深基坑变形的滚动预测方法  

Rolling prediction method of deep foundation pit deformation

作　　者：岳建伟[1,2] 仲豪磊 赵丽敏 邢旋旋[1] 黎鹏 张静 王自法[1] YUE Jianwei;ZHONG Haolei;ZHAO Limin;XING Xuanxuan;LI Peng;ZHANG Jing;WANG Zifa(School of Civil Engineering and Architecture, Henan University, Kaifeng, Henan 475004, China;Henan Rail Transit Intelligent Construction Engineering Center, Kaifeng, Henan 475004, China)

机构地区：[1]河南大学土木建筑学院,河南开封475004 [2]河南省轨道交通智能建造工程中心,河南开封475004

出　　处：《中国科技论文》2022年第6期639-646,共8页China Sciencepaper

基　　金：国家自然科学基金资助项目(51978634)。

摘　　要：为了探究在深基坑开挖阶段基坑监测点的数据是否超过预警值并预测变形规律,对传统BP算法进行改进,建立了基于附加动量法和时间效应的五步一出多步滚动BP神经网络预测模型,并将该方法预测结果与传统BP神经网络预测结果进行对比分析。结果表明,所提滚动优化BP神经网络方法对监测点坡顶沉降和水平位移预测的平均相对误差分别为0.80%和0.83%,达到了较高的精度。该方法在基坑变形预测中具有重要的应用价值。In order to explore whether the data of foundation pit monitoring points exceed the early warning value and predict the deformation law in the stage of deep foundation pit excavation,the traditional BP algorithm was improved.A one-output in five-step rolling BP neural network prediction model based on additional momentum method and time effect was established,and the prediction results of this method were compared with those of traditional BP neural network.The test results show that the average relative error of the method for predicting the slope top settlement and horizontal displacement of the monitoring point is 0.80%and 0.83%,respectively,reaching a relatively high accuracy.This method has important application value in the prediction of foundation pit deformation.

关 键 词：深基坑工程 BP神经网络 附加动量法 多步滚动预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]