基于LMD和SSA-SVM的电机故障诊断

Motor Fault Diagnosis Based on LMD and SSA-SVM

针对电机故障诊断问题,尤其电机轴承方面的诊断,提出了LMD分解和麻雀搜索优化算法(SSA)优化支持向量机(SVM)的故障诊断方法。第一步采取小波降噪和LMD算法相结合去处理原始信号,经过小波降噪后的原始故障信号会去掉一部分的干扰,再分解得到原始信号的一系列PF分量,接着使用相关性分析法选择出有效的PF分量进行信号重构,重构后的故障信号再次经过LMD分解得到的PF分量求出各自的能量熵,直接用能量图展现出来。接着将各个PF分量的能量熵组成一组组特征向量输入到支持向量机的故障诊断模型里。利用麻雀搜索算法在支持向量机(SVM)对于电机故障的分类的模型上进行惩罚参数和核参数的挑选和模拟,选择最合适的参数组合建立SSA-SVM故障诊断模型进行仿真实验,通过仿真实验验证该方法的故障诊断准确率高达99.2%,与PSO-SVM和SVM故障诊断模型进行比较分析,实验证明提出来的方案有着更适合的故障识别能力,对电机故障诊断有着很好的适应性和发展性。

A fault diagnosis method based on LMD decomposition and sparrow search optimizing algorithm(SSA)optimizing support vector machine(SVM)was proposed for motor fault diagnosis,especially for motor bearing diagnosis.The first step was to take the combination of wavelet noise reduction and LMD algorithm to process the original signal.After wavelet noise reduction,part of the interference of the original fault signal was removed,and then a series of PF components of the original signal was decomposed.Then,the correlation analysis method was used to select the effective PF components for signal reconstruction.The PF components of the reconstructed fault signals were decomposed by LMD again to calculate their energy entropies,which were directly displayed by the energy diagram.The energy entropy of each PF component was formed into a group of eigenvectors and input into the fault diagnosis model of the support vector machine.The sparrow search algorithm was used to select and simulate penalty parameters and kernel parameters on the classification model of motor faults by support vector machine(SVM).The most suitable parameter combination was selected to establish the SSA-SVM fault diagnosis model for simulation experiment.The simulation experiment verified that the fault diagnosis accuracy of this method was as high as 99.2%.Compared with PSO-SVM and SVM fault diagnosis model,the experiment proves that the proposed method has more suitable fault identification ability,and has good adaptability and development for motor fault diagnosis.