Thermal-Hydraulic Process Supervision and Monitoring Through a Data-Driven Digital Twin

Résumé fourni par la source

Abstract Real-time supervision of industrial processes is of high importance in order to maximize operational efficiency and prevent system failures. The main purpose of this is to implement predictive maintenance techniques in order to ensure optimal cost, reduce unplanned downtime, and extend the life of the machinery. However, this maintenance and supervision of industrial processes remains a persistent barrier to the efficiency of industrial solutions. Current technological advancements, having provided useful simulation tools coupled with powerful machine learning and AI-based algorithms, have the potential to facilitate the development of data-driven and numerical tools for predictive maintenance and supervision. This paper outlines the implementation of a fault detection and diagnosis algorithm as the decision component of a digital twin. The proposed pipeline uses a three-step approach for fault detection, localization, and estimation using data-driven approaches for detection and estimation through machine learning algorithms and one-dimensional Computational Fluid Dynamics simulations. The proposed Fault Detection and Diagnosis Framework has been validated first through simulations in both the noiseless and noisy cases, and finally in the physical setup of the hydraulic system. The preliminary results indicate a high degree of accuracy in fault identification and good accuracy in fault estimation in a noisy physical environment.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.