高强钢管高强混凝土短柱力学性能智能模型研究  

Study on a Neural Network Intelligent Model of Bearing Capacity of High-Strength Concrete Filled High-Strength Square Steel Tubular Stub Columns

作　　者：刘坚[1,2] 彭林苗 马宏伟[3] 刘长江 张专涛 陈原[1,2] 任达[1,2] 招渝 肖海鹏 杨勤鹏 戚玉亮 童华炜[1,2] LIU Jian;PENG Linmiao;MA Hongwei;LIU Changjiang;ZHANG Zhuantao;CHEN Yuan;REN Da;ZHAO Yu;XIAO Haipeng;YANG Qinpeng;QI Yuliang;TONG Huawei(College of Civil Engineering,Guangzhou University,Guangzhou 510006,China;Guangdong Provincial Engineering Technology Research Center for Complex Steel Structures,Guangzhou University,Guangzhou 510006,China;School of Civil Engineering and Transportation,South China University of Technology,Guangzhou 510006,China;Hangxiao Steel Structure(Guangdong)Co.,Ltd.,Zhuhai 519000,China;Guangzhou Construction Industry Research Institute Co.,Ltd.,Guangzhou 510663,China)

机构地区：[1]广州大学土木工程学院,广州510006 [2]广州大学广东省复杂钢结构工程技术研究中心,广州510006 [3]华南理工大学土木与交通学院,广州510006 [4]杭萧钢构(广东)有限公司,广东珠海519000 [5]广州建筑产业研究院有限公司,广州510663

出　　处：《建筑钢结构进展》2023年第11期12-20,共9页Progress in Steel Building Structures

基　　金：国家自然科学基金(51678168);广东省自然科学基金(2017A030313267);广州市科技计划项目(201607010107);广建装配式建筑成套技术研究(19C00001);装配式型钢混凝土新型组合结构体系成套技术体系研发(19C00001-2)。

摘　　要：为研究超出规范范围材料强度的高强钢管高强混凝土短柱承载力在规范范围外拓展应用的相关问题,搜集61组试验及有限元数据,利用神经网络智能技术研究了在不同的截面尺寸、高强材料、含钢率、偏心率和长细比等影响因素下的高强钢管高强混凝土短柱的极限承载力及荷载-位移曲线,将承载力预测结果与规范EC4、规范ACI及规范GJB 4142—2000的计算结果进行了对比分析,并预测了试件的荷载-位移曲线,对各阶段各模型的切线刚度进行了分析研究。研究结果表明:所述规范可以在一定程度上继续延伸使用,而神经网络智能模型不仅与试验数据吻合度较高、精确度高,还考虑了规范没有考虑的约束效应减弱及两种材料强度匹配程度对承载力的影响,承载力会随着混凝土强度和钢材强度的提高而减缓增强速度;神经网络智能模型非常适合用于预测高强钢管高强混凝土短柱承载力,可以为实际工程和规范延伸拓展提供设计参考;神经网络智能模型相比于现有组合构件应力-应变理论公式会更加简便准确,以切线刚度的对比结果可知,其可以更加准确地反映和再现轴压作用下高强钢管高强混凝土短柱荷载-位移曲线的特征,对构件的力学性能研究有一定的参考作用。In order to study the expansion and application of the bearing capacity of high-strength steel tube and high-strength concrete short columns that exceed the strength of materials in the specification,61 groups of tests and finite metadata were collected,and the ultimate bearing capacity and load-displacement curves of high-strength steel tube and highstrength concrete short columns under different section sizes,high-strength materials,steel ratio,eccentricity,slenderness ratio and other influencing factors were studied using neural network intelligent technology.The prediction results of bearing capacity are compared with the calculation results of code EC4,code ACI and code GJB 4142—2000,and the load-displacement curves of the specimens was predicted.The tangent stiffness of each model at each stage is analyzed and studied.The research results show that the code can be extended to a certain extent,and the neural network intelligent model does not only have high consistency and accuracy with the test data,the weakening of the constraint effect is not considered in the code and the influence of the strength matching degree of the two materials on the bearing capacity are also considered.The bearing capacity will slow down with the increase of concrete strength and steel strength.It is very suitable for predicting the bearing capacity of high-strength concrete filled high-strength square steel tubular stub columns,and can provide a design reference for actual projects and the extension of specifications.Compared with the existing stress-strain theoretical formula of composite members,the intelligent model is simpler and more accurate.Compared with the tangent stiffness,it can more accurately reflect and reproduce the characteristics of the load-displacement curves of high-strength concrete filled high-strength square steel tubular stub columns under axial compression,and provides a reference for the study of mechanical properties of members.

关 键 词：高强钢管 钢管高强混凝土 荷载-位移曲线 神经网络 极限承载力 

分 类 号：TU312.1[建筑科学—结构工程]