面向轨道车辆轮对智能运维的镟修辅助决策系统设计与实现

Design and Implementation of an Intelligent Reprofiling Decision Support System for Railway Wheelset Maintenance

摘要: 在轨道车辆轮对运维工作中，现有镟修方案制定仍较多依赖人工经验，存在磨耗状态评估效率低、镟修决策标准不统一、轮材利用率不足以及预测模型与现场作业流程衔接不紧密等问题。为提升轮对镟修决策的科学性、规范性和工程适用性，本文以某线路的轨道车辆轮对历史检测数据和现场镟修数据为基础，构建了融合轮对磨耗预测、镟修优化决策与可视化交互的辅助决策系统。首先，基于历史运维数据建立WOA-BP轮对磨耗预测模型，对目标里程下的车轮直径和轮缘厚度变化趋势进行预测；其次，以轮径切削总量最小为目标，建立考虑同轴、同转向架、同车厢和整列车多级轮径差约束的整车镟修优化模型，并进一步引入轮缘厚度恢复目标，形成兼顾安全约束与轮材经济性的综合镞修策略；再次，结合车辆段现场单车厢、多车厢联合等分批作业需求，构建面向不同镟修范围的分批决策模型；最后，基于Python开发轮对镞修辅助决策系统，实现数据录入、预测调用、参数配置、策略求解、结果展示与文件保存等功能。应用验证结果表明，系统可支持整车镟修、单车厢镟修和多车厢联合镟修等典型场景。在整车镟修案例中，优化方案可将仿真切削总量由37.47 mm降至13.47 mm；在考虑轮缘厚度恢复目标后，理论上可减少约33.1%的单次镟修轮径损耗；在单车厢分批镟修场景下，相较传统定额镟修方式可减少约47%的轮径损耗。研究结果表明，所开发系统能够实现从轮对状态预测到镟修方案生成的流程贯通，为轨道车辆轮对智能运维和预防性镟修决策提供技术支撑。

Abstract: In railway vehicle wheelset maintenance, the formulation of reprofiling schemes still relies heavily on manual experience, resulting in low efficiency in wear-state assessment, inconsistent decision-making standards, insufficient wheel-material utilization, and weak integration between prediction models and on-site maintenance workflows. To improve the scientific rigor, standardization, and engineering applicability of wheelset reprofiling decisions, an intelligent decision support system integrating wheelset wear prediction, reprofiling optimization, and visual interaction is developed based on historical inspection data and on-site reprofiling data from suburban EMU wheelsets. First, a WOA-BP wheelset wear prediction model is established to predict wheel diameter and flange thickness at the target mileage. Second, with the objective of minimizing the total wheel diameter cutting amount, a full-train reprofiling optimization model is constructed by considering multi-level wheel diameter difference constraints, including same-axle, same-bogie, same-car, and whole-train constraints. A comprehensive reprofiling strategy considering flange thickness restoration is further introduced to balance safety constraints and economic efficiency. Third, in response to practical maintenance needs, batch reprofiling models for single-car and multi-car scenarios are developed. Finally, a Python-based decision support system is implemented, integrating data input, prediction calling, parameter configuration, strategy solving, result visualization, and file saving. Application results show that the system can support full-train, single-car, and multi-car reprofiling scenarios. In the full-train case, the optimized scheme reduces the simulated total cutting amount from 37.47 mm to 13.47 mm. After introducing the flange thickness restoration target, the proposed strategy theoretically reduces the wheel diameter loss of a single reprofiling operation by about 33.1%. In the single-car batch reprofiling scenario, the wheel diameter loss is reduced by about 47% compared with the traditional fixed-amount reprofiling scheme. The results demonstrate that the developed system connects wheelset wear prediction with reprofiling decision-making and provides technical support for intelligent wheelset maintenance and preventive reprofiling.