基于LS-SVM多分类器融合决策的混合故障诊断算法

PDF(1622 KB)

振动与冲击 ›› 2013, Vol. 32 ›› Issue (19) : 159-164.

论文

基于LS-SVM多分类器融合决策的混合故障诊断算法

李鑫滨，陈云强，张淑清

作者信息 +

Hybrid Fault Diagnosis Algorithm Based on Fusion Decision of Multiple LS-SVM Classifiers

Li Xinbin, Chen Yunqiang, ZHANG Shuqing

Author information +

文章历史 +

摘要

故障诊断的关键是特征向量提取和分类器的选择, 提出一种综合运用多特征提取和多分类器组融合决策的故障诊断算法. 多特征提取选择小波包变换、总体平均经验模式分解方法(Empirical Mode Decomposition, EEMD)和改进小波能熵方法, 得到三组不同的故障特征信息; 将这三组特征信息输入由3个最小二乘支持向量机(Least Square Support Vector Machine, LS-SVM)组成的分类器组进行初步诊断; 采用自整定权值的决策模板法(Self-adjusting weighted Decision Templates, SWDT)进行多分类器诊断结果的融合决策. 实验证明, 该方法能实现轴承不同故障类型, 尤其是复合故障的可靠识别, 验证了该算法提取轴承故障特征信息的完备性, 以及分类器组融合决策的可靠性.

Abstract

The fault diagnosis key were the feature extraction and the classifier selection. This paper presented a hybrid diagnosis algorithm, which used features extraction and the fusion decision of multiple classifiers. The feature extraction method included the wavelet packet transform, EEMD, and the improved wavelet energy entropy, and each method may extract the feature information respectively. The features information was input to the classifiers group, which was consist of three LS-SVM classifiers, to make the initial diagnosis. SWDT was chosen to make the fusion decision of the diagnosis results. The experiments indicate that the method realize the reliable identification of different bearing fault, even the compound fault. It confirm the completeness of the different features information, and more reliability of the classifiers group fusion decision.

导出引用

李鑫滨;陈云强;张淑清. 基于LS-SVM多分类器融合决策的混合故障诊断算法[J]. 振动与冲击, 2013, 32(19): 159-164

Li Xinbin;Chen Yunqiang;ZHANG Shuqing. Hybrid Fault Diagnosis Algorithm Based on Fusion Decision of Multiple LS-SVM Classifiers[J]. Journal of Vibration and Shock, 2013, 32(19): 159-164