自回归和EWMA在齿轮箱早期异常检测中的应用  

Early Fault Diagnosis of Gearbox Based on Autoregression and EWMA

作　　者：李鑫 卢灿铭[1] 左洪福 柏宇星[1] LI Xin;LU Can-ming;ZUO Hong-fu;BAI Yu-xing(School of Automobile and Rail Transit,Nanjing Institute of Technology,Jiangsu Nanjing 211167,China;College of Civil Aviation,Nanjing University of Aeronautics and Astronautics,Jiangsu Nanjing 211106,China)

机构地区：[1]南京工程学院汽车与轨道交通学院,江苏南京211167 [2]南京航空航天大学民航学院,江苏南京211106

出　　处：《机械设计与制造》2024年第8期63-66,共4页Machinery Design & Manufacture

基　　金：中央高校基本科研业务费专项资金(NJ2020019);南京工程学院高层次引进人才科研启动基金资助项目(YKJ201843);国家自然科学基金重点基金项目(U193320003)。

摘　　要：针对齿轮箱传统故障分析方法难以及时发现早期故障的问题,进行基于自回归模型和统计过程控制方法的齿轮箱早期异常检测方法研究。采用时域同步平均算法(Time Synchronous Averaging,TSA)去除原始振动信号中的噪声,并建立自回归(Autoregressive Model,AR)模型计算残差;提取残差的标准差、峭度与均方根作为指标,初步判断齿轮早期故障;选取残差的标准差进行统计过程分析,分别建立均值控制图(Xbar)、单值-移动极差控制图(I-MR)、指数加权移动平均(EWMA)控制图来判断齿轮早期异常发生点。研究结果表明,EWMA控制图检测到第62个文件开始超出控制限,而其他两种控制图均从第65个文件出现异常,说明EWMA控制图相比其他方法能够更早的检测出齿轮早期故障发生的异常点,进一步验证了AR模型与EWMA控制图相结合进行早期异常检测的有效性。Aiming at the problem that traditional gearbox monitoring is difficult to find fault in time,the early abnormal detection of gearbox based on autoregressive model and statistical process control is studied.Time synchronous averaging(TSA)algorithm is used to remove the noise in the original vibration signal,then the autoregressive model(AR)is established and the residuals are obtained.The standard deviation,kurtosis and root mean square of the residuals are extracted to preliminarily judge the early fault of the gear.Based on the statistical analysis of residual standard deviation,the mean control chart,single value moving range difference control chart and exponential weighted moving average(EWMA)control chart are established respectively.The result shows that EWMA control chart detected the 62nd file began to exceed the control limit,and the other two control charts began to appear abnormal from the 65th file.The EWMA control chart detected abnormal points earlier,indicating that the combination of AR model and EWMA control chart can judge the early abnormal points more effectively.

关 键 词：时域同步平均 齿轮箱 早期故障诊断 自回归模型 统计过程控制 

分 类 号：TH16[机械工程—机械制造及自动化] TP18[自动化与计算机技术—控制理论与控制工程]