改进VPMCD法及其在机械故障诊断中的应用

提出了一种基于时序AR模型的VPMCD(基于变量预测模型的模式识别)故障诊断方法:利用时序分析方法对故障信号建立AR模型，以蕴含故障特征的自回归参数作为故障特征量,采用VPMCD方法训练得到各故障特征量的预测模型,并利用预测模型对待诊断样本的故障类型和工作状态进行分类和识别。对滚动轴承和齿轮的振动信号的分析结果证明了该方法的有效性，与基于EMD的VPMCD法和基于AR的KNN法的对比结果证明了所提方法的优越性。

A Pattern Recognition Method Based on Fusion of Time Series Analysis with VPMCD and Its Application in Machinery Fault Diagnosis

A pattern recognition method was proposed herein based on fusion of time series analysis AR model with VPMCD for fault  diagnosis. AR model of fault signals was established by using time series analysis,taking its autoregressive parameters that contain the fault features as the fault characteristic values, fusing VPMCD training to get the prediction models of fault characteristic values,and by using these predictive models to classify and recognize the faults  of sample types and working states.Analyses of rolling bearings  and gear vibration signals show the effectiveness of this method, comparison of the diagnosis method based on fusion of empirical mode decomposition(EMD) with VPMCD shows the superiority of this method.