Improving one class support vector machine novelty detection scheme using nonlinear features

作者：

Highlights：

• A systematic preprocessing is proposed to improve oc-svm novelty detection scheme.

• The effect of load, fault type and fault intensity factors, are compared.

• Only the nonlinear features can classify data corresponding to nonlinear systems.

• The proposed method leads to a high efficiency oc-svm novelty detection scheme.

摘要

•A systematic preprocessing is proposed to improve oc-svm novelty detection scheme.•The effect of load, fault type and fault intensity factors, are compared.•Only the nonlinear features can classify data corresponding to nonlinear systems.•The proposed method leads to a high efficiency oc-svm novelty detection scheme.

论文关键词：Novelty detection,OC-SVM,Nonlinear feature,Wavelet,Bearing vibration signal,Entropy

论文评审过程：Received 5 November 2017, Revised 2 April 2018, Accepted 3 May 2018, Available online 5 May 2018, Version of Record 23 May 2018.

论文官网地址：https://doi.org/10.1016/j.patcog.2018.05.002