Integrated system-level prognosis for hybrid systems subjected to multiple intermittent faults

Considering the condition that multiple intermittent faults occur sequentially or simultaneously in hybrid systems, this paper proposes an integrated system-level prognosis scheme. The scheme is built by integrating model-based fault estimation and data-driven system-level remaining useful life (RUL) prediction. As for fault estimation, the level-based learning swarm optimization is utilized to identify state/value changes of intermittently faulty components based on the constructed hybrid system model. With estimation results, the system-level prognosis framework can be designed as follows: 1) The abnormal duration index (ADI) is established to quantify the system degradation extent under multiple intermittent faults. In view of the stochasticity of intermittent fault appearance and disappearance, the ADI is assessed approximately by as pessimistic as possible and as optimistic as possible indicators. 2) A synthetical degradation evaluator is developed to determine the moment to activate predictor, which ensures the obtained fault features can sufficiently reflect the evolutionary trend of the intermittently faulty component when activating RUL predictor. 3) Considering sequential or simultaneous intermittent faults and variations of modes for hybrid systems, the system-level prognosis module is designed, where the optimized support vector regression is adopted for system RUL prediction. Finally, the experimental study is investigated to verify the proposed scheme.