摘要
为降低土体参数不确定性对基坑支护位移预测的影响,保障工程安全,用粒子群算法(PSO)和天牛须搜索算法(BAS)优化BP神经网络,设计土体参数反分析方法。基于苏州某基坑监测数据,经正交设计、有限元法获取训练样本,PSO、BAS优化参数,建立反演参数与位移关系来反演土体参数并代入模型计算位移。结果显示:在相同样本数据下,PSO-BP神经网络的预测精度高于BAS-BP神经网络;PSO-BP反分析所得土层参数用于有限元计算,测点水平位移计算值与实测值基本一致,说明该算法位移反分析准确性较好;依据工况2反分析结果预测下一工况的围护结构位移,与实测值接近,预测效果符合要求。因此基于PSO-BP算法的位移反分析能提升模拟精度,在不同施工阶段应用效果显著,可为类似工程计算提供参考。
In order to reduce the influence of the uncertainty of soil parameters on the displacement prediction of foundation pit support and ensure the safety of the project,the BP neural network is optimized by particle swarm optimization(PSO)and beetle antennae search algorithm(BAS),and the back analysis method of soil parameters is designed.Based on the monitoring data of a foundation pit in Suzhou,the training samples are obtained by orthogonal design and finite element method,and the parameters are optimized by PSO and BAS.The relationship between inversion parameters and displacement is established to invert the soil parameters and substitute them into the model to calculate the displacement.The results show that the prediction accuracy of PSO-BP neural network is higher than that of BAS-BP neural network under the same sample data.The soil parameters obtained by the PSO-BP back analysis are used for finite element calculation.The calculated horizontal displacement of the measuring point is basically consistent with the measured value,indicating that the accuracy of the back analysis of the displacement of the algorithm is better.According to the results of the second reverse analysis of the working condition,the displacement of the retaining structure under the next working condition is predicted,which is close to the measured value,and the prediction effect meets the requirements.Therefore,the displacement back analysis based on PSO-BP algorithm can improve the simulation accuracy,and the application effect is remarkable in different construction stages,which has important value.
作者
徐雷峰
田光辉
张志增
祝彦知
纠永志
XU Leifeng;TIAN Guanghui;ZHANG Zhizeng;ZHU Yanzhi;JIU Yongzhi(School of Intelligent Construction and Building Engineering,Zhongyuan University of Technology,Zhengzhou 450007,China;College of Geosciences and Engineering,North China University of Vater Resources and Electric Power,Zhengzhou 450045,China)
出处
《广东水利水电》
2025年第9期65-72,78,共9页
Guangdong Water Resources and Hydropower
基金
河南省自然科学基金面上项目(编号:222300420596)。
关键词
基坑开挖
土体参数
BP神经网络
粒子群算法
天牛须搜索算法
foundation pit excavation
soil parameters
BP neural network
particle swarm optimization algorithm
beetle whisker search algorithm
