钢管约束型钢混凝土圆柱承载力的神经网络分析模型研究  

Research on Neural Network Analysis Model of Bearing Capacity of Steel Tubed Steel Reinforced Concrete Cylinder

作　　者：刘坚[1,2] 招渝 王飞程 刘长江 曾嵘森 周观根 戚玉亮 任达 陈原[1,2] 肖海鹏 彭林苗[1] LIU Jian;ZHAO Yu;WANG Feicheng;LIU Changjiang;ZENG Rongsen;ZHOU Guangen;QI Yuliang;REN Da;CHEN Yuan;XIAO Haipeng;PENG Linmiao(College of Civil Engineering,Guangzhou University,Guangzhou 510006,China;Guangdong Engineering Technology Research Center for Complex Steel Structures,Guangzhou 510006,China;Zhejiang Southeast Space Frame Co.,Ltd.,Hangzhou 311200,China;Guangzhou Construction Industry Research Institute Co.,Ltd.,Guangzhou 510000,China)

机构地区：[1]广州大学土木工程学院,广州510006 [2]广东省复杂钢结构工程技术研究中心,广州510006 [3]浙江东南网架股份有限公司,杭州311200 [4]广州建筑产业研究院有限公司,广州510000

出　　处：《工业建筑》2022年第9期147-152,120,共7页Industrial Construction

基　　金：国家自然科学基金项目(51678168);广东省自然科学基金(2017A030313267);广州市科技计划项目(201607010107);广东省应用型科技研发重大专项资金项目(2015B020238014);广建装配式建筑成套技术研究项目(19C00001-2)。

摘　　要：根据轴心受压和偏心受压的钢管约束型钢混凝土(TSRC)圆柱承载力现有计算公式,提出了轴压和偏压TSRC圆柱承载力的神经网络分析模型。选取10个影响承载力的敏感参数来确定输入层的节点个数,以轴压或偏压TSRC圆柱承载力作为输出层;隐含层节点数采用试凑法,根据均方误差MSE与相关系数R确定为12,由此建立了N10-12-1神经网络分析模型。该神经网络分析模型对承载力的预测结果显示,最大误差仅为6.08%,说明建立的轴压和偏压TSRC圆柱承载力神经网络分析模型是一种较好的方法。最后基于Garson算法进行敏感性分析,得到了各输入参数对TSRC圆柱承载力的影响程度,可供工程设计参考。According to the existing calculation formula of the bearing capacity of steel tubed steel reinforced concrete(TSRC)columns under axial and eccentric compression,the neural network analysis model of the bearing capacity of TSRC columns under axial and eccentric compression was proposed.Ten sensitive parameters affecting the bearing capacity were selected to determine the number of nodes in the input layer,and the bearing capacity of TSRC cylinder was taken as the output layer.The number of nodes in the hidden layer was determined as 12 according to the mean square error MSE and correlation coefficient R by trial and error method,and the N10-12-1 neural network analysis model was established.The prediction results of the neural network analysis model show that the maximum error was only 6.08%,indicating that the established neural network analysis model for the bearing capacity of TSRC cylinder under axial compression and eccentric londing was a good method.Finally,sensitivity analysis based on Garson algorithm was carried out to obtain the influence degree of each input parameter on the bearing capacity of TSRC cylinder,which could be used for reference in engineering design.

关 键 词：圆钢管约束型钢混凝土 BP人工神经网络 轴压承载力 偏压承载力 神经网络分析模型 

分 类 号：TU398.9[建筑科学—结构工程] TP183[自动化与计算机技术—控制理论与控制工程]