岩溶地质条件下建筑变形数据的分析与预测  

The Analysis and Predict of Construction Deformation under Karst Geological Conditions

作　　者：赵拓 王彦芳 

机构地区：[1]河北建科唐秦建筑科技有限公司,石家庄050000

出　　处：《华北地震科学》2017年第B07期34-38,共5页North China Earthquake Sciences

摘　　要：通过对岩溶区与非岩溶区两栋高层A、B塔沉降观测数据的分析,得出:封顶前的建筑变形可以达到最终稳定时累计变形量的80%~83%,大部分地基沉降发生在封顶前的加载过程中;在封顶前随着荷载的增加沉降基本呈线性增长,封顶后荷载不变,沉降缓慢增加,曲线趋于收敛;应用BP神经网络对A塔观测数据进行预测分析,得到了当隐层神经元数为3时预测精度较高;BP神经网络预测模型预测结果曲线相对误差最小值为1%,最大值为2.08%,拟合程度较高,说明该模型的适用性较好。The settlement observation data of A,B two tall building in karst and non-karst area were analyzed.Building deformation before thecap can reach 80%-83% of accumulative deformation in the final.Most of the foundation settlement occurred in front of the cap.Settlement were linear growth with the increase of load in front of the cap,Settlement increased slowly after cap,Curve was basically completed,The predict of A tower observation data were analyzed using BP neural network,The best number of hidden layer neurons was 3,The predict result curve relative error minimum value was 1%,The maximum was 2.08%,Fitting degree was higher,Applicability of the model was good.

关 键 词：岩溶地质 沉降观测 BP神经网络 隐层神经元 

分 类 号：TU196[建筑科学—建筑理论]