一种基于概率盒—HGWO优化SVM的滚动轴承故障诊断方法

Application of the P-box theory and HGWO-SVM in the fault diagnosis of rolling bearings

针对滚动轴承故障振动信号在特征提取时出现的信息丢失、误动等不确定性问题以及故障诊断准确性不理想的问题,提出了一种基于概率盒理论和改进灰狼算法(grey wolf optimization,GWO)优化支持向量机(support vector machine,SVM)的混合智能机械故障诊断方法。利用直接建模的方法得到概率盒,再采用累积不确定性测量方法提取其特征,构建出用于故障诊断的特征向量集;利用改进的灰狼算法对支持向量机进行优化;利用优化后的支持向量机实现对特征集的分类诊断。所提方法充分利用了概率盒在处理不确定性问题的优势和支持向量机在解决小样本、非线性模式识别中优秀的分类性能,可对不同故障类型的振动信号进行更加精准的辨识。通过对滚动轴承振动信号的试验验证与对比试验分析表明,该方法在滚动轴承故障诊断方面具有一定的有效性。

A hybrid intelligent mechanical fault diagnosis method based on the probability box theory and improved grey wolf optimization(GWO)to optimize the support vector machine(SVM)was proposed to solve the problems of information loss,misoperation and other uncertainties in feature extraction of rolling bearing fault vibration signals and also the problem of poor accuracy of fault diagnosis.Firstly,the probability box was obtained by the direct modeling method,and the features were extracted by the cumulative uncertainty measurement method to construct the feature vector set for fault diagnosis.Then,the improved grey wolf optimization was used to optimize the support vector machine.Finally,the feature set was classified and used to diagnose by adopting the optimized support vector machine.The proposed method makes full use of the advantages of the probability box in dealing with uncertain problems and the excellent classification performance of the support vector machine in solving small sample and nonlinear pattern recognition,so that different fault types of vibration signals can be more accurately identified.The experimental verification and comparative analysis of rolling bearing vibration signals show that the method is effective in fault diagnosis of rolling bearings.