基于迁移VPMCD的滚动轴承故障诊断方法

Fault diagnosis method of roller bearing based on transfer VPMCD

针对滚动轴承诊断中难以获得大量故障样本的问题,拟结合迁移学习的思想,提出了一种基于迁移学习的多变量预测模型(TVPMCD)方法。该方法首先采用已知样本库建立基础变量预测模型(BVPM);然后利用少量的目标域已知样本更新基础变量预测模型,使得更新的基础变量预测模型能兼顾目标域已知样本的信息;同时,以目标域已知样本的判别误差最小为目标,剔除已知样本库中误识样本,建立迁移变量预测模型(简称TVPM);最后利用迁移变量预测模型对待测样本进行识别,从而可以有效地解决小样本的故障诊断问题。对滚动轴承数据的分析结果表明,适合于小样本的TVPMCD模式识别方法可以更快更准确地识别滚动轴承故障类型。

Aiming at the problem that it is difficult to obtain a large number of fault samples in roller bearing diagnosis,a transfer variable predictive model based class discriminate(TVPMCD)method is proposed combined with the idea of transfer learning in the paper.Firstly,the basic variable prediction model(BVPM)is established with the known sample base.Then a small number of known samples in the target domain is used to update the parameters of the BVPM,so that the updated BVPM can take the information of the known samples in the target domain to achieve the purpose of migration.Meanwhile,the minimum discrimination error of the known samples in the target domain is taken as the goal,and a transfer variable prediction model(TVPM)is established by eliminating the false samples in the known sample base.Finally,the TVPM is used to identify the tested samples,which can effectively solve the problem of small sample in fault diagnosis.The analysis results of roller bearings show that the TVPMCD pattern recognition method suitable for small samples can identify the fault types of rolling bearings more quickly and accurately.