Damage Prediction of Reinforced Concrete Slab under Underwater Near-Field Explosion Based on Machine Learning
-
摘要: 钢筋混凝土(RC)板是水下结构典型承重构件, 其水下爆炸损伤预测是工程防护领域的研究难点, 对结构健康状态监测、抗爆设计及安全性评估至关重要。文中提出一种将数值仿真与机器学习相结合的损伤预测新方法。首先, 基于LS-DYNA软件建立流固耦合模型, 采用任意拉格朗日-欧拉(ALE)算法模拟不同爆炸当量与距离下RC板的动态响应及损伤范围; 其次, 基于深度神经网络(DNN)构建损伤预测模型, 通过优化隐藏层结构与神经元配置, 显著提升精度(超98%)并避免过拟合; 进一步引入卷积神经网络(CNN)实现损伤图像自动识别, 大幅提高损伤预测效率。研究发现, RC板损伤区域在一定范围内呈现几何规律性。文中方法为水下爆炸损伤评估提供了新思路, 对防护工程设计具有一定参考价值。Abstract: Reinforced concrete(RC) slabs are typical load-bearing components of underwater structures, and accurately predicting their damage under underwater explosions remains a critical research challenge in protective engineering. Such prediction is essential for structural health monitoring, blast-resistant design, and safety assessment. This study proposed a damage prediction approach that integrated numerical simulation with machine learning. First, a fluid-structure coupling model was established in LS-DYNA software, and the arbitrary Lagrangian-Eulerian(ALE) algorithm was used to simulate the dynamic response and damage range of RC slabs subjected to different explosive charges and stand-off distances. Subsequently, a deep neural network(DNN)-based prediction model was developed, and its accuracy was significantly improved(exceeding 98%) through optimization of the hidden-layer architecture and neuron configuration, effectively avoiding overfitting. Furthermore, a convolutional neural network(CNN) was introduced for automatic recognition of damage images, greatly enhancing the efficiency of damage prediction. The results reveal that the damaged areas of RC slabs exhibit geometric regularity within a certain range. The proposed methodology provides a new perspective for underwater explosion damage assessment and offers valuable guidance for the design of protective engineering structures.
-
图 2 RC板水下近场爆炸有限元模型
Figure 2. Finite element model of underwater near-field explosion of reinforced concrete slab
图 5 炸药当量、爆炸距离、损伤范围三维数据库分布图
Figure 5. Three-dimensional database distribution map of explosive yield, explosion distance, and damage range
图 6 m=0.207 kg, d=0.32 m时RC板损伤预测结果与有限元结果对比
Figure 6. Comparison of RC slab damage prediction results with finite element results when m=0.207 kg and d=0.32 m
图 7 m=0.207 kg, d=0.32 m时RC板损伤轮廓
Figure 7. Damage outline of RC slab when m=0.207 kg, d=0.32 m
图 8 3种神经网络模型预测结果与有限元结果对比
Figure 8. Comparison between prediction results of three neural network models and finite element results
图 9 3种神经网络模型收敛效果对比
Figure 9. Comparison of convergence effects of three neural network models
图 10 不同神经元组合的模型预测结果MSE对比
Figure 10. Comparison of MSE prediction results for models with different neuronal combinations
图 11 3组工况下RC板损伤范围预测对照
Figure 11. Comparison of damage range predictions for RC slabs under three working conditions
图 12 3组工况RC板挠度预测对照
Figure 12. Comparison of deflection prediction for RC slabs under three working conditions
表 1 RHT本构模型材料参数
Table 1. RHT constitutive model of material parameters
ρ0 /(kg/m3) Fc /MPa Ft /MPa G A B N 2400 40.2 3.5 16 000 1.6 0.65 0.79 
下载: 导出CSV
表 2 JWL状态方程参数
Table 2. Parameter of JWL equation of state
ρ0 /(kg/m3) K1/MPa K2/MPa R1 R2 ω E0 /(J/kg) 1610 3.71×105 3.21×103 4.15 0.95 0.3 7×106
下载: 导出CSV
表 3 Gruneisen状态方程参数
Table 3. Parameter of Gruneisen equation of state
ρw0 /(kg/m3) c/ (m/s) S S1 S2 S3 1 000 1 480 0 2.56 −1.986 0.2268
下载: 导出CSV
-
[1] 马上, 王振清, 陈叶青, 等. 水中爆炸作用下混凝土重力坝损伤研究[J]. 混凝土, 2023(11): 63-71. [2] COLE R H. Underwater explosions[M]. Princeton: Princeton University Press, 1948. [3] WANG J, GUO J, YAO X L, et al. Dynamic buckling of stiffened plates subjected to explosion impact loads[J]. Shock Waves, 2017, 27: 37-52. doi: 10.1007/s00193-016-0638-z [4] HOLMQUIST T J, JOHNSON G R, COOK W H. A computational constitutive model for concrete subjected to large strains, high strain rates, and high pressures[J]. Proceedings of the Fourteenth International Symposium on Ballistics, 1993, 1: 591-600. [5] 周飞, 李贺东, 吴昊. 水中近场爆炸下钢筋混凝土板局部破坏模式数值模拟研究[J]. 浙江大学学报(英文版)(A辑:应用物理和工程), 2024, 25(8): 650-669.ZHOU F, LI H D, WU H, et al. Numerical simulation study on local failure mode of reinforced concrete slabs under near-field explosion in water[J]. Journal of Zhejiang University-SCIENCE A(Applied Physics & Engineering), 25(8), 650-669. [6] HOLMQUIST T J, JOHNSON G R. A computational constitutive model for glass subjected to large strains, high strain rates and high pressures[J]. Journal of Applied Mechanics, 2011, 109(12): 051003. [7] 刘丽滨, 李海涛, 刁爱民, 等. 水下爆炸船体梁总体响应特性数值模拟[J]. 高压物理学报, 2021, 35(6): 178-185.LIU L B, LI H T, DIAO A M, et al. Numerical simulation of overall response characteristics of ship beams under underwater explosion[J]. Chinese Journal of High Pressure Physics, 2021, 35(6): 178-185. [8] 朱玉富, 赵春风, 周志航. 基于机器学习的钢筋混凝土板在爆炸作用下的最大位移预测模型[J]. 高压物理学报, 2023, 37(2): 92-106.ZHU Y F, ZHAO C F, ZHOU Z H. Maximum displacement prediction model of reinforced concrete slab under blast loading based on machine learning[J]. Chinese Journal of High Pressure Physics, 2023, 37(2): 92-106. [9] Li Q L, Wang Z T, Chen W S, et al. Advancing blast fragmentation simulation of RC slabs: A graph neural network approach[J]. Engineering Structures, 2024, 308: 118009. doi: 10.1016/j.engstruct.2024.118009 [10] FARIS H. MIRJALILI S. ALJARAH I. Automatic selection of hidden neurons and weights in neural networks using grey wolf optimizer based on a hybrid encoding scheme[J]. International Journal of Machine Learning and Cybernetics, 2019, 10: 2901-2920. doi: 10.1007/s13042-018-00913-2 [11] 赵春风, 向思麒, 朱玉富. 基于机器学习的RC板爆炸两阶段损伤评估模型[J]. 建筑结构学报, 2025, 46(1): 212-222. [12] WANG W, ZHANG D, LU F Y, et al. Experimental study on scaling the explosion resistance of a oneway square reinforced concrete slab under a close-in blast loading[J]. International Journal of Impact Engineering, 2012, 49: 158-164. doi: 10.1016/j.ijimpeng.2012.03.010 [13] 刘靖晗, 唐廷, 韦灼彬, 等. 水下爆炸作用下高桩码头损伤特性数值模拟研究[J]. 兵器装备工程学报, 2024, 45(10): 53-60. [14] 胡锦华. 水下接触爆炸下钢筋混凝土板毁伤模式探究[J]. 工程爆破, 2022, 28(3): 17-23. [15] BILLINGS S A, VOON W S F. Least squares parameter estimation algorithms for non-linear systems. International[J] Journal of Systems Science, 1984, 15(6): 601-615. [16] HAGAN M T, MENHAJ M B. Training feed forward networks with the Marquardt algorithm[J]. IEEE Transactions on Neural Networks, 1994, 5(6): 989-993. doi: 10.1109/72.329697 [17] RUMELHART D E, HINTON G E, WILLIAMS R J. Learning representations by back-propagating errors[J]. Nature, 1986, 323(6088): 533-536. doi: 10.1038/323533a0 [18] HE K, ZHANG X, REN S, et al. Delving deep into rectifiers: Surpassing human-level performance on imagenet classification[C]//2015 IEEE International Conference on Computer Vision(ICCV). Santiago, Chile: ICCV, 2015: 1026-1034. [19] SCHMIDHUBER J. Deep learning in neural networks: An overview[J]. Neural networks, 2015, 61: 85-117. doi: 10.1016/j.neunet.2014.09.003 [20] ZHANG W, NIU L, ZHANG D, et al. HW-ADAM: FPGA-based accelerator for adaptive moment estimation[J]. Electronics, 2023, 12(2): 263. doi: 10.3390/electronics12020263 [21] WIANGKHAM A, ARIYARIT A, AENGCHUAN P. Prediction of the influence of loading rate and sugarcane leaves concentration on fracture toughness of sugarcane leaves and epoxy composite using artificial intelligence[J]. Theoretical and Applied Fracture Mechanics, 2022, 117: 103188. doi: 10.1016/j.tafmec.2021.103188 [22] REN S F, ZHAO P F, WANG S P, et al. Damage prediction of stiffened plates subjected to underwater contact explosion using the machine learning-based method[J]. Ocean Engineering, 2022, 266: 112839. doi: 10.1016/j.oceaneng.2022.112839 [23] 于玲玲, 张飞, 何文凯. 考虑地应力与先爆孔损伤的光面爆破数值模拟[J]. 地下空间与工程学报, 2024, 20(2): 636-644. [24] PENG X, HU Z, LI T. Neural network based combining prediction model and its application in ship motion prediction[C]//2010 8th World Congress on Intelligent Control and Automation. Jinan, China: IEEE, 2010: 5624-5627. [25] HINTON G E, SALAKHUTDINOV R R. Reducing the dimensionality of data with neural networks[J]. Science, 2006, 313(5786): 504-507. doi: 10.1126/science.1127647 [26] SRIVASTAVA N, HINTON G E, KRIZHEVSKY A, et al. Dropout: A simple way to prevent neural networks from overfitting[J]. The Journal of Machine Learning Research, 2014, 15(1): 1929-1958. -
点击查看大图
计量
- 文章访问数: 22
- HTML全文浏览量: 19
- PDF下载量: 2
- 被引次数: 0
