Application of adaptive MCKD method optimized by SSA based on mixed strategy in rolling bearing fault diagnosis

Abstract

Because of the non-obvious periodic impulses interfered by noise, harmonics and unexpected pulses, fault character extraction of rolling bearing is a difficult problem. Maximum correlated kurtosis deconvolution (MCKD) needs a tight choice of parameters, and any improper choice may greatly reduce the fault diagnosis performance of MCKD. To improve the fault diagnosis performance of MCKD, an adaptive MCKD method optimized by salp swarm algorithm based on mixed strategy (MSSSA) named MSSSA-MCKD is proposed. MSSSA is an improved salp swarm algorithm (SSA) based on mixed strategy to remedy the defects that SSA is apt to trap in local optimum and converges slowly. Through the analysis of impact of different improved strategies on the performance of SSA, performance comparison with other optimization algorithms and performance comparison with other improved SSA, the following conclusion can be reached: The optimization accuracy and convergence speed of MSSSA are superior. Whether the parameters are reasonable directly determines the performance of MCKD. The filter length and displacement number of MCKD are adaptively selected by MSSSA. Then the simulated fault signal and experimental fault signals of rolling bearing were processed by MSSSA-MCKD. The validity of MSSSA-MCKD was proved by the comparison of spectral analysis, MCKD with random parameters, empirical mode decomposition (EMD) and MSSSA-MCKD. Finally, the following conclusion can be reached: MSSSA-MCKD can precisely extract fault characters of rolling bearing.

Keywords

• maximum correlated kurtosis deconvolution
• salp swarm algorithm based on mixed strategy
• rolling bearing
• fault diagnosis
• adaptive parameter selection