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... [more]

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 m... [more]

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,... [more]

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.

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 accordi... [more]

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 traini... [more]

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,... [more]

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.... [more]

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 prop... [more]

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 Newt... [more]

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... [more]

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 comput... [more]

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 ar... [more]

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 proces... [more]

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 proces... [more]

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 lib... [more]

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 prop... [more]

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 t... [more]

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 perfo... [more]

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.... [more]

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-... [more]

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 mix... [more]

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 dir... [more]

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 pr... [more]

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 op... [more]