基于加权组合模型的基坑位移变形分析与预测  

Analysis and Prediction of Pit Displacement Deformation Based on Weighted Combination Model

作　　者：王永明[1] 李明峰[2] 欧江霞 石星照[2] 

机构地区：[1]武汉大学测绘学院,武汉430079 [2]南京工业大学地球空间信息研究中心,南京210009

出　　处：《地下空间与工程学报》2013年第S1期1564-1567,1573,共5页Chinese Journal of Underground Space and Engineering

基　　金：国家自然科学基金项目(41274009);南京市科技计划项目(201101069);江苏省测绘科研项目(JSCHKY201108);江苏省建设厅科技项目(JS2011JH23)

摘　　要：位移变形分析与预测可为基坑工程安全趋势评定提供量化依据。在论述回归分析、灰色模型与人工神经网络3种变形分析模型的基础上,以残差平方和最小为准则,提出了加权组合模型的建立思路。以南京滨江公寓基坑变形监测为例,分别将单一模型与组合模型应用于沉降监测数据处理过程中,采用最大绝对值、平均绝对值与标准差等残差指标对各方法建模精度及预测精度进行对比分析,通过3种模型特点与实用范围的总结,验证了加权组合模型的可行性与优越性。The quantitative basis for safety trend evaluation of foundation pit engineering can be provided by the analysis and prediction of displacement deformation. Based on the discussion on the three deformation analysis models such as the regression analysis,grey model and artificial neural network,the weighted combination model was proposed with the minimum sum of squared residuals as criterion. Taking the deformation monitoring of Binjiang residential foundation in Nanjing as an example,a single model and combination model were separately applied to the data processing stage of settlement monitoring. Using the residual indexes of the maximum and average absolute value and standard deviation,the characteristics and practical range of three models were summarized while the modeling precision and prediction accuracy of each model were compared. Then,the feasibility and superiority of the weighted combination model was verified.

关 键 词：变形预测 回归分析 灰色模型 人工神经网络 组合模型 

分 类 号：TU433[建筑科学—岩土工程]