深基坑施工变形预测VAR建模与应用分析

在实际施工过程中，基坑变形发展与多种因素有关，很难用力学模型进行定量分析，而根据已有的监测数据建立模型预测未来一段时间内的变形发展趋势，进而对当前施工方案进行评价、调整，方法更可靠。首先，分析深基坑施工实际监测数据的发展规律，由经济预向量自回归测VAR模型（vector auto regression，VAR）建立基坑施工变形预测VAR模型，并结合监测数据进行模型精度评价和预测比分析。根据分析结果，模型精度排序为：VAR模型＞新陈代谢模型＞神经网络模型＞GM(1, 1)模型，模型预测比排序为：VAR模型＞新陈代谢模型＞神经网络模型＞GM(1, 1)模型。

VAR modeling of construction deformation prediction of deep foundation pit and application

During the practical construction, deformation development of foundation pit is related to many factors. It is hard to make quantitative analysis with mechanical model. So we can make model based on measured value to predict the deformation trend in a short future time. This model can be used to evaluate and adjust the current construction scheme. Firstly, the characteristics of measured deformation are given. Then the vector auto regression (VAR) model of foundation pit deformation prediction is made by economy prediction VAR model. Lastly, the VAR model precision and prediction ratio are given with measured data. From the prediction analysis, the model precision sorting is VAR model, metabolism model, neural network model and GM(1,1) model. And the prediction ratio sorting is VAR model, metabolism model, neural network model and GM(1, 1) model.