A proposed internal model principle-based Kalman filter for fault detection

In this paper, we propose an internal model principle-based Kalman filter (IMP-KF) structure for use in fault detection. We show that the closed-loop structure of the IMP-KF is a necessary and sufficient condition for generating residuals upon which the fault detection process hinges. We advocate a residual generator structure similar to that used in standard Kalman filtering (KF), and judiciously exploit the non-robustness to model mismatch of the proposed IMP-KF scheme to detect faults in the presence of noise and disturbances. With no model mismatch, the KF residual's whiteness is exploited to derive a composite hypothesis testing that accounts for a low probability of false alarm and a high probability of correct decision for various reference inputs. The proposed scheme was successfully evaluated on both simulated and lab-scale physical systems