黄土路基沉降量预测方法的研究

归纳对比了评估指南所提出的几种常用沉降预测模型的优缺点，提出检验任一拟合方法是否是通过实测沉降数据预测最终沉降量的最优方法，除检验其拟合精度外，还必须检验其预测精度，并基于所提出的拟合指标及预测指标，分析郑西客运专线实测数据后提出了适用于黄土路基沉降预测的预测方法。并根据组合预测的思想，通过加权系数调整各单一预测模型的比例，基于最小二乘法进行求解，建立了组合预测模型。探讨了其拟合及预测精度，指出使用组合预测模型可综合利用各种方法所提供的信息，有效地减少单个预测模型受随机因素的影响，较大地提高预测精度，其拟合精度比任何单一预测模型高得多。但用其预测之前，必须将参与组合的模型进行筛选，预测精度与所选取的筛选准则是否合理息息相关，如何选择一个比较合理的筛选准则使得在保证被筛选用于最优组合模型的预测模型其权系数均大于0的前提下，误差平方和最小，仍需进一步研究。

Study of settlement prediction methods of loess subgrade

The advantages and disadvantages of some commonly used settlement prediction models are analyzed. comparatively An optimal method of settlement prediction by fitting precision and prediction accuracy is proposed. A better loess subgrade settlement prediction method is put forward based on the measured data of Zhengzhou-Xi’an passenger dedicated line.The fitting precision and the prediction accuracy of combination forecasting model are studied. It is concluded that the fitting precision and the prediction accuracy of combination forecasting model should be higher then any other prediction methods because of its comprehensive nature. But the members of combination forecasting model must be selected before prediction. Prediction accuracy and screening criteria are closely linked. How to choose better screening criteria which make square sum of error small in the prerequisite conditions of weight coefficient being greater than zero still need further study.