Scheduled Health Monitoring of Hybrid Systems With Multiple Distinct Faults

Abstract

Health monitoring of hybrid systems has attracted substantial attention in recent years. However, two issues exist in the monitored hybrid systems which could add to the difficulty of algorithm development: First, multiple faults with unknown distinct nature may occur simultaneously; and second, fault detection time and mode change time might be negatively correlated which obstructs the completion of fault estimation. To deal with these two problems, a scheduled health monitoring (SHM) method is developed in this paper. First, a unified fault modeling that enables the fault estimation, when the fault nature is unknown, is proposed. After that, to tackle the problem that the fault estimation cannot be completed when the time of fault detection is close to the following mode change time which makes the fault undetectable, a scheduling method is developed where the activation of fault estimator is controlled by the scheduler. For the purpose of fault estimation and prognosis under the SHM framework, a diversity enhanced particle filter (PF) is developed which alleviates the sample impoverishment problem in the traditional PF. Finally, experimental investigations are conducted to validate the proposed methodology.