基于粒子群优化核主元分析的轴承状态识别

轴承状态识别的准确率与特征提取紧密相关,而特征提取对轴承状态识别显得尤为重要.因时频域的各个特征对不同程度的故障信号敏感度各不相同,特征提取不当将会造成状态识别准确率下降.针对上述问题提出粒子群优化(PSO)核主元分析(KPCA),并利用该方法对轴承的复合特征集进行特征提取,提取后的特征向量构成识别特征集,由优化的支持向量机识别分类.选用美国凯斯西储大学滚动轴承试验台的振动数据进行处理分析,通过3种实验方案进行验证.结果表明,提出的方法明显改善了轴承状态识别的准确率.

Bearing State Recognition Based on Kernel Principal Component Analysis of Particle Swarm Optimization

The accuracy of bearing state identification is closely related to feature extraction.Affected by the strong noise,feature extraction is very important for bearing state recognition.Because the fault signal characteristics of different degrees of sensitivity are different in the time frequency domain,improper feature extraction will decrease state recognition accuracy.To solve these problems,the composite feature of particle swarm optimization (PSO) kernel principal component analysis (KPCA) was proposed.The composite feature of the bearing was extracted to constitute the feature set.The state of the bearing was classified and recognized by optimized support vector machine.The vibration data of the rolling bearing test rig of Case Western Reserve University were analyzed and verified by three experimental schemes.The results show that the proposed method obviously can improve the accuracy of bearing state identification.