重载铁路钢轨磨耗廓形的快速预测与分析OA

[Objective]Rail wear is influenced by multiple factors,and changes in rail profile caused by wear significantly affect the safety and smoothness of train operations.Conducting field tests is labor-and resource-intensive,and frequent adjustments to track parameters on operational lines are not feasible.Numerical simulation methods for calculating worn rail profiles are costly and make rapid prediction difficult.Additionally,existing research on wear prediction typically relies on qualitative indicators,lacking the ability to quantitatively analyze the precise distribution of wear.[Methods]To address these gaps in current research,a rapid prediction model for rail wear profiles was developed based on vehicle-track coupled dynamics.Using simulation datasets,the model applied a multi-layer stacking strategy in combination with K-fold cross-validation to predict rail wear profiles and quantitatively analyze wear distribution.This model provided an efficient and cost-effective solution for complex wear problems.[Results]The results of the rapid prediction model showed that,for predicting rail wear on a curve with a radius of 600 m,the mean absolute error was less than 0.03 mm,and the computation speed was approximately 31 times faster than numerical simulations.Under the maximum accumulated gross tonnage condition,the median prediction error ranged between 0.1 and 0.2 mm,achieving rapid prediction of rail wear profiles.Quantitative analysis of wear distribution revealed that the nonlinear characteristics of rail profile evolution were primarily concentrated at the gauge corner.As the running speed increased,the wear distribution range expanded,and the wear peak shifted inward toward the rail head.Increasing the transition curve length significantly reduced both the total wear volume and wear peak,while having minimal impact on the wear distribution range.An increase in the friction coefficient expanded the contact patch area,made the stress distribution more complex,and increased the number of peak positions.[Conclusion]The research methods and results can serve as a reference for similar research.