基于生成对抗网络的自动框架结构设计  

Intelligent Generation Method Of Concrete Frame Structures

作　　者：龙丹冰[1] 雷昕 方长建[2] 康永君[2] Long Danbing;Lei Xin;Fang Changjian;Kang Yongjun(School of Civil engineering,Southwest Jiao tong University,Chengdu 610031,China;China Southwest Architecture,Chengdu 610042,China)

机构地区：[1]西南交通大学土木工程学院,成都610031 [2]中国建筑西南设计研究院有限公司,成都610042

出　　处：《土木建筑工程信息技术》2024年第3期104-108,共5页Journal of Information Technology in Civil Engineering and Architecture

基　　金：中国建筑西南设计研究院科研课题项目(编号:R-2021-14-BIM-A-2023)。

摘　　要：本文在结构设计中引入人工智能方法,通过提出建筑结构特征表达方法,构建以生成对抗网络为核心的框架结构自动设计方法。在数据前处理阶段,通过对建筑结构图纸分析,提出建筑特征表达方法与结构特征表达方法。在构建算法模型阶段,实现在有限的数据量下训练生成对抗算法学习框架结构布置,构建了依据建筑信息自动生成含有构件尺寸信息的框架结构自动布置模型,并提出了评价指标量化评价模型的结构设计能力。在案例中使用本文构建的框架结构自动布置模型完成一栋实际建筑的自动设计,验证了本文提出方法对结构设计效率的提高。This paper introduces artificial intelligence methods in structural design,and further proposes a method for expressing building structural features,and constructs an automatic design method for frame structures centered on generating adversarial networks.In the data preprocessing stage,this paper proposes building feature expression methods and structural feature expression methods by analyzing building structural drawings.During constructing algorithm models,the adversarial algorithm is trained and generated under limited data to learn framework structure layout.A framework structure automatic layout model containing component size information is constructed based on building information,and evaluation indicators are proposed to quantify the structural design ability of the model.In the case study in this paper,the framework structure automatic layout model constructed is used to complete the automatic design of an actual building,which verifies the improvement of the proposed method in structural design efficiency.

关 键 词：框架结构自动设计 生成对抗网络 人工智能 机器学习 

分 类 号：TU17[建筑科学—建筑理论]