基于改进深度神经网络的桩身内力测试数据自动定时采集算法  

Automatic Timed Acquisition Algorithm for Pile Internal Force Test Data Based on Improved Deep Neural Networks

作　　者：艾启胜 王瑞杰 陈文德 任旭 彭亮 张超超 AI Qisheng;WANG Ruijie;CHEN Wende;REN Xu;PENG Liang;ZHANG Chaochao(Geophysical Exploration Brigade of Hubei Geological Bureau,Wuhan 430056,China;Hubei Shenlong Engineering Testing Technology Co Ltd,Wuhan 430030,China)

机构地区：[1]湖北省地质局地球物理勘探大队,湖北武汉430056 [2]湖北神龙工程测试技术有限公司,湖北武汉430030

出　　处：《土木工程与管理学报》2024年第5期51-56,共6页Journal of Civil Engineering and Management

摘　　要：为了提高桩身内力测试数据的采集性能,进一步保证深基坑施工效果,提出一种基于改进深度神经网络的桩身内力测试数据自动定时采集算法。根据改进深度神经网络的结构,建立桩身位移在每一层神经元的输出函数,利用映照关系,判断深基坑施工中桩身的位置,预测桩身位移,利用地基抗力系数模型,获得桩身位移预测结果,初步确认桩身在工作阶段和张拉阶段的受力情况,搭建桩身的挠曲线控制方程,构建桩身受力分析模型,完成桩身的受力分析,通过桩身内力测试数据采集总量的最大化约束,建立跳数与采集节点接收数据总量的函数关系,利用最短采集路径优化的目标函数,实现桩身内力测试数据自动定时采集。实验结果表明,文中方法在采集桩身内力测试数据的冗余度不超过5%,实时性方面表现较好,可以有效提高桩身内力测试数据质量。To enhance the efficiency of pile internal force test data collection and safeguard the efficacy of deep foundation pit construction,an automatic,timed collection algorithm has been devised using an improved deep neural network.This algorithm leverages the network’s structure to establish displacement output functions for each layer of neurons in the pile body.By utilizing reflection relationships,it determines the pile body’s position in the foundation pit,predicts displacement,and utilizes a foundation resistance coefficient model to forecast pile body displacement.This allows for initial confirmation of pile body forces during operation and tensioning stages.An analysis model for pile body bearing force is then constructed,utilizing the foundation resistance coefficient model.After analyzing the pile body’s bearing force,a functional relationship between the number of hops and the total data received by the collection node is established,maximizing the total data collected from the Pile Internal Force test.The algorithm further optimizes the shortest collection path,enabling automatic and regular collection of pile internal force test data.Experimental results indicate that this method reduces data collection redundancy to less than 5%while improving real-time performance,significantly enhancing the quality of Pile Internal Force test data.

关 键 词：深度神经网络 定时采集 测试数据 位移预测 桩身内力 受力分析 

分 类 号：TG146[一般工业技术—材料科学与工程]