基于改进AFSA的参数优化VMD和ELM的轴承故障诊断

Bearing Fault Diagnosis Based on Parameter Optimized VMD with Improved AFSA and ELM

针对滚动轴承早期故障信号微弱、故障特征难以提取,导致故障分类效果差的问题,提出了一种基于改进人工鱼群(AFSA)进行参数优化的变分模态分解(VMD)和多特征向量融合的极限学习机(ELM)的滚动轴承故障诊断方法。首先,将改进后的AFSA对VMD算法的重要参数(分解个数K和惩罚因子α)进行自动寻优,适用度函数采用最小包络谱熵;其次,提取经AFSA-VMD分解后的包络谱熵最小的内蕴模态函数(IMF)分量作为最优分量,通过计算最优IMF分量的均方根值和峰值构造第一层特征值向量,计算最优IMF分量的样本熵、峭度和均方根构造第二层特征值向量;最后,将特征值向量送入极限学习机ELM进行滚动轴承故障的训练分类。试验结果表明本文算法具有良好的故障诊断效果且最终可实现98.25%的分类准确率和93.36%的实际诊断精度。

Aiming at the problem that the initial fault signal of rolling bearing is weak and the fault characteristic is difficult to extract,the paper proposed a method for rolling bearing fault diagnosis based on parameter optimized variational mode decomposition(VMD)with artificial fish swarm algorithm(AFSA)and extreme learning machine(ELM).Firstly,the improved AFSA optimizes the important parameters(decomposition number K and penalty factorα)of VMD algorithm,and the fitness function adopts the minimum envelope entropy.Secondly,the best intrinsic modal function(IMF)component after AFSA-VMD decomposition is extracted.The first-level eigenvector is constructed by calculating the root mean square and peak value of the optimal IMF component,and the second-level eigenvector is generated by computing the sample entropy,kurtosis and root mean square of the optimal IMF component.Finally,the eigenvectors are sent to the extreme learning machine ELM to train and classify the rolling bearing faults.The experimental results show that the algorithm in this paper which can finally achieve 98.25%classification accuracy and 93.36%actual diagnosis accuracy has an excellent fault diagnosis effect.