Causality in Process Systems Engineering: Fundamentals, Applications, and Emerging Trends

Abstract

The increasing availability of high-dimensional data from chemical and industrial processes has enabled the widespread adoption of machine learning and deep learning methods. However, their black-box nature raises critical concerns about reliability, ethics, and security in safety-critical industrial applications, highlighting the need for Explainable Artificial Intelligence (XAI) solutions. In this context, Causality analysis emerges as a foundational approach within XAI, moving beyond correlations to uncover genuine cause-and-effect relationships that are essential for reliable decision-making.

Despite its potential, the adoption of causal reasoning in Process Systems Engineering (PSE) is still incipient. Therefore, in this work, we establish the crucial role of formal causal analysis as both a theoretical framework and a practical toolkit for addressing core challenges in PSE. We systematically present the fundamental concepts and methods of causal analysis, including do-calculus, causal discovery, and causal inference, providing the necessary fundamentals for PSE researchers and practitioners entering this field. Furthermore, we emphasize the integration of these causal data-driven techniques with domain knowledge, such as process diagrams, hazard studies, and first principles, to address inherent industrial complexities, including nonlinearities, multi-mode operations, feedback control loops, and dynamic behavior. The practical value of causality is illustrated in several application fields, and recent emerging trends are also covered.