基于噪声辅助多元经验模态分解和多尺度形态学的滚动轴承故障诊断方法

摘要

A rolling bearing fault diagnosis method was proposed based on the noise assisted multivariate empirical mode decomposition (NAMEMD)and the mathematical morphology.NAMEMD,as a noise assisted data analysis-based method,can effectively avoid shortcomings of ensemble empirical mode decomposition,such as,mode mixing and heavy computation,thus it is superior to the traditional noise assisted data analysis-based method to a certain extent.Here, NAMEMD was combined with the multiscale morphology to be used for rolling bearing fault diagnosis.NAMEMD was used to adaptively decompose multi-component FMand AMfault signals into a series of IMF components,the high-energy IMFs were selected to be summed for signal reconstruction.Then a multiscale morphological difference filter was employed to extract the fault characteristic frequency of signals.In order to verify the correctness of the proposed method,simulation tests and bearing fault ones were performed,the results were compared with those of EEMD and envelope demodulation-based methods.It was shown that the proposed method can further alleviate mode mixing effects,significantly improve the computation speed,bring about higher detection accuracy for the faults in outer race,inner race and roller in rolling bearings,and clearly extract the characteristic frequencies of fault signals.%为了从强噪背景中提取滚动轴承微弱故障特征，提出一种基于噪声辅助多元经验模态分解（Noise Assis-ted Multivariate Empirical Mode Decomposition，NAMEMD）和数学形态学的滚动轴承故障诊断方法。NAMEMD 是新提出的一种基于噪声辅助数据分析方法，其克服了集成经验模态分解的模态混淆和运算量大等问题。将 NAMEMD 与多尺度形态学相结合应用于滚动轴承故障诊断。该方法首先利用 NAMEMD 将多分量调频调幅故障信号自适应分解为一系列 IMF分量；其次，选取能量高的 IMF 分量求和重构；最后利用多尺度形态学差值滤波器提取信号的故障特征频率。为了验证理论的正确性，进行了仿真试验和轴承故障试验，并与 EEMD 和包络解调进行了比较，结果表明该方法在进一步降低模态混叠效应的同时，明显提高了运算速度，对滚动轴承外圈、内圈和滚子故障的检测精度更高，能够清晰地提取出故障信号的故障特征频率。