基于偏离度优化的高速列车轴箱轴承温度预测方法

High-speed train axle box bearing temperature prediction methodbased on optimization of deviation degree

根据灰色二次回归与GM(1,1)两个模型预测值与实际值的偏差,确定模型各自的权重系数进行组合预测。首先对模型的建模数、迭代数进行合理选择,对原始轴温数据进行预处理,然后分别分析了灰色二次回归与GM(1,1)模型的特点,最后为避免均值加权不能很好地适应轴温波动情况,采用偏离度的权重系数来将两个模型组合起来。通过权值分配的方式,不断对上一阶段单一模型的预测偏差校正,减小组合模型的预测偏差。经实际验证表明优化作用有效。

The prediction of axle temperature has practical application value to ensure the safe operation of high-speed train.In this paper,according to the deviation between the predicted value and the actual value of grey quadratic regression and GM(1,1),the weight coefficients of each model are determined for combined prediction.Firstly,the model is constructed with reasonable modulus and overlapping algebra,and the original axial temperature data is preprocessed.Then the features of grey quadratic regression and GM(1,1)model are analyzed respectively.Finally,in order to avoid the mean weight can not adapt to the fluctuation of axial temperature,the two models are combined by the weight coefficient of deviation degree.Through weight allocation,the prediction deviation correction of the first mock exam model is continuously reduced,and the prediction bias of the combined model is reduced.Based on the axle temperature data of a certain type of high-speed train,the model used in this paper is used to predict the axle temperature of a certain line vehicle.The model is verified by the model evaluation index,and the results show that the optimization effect is effective.