基于差分自回归—随机森林的动车组轮对旋修策略优化研究  

Research on Optimization of Wheel Repair Strategy Method Based on ARIMA-random Forest for EMU

作　　者：刘成 朱腾飞 王紫光 沙智华 张生芳 LIU Cheng;ZHU Tengfei;WANG Ziguang;SHA Zhihua;ZHANG Shengfang(School of Traffic and Transportation Engineering,Dalian Jiaotong University,Dalian 116028 Liaoning,China)

机构地区：[1]大连交通大学交通运输工程学院,辽宁大连116028

出　　处：《铁道机车车辆》2024年第5期132-139,共8页Railway Locomotive & Car

基　　金：辽宁省教育厅科学研究项目(LJKZ0506);大连市科技创新基金(2021JJ12GX009)。

摘　　要：基于动车组运行里程和轮对尺寸退化过程为非平稳时间序列的特点,将差分自回归移动平均模型(ARIMA)与随机森林算法相结合,对关键尺寸退化趋势影响下的轮对旋修策略优化进行研究。利用ARIMA对运行里程数据进行差分处理,运用基尼系数划分特征构建随机森林决策树,将轮对历史检测数据划分为训练集和测试集进行训练,以预测均值确定轮对尺寸预测值。以轮对几何尺寸和动力学性能为约束条件,以最长使用寿命、最少旋修次数和平稳性指标为优化目标,构建轮对旋修策略优化模型,并对轮对旋修量和旋修后轮径值进行预测。结果表明,当轮径旋修量为2.5 mm,轮缘厚度在HAi=28.5 mm和HBi=30 mm时旋修策略最佳,轮对寿命可提高31.4%。研究成果可为动车组轮对旋修策略优化提供理论支持。Based on the characteristics of EMU running mileage and the degradation of wheel as non-stationary time series,the CR400BF was used as the research object.The ARIMA and random forest algorithm are combined to carry out the wheel degradation prediction and lathing strategy optimization research.ARIMA is used to carry out differential processing on the running mileage data,construct random forest decision tree based on Gini coefficient partition characteristics,input wheel detection history data into training set and test set for training,determine the predicted value of wheels size with the predicted mean value.Constrained by the geometric size and dynamic performance of the wheel set,the longest service life of wheel and the minimum number of lathing and ride index as the optimization objectives,establish the optimization model of wheels lathing strategy,and the lathing amount of wheels and the wheel diameter value are predicted.The results show that when the wheel repair diameter amount is 2.5 mm,the edge of wheel thickness is H_(Ai)=28.5 mm and H_(Bi)=30 mm,the lathing strategy is the best,and the wheels life can be increased by 31.4%.The research results can provide theoretical support for the optimization of wheelset rotation repair strategy of EMU.

关 键 词：动车组 轮对旋修 差分自回归移动平均模型 随机森林算法 策略优化 

分 类 号：U266[机械工程—车辆工程]