To address the high computational cost of the traditional finite element method in the multi-scenario assessment of aging building blast damage, a real-time simulation model of blast damage based on graph sample and aggregate (GraphSAGE) and long short-term memory (LSTM) is proposed. Firstly, the spatial three-dimensional grid is used to segment the building data, and the decomposed building grid is imported into Abaqus finite element simulation software to construct a high-precision dataset of blast damage. Secondly, GraphSAGE is employed to capture building spatial features, while LSTM is used to model the temporal features evolution of blast damage. Finally, a multi-task decoder is introduced to achieve joint temporal prediction of displacement fields and failure states. Experimental results demonstrate that the proposed model maintains building displacement prediction accuracy above 96% and damage identification PR-AUC above 95%, while achieving a computational speed nearly 40 times higher than that of the finite element method.
为解决上述问题,本文提出一种结合GraphSAGE和LSTM的爆炸毁伤实时预测模型。首先,利用GeoSOT-3D全球立体网格[14]对建筑信息模型(Building Information Modeling, BIM)进行剖分,将剖分后的建筑网格导入Abaqus软件中进行仿真计算,得到高保真建筑爆炸数据集;其次,结合GraphSAGE的空间拓扑建模能力与LSTM的时序建模能力,实现了对建筑物局部与整体响应特征的刻画以及位移场与失效状态的动态演化预测;最后,建立多任务解码器,实现位移场与失效状态的联合时序预测。通过对比有限元仿真算法和当前主流GNN与RNN网络,验证本文模型在位移场精度、失效场精度、训练速度以及推理速度上的优势。本文研究可为老旧房屋爆破拆除过程中的安全评估、多方案比选、方案优化与快速决策提供可靠且高效的技术支撑。
PARKD, LEET H, LEEY, et al. Blast simulations of a reinforced concrete slab using the continuous surface cap model (CSCM)[J]. Journal of Building Engineering, 2024,96:No.110603.
PFAFFT, FORTUNATOM, SANCHEZ-GONZALEZA,et al. Learning mesh-based simulation with graph networks[C]//Proceedings of the 9th International Conference on Learning Representations. Appleton, USA: ICLR, 2021. DOI: 10.48550/arXiv.2010.03409 .
[5]
SANCHEZ-GONZALEZA, GODWINJ, PFAFFT, et al. Learning to simulate complex physics with graph networks[C]//Proceedings of the 37th International Conference on Machine Learning. New York, USA: ACM, 2020:8459-8468.
[6]
KIPFT N, WELLINGM. Semi-supervised classification with graph convolutional networks[C]//Proceedings of the 5th International Conference on Learning Representations. Appleton, USA: ICLR, 2017. DOI: 10.48550/arXiv.1609.02907 .
[7]
VELICKOVICP, CUCURULLG, CASANOVAA, et al. Graph attention networks[C]//Proceedings of the 6th International Conference on Learning Representations. Appleton, USA: ICLR, 2018. DOI: 10.48550/arXiv.1710.10903 .
[8]
HAMILTONW L, YINGR, LESKOVECJ. Inductive representation learning on large graphs[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook, USA: Curran Associates Inc., 2017:1025-1035.
[9]
CHOK, VAN MERRIENBOERB, GULCEHREC, et al. Learning phrase representations using RNN encoder-decoder for statistical machine translation[C]//Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing. Stroudsburg, USA: ACL, 2014:1724-1734.
[10]
LEAC, FLYNNM D, VIDALR, et al. Temporal convolutional networks for action segmentation and detection[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, USA: IEEE, 2017:1003-1012.
[11]
HOCHREITERS, SCHMIDHUBERJ. Long short-term memory[J]. Neural Computation, 1997,9(8):1735-1780.
[12]
WANGZ Q, PENGJ Z, HUJ, et al. BlastGraphNet: an intelligent computational method for the precise and rapid prediction of blast loads on complex 3D buildings using graph neural networks[J]. Engineering, 2025,49:205-224.
[13]
LIY G, YUR, SHAHABIC, et al. Diffusion convolutional recurrent neural network: data-driven traffic forecasting[C]//Proceedings of the 6th International Conference on Learning Representations. Appleton, USA: ICLR, 2018. DOI: 10.48550/arXiv.1707.01926 .
HYDED W. Microcomputer programs CONWEP and FUNPRO, applications of TM 5-855-1, ‘fundamentals of protective design for conventional weapons’ (user’s guide)[R]. Vicksburg, USA: US Army Engineer Waterways Experiment Station, 1988:1-22.
GREFFK, SRIVASTAVAR K, KOUTNÍKJ, et al. LSTM: a search space odyssey[J]. IEEE Transactions on Neural Networks and Learning Systems, 2017,28(10):2222-2232.
[18]
DAW A, KARPATNEA, WATKINSW D, et al. Physics-guided neural networks (PGNN): an application in lake temperature modeling[M]//Knowledge-guided machine learning. Boca Raton, USA: Chapman and Hall/CRC, 2022:353-372.
[19]
WILLMOTTC J, MATSUURAK. Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance[J]. Climate Research, 2005,30:79-82.
[20]
BISHOPC M, NASRABADIN M. Pattern recognition and machine learning[M]. New York, USA: Springer, 2006:192-195.
[21]
LINT Y, GOYALP, GIRSHICKR, et al. Focal loss for dense object detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020,42(2):318-327.
[22]
CIPOLLAR, GAL Y, KENDALLA. Multi-task learning using uncertainty to weigh losses for scene geometry and semantics[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway, USA: IEEE, 2018:7482-7491.
[23]
HYNDMANR J, KOEHLERA B. Another look at measures of forecast accuracy[J]. International Journal of Forecasting, 2006,22(4):679-688.
[24]
DAVISJ, GOADRICHM. The relationship between precision-recall and ROC curves[C]//Proceedings of the 23rd International Conference on Machine Learning. New York, USA: ACM, 2006:233-240.