Unknown Fault Diagnosis for Nonlinear Hybrid Systems Using Strong State Tracking Particle Filter

A strong state tracking particle filter (SST-PF) is put forward for unknown fault diagnosis of hybrid system. SST-PF overcomes the problem of sample impoverishment for tracking the state of nonlinear hybrid system by setting permanent transition probabilities from one mode to another. Meanwhile threshold logic of normalization factor based on the statistics is built to detect unknown-faults, which is more accurate and reasonable for tiny mode differences of hybrid system. Simulation experiments are carried out to analyze the effects of SST-PF, and it is shown that our algorithm has strong tracking ability for states and pretty detection ability for both known and unknown faults.