地铁车辆齿轮箱故障诊断方法探究  

Analysis of Fault Diagnosis Methods for Gearboxes in Subway Vehicles

作　　者：周峻峰 Zhou Junfeng(Shenzhen Metro Operation Group Co.,Ltd.,Shenzhen,Guangdong 518000)

机构地区：[1]深圳地铁运营集团有限公司,广东深圳518000

出　　处：《现代工程科技》2024年第21期117-120,共4页Modern Engineering Technology

摘　　要：为探讨地铁车辆齿轮箱故障诊断方法,基于油液磨粒图像(oil abrasive particle image,OAPI)设计齿轮箱故障诊断系统,凭借图像采集模块获取故障参数,并依次完成图像预处理、边缘检测以及尺寸标定,结合融入元学习算法(model-agnostic meta-learning,MAML)的卷积神经网络(convolutional neural networks,CNN)模型,实现OAPI的图像分类、识别。研究结果表明,该方法相较于基于干预训练法的故障诊断技术,能够提高近23%的故障识别准确度。Analysis and discussion will be conducted on the fault diagnosis methods of subway vehicle gearboxes.A gearbox fault diagnosis system will be designed based on oil abrasive particle image(OAPI).The fault parameters will be obtained through an image acquisition module,and image preprocessing,edge detection,and size calibration will be completed in sequence.Combined with convolutional neural networks(CNN)models incorporating model agnostic meta learning(MAML)algorithms,OAPI image classification and recognition will be achieved.According to research findings,this method can improve fault recognition accuracy by nearly 23%compared to fault diagnosis techniques based on intervention training.

关 键 词：图像特征提取 齿轮箱故障诊断 卷积神经网络模型 油液磨粒图像 

分 类 号：U279.323[机械工程—车辆工程]