基于EMD瞬时功率谱熵的神经网络滚动轴承故障诊断

滚动轴承在发生故障时,其动力学特性往往呈现出复杂性和非线性,振动信号也会随之表现出非平稳性。为此,提出一种基于EMD瞬时功率谱熵的滚动轴承特征提取方法。该方法将轴承信号进行EMD分解,得到有限个IMF分量,对这些分量进行功率谱处理,计算其功率谱的信息熵。EMD瞬时功率谱熵作为特征向量,采用神经网络进行故障分类,实验结果表明,此方法的分类准确率可达96.25%。

PNN Fault Diagnosis for Rolling Bearing Based on EMD and Instantaneous Power Spectral Entropy

Considering the fact that dynamic characteristics of faulted rolling bearings are complex and nonlin-ear and the fault signals show up non-stationarity,a rolling bearing feature extraction method based on empiri-cal mode decomposition (EMD)and instantaneous power spectral entropy was proposed.In which,having EMD adopted to decompose bearing signals into a finite number of IMF components,and then having these components processed with power spectrum and having information entropy of the power spectrum calculated. Taking the power spectrum entropy as the characteristic vector and then employing probabilistic neural network (PNN)to classify the failures into different types,the experimental results show that the classification preci-sion can reach 96.25%.