Research on wheel-rail vertical load detection based on digital speckle pattern

被引：0

作者：

Chen, Jingwei ^{[1
]}

Jiang, Man ^{[1
]}

Yang, Yue ^{[1
,2
]}

机构：

[1] School of Traffic and Transportation Engineering, Central South University, Changsha,410075, China

[2] The State Key Laboratory of Heavy Duty AC Drive Electric Locomotive Systems Integration, CRRC Zhuzhou Locomotive Co. Ltd., Zhuzhou,412000, China

来源：

Journal of Railway Science and Engineering | 2024年 / 21卷 / 02期

关键词：

Learning systems - Numerical methods - Rails - Speckle - Stresses;

D O I：

暂无

中图分类号：

学科分类号：

摘要：

Fast and effective detection of wheel/rail vertical load is of great significance for ensuring the reliability and safety of rail vehicles in service. Although existing wheel/rail load detection methods can effectively detect wheel/rail vertical loads, their measurement accuracy is limited by sensor deployment schemes and system calibration, and can only achieve fixed point detection. Therefore, this study introduced digital speckle image correlation technology to provide a new technical solution for rapid and non-fixed point detection of wheel/rail vertical loads. In order to achieve non-contact and rapid measurement of the stress field on the rail side, a sequence of rail speckle images under vertical load was synchronously collected by left and right cameras, and the sub-region pixel point matching and stress field calculation before and after deformation were carried out through image related theoretical models. In order to identify the wheel/rail vertical load using the rail side stress distribution obtained by digital speckle detection, a wheel/rail vertical load identification method based on the Extreme Learning Machine (ELM) was proposed. This method used Workbench finite element software to establish a numerical model of the steel rail. Through the numerical simulation results, areas that are relatively sensitive to stress changes were selected as the interest domain for wheel/rail vertical load detection. Based on the stress field dataset and corresponding load dataset, ELM network parameters were designed to achieve automatic recognition of wheel/rail vertical load. To verify the effectiveness of the proposed wheel/rail vertical load identification model, a digital speckle detection experimental platform for wheel/rail vertical load based on XT-DIC was established. The experimental results show that the ELM model constructed by inputting mode stress in the Y and Z directions has the best recognition performance, with a vertical load prediction error of only 5.357%. The research results provide a new and reliable way approach for non fixed point rapid detection of wheel/rail vertical load. © 2024, Central South University Press. All rights reserved.

引用

页码：851 / 859