Process Fault Detection and Diagnosis by Synchronous and Asynchronous Decentralized Kalman Filtering using State-Vector Fusion Technique

This paper addresses the problem of improving state estimation of dynamic industrial processes in real time for single, double, triple and quadruple fault detection and diagnosis purposes using multi-sensor data fusion strategy. The proposed monitoring systems track the process states to infer its operating conditions utilizing a decentralized Kalman filtering methodology based on state-vector fusion technique. The paper considers both the synchronous and asynchronous multi-sensor scenarios to explore their relevant data fusion problems. The performances of the resulting monitoring systems are investigated under the two possible cases of time-delayed measurements due to communication delay and multi-rate sensors. The state-vector data fusion technique is also adopted to integrate the individual state feature coming from the distributed extended Kalman filter (EKFs) so as to extract the necessary global detection and diagnostic information. The feasibility and effectiveness of the presented algorithms are demonstrated through simulation studies on a continuous stirred tank reactor (CSTR) benchmark problem.