基于数据驱动的轨道车辆焊缝疲劳寿命预测

Fatigue life prediction of rail vehicle weld based on data driven

针对实际运行环境下轨道车辆受力情况复杂、寿命难以预测,而传统寿命预测方法耗时耗力的问题,文章提出了一种基于数据驱动的疲劳寿命预测方法。利用有限元仿真软件建立某地铁车辆三人位座椅的有限元模型,通过结构应力法计算得到座椅的一条关键焊缝在不同载荷下的多组疲劳寿命数据,并以此作为训练数据来训练模型,对座椅的关键焊缝进行试验验证。将使用结构应力法计算得到的疲劳寿命结果与文中所提方法预测的结果做对比,通过均方误差、百分比误差和决定系数三项评价指标对预测结果进行量化,最终结果显示,均方误差均在可接受范围内,百分比误差低至2.87%,决定系数均在0.9以上,表明该方法在轨道车辆焊缝疲劳寿命预测领域可行。

Aiming at the problems that the stress condition of rail vehicle is complex and the life is difficult to predict in the actual operating environment,and the traditional life prediction method is time-consuming and laborintensive,this paper proposes a data driven fatigue life prediction method.The finite element model of a threeperson seat of a metro vehicle is established by using the finite element simulation software.The fatigue life data of a key weld of the seat under different loads is calculated by the structural stress method,and the model is trained as the training data to verify the key weld of the seat.The experimental results of the structural stress method are compared,and the prediction results are quantified by three evaluation indexes:mean square error,percentage error and determination coefficient.The final results show that the mean square error is within the acceptable range,the percentage error is low to 2.87%,and the determination coefficient is above 0.9,which indicates that the method is feasible in the field of rail vehicle weld fatigue life prediction.