高速列车齿轮箱箱体材料拉伸损伤的声发射信号表征识别及寿命预测

Characteristic Identification and Life Prediction for Acoustic Emission Signal of Tensile Damage for High-Speed Train Gearbox Shell Materials

为保障高速列车运行安全,需要对其齿轮箱箱体材料的拉伸损伤进行实时无损监测,传统的力学性能试验不能满足这种要求。为此,利用声发射技术,针对某型号高速列车齿轮箱箱体材料进行拉伸试验,采集拉伸过程中的声发射信号进行参数分析,并利用支持向量机(SVM)分类器对提取的特征参数进行识别分类。在此基础上改进SVM分类器,应用加权支持向量机(WSVM)方法有效减少SVM分类器的误判,并通过研究声发射特征值与拉伸寿命之间关系的规律,建立齿轮箱箱体材料拉伸过程的退化模型。结果表明:声发射信号的对数撞击计数增长速率和对数幅值增长速率可以较好地表征材料拉伸过程所处的阶段,可用于对箱体材料拉伸过程的损伤识别;应用WSVM方法使不平衡数据分类准确率提升至94%以上;建立的退化模型实现了对高速列车齿轮箱箱体材料在拉伸过程中的损伤识别,以及对其剩余寿命的预测。

In order to ensure the safe operation of highspeed railway,the real time nondestructive monitoring was needed for the tensile damage of highspeed train gearbox shell materials,but the traditional mechanical property test could not meet this requirement.Therefore,the acoustic emission technology was used to conduct the tensile test on the material of a certain type of highspeed train gearbox shell,and the acoustic emission signals during the tensile process were collected for parameter analysis.The support vector machine(SVM)classifier was used to identify and classify the extracted feature parameters.On this basis,the SVM classifier was improved,and the weighted support vector machine(WSVM)method was used to effectively reduce the misjudgment of the SVM classifier.By studying the relationship between acoustic emission eigenvalues and tensile life,the degradation model of the gearbox shell material during the tensile process was established.The results show that the stage of the material tensile process can be better characterized by the logarithmic impact count growth rate and the logarithmic amplitude growth rate of acoustic emission signals,which can be used to identify the damage of the box material during tensile process.The classification accuracy of unbalanced data is improved to more than 94%using WSVM method.The damage identification of the highspeed railway gearbox shell material in the tensile process and the prediction of its remaining life are realized by the established degradation model.