【摘要】 为实现复杂建筑平面的结构拓扑自动生成，提出了一种基于图生成模型的结构拓扑智能设计方法，能够辅助设计人员实现快速、高效的结构骨架生成。首先采用无向图数据结构描述结构平面图，将结构拓扑生成问题转化为无向图的邻接矩阵生成问题；然后收集了200套复杂建筑平面布置图和结构平面布置图，采用数字图像处理方法识别结构平面布置图并转化为对应的无向图数字表征；进而利用对抗生成网络架构进行邻接矩阵生成，使用多层感知机作为生成器，并使用图神经网络作为判别器以判断生成的邻接矩阵的真实性。数值试验结果表明，该结构拓扑智能设计方法较为合理，可提供良好的结构初步设计方案，为类似设计研究提供了参考与借鉴。更多还原

【Abstract】 This article presents a novel method for automatic generation of structural topologies for complex building architectural layouts by a graph generative model, which assists the designers in rapid and effectively completing the structure skeleton. Firstly, the structural layouts is represented by undirected graph data structures and the structural topology generation problem is transformed into the adjacency matrix generation problem of undirected graphs;Then, 200 sets of complex architectural layouts and structural layouts were collected. The structural layouts were transformed into the corresponding undirected graph representations by digital image processing methods; By a multilayer perceptron(MLP) generator, the adjacency matrices is generated through an adversarial generative network model. The matrices are verified by a graph neural network discriminator. Numerical experiments demonstrate that the proposed method can efficiently generate preliminary structural topologies, providing significant assistance for similar complex architectural layouts.更多还原