SOM与EWMA在滚动直线导轨故障预测中的应用  

Application of SOM and EWMA in Fault Prediction of Rolling Linear Guideway

作　　者：钟健康 陈元华[1] 张瑞宾[1] ZHONG Jiankang;CHEN Yuanhua;ZHANG Ruibin(School of Automobile Engineering, Guilin University of Aerospace Technology, Guilin 541004, Guangxi, China)

机构地区：[1]桂林航天工业学院汽车工程学院,广西桂林541004

出　　处：《机械科学与技术》2022年第2期278-287,共10页Mechanical Science and Technology for Aerospace Engineering

基　　金：2019年度广西高校中青年教师科研基础能力提升项目(2019KY0819)。

摘　　要：提出了一种基于机器学习的滚动直线导轨故障预测集成方法。首先,通过寿命试验,对由三轴加速度传感器采集的振动信号进行小波包分解,提取分部能量作为信号特征;其次,运用提取的特征训练自组织映射(Self-organizing map,SOM)神经网络,应用训练后的SOM识别线轨健康状态;最后,使用最小量化误差与指数加权移动平均控制图(Exponentially weighted moving-average,EWMA)实现动态故障预警。该方法将SOM与小波包分解相结合,选用最小量化误差构建EWMA控制图,解决了线轨状态监测可视化与疲劳程度数值评定问题,验证了该集成方法用于直线导轨故障预测的有效性。This paper presents a machine learning based fault prediction integration method for rolling linear guideway.Firstly,through the life test,the vibration signals collected by the triaxial acceleration sensor have been decomposed by wavelet packet,and the partial energy has been extracted as the signal feature.Secondly,the SOM(self organizing map)neural network has been trained by the extracted feature,and the trained SOM has been applied to identify the health status of the railway.Finally,the minimum quantization error and EWMA(exponential weighted moving average control chart)have been used to realize dynamic fault early warning.This method combines SOM with wavelet packet decomposition,selects the minimum quantization error to build EWMA control chart,solves the problems of visual monitoring and numerical evaluation of fatigue degree,and the effectiveness of the integrated method for fault prediction of linear guideway was verified.

关 键 词：滚动直线导轨 故障预测 机器学习 小波包分解 自组织映射 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]