A novel fault prognostic approach based on particle filters and differential evolution

This paper proposes an improved fault prognostic approach based on a modified particle filter with a built-in differential evolution characteristic. The main methodological contribution of this study is to handle the problem of sample impoverishment faced by particle filters when only a few particles are resampled. This is done by incorporating modified mutation and selection operators for differential evolution into the proposed particle filter. The proposed method is performed to deal with two real applications of condition monitoring and fault prognosis, namely an accelerated degradation of bearings under operating conditions from the platform PRONOSTIA and a high-speed computer numerical control (CNC) milling machine 3-flute cutters.