基于本征时间尺度分解和变量预测模型模式识别的机械故障诊断

基于变量预测模型的模式识别(variable predictive model based class discriminate,VPMCD)方法是一种充分利用特征值之间相互内在关系进行多分类模式识别的新方法。对VPMCD算法进行了研究,并采用交叉验证法来选择VPMCD模型。针对机械故障振动信号的特征值之间的相互内在关系,结合本征时间尺度分解(intrinsic time-scale decom-position,ITD),提出了一种基于本征时间尺度分解和VPMCD的机械故障诊断方法。该方法首先利用ITD方法将原始信号分解若干个PR(proper rotation,PR)分量,然后提取第一个PR分量的无量纲时域统计参数组成特征向量,最后采用VPMCD方法进行机械故障诊断。通过滚动轴承故障诊断实验验证了该方法能有效地应用于小样本多分类机械故障诊断。

Machine Fault Diagnosis Method Based on ITD and Variable Predictive Model Based Class Discriminate

Variable predictive model based class discriminate (VPMCD) is a new multivariate classification approach for pattern recognition applications, which takes full advantage of the inhere relation between the features .The generalized VPMCD algorithm was studied and the VPMCD model was selected via cross validation method in the paper. In view of inter-relations among the feature vectors of machine fault vibration signal, a new machine failure diagnosis method based on ITD (intrinsic time-scale decomposition, ITD) and VPMCD was proposed with the combination of the advantages of ITD method. Firstly, non-stationary original signal was decomposed into a set of proper rotation components. Secondly，five traditional time domain dimensionless parameters of first proper rotation component were extracted as eigenvectors .Lastly, VPMCD was exploited to identify fault type . It is demonstrated that the proposed method can be effectively applied to diagnose machine failure in case of small sample multivariate classification by fault diagnosis experiment of roller bearing.