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 operatin... [more]

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 econ... [more]

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... [more]

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 pressu... [more]

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 optim... [more]

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 usi... [more]

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 byp... [more]

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 scenario... [more]

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.

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 en... [more]

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.

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.

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... [more]

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 serv... [more]

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... [more]

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 be... [more]

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 al... [more]

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 a... [more]

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 re... [more]

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.

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. W... [more]

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.

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 c... [more]

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... [more]

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 s... [more]