基于PSO-BP神经网络的基坑周边地面沉降预测方法研究

基坑工程施工过程中的周边地面沉降直接关系到周围建筑物的安全，本文根据上海前滩地区某基坑工程的历史监测数据、施工工况和周边地层参数等多源数据对基坑周边地面沉降进行监测和预测。以PSO-BP神经网络为基础，通过将基于时序和基于沉降影响因素的网络模型对比发现：二者预测结果误差较小且基于时序的神经网络预测精度更高，说明利用PSO-BP神经网络能够很好地对基坑周边地面沉降进行分析与预测。为了综合考虑时间效应和空间效应的影响，在基于沉降影响因素的预测模型的基础上加入历史监测数据作为模型输入层进行优化，结果表明：优化后的PSO-BP神经网络模型具有更小的相对误差范围和更高的预测精度，在基坑周边地面沉降预测中有很好的应用前景。

Study on the Prediction Methods of Ground Settlement Surrounding the Foundation Pit Based on PSO-BP Neural Network

The surrounding ground settlement in the process of the foundation pit construction is directly related to the safety of the surrounding buildings. In this paper, the ground settlement surrounding the foundation pit was monitored and predicted according to the historical monitoring data, the construction conditions and the surrounding stratum parameters of a foundation pit project in Qiantan district of Shanghai. Based on PSO-BP neural network, this paper compares the network model based on the time series with that based on the settlement influence factors. It is found that the prediction error of these 2 models is small and the prediction precision of neural network based on the time series is higher, which means that PSO-BP neural network can be used to analyze and predict the ground settlement surrounding the foundation pit. In order to comprehensively consider the time effect and space effect, the historical monitoring data is added as the input layer of prediction model for optimization on the basis of prediction model of settlement influencing factors. The results show that the optimized PSO-BP neural network model has a smaller relative error range and a higher prediction precision, and it has good application prospect in the prediction of ground settlement surrounding the foundation pit.