Data-Driven Process Monitoring Approach for Closed-Loop Cascade Systems

This paper proposes a data-driven process monitoring approach for closed-loop cascade systems. Each subsystem is operated under the feedback control in the cascade connection. For each subsystem, the external input is proved to be represented by the linear combination of inputs and outputs from its upstream subsystems. In addition, a data-driven residual generator is proposed via the LQ decomposition for process monitoring and takes into account the feedback controller. Furthermore, the fault detection strategy is developed in the context of probability. The proposed methods are illustrated and verified on a cascade system consisting of five discrete-time LTI subsystems.