基于布谷鸟搜索算法和最大二阶循环平稳盲解卷积的滚动轴承故障诊断方法

Research on Fault Diagnosis Method of Rolling Bearings Based on Cuckoo Search Algorithm and Maximum Second Order Cyclostationary Blind Deconvolution

针对最大二阶循环平稳盲解卷积(Maximum second order cyclostationary blind deconvolution,CYCBD)的轴承故障诊断效果取决于选取的故障特征频率的精度以及滤波器的长度的问题,提出了用布谷鸟搜索算法(Cuckoo search algorithm,CSA)优化CYCBD,并以改进的最大谐波显著性指标(Improved maximum harmonic significance index,IHSI)为优化依据的诊断方法。该方法首先要预估故障特征频率以及滤波器长度的搜索范围,然后利用CSA比较不同故障特征频率以及滤波器长度下解卷积信号的IHSI值,并选取最大IHSI值对应的故障特征频率和滤波器长度作为CYCBD的输入参数,最后对解卷积后的信号进行平方包络来提取故障特征。仿真和实验结果表明,CSA能够高效地寻找出精确的故障特征频率以及合适的滤波器长度,从而确保CYCBD的解卷积效果,而CYCBD与最小熵解卷积(Minimum entropy deconvolution,MED)、最大相关峭度解卷积(Maximum correlation kurtosis deconvolution,MCKD)的比较显示,CYCBD拥有更强的故障特征提取能力。

Aiming at the problem that the effectiveness of bearing fault diagnosis for maximum second order cyclostationary blind deconvolution(CYCBD) depends on the accuracy of the selected fault feature frequency and the length of the filter, we propose a diagnosis method that uses cuckoo search algorithm(CSA) to optimize CYCBD and uses the improved maximum harmonic significance index(IHSI) as the optimization basis. Firstly, The search range of fault feature frequency and filter length is estimated. then, CSA was used to compare the IHSI values of the deconconvolution signals at different fault feature frequencies and filter lengths, and the fault feature frequencies and filter lengths corresponding to the maximum IHSI values were selected as the input parameters of CYCBD. Finally, The square envelope of the deconvolve signal is used to extract fault features. Simulation and experiment show that CSA can efficiently find the accurate fault characteristic frequency and the appropriate filter length to ensure the deconvolution effect of CYCBD, and the comparisons of CYCBD with minimum entropy deconvolution(MED) and maximum correlation kurtosis deconvolution(MCKD) show that CYCBD has stronger fault feature extraction capability.