A robust fault-detection approach with application in a rolling-mill process

This paper proposes a new approach to robust fault detection. The proposed scheme adopts an observer-based fault-detection scheme and is characterized by a structure of two observers. One observer estimates the plant output driven by the plant normal input. The other observer estimates the output driven by the fault. To demonstrate the proposed fault-detection scheme, a rolling mill case study is considered. This task involves the development of models for the most common occurring faults on a rolling mill - the top back-up roll eccentricity, and construction of two observers. Simulation results illustrate the applicability of the fault-detection scheme to the rolling mill example.