关键词: 单类基于变量预测模型的模式识别, 本征时间尺度分解, 排列熵, 滚动轴承, 故障诊断

Abstract:

Variable predictive model based class discriminate (VPMCD) is a way to pattern recognitions. It made full use of the inner relations among characteristic values extracted from those original data to recognize models and classified the faults by minimum prediction error sum of squares value. Based on that, the paper proposed a new one-class classification method—OC-VPMCD and combined OC-VPMCD with ITD and applied into the rolling bear fault diagnosis. Firstly, rolling bearing vibration signals would be adaptively decomposed by ITD and the permutation entropy of proper rotations (PR) which contain the main fault information would be extracted as characteristic values. Secondly, OC-VPMCD classifier would be trained and determined the prediction error sum of squares threshold value. Finally, the OC-VPMCD classifier would be used to complete pattern recognitions; according to the pattern recognitions results the working states of the rolling bearing were judged. The experimental results show that this method can be applied to rolling bearing fault diagnosis effectively.

Key words: one-class variable predictive model based class discriminate(OC-VPMCD), intrinsic time-scale decomposition(ITD), permutation entropy, rolling bearing, fault diagnosis