Early Detection of Thermal Runaway in Chemical Reactors: From Classical Methods to Artificial Intelligence
Year: 2026
Type: conference-paper
Abstract: Thermal runaway is a rare event. It occurs when there is an imbalance between the heat generated and the heat removed. The temperature then rises, which accelerates the reaction and can make the reactor out of control. The eraly detection provide to the control systems more time to respond before the critical conditions are reached. Several approaches have been proposed for detecting thermal runaway. Classical criteria based on physical principles are considered first, followed by model based methods and more recent artificial intelligence approaches. Their ability to provide an early warning is discussed together with practical aspects such as interpretability, model uncertainty, data quality and robustness. Rather than pointing to one universally suitable method, the available studies suggest that combining process knowledge with data driven techniques may offer a useful direction for future work.
Source: HAL
Institutions: Université Le Havre Normandie, Groupe de Recherche en Electrotechnique et Automatique du Havre, Laboratoire de Mathématiques Appliquées du Havre, Institut de Recherches sur la Catalyse et l'Environnement de Lyon, Laboratoire de Sécurité des Procédés Chimiques, Université de Rouen Normandie
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
Funder: Agence Nationale de la Recherche
OpenAlex ID: W7214566443
