基于改进多项式响应面的VPMCD方法及其在滚动轴承故障诊断中的应用

摘要

基于变量预测模型的模式识别（Variable predictive model based class discriminate，简称VPMCD）方法在训练过程中是用多项式响应面（Polynomial Response Surface，简称PRS）法来建立预测模型的，然而PRS法的模型拟合精度不能随训练样本容量的增加而显著提高。针对这一缺陷，将原方法中的PRS方法进行了改进，提出了基于改进多项式响应面（Improved Polynomial Response Surface，简称IPRS）的VPMCD方法，并将其应用于滚动轴承故障诊断。通过实验，将原方法和改进方法在训练样本容量不同情况下的模式分类精度进行对比，结果表明，相对于原VPMCD 方法，改进的VPMCD方法不仅具有更好的模式分类效果，而且其分类精度随训练样本容量的增加提高得更明显。%In training process of the variable predictive model-based class discrimination (VPMCD)approach,the predictive model is built with the polynomial response surface (PRS)method.However,the model fitting precision with PRS can’t be significantly improved with increase in training samples.Aiming at this defect,the PRS method was improved here and the method of VPMCD based on improved polynomial response surface (IPRS)was proposed and used to diagnose rolling bearing faults.Through tests,the pattern classification accuracies of the original method and the improved method were compared under the condition of different training samples.The results showed that compared with the original VPMCD method,the improved VPMCD method not only has a better ability of pattern classification,but also its classification precision rises more obviously with increase in training samples.