基于信号局部均值分解的机械故障诊断研究

Research on mechanical fault diagnosis based on signal local mean decomposition

基于信号局部均值分解法拓展原始机械信号的两端,优化经典局部均值分解法的端点效应,避免分解后原始信号失真。阐述了旋转机械故障诊断的基本流程,引入Fast ICA(快速独立分析因子)联合降噪,并基于能量系数和偏度系数构建特征向量,用相关系数增强分量包络谱的可靠性,提高了故障诊断与识别的准确率。实验结果表明,提出的故障诊断方法的诊断准确率在特征识别与模式识别下均较高,同时可以将均方误差值控制在一个较低的水平。

Based on the signal local mean decomposition method,the two ends of the original mechanical signal are expanded to improve the endpoint effect of the classical local mean decomposition method,and avoid the distortion of the original signal after decomposition.The basic process of rotating machinery fault diagnosis is given,the fast-independent analysis factor is introduced to joint noise reduction,the eigenvector is constructed based on the energy coefficient and skewness coefficient,and the correlation coefficient is used to enhance the component.The reliability of envelope spectrum can improve the accuracy of fault diagnosis and recognition.Experimental results show that the diagnosis accuracy of the proposed fault diagnosis method has advantages in feature recognition and pattern recognition,and the mean square error can be controlled at a lower level.