On-line crack propagation prognostics using a resample-move particle filter

Accurate prognosis of crack propagation is of great importance for prognostics and health management (PHM) to improve the safety of aircraft structure and reduce the operation and maintenance cost. Nonetheless, fatigue crack propagation is a complicated random process influenced by several uncertainties such as intrinsic material properties, loading conditions and environment. The particle filter (PF) based method is suitable for nonlinear and non-Gaussian problems, so as to be one of the most advanced and powerful methods to solve crack propagation prognostic problems. Under the PF framework, the crack measurements are fused to estimate model parameters and the current crack state, so as to achieve on-line prognostics of the future moment. However, the basic PF usually suffers from particle diversity impoverishment, which reduces the prognostics performance. To solve this problem, an on-line crack propagation prognostics based resample-move particle filter is proposed in this paper, which increase the particle diversity by a moving step of crack propagation model parameters after resampling. The proposed method is verified on the fatigue test of aircraft attachment lugs.