Event-Triggered Discrete Component Prognosis of Hybrid Systems Using Degradation Model Selection

In this article, a new event-triggering mechanism based prognosis scheme is proposed for discrete components in hybrid systems using degradation model selection. As an advantage over existing prognosis methods for hybrid systems, a more challenging case that multiple discrete components suffer from intermittent faults with unknown degradations is investigated. By introducing a hybrid structure where a centralized structure is used for fault detection and isolation and a distributed structure is adopted for fault estimation, the fault isolation performance under multiple discrete faults condition can be enhanced using the IMCSM and the fault estimation can be implemented with less computational burden using the decomposed submodels. With the aid of tumbling window, the total duration of intermittent fault in tumbling window can be treated as the intermittent fault feature. Since the degradation model describing the feature evolution is unknown in practice and may vary with the usage condition, an event-triggered prognosis method is proposed where a degradation model selection method is developed to find the best fit model under various usage conditions and thus remaining useful life prediction can be achieved. Experimental results on an electrical hybrid system validate the developed scheme.

Automated summary

A new event-triggering mechanism based prognosis scheme is proposed for discrete components in hybrid systems using degradation model selection. The scheme introduces a hybrid structure for fault detection and isolation, and fault estimation, enhancing fault isolation performance under multiple discrete faults condition. The scheme also proposes an event-triggered prognosis method with a degradation model selection to achieve remaining useful life prediction.