市政技术2024，Vol.42Issue(3)：117-123,7.DOI:10.19922/j.1009-7767.2024.03.117
基于BP神经网络的矩形沉井下沉土体参数反演研究
Study on Parameters Inversion of Rectangular Caisson Sinking Soil Based on BP Neural Network
摘要
Abstract
Various artificial neural network intelligent algorithms have been widely used in geotechnical engineering parameter inversion and prediction.Most of them focus on dam bodies,pile foundations and tunnel engineering.They are rarely used in projects in the form of open caisson foundations.The BP neural network algorithm is suscep-tible to the spatiotemporal changes in geotechnical structure parameters during the construction process so that it is suitable for studying foundation forms of open caisson that experienced significant geological disturbances and dy-namic displacements.Relying on the anchorage open caisson foundation project in the southern part of the Longtan Yangtze River Bridge,the model was built by finite element parameter.The dynamic caisson sinking was divided into multiple working conditions.Without affecting the simulation accuracy,the calculation efficiency is improved to get a data set between multiple sets of soil parameters and the stress response of the open caisson structure.The data sets was used to train BP neural network models.The caisson stress monitored on site was input into the BP neural network model to obtain the corresponding soil parameters by inversion.The parameter change rules were analyzed,and the dynamic stress characteristics of the open caisson were summarized.The research improves the sinking aid scheme and plays a crucial role in solving problems of sudden settlement and resistance to sinking.关键词
沉井下沉/土体参数/有限元模拟/数据训练/BP神经网络/参数反演Key words
caisson sinking/soil parameters/finite element simulation/data training/BP neural network/param-eter inversion分类
信息技术与安全科学引用本文复制引用
王峻科,郑华凯..基于BP神经网络的矩形沉井下沉土体参数反演研究[J].市政技术,2024,42(3):117-123,7.基金项目
国家重点研发计划(2021YFB1600300) （2021YFB1600300）
