ABSTRACT
A Hybrid Dynamical System (HDS) includes a set of continuous dynamics in which a particular continuous dynamics is activated at a particular set of discrete events, termed as mode of the system. Thus, a different mode triggers a different continuous dynamics. Degradation evolutions of the components of HDS depend on the operating mode of the system. Thus, the existing approaches for continuous systems are not suited for Remaining Useful Life (RUL) prediction for HDS. In addition, a discrete mode fault may be possible besides the parametric faults (abrupt or progressive nature). This article presents an integrated approach to Fault Diagnosis (FD) and RUL prediction of multiple deteriorating components in an HDS. For improving FD scheme, dynamic fault signature matrices are utilized for parametric and discrete mode fault isolation, which minimize the possible suspected faults by using the possible deviation direction of the faults. If the detected fault is progressive in nature, then the FD scheme is further utilized to point out the severity change points of the degradation. Using the knowledge of each severity change points and the deviation direction of progressive fault, constrained parameter estimation method with dynamically updated parameter’s bound is proposed for fast degradation states estimation. The estimates of the degradation at different time instances in a respective operating mode are further utilized for mode-dependent degradation model identification and RUL prediction. An online degradation model selection scheme is proposed for degradation model identification in different operating modes. The proposed method is able to identify the degradation model of multiple degrading components in a real time at different operating modes and can be adapted with new information of their degradation states estimated by the constrained parameter estimation during continuous monitoring. The proposed approach is demonstrated through numerical simulation of an example hybrid dynamical system.
