基于GM(1,1)与AR模型的轨道不平顺状态预测

Prediction of track irregularity state based on grey GM(1,1) and stochastic linear AR model

反映轨道几何状态变化的轨道几何检测数据是一个随时间变化的时间序列,具有随机性特点.本文将经过普遍适应性改进的灰色GM(1,1)与随机线性AR模型相结合,研究轨道水平不平顺状态在点、单元区段范围随时间变化趋势,并对GM(1,1)预测的残差进行修正,从轨道水平的几何状态变化的随机数据序列中找寻变化规律.用得到的几何状态变化模型分别对轨道的短期、中长期状态进行预测分析,预测结果表明模型是有效的,满足预定精度的要求.

Track geometry inspection data can reflect the change of track geometry state. It is a time series change with time and has random characteristics. In this paper, combination of a generally adaptability improved grey GM （1, 1） model with residual error correction and the stochastic linear AR model are applied to analyze track irregularity of cross level in designated point and unit section, and we find the law from the random data sequence of the track geometry state changes of cross level and predict short-term, long-term track state. The results show that the model is valid and can meet the intended accuracy.