Failure prognostics using improved kernel smoothing auxiliary particle filtering

A failure prognostics approach based on improved kernel smoothing auxiliary particle filtering (IKS-APF) is proposed for systems with multiple unknown parameters. Firstly, a gain factor and an acceleration factor are employed in the kernel smoothing APF to bi-directionally control the parameter variance and accelerate the parameter convergence. Secondly, the IKS-APF method is used to jointly estimate the states and parameters. In order to ensure the accuracy of parameter estimation and reduce its uncertainty, an adaptive control scheme for the particle variance of parameters is presented, combining the variance monitoring and the short-term prediction errors. Finally, iterative prediction is implemented based on the latest estimated state and parameter particles, and then the remaining useful life (RUL) is calculated by the time of each propagated state particle first entering the failure zone. Simulation results demonstrate the effectiveness and superiority of the proposed approach.