摘要
传统的机械驱动主轴异常检测方法通常仅提取主轴状态数据的单一特征,导致检测效果不佳,为此研究车辆段不落轮镟床作业机械驱动主轴异常检测方法。首先利用传感器采集主轴状态数据,并对其进行归一化操作和滤波处理;然后对预处理后的状态数据进行特征提取,并利用代价函数进行特征分类,以筛选出异常特征;最后通过计算待测数据与异常特征数据间的相似度,实现对机械驱动主轴的异常检测。实验结果表明,该方法在实际应用中ROC曲线的面积大,异常检测效果好。
Traditional mechanical driven spindle anomaly detection methods usually only extract a single feature of spindle state data,resulting in poor detection performance.To this end,a method for detecting abnormalities in the mechanical drive spindle of the non falling wheel lathe operation in the vehicle depot is studied.Firstly,it uses sensors to collect spindle status data,and normalize and filter it.Secondly,feature extraction is performed on the preprocessed state data,and cost functions are used for feature classification to filter out abnormal features.Finally,anomaly detection of the mechanical drive spindle is achieved by calculating the similarity between the tested data and abnormal feature data.The experimental results indicate that the design method has a larger area of ROC curve and better anomaly detection performance in practical applications.
作者
陈津龙
石航
Chen Jinlong;Shi Hang(Suzhou Rail Transit Group Co.,Ltd.,Jiangsu Suzhou,215004,China;China Railway Siyuan Survey and Design Group Co.,Ltd.,Hubei Wuhan,430063,China)
出处
《机械设计与制造工程》
2023年第12期83-86,共4页
Machine Design and Manufacturing Engineering
关键词
车辆段不落轮镟床作业
机械驱动主轴
主轴异常
异常检测方法
方法设计
vehicle non falling wheel lathe operation
mechanical drive spindle
abnormal spindle
anomaly detection methods
method design