Fault diagnosis based on intelligent particle filter

Abstract

Practical production systems are usually complex, nonlinear and non-Gaussian. Different from some other fault diagnosis methods, particle filter can applied to nonlinear and non-Gaussian systems effectively. The particle impoverishment problem exists in the traditional particle filter algorithm, which influences the results of state estimation. In this paper, we conclude that the general particle impoverishment problem comes from the impoverishment of particle diversity by analyzing the particle filter algorithm. We then design an intelligent particle filter(IPF) to deal with particle impoverishment. IPF relieves the particle impoverishment problem using the genetic strategy. In fact, the general PF is a special case of IPF relieves the particular parameters. Experiment on 160 MW unit fuel model shows that the intelligent particle filter can increase the particles diversity and improve the state estimation results.