基于神经网络的组合模型在基坑沉降预测中的应用研究  

Study for Application of Combined Model Based on Neural Network on Foundation Pit Settlement Prediction

作　　者：钱大林[1] 张志伟[1] 靳璐岩 

机构地区：[1]东南大学交通学院,江苏南京210096 [2]山西省科学技术情报研究所,山西太原030001

出　　处：《现代测绘》2014年第5期8-11,共4页Modern Surveying and Mapping

摘　　要：通过对基坑沉降发展规律及其沉降曲线特点进行的研究,在多种S型单项预测模型基础上引入了组合预测的思想,本文先用4种S型增长曲线模型分别对基坑开挖周边地表沉降值进行拟合和预测,然后基于各单一模型预测数值通过神经网络进行组合建立组合模型进行预测。通过实例,对模型的预测结果进行了分析和检验,证明了在沉降变形分析中应用此组合预测法的可行性。Through a study on the settlement development regularity of foundation pit settlement and its settlement curve characteristics, the paper brings in the idea of combination prediction based on several S-type curve forecasting models. This paper uses 4 S-type curve models to simulate and forecast dates of the ground surface settlement nearby around the excavated foundation pit, then found combined forecasting model of neural network based on predicted dates of each single model to predict. Through example, analysis and testing are done on prediction results of the model. The result proves that using this combination forecast method in settlement deformation analysis is of feasibility.

关 键 词：基坑沉降 S型增长模型 BP神经网络 组合模型 预测精度 

分 类 号：TU196[建筑科学—建筑理论]