基于LS-SVM的深基坑工程位移反分析  

Back Analysis of the Displacement of a Foundation Pit Based on the LS-SVM

作　　者：张庆[1] 邹正盛 

机构地区：[1]河南理工大学土木工程学院,焦作454003 [2]中铁十二局集团第二工程有限公司,太原030032

出　　处：《科学技术与工程》2012年第33期8943-8946,8955,共5页Science Technology and Engineering

基　　金：河南理工大学博士基金项目(648198)资助

摘　　要：由于传统的位移反分析方法计算复杂,精度不高,不易迅速掌握等,则基于最小二乘支持向量机的反分析技术孕育而生。将探索该方法在深基坑工程位移反分析中的应用。首先结合相关规范和地区经验,通过正交设计方法获得多组参数组合,用来做数值模型的参数输入。再通过有限差分软件FLAC3D进行数值分析获得各组参数对应的位移信息,从中提取支持向量机的训练样本和测试样本,获得支持向量机模型。最后再以实际监测位移值带入模型获得反演计算的土层参数,并且通过数值模拟进行验证,取得了很好的效果。获得了与实际非常相似的数值模型。反演和验证结果反映了最小二乘支持向量机在基坑工程参数反演分析中的应用是可行的和可靠的。Due to the fact that the conventional displacement back analysis methods have disadvantages of complicated computing, hard to master, and not high precision solutions, someone Least Squares Support Vector Machine are applied to solve the displacement back analysis problem. The new method is applied in the deep foundation pit. The main point of the method is to use LS-SVM and FDM program FLAC3D to get train sample and test sample, and then through the parameter optimization method to get the best SVM model, in the forward process and to map the nonlinear relation of samples and reduce computed data. The LS-SVM is used in the backward process a high precision soil parameter can be achieved. Finally, the back analyses result is used to FLAC model of the foundation pit, the simulation results show that the LS-SVM method is not only feasible but effective in the field of foundation pit.

关 键 词：深基坑 支持向量机 最小二乘法 参数反演 正交设计 FLAC3D 

分 类 号：TU473.2[建筑科学—结构工程]