改进的车轮踏面磨损量最优非负变权组合预测模型

An improved optimal non-negative variable weight combination model for prediction of wheel tread wear

为准确获取车辆轮对踏面随列车运行里程的磨损量变化情况,以灰色离散GM(1,1)模型、指数平滑模型和线性回归模型为单项预测模型,以灰色关联度最大为最优准则,建立车轮踏面磨损趋势的最优非负变权组合预测模型。与各单项模型、以组合预测误差平方和最小为最优准则的变权组合模型进行对比。研究结果表明,变权组合模型的局部预测精度和全局预测精度均高于各单项模型,而本文模型有效抑制了基于误差平方和最小的变权组合模型误差"放大"的缺陷,预测精度及稳定性更高。

In order to obtain the wheel tread wear loss along with the changes of running distances,an optimal non-negative variable weight combination forecasting model was proposed based on grey correlation degree of maximum. The single forecasting models included grey discrete GM（1,1） model, exponential smoothing model and linear regression model. The calculation results were compared with three single models and the variable weight combination model based on the error square of minimum. The results show that the partial and global forecasting accuracy of variation weight combination model are distinctly higher than each single model. Moreover, the new model can effectively control the error defect of the variation weight combination model based on the error square of minimum, which has higher forecasting accuracy and stability.