粒子群神经网络在大跨钢结构挠度监测中的应用  

Application of PSO-BP Neural Networks to Deflection Monitoring of Large-span Steel Structure

作　　者：肖兴国[1] 袁长征[1,2] 李超[1,2] 

机构地区：[1]重庆市勘测院,重庆401121 [2]重庆市智能感知大数据产业技术协同创新中心,重庆401121

出　　处：《城市勘测》2016年第5期135-139,共5页Urban Geotechnical Investigation & Surveying

基　　金：住房和城乡建设部科学技术计划项目(2015-k8-012)

摘　　要：针对传统BP神经网络拓扑结构不确定、收敛效率低、容易陷入局部最优解的缺陷,引入粒子群(PSO)算法替代BP神经网络中基于误差函数梯度下降的学习训练规则,构建了粒子群神经网络(PSO-BP)模型,并以重庆某大型场馆安全监测项目为例,通过对比分析验证了粒子群神经网络模型在大跨钢结构挠度监测中的可行性以及与传统BP神经网络模型相比在效率和精度方面的优越性。The traditional BP neural network model’ s topological structure is difficult to determine,the rate of con-vergence is slow and its solution is likely to be a local optimal solution. In order to overcome these shortcomings,the PSO-BP neural network model was constructed by combining the PSO algorithm with the BP neural network,that is,using the PSO algorithm as the learning and training rules of BP neural network instead of the traditional one which based on the error function gradient descent guidelines. On this basis,the deflection monitoring data of a certain venues in Chongqing was taken as an example to verify the feasibility of the model in the deflection monitoring of large-span steel structure and its superiority on efficiency and accuracy compared with the traditional BP neural network.

关 键 词：BP神经网络 粒子群算法 大跨钢结构 挠度监测 

分 类 号：P258[天文地球—测绘科学与技术] TU196[建筑科学—建筑理论]