建筑物基坑沉降变形预测方法

Prediction Method of Settlement Deformation of Building Foundation Pit

为了对建筑物基坑沉降变形进行准确有效的预测,结合SVM与ARIMA拟合预测模型的优势,提出一种SVM-ARIMA组合模型的时间序列预测模型。该组合模型将建筑物基坑沉降监测时间序列分为非线性部分和线性部分,针对非线性部分,使用SVM模型进行单步滚轮预测;针对线性部分,则基于AIC与BIC模型选取最优ARIMA模型进行单步滚轮预测。结合建筑物基坑沉降实测数据进行模型效果试验的结果表明,相比于NAR神经网络模型和SVM模型,组合模型的预测效果更好,预测精度更高,且能够较好地反映建筑物基坑沉降的变形趋势。

In order to predict the settlement deformation of building foundation pit accurately and effectively,a time series prediction model named SVM ARIMI combined model is proposed,which take the advantages of SVM and ARIMA fitting prediction models.The combined model divides the monitoring time series of building foundation pit settlement into nonlinear part and linear part.For the nonlinear part,SVM model is used for one step prediction;For the linear part,the optimal ARIMA model is selected based on AIC and BIC models for one step roller prediction.Combined with the measured data of building foundation pit settlement,the model effect experiment is carried out.The results show that compared with NAR neural network model and SVM model,the combined model has better prediction effect and higher prediction accuracy,and can better reflect the deformation trend of building foundation pit settlement.