摘要
为了探索土岩复合地层深基坑变形规律,预测地表沉降,防止因地表沉降造成不必要损失,针对传统算法效率低、易陷入局部最优、抵抗波动能力弱、预测精度低等特点,以长春地铁5号线某车站深基坑工程为依托,利用时间序列获取训练数据,将GA算法嵌入到PSO算法进行参数优化,结合灰色理论建立基于灰色最小二乘支持向量机的基坑变形时序预测模型。研究结果表明:GA-PSO-GLSSVM基坑变形预测模型预测结果平均绝对误差、均方根误差和平均百分比误差分别为0.1284、0.1658、2.0768%,优于传统BP神经网络预测模型和GA-PSO-LSSVM模型;经过GM(1,1)预处理的GA-PSO-GLSSVM模型,可以更好地预测波动较大的数据,更加适用于土岩复合地层深基坑变形的预测。
To explore deep foundation pit deformation law in soil-rock compound formation,predict the surface subsidence,to prevent unnecessary loss caused by the surface subsidence,in view of the low efficiency of traditional algorithm,easy to fall into local optimum,weak ability to resist fluctuation,low precision,etc.,the time series were used to collect training data,the GA algorithm was embedded into PSO algorithm for parameter optimization,and a time series prediction model of foundation pit derformation was established based on grey least squares support vector machine combined with grey theory,by taking the deep foundation pit engineering of Changchun subway line 5 station as the background.The results show that the average absolute error,root mean square error and average percentage error of GA-PSO-GLSSVM-based prediction model of foundation pit derformation are 0.1284,0.1658 and 2.0768%,respectively,which are better than the traditional BP neural network prediction model and GA-PSO-LSSVM model.It is found that the GA-PSO-GLSSVM model pretreated by GM(1,1)can better predict the data with large fluctuations,and is more suitable for the predicting the deformation of deep foundation pit in soil-rock composite stratum.
作者
薛艳杰
XUE Yanjie(The 4th Engineering Company,China Railway No.3 Engineering Group Co.Ltd.,Beijing 102300)
出处
《现代隧道技术》
CSCD
北大核心
2022年第S02期77-85,共9页
Modern Tunnelling Technology
基金
吉林省科技发展计划项目(20220203058SF)
中铁三局集团科技研发项目(ZTSJCT5-2021-JY001)
作者简介
薛艳杰(1980-),男,本科,高级工程师,主要从事建筑施工技术管理工作,E-mail:33697860@qq.com.
