Optimal Detection Schemes for Multiplicative Faults in Uncertain Systems With Application to Rolling Mill Processes

In rolling mill processes, changes of operation conditions, mismatching of controller parameters, and aging of system components will cause multiplicative faults, which may considerably affect the system operation and the quality of steel products. The main objective of this paper is to address optimal design approaches for detecting multiplicative faults in uncertain industrial systems. To be specific, the design schemes of observer-based fault detection (FD) systems, including residual generators and threshold setting, are investigated first in the closed-loop system setup, aiming at enhancing fault detectability and system robustness against model uncertainties simultaneously. Then, an optimal design scheme of observer-based FD systems for open-loop processes is investigated. The application to an industrial rolling mill benchmark is given to demonstrate the efficiency of the proposed approaches. The task involves the detection of the parameter faults in hydraulic reduction devices in both the closed-loop and open-loop configurations.