基于软性核凸包支持向量机算法滚动轴承故障诊断分析

Rolling Bearing Fault Diagnosis Analysis Based on Soft Kernel Convex Hull Support Vector Machine Algorithm

滚动轴承运行过程中形成局部故障如不及时做到诊断会造成严重后果。为了提高滚动轴承故障诊断精度,设计了一种软性核凸包支持向量机(SCH-SVM)测试方法。设计实验测试了滚动轴承故障的热成像结果,完成本方法的有效性验证。研究结果表明:提高软性因子后,诊断结果都表现为先升高再降低变化,将软性因子λ都设定在最优诊断分类状态为1.1。张量分类器SCH-SVM与STM相对向量分类器SVM与FCH获得了更优诊断性能,说明SCH-SVM对于机械故障诊断满足可靠性要求。逐渐增加训练样本数量后,形成了更相近的诊断精度,采用热成像作为输入时可以获得比振动信号作为输入时更高的精度。该研究能够有效提高滚动轴承故障诊断效率,可推广到其它的机械传动领域,具有很好的理论研究价值。

Local faults in the running process of rolling bearings will cause serious consequences if they are not diagnosed in time.In order to improve the fault diagnosis accuracy of rolling bearings,a soft kernel convex hull support vector machine(SCH-SVM)test method is designed.Experiments are designed to test the thermal imaging results of rolling bearing faults,and the validity of this method is verified.The results showed that after the soft factor was increased,the diagnosis results were increased first and then decreased,and the soft factor X was set at the optimal diagnostic classification state of 1.1.Compared with vector classifiers SVM and FCH,tensor classifiers SCH-SVM and STM obtain better diagnostic performance,indicating that SCH-SVM meets the reliability requirements for mechanical fault diagnosis.When the number of training samples is gradually increased,more similar diagnostic accuracy is formed,and higher accuracy can be obtained when thermal imaging is used as input than when vibration signal is used as input.This research can effectively improve the fault diagnosis efficiency of rolling bearings,and can be extended to other mechanical transmission fields,which has a good theoretical research value.