铁道标准设计2025，Vol.69Issue(6)：205-213,9.DOI:10.13238/j.issn.1004-2954.202309280002
基于Kriging代理模型的车轮多边形识别方法研究
Research on Wheel Polygon Recognition Method Based on Kriging Surrogate Model
摘要
Abstract
To address the limitations of current dynamic detection methods for wheel polygonal wear,such as the inability to quantitatively identify polygonal patterns and low recognition accuracy,this study proposed a novel identification method based on the Kriging Surrogate Model(KSM).The method leveraged the relationship between the vertical axle-box acceleration and the wheel roughness level to inversely estimate the roughness,thereby enabling a quantitative assessment of wheel polygonal wear severity.First,a vehicle dynamics model incorporating flexible wheels was established.Extensive simulations were conducted under various vehicle speeds,polygonal orders,and roughness levels,with polygonal wheels serving as excitation sources.A response surface was then constructed using the Kriging surrogate model,describing the vibration level of vertical axle-box acceleration as a function of speed and wheel roughness level for each polygonal order.Measured axle-box vibration levels and vehicle speed were input into the KSM-based response surface to predict the wheel roughness level.The severity of polygonal wear was subsequently assessed in accordance with ISO 3095:2013.Results showed that for higher-order polygonal wear,the predicted wheel roughness levels closely matched the actual values,with a minimum absolute error of 0.14 dB/μm and a maximum absolute error below 1.50 dB/μm.Moreover,the study revealed that the vibration level exhibited an approximately linear relationship with the wheel roughness level,indicating that vibration level was a reliable indicator for reflecting the interaction between axle-box acceleration and wheel surface conditions.In contrast,the relationship between vibration level,polygonal order,and speed was nonlinear and complex,suggesting that both factors should be considered when setting polygonal wear threshold criteria.This paper presents a method capable of quantitatively identifying wheel polygonal wear,which offers both theoretical insight and practical value for fault diagnosis in railway vehicle systems.关键词
铁道车辆/车轮多边形/Kriging代理模型/轴箱振动/车轮粗糙度级/识别方法Key words
railway vehicle/wheel polygon/Kriging surrogate model/axle box vibration/wheel roughness level/recognition methods分类
交通工程引用本文复制引用
戴成昊,许文天,彭佳宁,崔利通,梁树林,池茂儒..基于Kriging代理模型的车轮多边形识别方法研究[J].铁道标准设计,2025,69(6):205-213,9.基金项目
中国国家铁路集团有限公司科技研究开发计划重点课题(N2023J043) （N2023J043）
