a plastic injection molding process to identify the root cause of warpage.

As industries pursue more sustainable and flexible manufacturing strategies, batch processes continue to play a vital role across a wide range of applications. Batch operations offer the ability to handle diverse feedstocks and accommodate varying product specifications. These processes are broadly used in sectors such as pharmaceuticals, specialty chemicals, food production, and bioprocesses, where precise control over reaction conditions and product quality is essential. However, maintaining optimal conditions in a batch process can be challenging due to the minimal opportunities for mid-batch interference. This work focuses on a real industrial batch process that frequently sees batches with poor yields resulting in large financial losses. Despite utilizing a mid-infrared spectrometer analyzing the batch medium in real-time, the reduced product accumulation observed in faulty batches is not evident until over a third of the batch time has passed, by which point the batch is not economically viable to abort. This study applies and compares various machine learning based fault detection strategies, including multiple Principal Component Analysis (PCA) variants and autoencoder variants, to identify faulty batches as soon as possible, since resetting reset the batches earlier can maximize overall productivity. The findings of this study offer faster and more robust fault detection than typical PCA for this industrial batch process, reducing time lost to faulty batches and improving overall productivity. This work supports the transition towards autonomous and digitalized batch manufacturing while providing an in-depth comparison between several online fault detection strategies on real industrial data.

Modern semiconductor manufacturing requires extreme precision as yield margins narrow in the “More-than-Moore” era. While physics-based models (PBMs) provide high-fidelity insights into plasma etching, their computational intensity—often requiring hours per simulation—renders them impractical for direct iterative optimization. This work demonstrates a hybrid framework that utilizes data-driven surrogate models to enable rapid, cost-effective process optimization. A 2D axisymmetric fluid model of an inductively coupled O2 plasma (ICP) reactor was developed to generate a training dataset for two neural architectures: a Multi-Layer Perceptron (MLP) and a Kolmogorov-Arnold Network (KAN). These surrogates predict radial etching rates across a wide operating window of power, pressure, gas flow, and bias voltage. By replacing the expensive PBM with these high-speed surrogates, derivative-free optimization algorithms (Nelder-Mead and Powell) successfully identified a profit-maximizing operating point (2000 W, 10 mTorr) orders of magnitude faster than direct physical simulation. The results confirm that surrogate-based optimization effectively captures dominant physical trends, such as ion-flux limited regimes, while providing a “Confidence Gap” through model disagreement to flag epistemic uncertainty. This methodology offers a scalable blueprint for reducing the computational burden of process design, transitioning from expensive trial-and-error to efficient, physics-validated autonomous discovery.

Reliable process operation under uncertainty remains a fundamental challenge in chemical and pharmaceutical manufacturing. Variability arising from feed fluctuations, material properties, external disturbances, and uncertain model parameters can significantly impact feasibility and performance when operating ranges are designed under nominal assumptions. Design Space Identification (DSI) addresses these limitations by defining a Probabilistic Design Space (PDS), within which process constraints are satisfied with a prescribed confidence level. However, identifying and extracting robust and practically usable nominal operating ranges (NORs) from complex, high-dimensional, and nonconvex PDSs remains computationally demanding. This work proposes a novel reinforcement learning (RL)–based framework for automated identification of the largest robust hyper-rectangular NOR fully contained within a given PDS. The design centering problem is formulated as a sequential decision-making task and solved using the Proximal Policy Optimization (PPO) algorithm. By interacting with a probabilistic feasibility environment, the RL agent adaptively adjusts the location and size of the NOR to maximize its area while enforcing probabilistic constraint satisfaction. Deep neural network policies enable efficient exploration of nonlinear feasibility boundaries without exhaustive sampling or gradient-based optimization. The approach is demonstrated on a tablet lubrication case study with parametric uncertainty. Results show rapid convergence to well-located, near-maximal NORs using substantially fewer probabilistic feasibility evaluations than conventional grid-based or Monte Carlo methods. The proposed PPO-based framework provides a scalable and data-efficient solution for robust operating range identification in uncertainty-aware process design.

Physics-Informed Neural Networks (PINNs) offer a promising approach for integrating first-principles modeling with data-driven methods, especially in dynamic thermal systems. This study introduces a hybrid PINN framework for a one-dimensional heating rod governed by heat transfer equations. Unlike traditional PINNs that rely on time-dependent automatic differentiation, this approach employs numerical derivatives to bypass gradient saturation and enhance robustness. The proposed model demonstrates accurate extrapolation and generalization with limited training data and is effectively used as a surrogate in a Model Predictive Control (MPC) framework for rod-tip temperature regulation. Additionally, a plan is outlined to apply physics-informed dimensionality reduction and model order reduction to improve computational efficiency and enable real-time application. The findings affirm PINNs’ potential as control-oriented reduced models for thermal processes.

With the ongoing advances in generative artificial intelligence (GenAI), the initial skepticism surrounding its tools is gradually diminishing. In fact, tools such as ChatGPT, Copilot and similar, are often used in everyday tasks, both in our personal lives and in educational contexts. Educators may use them for content creation, grading exams, or automating repetitive tasks. Students resort to them to better understand a topic, get feedback on an assignment and brainstorm ideas. Research has shown that, if used correctly, these tools can spur and support both teaching and learning. However, these continuous advancements and the increasing number of available tools also require more research to benchmark all these models and, if possible, provide quantifiable indications of which tool is better to use for which specific subtopic. As such, we created FermBench, a dataset of fermentation knowledge, which can be used to benchmark various large language models (LLMs). The models selected for the experiment are: GPT-5 mini, Claude Sonnet 4.5, Gemini 3 Flash, Mistral Small 3.2, and DeepSeek-V3.2. The performance of the models is scored using the LLM-as-a-Judge approach, with GPT 5.2 and Claude Opus 4.5 as judges. The LLMs agree that the model used in the free tier of ChatGPT is the most factually accurate. DeepSeek and Gemini outperform the other models in readability and helpfulness, while Claude and Gemini provide the most concise answers. Overall, we conclude that the quality of the responses generated is satisfactory for educational contexts. We also provide students with practical guidelines on how to be more critical of responses generated by LLMs and to improve their prompts. All code and curated data are open-source.

Workflow automation platforms connect applications and services to automate data transfer and multistep processes. Although widely used in engineering research and institutional administration, including engineering institutions, they are rarely integrated into undergraduate engineering curricula, and their educational adoption introduces ethical, technical, and pedagogical risks. This paper proposes a practical framework for developing, deploying, and assessing workflow-automation-enabled learning tools coupled with generative AI, with explicit attention to institutional constraints and learning outcomes. As a conceptual case study, we present it in a second-year chemical engineering course (Heat and Mass Transfer) to support learning of heat conduction. The platform includes instructor-approved assets such as content slides, solved problems, pre-prompts, and a validated question database, through an automation pipeline that issues structured API calls to a generative AI system and returns explanations and worked examples. A subject matter expert (SME) authors and tests prompt modules aligned with learning outcomes mapped to Bloom’s taxonomy levels (understand, apply, analyze, evaluate) and aligned with accreditation expectations. Effectiveness metrics are proposed to estimate learning gains, and a tool quality audit and an SME review are used to verify conceptual correctness and computational validity. To assess risks, we apply a likelihood–impact risk matrix and define five categories: data privacy and protection, transparency, pedagogical impact, institutional compliance, and ethical considerations. Preliminary analysis suggests potential concept gains from structured prompt/content integration but indicates that numerical problem reliability depends strongly on question-database coverage. The highest perceived risks relate to privacy and transparency, followed by pedagogical impact, motivating mitigations such as anonymization protocols, restricted data pathways, and human verification throughout the lifecycle.

As Industry 5.0 (I.D. 5.0) reshapes the engineering education landscape, Higher Education Institutes (HEIs) have evolved to integrate Generative Artificial Intelligence (GenAI) via strategic curriculum revamps to meet Education 5.0 (E.D. 5.0) competencies. EN.540.202 (Introduction to Chemical & Biological Process Analysis) is the first core course at Johns Hopkins University and was revamped in Fall 2025 to create more rigorous course content and the conscious creation of new weekly graded problem sets, which did not rely on prior course content/textbook-based solved examples. Problem sets were fed as Effective Prompt Engineering (EPE) inspired prompts to ChatGPT, and AI-elicited responses were compared. AI was able to perform fundamental calculations, offer detailed explanations, unit conversions/checks, proactive information (outside the problem scope), and graphical information. Key challenges and pitfalls observed were terminology misinterpretation, lack of visual representation, data acquisition issues, and ambiguity when tackling open-ended problem prompts, which pose a potential threat of incorrect learning if students exclusively rely on ChatGPT. Quantitative benchmark comparisons of AI performance spanning across various prompt categories indicate a significant 73% performance gap. While AI demonstrates 88% accuracy on standard problems, the success rate plunges to 15% on complex mass/energy balances involving non-ideal recycle streams and constraints. Introduction of EPE and Chain-of-Thought (CoT) strategies mitigated these risks, restoring AI accuracy to 91% in instructor-led trials. To ensure academic integrity, a dual-layer Quality Assurance (QA) audit was implemented using Turnitin’s AI-writing detection; analysis shows a mean AI-probability of <12% in derivation sections, suggesting students successfully used AI for structural brainstorming, while maintaining technical ownership of engineering execution. Anonymous student-filled instructor teaching evaluations capture a strongly positive perception of preparedness, better student engagement, and recognition of potential risks of over-reliance on AI-led Large Language Models (LLMs), potentially leading to AI-led misdirected learning (unless students can distinguish between correct/incorrect AI responses). Despite the intellectual challenge of the course being rated as high, overall course quality and instructor teaching effectiveness metrics were ranked high, signalling a successful case of AI-led curriculum revamps, capturing a thoughtful integration of EPE into the undergraduate curricula at HEIs. This work provides a scalable framework for HEIs to modernize engineering pedagogy, positioning AI as a collaborative tool that augments and empowers, rather than replaces, critical human-centric engineering judgment.

Data extraction and processing constitute the cornerstone of quantitative management and operational analysis in industrial process plants. However, most manufacturing facilities currently lack efficient data extraction systems, relying instead on engineers to manually write and execute database queries, which is time-consuming, error-prone, and inflexible when handling diverse data formats or large-scale equipment networks. To address these limitations, this work presents a novel Large Language Model (LLM)-based intelligent framework designed for data extraction and basic data of industrial equipment. The system integrates natural language understanding capabilities with process database schemas, enabling users to perform complex data queries and analyses through natural language prompts. Specifically, it can perform data mining, time-dependent analyses, equipment comparisons, and cross-period performance evaluations when process information is provided. By integrating the process context, the system can further interpret the analysis results to suggest potential operational conditions or abnormalities occurring in the system. Developed in collaboration with a process industry partner, the proposed system was evaluated using six months of real-world operational data from 10 heat exchangers. Experimental results demonstrate that the framework achieves an extraction accuracy exceeding 99% in the online mode and ranging from 92% to 96% in the privacy-preserving offline mode, with a reduced retrieval latency of approximately 2.64 seconds for local inference. Unlike generic chatbots, this framework targets process-specific reasoning, acting as a cognitive assistant that empowers users to interact intelligently with industrial systems. The findings confirm that LLM-assisted extraction effectively addresses the rigidity and inefficiency of conventional approaches while maintaining high reliability and scalability for industrial analytics.

The Imperial College Integrated Design Project reframes the chemical engineering capstone as a structured educational journey that develops professional competence rather than simply delivering a final technical report. The programme is grounded in four pedagogical pillars—authenticity, integration, impact, and reflection—which align with the graduate attributes required by the Institution of Chemical Engineers. Authenticity is achieved through open-ended problems drawn from industrial partners and emerging research needs; integration connects knowledge from across the curriculum into a coherent systems perspective; impact emphasises user-centred, sustainable solutions; and reflection cultivates metacognitive awareness of decision making and learning from failure. A mentored-autonomy model supports student teams through weekly checkpoints, skills workshops, and access to disciplinary experts. Assessment deliberately balances artefact quality with evidence of process, rewarding reasoning under uncertainty, ethical judgement, and communication alongside technical performance. Ethics, safety, and sustainability are embedded through lifecycle analysis and HAZOP requirements rather than treated as add-ons. The four-phase structure—scoping, feasibility, detailed design, and deployment—scaffolds increasing complexity while maintaining measurable progress and accountability. A 2024 project on renewable-powered green ammonia illustrates how students integrate decision-making methods, simulation tools, and stakeholder negotiation within a realistic professional context. Educational outcomes include strengthened systems thinking, collaboration, project leadership, and the ability to justify design choices in social and environmental terms. Institutional benefits include deeper industry engagement and improved graduate readiness. The paper argues that capstones should be judged by the quality of learning processes as much as by final designs, and that this model produces reflective engineers equipped for interdisciplinary practice and responsible innovation.

A survey was conducted of 103 students and lecturers who had recently participated in chemical engineering design courses concerning their opinions on the use of Generative Artificial Intelligence (Gen-AI) in their capstone design education. Participants were at universities in Europe, the Middle East, North America, and South America, from at least eight different language groups. The survey found little difference in responses between students and lecturers, except for uptake, in which students reported higher rates of familiarity and adoption of Gen-AI tools than instructors. Both groups were net-positive generally on the use of Gen-AI in the classroom, reporting relatively high confidence in the ability to assess results, the general positive benefits of using Gen-AI in their chemical process design education, and the likelihood of using them in the future. However, participants reported that their trust in the results of Gen-AI tools was relatively low.

Computational tools are widely used in solving engineering problems and are now embedded within chemical engineering education. At the UCL Department of Chemical Engineering, students are taught gPROMS ModelBuilder in modules requiring coding; however, many choose alternative tools such as MATLAB, Python, or Polymath for coursework and capstone design project reactor design. This study investigates the reasons behind these preferences using a survey of fourth-year students who had completed their third-year design project. The results show that perceived ease of use, availability of external resources, and ease of debugging could strongly influence tool selection. The findings highlight the importance of accessibility, community support, and perceived relevance in shaping sustained student engagement with computational tools.

Electrolysis systems must be constructed from multiple stack units. This modular system of stack units (modular electrolysis system) requires systematic optimization of its composition. This optimization depends on numerous, and sometimes conflicting, criteria. While such criteria have already been evaluated for modular plants in the process industry, they must be re-assessed in the context of modular electrolysis systems. To address this challenge, an expert survey was conducted within two research networks (H2Giga and DECHEMA e.V. Research Network) and the VDMA P2X4A.The evaluation followed the two-stage process: After ranking the categories costs, flexibility, process engineering and time-to-process according to their overall importance a ranking of individual criteria within each category was conducted. The survey reveals a clear prioritisation: costs are in first place with 35.9%, followed by flexibility (25.4%), process technology (23.2%) time to process (15.4%). This ranking provides a structured overview of the most important factors guiding the optimisation of modular electrolysis systems. Another finding of the survey is how electrolysis systems should be operated in the future. Here, the focus was primarily on electrolysis systems related to fluctuating renewable energies, with a top-2-box result of 81% in each case. Based on this prioritisation, weightings for possible application in multi-criteria optimisation can now be derived. This makes it possible to determine suitable compositions of electrolysis stack unit systems when scaling systems and to refine optimisation efforts. The expert survey thus contributes to the optimised design of electrolysis systems and thereby supports the ramp-up of hydrogen production.

The rapid shift toward Industry 4.0 and data driven manufacturing has prompted universities to reshape chemical engineering programmes, yet the scope and coherence of “Digital Chemical Engineering” (DCE) within UK curricula remain unclear. This study qualitatively maps DCE provision across five IChemE accredited MEng degrees to identify which digital skills are taught, how they progress across programme stages, how skills are distributed between core and elective content or taught versus applied learning, and how well provision aligns with accrediting frameworks. The analysis is structured around eight domains: (D1) programming and computation; (D2) data literacy and statistics; (D3) process modelling and simulation; (D4) optimisation and process systems engineering; (D5) control, automation and instrumentation; (D6) AI, machine learning and digital twins; (D7) software engineering practices; and (D8) data governance, ethics and cybersecurity. Results show institution dependent digital skills coverage. Some programmes introduce a broad digital skill set from Year 1, while others concentrate digital content in later years or adopt a more progressive distribution. Across institutions, D7 is consistently underrepresented. Foundational skills in D1, D2 and D8 are typically embedded throughout programmes, whereas more advanced domains D3 to D6 are introduced at later stages. Domains D1 to D4 demonstrate a stronger balance between teaching and application, while D5 to D8 are predominantly taught, with application increasing in later years. Optional fourth year modules play a significant role in shaping graduate digital profiles. Overall, DCE provision broadly aligns with IChemE and AHEP4 guidance, where expectations for D1 to D4 are more explicitly articulated

As high-throughput experimentation (HTE) becomes a cornerstone of modern materials research, undergraduate and postgraduate curricula increasingly require students to possess Python programming skills to operate automated liquid-handling robots, such as the Opentrons OT-2 and Flex. However, the high cost of this hardware often necessitates shared equipment use during hands-on lab sessions, creating a significant pedagogical barrier: students lack the individual time required to iteratively test and debug their protocols on physical robotic platforms for automated material synthesis. Furthermore, we observe that the scarcity of robotic platforms creates an imbalance in group dynamics, where students with more coding experience often lead protocol development, while those with less experience remain disengaged. To address these challenges, we developed an interactive simulator that translates Python protocols into 2D animations on personal laptops. Using gold nanoparticle (AuNP) synthesis as a case study, we demonstrated how this platform enabled visualisation, captured common programming mistakes, and refined the efficiency of the code before implementation using the physical robot. By shifting the ‘trial-and-error’ phase from the lab to the laptop, the simulator showed potential to democratise students’ access to automation education and ensure equitable skill application across the cohort.

The implementation of processes that use renewable energy requires that a techno-economic analysis be performed beforehand to determine its economic and technical feasibility. A techno-economic analysis was performed proposed wind farms in Trinidad and Tobago using the System Advisor Model simulation software. Metrics included the annual energy production in kWh, capacity factor, net present value in US$ and internal rate of return. From the above, the number of households that can be powered each month by the farms were calculated. The results showed that rotor diameter, which defines the swept area has a significant impact on annual energy production as a 33 m difference translated into a 27.3 GWh and 22.9 GWh difference in output. The results are promising and show that the oil and natural gas-based economy can be diversified.

Advancing student understanding of process design requires a balanced integration of theoretical knowledge with real-world industrial applications. This study introduces system design thinking-based learning through an engineering clinic approach that bridges the gap between classroom concepts and chemical engineering practice. Using an industrial multiproduct oil pipeline operation as a case study, students are exposed to real-world industrial systems, identify bottlenecks in a process, draw similarities between systems at different scales, and implement control strategies to address a practical industrial problem. In this study, we highlight the collaborative efforts between faculty, students, and industry partners to provide experiential learning in process design, modeling, and control to address the challenge of minimizing product loss during flushing operations in multiproduct petroleum pipeline systems.

The accelerating social, environmental, and economic challenges of the twenty-first century call The growing complexity of sustainability challenges calls for educational approaches that integrate technical analysis with multi-stakeholder decision-making. Industrial ecology (IE) provides a relevant framework by combining systems thinking, resource flow analysis, and socio-environmental considerations. However, it is still predominantly taught through traditional lecture-based methods, limiting students’ ability to engage with real-world complexity. This paper proposes and evaluates an experiential pedagogical framework based on industrial ecology, combining stakeholder role-play, industrial symbiosis scenario design, and multi-criteria decision analysis (MCDA). Implemented in a semester-long course, the framework enables students to collaboratively design and evaluate resource-exchange networks while representing different stakeholder perspectives. Results show significant improvements in both technical skills (systems analysis, resource flow evaluation) and transversal competencies (communication, negotiation, reflexivity). The approach enhances student engagement and highlights the role of value judgments in sustainability decision-making. The framework offers a transferable methodology for sustainability education in engineering and related disciplines.

Digital competency is now core to chemical engineering practice, yet the extent and coherence of digitalisation skills provision across undergraduate curricula remain uneven. This study maps qualitatively and qualitatively digital learning outcomes across undergraduate chemical engineering programmes at the University of Sheffield, against a digital skills framework (data analysis, process simulation, process automation & control, reproducible workflows, programming, data governance). In recent years, digital skills education within chemical engineering education has advanced considerably, driven by the broader industrial shift toward Industry 4.0 and reinforced by the global challenges. Academic institutions have begun to integrate digitalisation-related content more deliberately within syllabus, in alignment with degree programme accreditation requirements and industry needs. Beyond introductory spreadsheet manipulation and basic programming, many courses are now embedding more advanced digital capabilities, including data engineering, scientific programming, process analytics, and optimisation techniques. Results of this study reveal a scaffolded digital pathway, where students are exposed gradually through core taught content and applied knowledge activities to a variety of digital tools across their years of study in a by-design approach at earlier years that becomes organic later on. From limited exposure and credit correlation in Year 1, students complete Year 4 having at least quadrupled their exposure to digital tools, with higher credits awarded to “digital” in the case of particular elective modules or research projects. This study offers a methodology and a worked example for chemical engineering programmes seeking measurable uplift in Industry 4.0 readiness and impact on chemical engineering graduate outcomes.