Keywords: Supervisory Control
Abstract:

How can you control a complex plant effectively using simple elements with a minimal amount of modelling? How can you put optimization into the control layer? Industry has been using simple and effective as “advanced regulatory control” schemes for almost 100 years. The objective of the workshop is to provide a systematic approach for designing such control systems. The approach is illustrated on numerous real industrial applications. The target audience includes both practicing control engineers as well as PhD students and teachers from academia.

Keywords: Fault Detection, Manufacturing Plant Control, Plant-Wide Optimization
Abstract: The theory and practice of process monitoring are diverging. Practical process monitoring requires fault detection, identification, diagnosis, and the implementation of process recovery actions. Algorithmic design and automation of all these steps are required if the future promise of more autonomous plants is to be realised. In contrast, theoretical research in process monitoring is overwhelmingly focused on fault detection. End-to-end process monitoring is introduced as a term that encompasses the fully automated workflow from detection to implementation of the appropriate corrective action in a process. This workshop is aimed at discussing and demonstrating a framework for improved experimentation of end-to-end process monitoring approaches. The workshop is hands-on, making use of open access case studies and example code (implementations available in both MATLAB and Python). The workshop will cover the motivation for an end-to-end process monitoring simulation framework, and give participants the opportunity to practice implementing the various components that make up the framework, including modular designs for process, control, sensors, actuators, process monitoring, operator actions, and economic performance assessment. The importance of variability and uncertainty for robust process monitoring design will be discussed and demonstrated in the case studies. The workshop is aimed at graduate students new to the fields of process monitoring, fault tolerant control, as well as being aimed at to experienced researchers in these fields (as well as the field of reinforcement learning) looking to expand the complexity and industrial relevance of case studies used to demonstrate their techniques.

Keywords: Optimal Control, Optimization under Uncertainties, Model Predictive Control
Abstract: This workshop will be an interactive tutorial on how to model complex nonlinear, continuous-time optimal control problems via InfiniteOpt.jl and JuMP.jl. Implementing my unifying abstraction for infinite-dimensional optimization (InfiniteOpt) problems, InfiniteOpt.jl is a Julia-based open-source software package that builds upon JuMP.jl to provide an intuitive symbolic modeling environment for many problem classes in dynamic, PDE-constrained, and stochastic optimization. Moreover, its extensibility allows researchers to make their cutting-edge techniques accessible to a wide audience of individuals. All these aspects make InfiniteOpt.jl a powerful tool for tackling advanced optimal control problems.

Keywords: Model Predictive Control, Data-Driven Optimization, Adaptive and Learning Systems
Abstract: The aim of this workshop is to provide hands-on experience for applications of machine learning tools in connection with control theory. Specifically, we introduce several possibilities as to how the MPC can be accelerated. We will provide a general overview and introduction to machine learning case studies suggesting suitable architectures and data preparation. Next, the workshop will introduce traditional MPC-mimicking procedures to generate approximate explicit control laws and procedures to general controllers posed as neural networks with stability guarantees. Case studies and examples will be tailored to chemical processes and problems arising in this domain. The workshop is primarily intended for graduate students and early-stage researchers and, as well as for all community members interested in hands-on applications of machine learning tools in process control. Furthermore, the workshop will be organized in interactive mode; hence, participants will gain experience in experiment preparation and coding examples.

Keywords: Machine Learning Assisted Modeling, Estimation and Robust Estimation, Adaptive and Learning Systems
Abstract: This half-day workshop aims to educate attendees on real-world experience in developing and implementing data science models to solve industry-relevant problems. Specifically, we will dive into practical data science workflow from visualization, data preprocessing, model development to implementation and then summarize best practices on how to frame business opportunity as analytics problem to deliver value. After that, we will highlight several industry case studies to illustrate these concepts. To wrap up, we will discuss some common misconceptions on data science and share future outlooks related to generative AI & the use of Large Language Models (LLMs) in industry.

Keywords: Resilient, Safe, and Cyber-Secure Systems, Fault Detection, Machine Learning Assisted Modeling
Abstract: In the dynamic landscape of complex industrial facilities, effective alarm management is paramount to ensuring operational efficiency, safety, and overall system reliability. However, in real industries, alarm systems usually suffer from degraded performance due to alarm overloading and the presence of alarm floods. Thus, there is a strong need to develop and deploy intelligent alarm management tools, so as to improve the alarm system performance and reduce operators’ workload. This workshop is meticulously crafted to address the unique challenges posed by the intricacies of modern industrial alarm systems. The workshop's primary objective is to empower industry professionals with the data analytic tools to navigate the complexities of alarm monitoring, ensuring not only compliance with industrial standards but also enhancing operational efficiency. The workshop provides participants with a comprehensive understanding of intelligent alarm management strategies for complex industrial facilities. Participants will explore the nuances of alarm overloading, gaining insights into factors contributing to alarm overloading. Through an exploration of real-world scenarios, the workshop aims to highlight the tangible consequences of ineffective alarm management while emphasizing the need for intelligent solutions, such as pattern mining and deep learning for improvement of industrial alarm systems. The workshop will cover key topics, including the design of univariate/multivariate alarm systems, identification and reduction of nuisance alarms, root cause analysis by causality inference, pattern mining of alarm floods, alarm data visualization, and alarm system performance evaluation. The talks will be accompanied by industrial case studies to demonstrate the practical utility of these intelligent methods. The workshop aims to foster a collaborative learning environment, where participants can exchange experiences, strategies, and insights, ultimately leaving with the expertise needed to revolutionize alarm management practices in their industrial contexts.