1.武汉理工大学土木工程与建筑学院,湖北 武汉 430070
2.中交武汉港湾工程设计研究院有限公司,湖北 武汉 430040
曹鸿猷(1986年生),男;研究方向:结构智能化设计;E-mail: caohongyou@whut.edu.cn
收稿:2025-07-24,
录用:2025-08-25,
网络出版:2025-09-30,
纸质出版:2025-11-25
移动端阅览
曹鸿猷,张廷曌,曾卓等.基于图神经网络的桁架结构响应代理模型[J].中山大学学报(自然科学版)(中英文),2025,64(06):102-110.
CAO Hongyou,ZHANG Tingzhao,ZENG Zhuo,et al.Graph neural network-based surrogate modeling for truss structures response[J].Acta Scientiarum Naturalium Universitatis Sunyatseni,2025,64(06):102-110.
曹鸿猷,张廷曌,曾卓等.基于图神经网络的桁架结构响应代理模型[J].中山大学学报(自然科学版)(中英文),2025,64(06):102-110. DOI: 10.13471/j.cnki.acta.snus.ZR20250140.
CAO Hongyou,ZHANG Tingzhao,ZENG Zhuo,et al.Graph neural network-based surrogate modeling for truss structures response[J].Acta Scientiarum Naturalium Universitatis Sunyatseni,2025,64(06):102-110. DOI: 10.13471/j.cnki.acta.snus.ZR20250140.
本文提出了一种基于图神经网络(GNN)的代理模型用于桁架结构响应预测。基于图论,利用图表征方法描述桁架结构的拓扑特征和物理信息,通过图卷积建立变量间的相关性实现基于结构几何与物理特征的训练空间降维;此外,在模型中引入注意力机制有效计入节点和边特征,并基于单元刚度自适应地分配邻居节点及边的特征权重,进而实现在训练过程中准确模拟结构中的力学传递路径。最后,以3个具有10、19和27个变量的桁架结构为例,对所提出模型预测精度进行了评估。结果表明:随着训练样本数量的增加,GNN模型的预测误差能显著降低;且在样本量合适的前提下,模型对于高维问题仍能获得极高的预测精度,克服了传统代理模型在样本数量增长和处理高维问题时预测精度急剧下降的弊端。
A surrogate model based on graph neural networks(GNN) is proposed for truss structure response prediction.Based on graph theory,a graph representation method is employed to describe the topological characteristics and physical information of truss structures.Graph convolution is used to establish correlations among variables,achieving dimensionality reduction of the training space through structural geometric and physical features.Furthermore,an attention mechanism is integrated into the model to effectively incorporate node and edge features,with feature weights of neighboring nodes and edges adaptively assigned according to element stiffness,thereby accurately simulating force transmission paths during training.Three truss structures with 10,19,and 27 design variables,respectively,serve as examples for evaluating prediction accuracy using the mean absolute percentage error metric.Comparative analysis indicates that the prediction error of the GNN model decreases significantly with an increasing number of training samples and remains exceptionally.
鲍跃全 , 李惠 , 2019 . 人工智能时代的土木工程 [J]. 土木工程学报 , 52 ( 5 ): 1 - 11 .
樊健生 , 杨晨 , 张翀 , 等 , 2024 . 工程结构超多维变量可微分智能优化方法研究 [J]. 建筑结构 , 54 ( 19 ): 59 - 65 .
冯德成 , 吴刚 , 2022 . 混凝土结构基本性能的可解释机器学习建模方法 [J]. 建筑结构学报 , 43 ( 4 ): 228 - 238 .
韩小雷 , 冯润平 , 季静 , 等 , 2021 . 基于深度学习的RC梁集中塑性铰模型参数研究 [J]. 工程力学 , 38 ( 11 ): 160 - 169 .
毛凤山 , 陈昌富 , 朱世民 , 2020 . 代理模型方法及其在岩土工程中的应用综述 [J]. 地基处理 , 2 ( 4 ): 295 - 306 .
宋凌寒 , 王琛 , 樊健生 , 2024 . 理论驱动的弹性结构体系图神经网络计算模型 [J]. 工程力学 , 41 ( 2 ): 25 - 34 .
王胜利 , 刘华 , 郑山锁 , 等 , 2024 . 一般大气环境下钢筋锈蚀深度的RBF神经网络预测模型研究 [J]. 地震工程学报 , 46 ( 2 ): 269 - 277 .
CAO H , LI M , NIE L , et al , 2024 . Vertex-based graph neural network classification model considering structural topological features for structural optimization [J]. Comput Struct , 305 : 107542 .
GILMER J , SCHOENHOLZ S S , RILEY P F , et al , 2017 . Neural message passing for quantum chemistry [C]// International Conference on Machine Learning . Sydney, NSW, Australia : 1263 - 1272 .
GORI M , MONFARDINI G , SCARSELLI F , 2005 . A new model for learning in graph domains [C]// IEEE International Joint Conference on Neural Networks . Montreal, QC, Canada : 729 - 734 .
HAMILTON W , YING Z , LESKOVEC J , 2017 . Inductive representation learning on large graphs [C]// International Conference on Neural Information Processing Systems . Long Beach, California, USA : 1025 - 1035 .
KAVEH A , 2024 . Optimal Design of Transmission Towers Using Genetic Algorithm and Neural Networks [M]. Cham : Springer: 285- 310 .
KIPF T N , WELLING M , 2016 . Semi-supervised classification with graph convolutional networks [J]. arXiv:1609 . 02907 .
LI Q , WANG Z , LI L , et al , 2023 . Machine learning prediction of structural dynamic responses using graph neural networks [J]. Comput Struct , 289 : 107188 .
LIU H , SONG L , XU L , et al , 2024 . Response prediction and probabilistic analysis of the vehicle-ballasted track system considering track irregularity based on long-short term memory neural network [J]. Eng Appl Artif Intell , 133 : 108604 .
LYU F , FAN X , DING F , et al , 2021 . Prediction of the axial compressive strength of circular concrete-filled steel tube columns using sine cosine algorithm-support vector regression [J]. Compos Struct , 273 : 114282 .
SCARSELLI F , GORI M , TSOI A C , et al , 2008 . The graph neural network model [J]. IEEE Trans Neural Netw , 20 ( 1 ): 61 - 80 .
SPERDUTI A , STARITA A , 1997 . Supervised neural networks for the classification of structures [J]. IEEE Trans Neural Netw , 8 ( 3 ): 714 - 735 .
VELIČKOVIĆ P , CUCURULL G , CASANOVA A , et al , 2017 . Graph attention networks [J]. Stat , 1050 ( 20 ): 10 - 48550 .
XIA J , LIAO W , HAN B , et al , 2025 . Intelligent co-design of shear wall and beam layouts using a graph neural network [J]. Autom Constr , 172 : 106024 .
ZENG J , KIM Y H , QIN S , 2023 . Bayesian model updating for structural dynamic applications combing differential evolution adaptive metropolis and Kriging model [J]. J Struct Eng , 149 ( 6 ): 04023070 .
ZHAO Y , LIU Y , LIN Q , et al , 2024 . A conditional generative model for end-to-end stress field prediction of composite bolted joints [J]. Eng Appl Artif Intell , 134 : 108692 .
0
浏览量
25
下载量
0
CSCD
关联资源
相关文章
相关作者
相关机构
京公网安备11010802024621
