Job Safety Analysis (JSA) is critical for proactively identifying workplace hazards, assessing their potential consequences, and implementing effective control measures. However, traditional JSA methods can be inefficient and prone to errors, particularly in complex industrial environments. This paper introduces AutoJSA, a knowledge-enhanced framework that leverages large language models (LLMs) to automate and optimize the JSA process. We collected 73 high-quality JSA reports from a chemical engineering company and divided the JSA workflow into three key tasks: hazard identification, consequence identification, and control measure generation. Two approaches - fine-tuning and retrieval-augmented generation (RAG) - were employed on a base LLM (GLM-4-9B-Chat) to adapt it for these domain-specific tasks. Experimental results demonstrate that both fine-tuning and RAG significantly improve task performance relative to the unmodified model, with fine-tuning generally providing larger gains. W... [more]

The growing size and complexity of energy system optimization models, driven by high-resolution
spatial data, pose significant computational challenges. This study introduces methods to reduce model’s size and improve computational efficiency while preserving solution accuracy. First, a composite-curve-based approach is proposed to aggregate granular data into larger resolutions without averaging out specific properties. Second, a general clustering method groups geographically proximate fields, replacing multiple transportation arcs with a single arc to reduce transportation-related variables. Lastly, a two-step algorithm that decomposes the supply chain design problems into two smaller, more manageable subproblems is introduced. These methods are applied to a case study of switchgrass-to-biofuels network design in eight U.S. Midwest states, demonstrating their effectiveness with realistic and detailed spatial data.

Three different implementations of the flipped class paradigm were used to teach Chemical Engineering students at Imperial College London (ICL) in the 2023-24 academic year: (1) The 3rd year elective course Introduction to Numerical Methods (INM) taught in its entirety in flipped format (the "good"); (2) The 2nd year core course on Process Dynamics and Control (PDC), with the first half of the course on process dynamics taught in traditional lecture format, and the second half on process control taught in flipped format (the "bad"); and (3) a one-week workshop on heat integration, taught as part of a 3rd year core course on Process Design (PD), taught in flipped format (the "surprising"). This paper describes these three implementations in detail and presents and analyzes the responses from student surveys intended to ascertain students' perceptions about the level of their satisfaction with the flipped class approach and the degree to which they achieved mastery of the taught... [more]

Maintaining sufficient amounts of dissolved oxygen throughout a microbial cultivation is a classic control task in bioprocess engineering to avoid negative effects onto cell physiology and productivity. But traditional PID-based algorithms struggle when faced with pulsed substrate additions and the resulting sudden surge of oxygen uptake. In this work a nonlinear MPC is employed and compared to a PID setup for the cultivation of an E. coli strain exposed to intermittent feeding. Both controllers are tuned for a fast pulse response combined with efficient and robust control action. Their performance was tested in-silico with isolated feed pulses, as well as throughout a full cultivation run. Further, the effects of parameter uncertainty were investigated to assess the impact of a model-plant mismatch. The results showed that the predictive nature of the MPC is well suited for maintaining the dissolved oxygen levels above a threshold and outperforms the PID in almost all investigated sim... [more]

Optimization under uncertainty is inherent to many PSE applications ranging from process design to RTO. Reaching process true optima often involves learning from experimentation, but actual experiments involve a cost (economic, resources, time) and a budget limit usually exists. Finding the best trade-off on cumulative process performance and experimental cost over a finite budget is a Partially Observable Markov Decision Process (POMDP), known to be computationally intractable. This paper follows the nonmyopic Bayesian optimization (BO) approximation to POMDPs developed by the machine-learning community, that naturally enables the use of hybrid plant surrogate models formed by fundamental laws and Gaussian processes (GP). Although nonmyopic BO using GPs may look more tractable, evaluating multi-step decision trees to find the best first-stage candidate action to apply is still expensive with evolutionary or NLP optimizers. Hence, we propose modelling the value function of the first-st... [more]

Active learning is widely recognized as an effective teaching approach that can improve classroom outcomes. This is enabled by providing the time for students to apply new knowledge, make mistakes, correct them, and repeat the process until mastery is achieved. One way to implement active learning is through the flipped classroom paradigm. However, to be effective, active learning depends on providing students with a variety of open-ended problems, ranging in difficulty from introductory to advanced levels. This paper presents four food-themed problems for use in numerical methods and process control courses:
1. Formulating Willy Wonka’s new chocolate bar: An introductory linear programming problem focused on translating verbal descriptions into mathematical models.
2. Optimal production for the Matrix Pizza company: A more advanced mixed-integer linear programming problem involving multiple scheduling scenarios.
3. Optimal frying time for fried ice cream production: A transient hea... [more]

Graph neural networks (GNNs) have proven state-of-the-art performance in molecular property prediction tasks. However, a significant challenge with GNNs is the reliability of their predictions, particularly in critical domains where quantifying model confidence is essential. Therefore, assessing uncertainty in GNN predictions is crucial to improving their robustness. Existing uncertainty quantification methods, such as Deep ensembles and Monte Carlo Dropout, have been applied to GNNs with some success, but these methods are limited to approximate the full posterior distribution. In this work, we propose a novel approach for scalable uncertainty quantification in molecular property prediction using Stochastic Gradient Hamiltonian Monte Carlo (SGHMC). Additionally, we utilize a cyclical learning rate to facilitate sampling from multiple posterior modes which improves posterior exploration within a single training round. Moreover, we compare the proposed methods with Monte Carlo Dropout a... [more]

These are the slides presented at the ESCAPE 35 conference on Monday July 7, 2025, in the talk with the same name. They briefly introduce the concept of exergy with a basic overview, and provide seven easy examples that professors can use in their courses. The topics include heating systems, pinch analysis, energy efficiency, energy integration, steam generation, utilities, heat pumps, organic Rankine cycles, direct air capture of CO2, and CO2 compression and sequestration. See the linked conference paper for more information.

Maintenance is critical for industrial plants to ensure operational reliability and worker safety. In process industries, fouling, the accumulation of solid residues in equipment, poses a significant challenge, causing inefficiencies and productivity losses. Effective modeling of fouling evolution over time is essential for maintenance planning to prevent equipment from operating under suboptimal conditions. Traditional approaches to fouling prediction include equation-based models, which offer high precision but may struggle with continuously changing process bound-aries, and machine learning techniques, which are more adaptable but less effective at capturing rapidly evolving trends driven by complex underlying physics. This study introduces an innova-tive hybrid machine learning approach for predictive maintenance, combining the strengths of both methods. Pressure differential is modeled using an equation-based approach that links pressure data with fouling thickness, while the foul... [more]

In this paper, we present our recent advances and achievements in automatic control course in the engineering study of cybernetics at the Faculty of Chemical and Food Technology STU in Bratislava. We describe the course elements and procedures used to improve teaching, learning, and administration experience. We discuss on-line learning management system, various teaching aids like e-books with/without solutions to practice examples, computer generated questions, video lectures, choice of computation and simulation tools.
The course is provided in the presence form of study for about 20 students, but it relies on on-line tools and methods. Starting from this academic year, flipped design of the course was designed. We describe our experience in the preparation of such a change and some initial feedback from the students.
The course concentrates on input/output linear approximation of processes in chemical and food technology and discusses poles/zeros, process dynamics, frequency and... [more]

Distillation processes account for a substantial share of the industrial energy demand. Yet, these energy requirements can be reduced by a variety of energy integration methods, including various forms of direct heat integration, multi-effect distillation, thermal coupling and vapor recompression. Consequently, these intensification methods should be evaluated quantitatively in comparison to each other for individual separation tasks, instead of benchmarking single options with conventional sequences or relying on simplified heuristics. In order to overcome the computational burden of a broad assessment of a large number of process alternatives, a computationally-efficient framework for the energetic and economic evaluation of such energy integrated distillation processes is presented, which builds on thermodynamically-sound shortcut models that do not rely on constant relative volatility and constant molar overflow assumptions.

A new methodology for the assessment of the performance loss in catalytic reactors due to deactivation was developed and applied to fixed- and fluidized-bed CO methanation, with catalyst subject to coking. The methodology is based on the solution of heat and mass balances, by decoupling the reactor and deactivation dynamics. This is possible by using consecutive 1D, steady-state calculations for the characterization of the reactor performance. In this way, the progressively lower values of catalyst activity along the time on stream are computed with the integration of a dedicated dynamic model. This method has shown promising results in the characterization of the loss of performance of the reactor over time. The model correctly describes a progressive deactivation of the catalyst in fixed-bed reactors, while it shows that the decrease in activity is sudden for the whole reactor volume in fluidized bed reactors and occurs after a critical time-on-stream. Besides, it was observed that t... [more]

Hybrid modelling is widely employed in chemical engineering to generate highly accurate predictions. Such an approach merges first-principle modelling with machine learning techniques to identify and model the epistemic uncertainty from experimental data. Despite its advantages, this still requires cross-domain competencies that are difficult to find in the chemical industry and high human involvement. The possibility of automating the identification and training model would be significantly beneficial for the widespread adoption of hybrid modelling methodology within the chemical industry. This work presents a novel algorithm for the automatic identification of hybrid models (HMs) starting from the first-principle representation of the system, described by differential equation sets. The methodology formulates the problem as mixed-integer programming, identifying the equation running under uncertainty, identifying the machine learning model hyperparameters, and training the latter. Th... [more]

The introduction, peer review policy, and International Scientific Committee for Systems and Control Transactions volume 4 (ESCAPE 35 Proceedings)

This is the cover page and front matter for Systems and Control Transactions volume 4 (ESCAPE 35 Proceedings)