铁道车辆扁疤识别与定量估计
李大柱1,吴兴文2,池茂儒1,梁树林1,许文天1
(1.西南交通大学 牵引动力国家重点实验室,四川 成都 610031;2.西南交通大学 机械工程学院,四川 成都 610031)
摘要:扁疤是铁道车辆车轮踏面常见的故障之一,扁疤的出现对列车正常运行有较大的危害,目前列车车轮无损检测以人工扫描为主,此方法检修周期长且效率低。为实现对车轮踏面扁疤的实时监测,本文提出了一种基于奇异值差分谱降噪与BP神经网络相结合的方法。该方法可通过轴箱振动加速度来识别车轮扁疤故障并对扁疤长度进行定量估计。首先对采集到的轴箱振动加速度信号进行奇异值差分谱降噪、包络、快速傅里叶变换,根据频谱中是否存在与车轮扁疤引起轴箱振动加速度特征频率相关的频率成分来定性的识别车轮扁疤故障。若诊断存在扁疤故障则从频谱图中提取该速度下车轮扁疤引起轴箱振动加速度特征频率的1~4倍频所对应的幅值,将车速以及提取出的幅值输入到训练后的BP神经网络模型中来对扁疤长度估计。通过仿真实验验证该方法能够快速准确地识别出车轮扁疤故障,且对车轮扁疤长度估计误差在3.5 mm内。
关键词:车轮扁疤故障;奇异值差分谱降噪;BP神经网络;轴箱振动加速度;幅值;特征频率 
中图分类号:U279.2 文献标志码:A doi:10.3969/j.issn.1006-0316.2022.06.006
文章编号:1006-0316 (2022) 06-0039-07
Detection and Quantitative Estimation of Railway Wheel-Flats
LI Dazhu1,WU Xingwen2,CHI Maoru1,LIANG Shulin1,XU Wentian1
( 1.State Key Laboratory of Traction Power, Southwest Jiaotong University, Chengdu 610031, China; 2.School of Mechanical Engineering, Southwest Jiaotong University, Chengdu 610031, China )
Abstract:Wheel flat is one of the common faults of wheel tread of railway vehicles. It does great harm to the operation of the train. At present, the nondestructive testing of train wheels mainly lies on manual scanning, which has the problem of long maintenance cycle and low efficiency. In order to realize the real-time monitoring of wheel flat, a method based on singular value difference spectrum noise reduction and BP neural network is proposed. This method can detect wheel flat by axle box vibration acceleration and quantitatively estimate the length of it. Firstly, the collected axle box vibration acceleration signal is subjected to singular value difference spectrum noise reduction, envelope and fast Fourier transform. The wheel flat is qualitatively identified according to whether there is a frequency component related to the characteristic frequency of wheel flat in the spectrum. If a wheel flat is detected, the amplitude corresponding to the 1~4 times of the characteristic frequency of the wheel flat at this speed is extracted from the spectrum diagram, and the vehicle speed and the extracted amplitude are input into the trained BP neural network model to estimate the length of it. The simulation results show that this method can quickly and accurately detect the wheel flat, and the estimation error of the wheel flat length is within 3.5 mm.
Key words:wheel flat;singular value difference spectrum noise reduction;BP neural network;axle box vibration acceleration;amplitude;characteristic frequency
———————————————
收稿日期:2021-11-25
基金项目:国家自然科学基金(51805450);国家重点研发计划(2018YFE0201401-01);中国科协青年人才托举工程项目(2019QNRC001);四川省基础研究计划(2020YJ0075)
作者简介:李大柱(1996-),男,甘肃武威人,硕士研究生,主要研究方向为铁道车辆智能运维,E-mail:lau_27345145539@163.com。
 

 

设为首页  |  加入收藏    |   免责条款
《机械》杂志版权所有     Copyright©2008-2012 Jixiezazhi.com All Rights Reserved 

  电话:028-85925070    传真:028-85925073    E-mail:jixie@vip.163.com

地址:四川省成都锦江工业开发区墨香路48号   邮编:610063

蜀ICP备08103512号

Powered by PageAdmin CMS