论文：Data enhancement for generative AI design of shear wall structures incorporating structural optimization and diffusion modelsDOI：https://doi.org/10.1177/13694332251353614 0太长不看版生成式结构设计的核心思路是AI可以从图纸中学习设计经验。然而，实际收集到的人工设计图纸往往质量良莠不齐，严重影响了生成式AI的学习效果。因此，我们提出：先全面优化设计图纸，提升训练集的数据质量，再让AI学习。通过结构优化得到高质量的设计图纸，训练出高质量的AI模型，从而生成高质量的设计结果。 1研究背景 课题组前期提出了系列基于生成式AI的建筑结构智能设计方法（生成对抗网络、图神经网络、扩散模型等），其核心思路都是让生成式AI学习既有图纸、文本中的设计经验，而后用户给定户型布置图和设计条件后，AI生成结构布置图。从而实现从无到有、从零到一的结构设计智能生成。 我们知道，AI的性能高度依赖于训练数据的质量。在前期的工作中，我们从不同渠道收集了大量人工设计的图纸，作为训练数据。然而，收集到的数据往往质量良莠不齐，低质量数据会给AI带来不利影响。这就好比学生学习的教案都有错误，自然没法写对作业。同时，既有的数据清洗技术主要关注图像、文本、表格、时序数据等通用数据，难以直接适用于高度专业化的结构设计数据。 2数据增强的生成式结构设计工作流 对此，我们提出了数据增强的生成式结构设计工作流，包括数据准备、数据增强、模型训练、模型预测与评估四个模块。相比传统工作流有三点改进，一是在数据准备中通过回归公式补充缺失的设计信息，得到完整信息；二是在数据增强中引入结构优化，提升设计数据质量，使其更加安全和经济；三是在模型评估中根据精细有限元得到设计指标（而非传统的IoU等相似性指标），进行更全面与合理的评价。 3数据集与回归分析以课题组研究较为充分的剪力墙结构为例开展数据增强研究。首先，构建了一个包含三百多个案例的剪力墙结构设计数据集。训练集包括设计条件、建筑设计、结构设计，用于训练扩散模型；还包括每个标准层的剪力墙厚度，材料等级，层高等必要设计信息，用于构建回归模型。测试集仅包括设计条件和建筑设计，用于测试模型。 目前，扩散模型等生成式AI仅关注剪力墙设计的核心部分——剪力墙布置。而要形成一个完整的设计方案，从而开展建模计算和优化，需要补充其他的必要设计信息。例如标准层的数量和划分，层高，材料等级等。我们对收集到的设计数据进行了统计分析和回归分析，从而确定了上述参数的初始取值。 类似地，对于连梁和框架梁的高度，我们统计了三千多根梁，结合跨高比要求和统计得到的高度范围，给出了初始设计阶段，梁高的确定公式。 4基于结构优化的设计数据质量增强为了对剪力墙设计数据进行质量提升，我们提出了一套结构优化算法。通过调整剪力墙长度，使设计方案满足合规性和经济性要求。具体而言，基于K-Means聚类算法，显著减少了优化变量的数量；通过合理的罚函数设置，同时考虑了多个关键的力学响应指标和材料成本指标。 在优化算法方面，采用了YJK-GAMA的Online Learning，这是一种有效的代理辅助进化算法。为了确定算法的超参数，选取了8个典型的剪力墙结构案例，开展了优化试算。考察了有限元计算次数与罚函数下降的关系，发现当有限元计算次数为100时能够最好地达到优化效果与耗时的平衡。 这里给出了3个典型的训练数据，优化前存在多种问题、数据质量低，优化后在位移角、剪重比、扭转周期比等方面合规性提升，或者材料用量减少。 总体上看，开展结构优化之后，训练集的数据质量明显提升。原来不合规的案例，力学罚函数平均降低14%；原来已合规的案例，材料用量平均降低3%。 5基于扩散模型的生成式结构设计基于增强后的数据集，训练了一个生成式AI。我们采用了课题组前期提出的扩散模型（Diffusion Model智能设计原理揭秘 | 论文和发明专利：基于扩散模型的剪力墙结构智能设计），根据建筑布置和设计条件来生成初步的剪力墙布置设计。相比生成对抗网络、变分自编码器等传统生成式AI，扩散模型能够生成更加高质量和多样化的结果。 对照实验表明，优质数据带来了测试集上AI设计质量的提升，原本不合规的案例，力学罚函数平均降低23%；原本已合规的案例，材料用量平均降低0.5%。 可以发现，材料用量的降低相比力学罚函数并不明显，这是因为设计中要优先满足安全性，然后才能考虑经济性。 典型案例研究表明，采用了本研究提出的数据增强方法之后，AI设计的合规性和经济性均有提升。左图案例的力学响应从位移角不合规变为合规，右图案例的材料用量节省了2.3%。可见优质的训练数据确实带来了更优质的AI设计。 6结语针对生成式结构设计普遍面临的数据质量问题，本研究提出了基于结构优化的数据增强方法，成功应用于基于扩散模型的剪力墙布置设计。通过提升用于训练AI的设计数据的质量，提高了AI设计的水平。本研究是对“AI+优化”的又一次尝试，尚存在诸多不足，欢迎各位专家批评指正！相关论文Liao WJ, Lu XZ, Huang YL, Zheng Z, Lin YQ, Automated structural design of shear wall residential buildings using generative adversarial networks, Automation in Construction, 2021, 132: 103931. DOI: 10.1016/j.autcon.2021.103931.Lu XZ, Liao WJ, Zhang Y, Huang YL, Intelligent structural design of shear wall residence using physics-enhanced generative adversarial networks, Earthquake Engineering & Structural Dynamics, 2022, 51(7): 1657-1676. DOI: 10.1002/eqe.3632.Zhao PJ, Liao WJ, Xue HJ, Lu XZ, Intelligent design method for beam and slab of shear wall structure based on deep learning, Journal of Building Engineering, 2022, 57: 104838. DOI: 10.1016/j.jobe.2022.104838.Liao WJ, Huang YL, Zheng Z, Lu XZ, Intelligent generative structural design method for shear-wall building based on “fused-text-image-to-image” generative adversarial networks, Expert Systems with Applications, 2022, 118530, DOI: 10.1016/j.eswa.2022.118530.Fei YF, Liao WJ, Zhang S, Yin PF, Han B, Zhao PJ, Chen XY, Lu XZ, Integrated schematic design method for shear wall structures: a practical application of generative adversarial networks, Buildings, 2022, 12(9): 1295. DOI: 10.3390/buildings1209129.Fei YF, Liao WJ, Huang YL, Lu XZ, Knowledge-enhanced generative adversarial networks for schematic design of framed tube structures, Automation in Construction, 2022, 144: 104619. DOI: 10.1016/j.autcon.2022.104619.Zhao PJ, Liao WJ, Huang YL, Lu XZ, Intelligent design of shear wall layout based on attention-enhanced generative adversarial network, Engineering Structures, 2023, 274: 115170. DOI: 10.1016/j.engstruct.2022.115170.Zhao PJ, Liao WJ, Huang YL, Lu XZ, Intelligent beam layout design for frame structure based on graph neural networks, Journal of Building Engineering, 2023, 63, Part A: 105499. DOI: 10.1016/j.jobe.2022.105499.Zhao PJ, Liao WJ, Huang YL, Lu XZ, Intelligent design of shear wall layout based on graph neural networks, Advanced Engineering Informatics, 2023, 55:101886, DOI: 10.1016/j.aei.2023.101886Liao WJ, Wang XY, Fei YF, Huang YL, Xie LL, Lu XZ, Base-isolation design of shear wall structures using physics-rule-co-guided self-supervised generative adversarial networks, Earthquake Engineering & Structural Dynamics, 2023, 52(11): 3281-3303. DOI:10.1002/eqe.3862.Feng YT, Fei YF, Lin YQ, Liao WJ, Lu XZ, Intelligent generative design for shear wall cross-sectional size using rule-embedded generative adversarial network, Journal of Structural Engineering-ASCE, 2023, 149(11). 04023161. DOI:10.1061/JSENDH.STENG-12206.Fei YF, Liao WJ, Lu XZ, Guan H, Knowledge-enhanced graph neural networks for construction material quantity estimation of reinforced concrete buildings, Computer-Aided Civil and Infrastructure Engineering, 2024, 39(4): 518-538. DOI: 10.1111/mice.13094.Zhao PJ, Fei YF, Huang YL, Feng YT, Liao WJ, Lu XZ, Design-condition-informed shear wall layout design based on graph neural networks, Advanced Engineering Informatics, 2023, 58: 102190. DOI: 10.1016/j.aei.2023.102190.Fei YF, Liao WJ, Lu XZ, Taciroglu E, Guan H, Semi-supervised learning method incorporating structural optimization for shear-wall structure design using small and long-tailed datasets, Journal of Building Engineering, 2023, 79: 107873. DOI：10.1016/j.jobe.2023.107873Liao WJ, Lu XZ, Fei YF, Gu Y, Huang YL, Generative AI design for building structures, Automation in Construction, 2024, 157: 105187. DOI: 10.1016/j.autcon.2023.105187Zhao PJ, Liao WJ, Huang YL, Lu XZ, Beam layout design of shear wall structures based on graph neural networks, Automation in Construction, 2024, 158: 105223. DOI: 10.1016/j.autcon.2023.105223Qin SZ, Liao WJ, Huang SN, Hu KG, Tan Z, Gao Y, Lu XZ, AIstructure-Copilot: assistant for generative AI-driven intelligent design of building structures, Smart Construction, 2024, DOI: 10.55092/sc20240001Gu Y, Huang YL, Liao WJ, Lu XZ, Intelligent design of shear wall layout based on diffusion models, Computer-Aided Civil and Infrastructure Engineering, 2024, 39(23):3610-3625. DOI: 10.1111/mice.13236Fei YF, Liao WJ, Zhao PJ, Lu X*, Guan H, Hybrid surrogate model combining physics and data for seismic drift estimation of shear-wall structures, Earthquake Engineering & Structural Dynamics, 2024, 53(10): 3093-3112. DOI: 10.1002/eqe.4151Han J, Lu XZ, Gu Y, Cai Q, Xue HJ, Liao WJ, Optimized data representation and understanding method for the intelligent design of shear wall structures, Engineering Structures, 2024, 315: 118500. DOI: 10.1016/j.engstruct.2024.118500Qin SZ, Guan H, Liao WJ, Gu Y, Zheng Z, Xue HJ, Lu XZ, Intelligent design and optimization system for shear wall structures based on large language models and generative artificial intelligence, Journal of Building Engineering, 2024, 95: 109996. DOI: 10.1016/j.jobe.2024.109996Wang ZH, Yue Y, Chen Y, Liao WJ, Li CS, Hu KG, Tan Z, Lu XZ. Expert experience-embedded evaluation and decision-making method for intelligent design of shear wall structures. Journal of Computing in Civil Engineering-ASCE, 2025, 39(1). DOI: 10.1061/JCCEE5.CPENG-6076Tan Z, Qin SZ, Hu KG, Liao WJ, Gao Y, Lu XZ, Intelligent generation and optimization method for the retrofit design of RC frame structures using buckling-restrained braces, Earthquake Engineering & Structural Dynamics, 2025, 54(2): 530-547. DOI: 10.1002/eqe.4268Yu Y, Chen Y, Liao WJ, Wang ZH, Zhang SL, Kang YJ, Lu XZ, Intelligent generation and interpretability analysis of shear wall structure design by learning from multidimensional to high-dimensional features, Engineering Structures, 2025, 325: 119472. DOI: 10.1016/j.engstruct.2024.119472Qin SZ, Liao WJ, Huang YL, Zhang Shulu, Gu Y, Han J, Lu XZ, Intelligent design for component size generation in reinforced concrete frame structures using heterogeneous graph neural networks, Automation in Construction, 2025, 171: 105967. Xia JK, Liao WJ, Han B, Zhang SL, Lu XZ, Intelligent co-design of shear wall and beam layouts using a graph neural network, Automation in Construction, 2025, 172: 106024. Qin SZ, Liao WJ, Tan Z, Hu KG, Gao Y, Lu XZ, Comparative analysis of intelligent retrofit design methods of RC frame structures using buckling-restrained braces. Bulletin of Earthquake Engineering, 2025, DOI: 10.1007/s10518-025-02164-3Liao WJ, Zhang ZL, Liu B, Lu XZ, Liu DF, Liu Q, Duan ZJ, Liu C, Intelligent zoning design of concrete-faced rockfill dams using image-parameter fusion enhanced generative adversarial networks, Engineering Structures, 2025, 339: 120662. DOI: 10.1016/j.engstruct.2025.120662Qin SZ, Fei YF, Liao WJ, Lu XZ*, Leveraging data-driven artificial intelligence in optimization design for building structures: A review, Engineering Structures, 2025, 341: 120810. DOI: 10.1016/j.engstruct.2025.120810 来源：陆新征课题组
