基于小波包分解与广义隐马尔科夫模型的机车牵引座裂纹状态识别  

Crack state identification of the locomotive’s traction seat based on wavelet packet decomposition and the generalized hidden Markov model

作　　者：谢锋云 闫少石 冯春雨 王二化[2] 刘翊 肖乾 XIE Feng-yun;YAN Shao-shi;FENG Chun-yu;WANG Er-hua;LIU Yi;XIAO Qian(School of Mechatronics and Vehicle Engineering,East China Jiaotong University,Nanchang 330013;School of Mechatronics Engineering,Changzhou College of Information Technology,Changzhou 213164;CSR Zhuzhou Electric Locomotive Co.,Ltd.,Zhuzhou 412000)

机构地区：[1]华东交通大学机电与车辆工程学院,江西南昌330013 [2]常州信息职业技术学院机电工程学院,江苏常州213164 [3]中车株洲电力机车有限公司,湖南株洲412000

出　　处：《机械设计》2022年第6期35-41,共7页Journal of Machine Design

基　　金：国家自然科学基金资助项目(51805168,51565015);江西省教育厅项目(GJJ180301,GJJ190307);常州高技术重点实验室项目(CM20183004)。

摘　　要：牵引座作为连接机车车体和转向架、承受传递机车纵向力的重要部件,其状态影响着机车的安全。文中针对机车牵引座正常、小裂纹故障、大裂纹故障等3种状态,提出了一种基于小波包分解(WPD)与广义隐马尔科夫模型(GHMM)的状态识别方法:通过db5小波包分解方法对机车牵引座的振动信号进行特征提取,同时结合时域敏感特征组成时频敏感特征向量,由模态区间不确定性分析方法构建模态区间特征向量,最后通过模态区间与隐马尔科夫模型组成的广义隐马尔科夫模型进行状态识别。识别结果表明,针对牵引座的3种不同状态,广义隐马尔科夫模型识别方法相较于传统的隐马尔科夫模型有更高的识别率,而且由于输出结果的模态区间形式包含更多的信息,使机车牵引座裂纹状态结果更加可靠。The traction seat is an important part that connects the locomotive’s car body with the bogie and bears the transmi-ssion of the longitudinal force of the locomotive,and its state affects the vehicle’s safety.In this article,a method of state identification is proposed based on wavelet packet decomposition(WPD)and the generalized hidden Markov model(GHMM)for the three states of the locomotive’s traction seat:normal,small crack failure and large crack failure.the vibration signal of the locomotive’s traction seat is subject to feature extraction with the help of db5 wavelet packet decomposition and in combination with the time-domain sensitive feature,the time-frequency sensitive feature vector is identified;then,the modal interval’s feature vector is worked out by means of the analysis on the modal interval’s uncertainty.Finally,the state identification is carried out through the generalized hidden Markov model which is composed of the modal interval and the hidden Markov model.It’s shown that for the three different states of the traction seat,the identification results obtained from the generalized hidden Markov model have a higher identification rate than those obtained from the traditional hidden Markov model.Besides,for the output results,the form of the modal interval contains more information,so the results of crack state identification of the locomotive’s traction seat are more reliable.

关 键 词：牵引座 小波包 模态区间 广义隐马尔科夫模型 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程] TH133[自动化与计算机技术—控制科学与工程]