基于深度学习的水下爆炸冲击响应谱求解器

王爽; 吕峰; 马峰; 陈思; 朱炜; 韩峰; 黄沁怡

doi:10.11993/j.issn.2096-3920.2024-0144

基于深度学习的水下爆炸冲击响应谱求解器

doi: 10.11993/j.issn.2096-3920.2024-0144

王爽¹,
吕峰²,
马峰^1,,
陈思¹,
朱炜¹,
韩峰¹,
黄沁怡¹

1.
北京理工大学爆炸科学与技术国家重点实验室, 北京, 100081
2.
中国人民解放军32398部队, 北京, 100026

基金项目: 国家自然基金重点项目资助(U20A2071); 爆炸科学与安全防护全国重点实验室自主课题重点项目(ZDKT24-01).

详细信息

作者简介:
王爽：王　爽(1995-), 男, 博士研究生, 主要研究方向为水中爆炸、舰艇冲击毁伤等

通讯作者:
马　峰(1973-), 男, 教授, 博士生导师, 主要研究方向为水中爆炸, 舰艇冲击毁伤等.

中图分类号: U674.7
计量
- 文章访问数: 35
- HTML全文浏览量: 14
- PDF下载量: 6
- 被引次数: 0
出版历程
- 收稿日期: 2024-10-13
- 修回日期: 2024-11-22
- 录用日期: 2024-12-04
- 网络出版日期: 2024-12-27

A Deep Learning-Based Solver for Underwater Explosion Shock Response Spectrum

1.
State Key Laboratory of Explosion Science and Technology, Beijing Institute of Technology, Beijing 100081, China
2.
Unit32398S1, the Liberation Army of China, Beijing 100026, China

摘要

摘要: 由于冲击响应具有短时性和复杂性, 常使用冲击响应谱(SRS)作为分析冲击响应的工具。为克服传统冲击响应谱求解方法存在的计算速度与计算精度之间的矛盾, 文中提出基于深度学习的冲击响应谱快速求解器, 并根据冲击响应谱的特点设计自适应阈值选择机制, 提升求解器计算精度。对比求解器得到的冲击响应谱与采用传统方法计算的结果, 两者显示出高度一致性, 从而验证了求解器的有效性。此外, 文中在求解过程中引入L2正则化技术, 有效避免了过拟合现象的发生, 进一步增强了求解器的鲁棒性。
- 舰船 /
- 水下爆炸 /
- 冲击响应谱 /
- 深度学习
Abstract: Due to the short-duration and complexity of shock responses, Shock Response Spectrum(SRS) is commonly used as a tool for analyzing these responses. To address the trade-off between calculation speed and accuracy inherent in traditional SRS solving methods, this paper proposes a deep learning-based fast solver for shock response spectra. An adaptive threshold selection mechanism tailored to the characteristics of shock response spectra is designed to improve the solver's accuracy. A comparison between the SRS obtained by the proposed solver and the results calculated using traditional methods demonstrates a high degree of consistency, validating the effectiveness of the solver. Additionally, L2 regularization is introduced in the solution process, effectively preventing overfitting and further enhancing the robustness of the solver.
- ship /
- underwater explosion /
- shock response spectrum /
- deep learning

HTML全文

图 1 时、频域下的冲击信号

Figure 1. Shock response in time and frequency domains

下载: 全尺寸图片幻灯片

图 2 冲击谱计算示意图

Figure 2. Illustration of shock spectrum calculation

下载: 全尺寸图片幻灯片

图 3 求解器模型架构

Figure 3. Solver model architecture

下载: 全尺寸图片幻灯片

图 4 自适应阈值选择机制对比实验

Figure 4. Comparison experiment on adaptive threshold selection mechanism

下载: 全尺寸图片幻灯片

图 5 不同方法结果对比

Figure 5. Comparison of results using different methods

下载: 全尺寸图片幻灯片

图 6 采用正则化方法对过拟合影响比较

Figure 6. Comparison of the impact of overfitting using regularization methods

下载: 全尺寸图片幻灯片

表 1 自适应阈值选择机制对比实验

Table 1. Comparison experiment on adaptive threshold selection mechanism

是否使用阈值选择机制	RE/%	RMSE
是	5.64	0.135
否	15.8	0.641

下载: 导出CSV

表 2 数值仿真工况表

Table 2. Table of numerical simulation conditions

序号	水深/m	爆距/m	装药/kg	方位
1	100	7	295	侧前方
2	100	7	295	正下方
3	100	7	295	侧后方
4	100	7	45	正下方

下载: 导出CSV

表 3 模型数据集信息

Table 3. Model Dataset Information

模型输入/输出数据信息		数据集划分信息
输入	冲击载荷时域加速度	训练集	80%
输入	冲击载荷时域加速度	测试集	20%
输出	感兴趣点冲击响应谱	训练集	80%
输出	感兴趣点冲击响应谱	测试集	20%

下载: 导出CSV

表 4 求解器与传统冲击谱计算方法性能对比

Table 4. Comparison between solver and traditional shock spectrum calculation methods

自然频率	指标	求解器	直接积分法	数字滤波器	龙格库塔法
100	核时	1.2 s	107 s	62 s	78 s
100	RE	5.54 %	−	15.48 %	8.29 %
500	核时	1.5 s	310 s	185 s	206 s
500	RE	5.71 %	−	15.93 %	8.54 %
1000	核时	1.6 s	539 s	249 s	320 s
1000	RE	6.18 %	−	16.84 %	8.71 %

下载: 导出CSV

参考文献(15)

[1]	姚熊亮, 侯明亮, 李青, 等. Y型舷侧结构抗冲击性能数值仿真实验研究[J]. 哈尔滨工程大学学报, 2006, 27(6): 796-801. doi: 10.3969/j.issn.1006-7043.2006.06.002 YAO X L, HOU M L, LI Q, et al. Numerical simulation research on counter-impingement capability of Y-shape shipboard structure[J]. Journal of Harbin Engineering University, 2006, 27(6): 796-801. doi: 10.3969/j.issn.1006-7043.2006.06.002
[2]	尹群, 陈永念, 张健, 等. 水下爆炸载荷作用下舰船结构动响应及新型防护结构[J]. 中国造船, 2007, 48(4): 42-52. YIN Q, CHEN Y N, ZHANG J, et al. Dynamic response of ship structures under underwater explosion loads and new protective structures[J]. Shipbuilding of China, 2016, 11(4): 51-58.
[3]	张振华, 牛闯, 钱海峰, 等. 六层金字塔点阵夹芯板结构在水下近距爆炸载荷下的冲击实验[J]. 中国舰船研究, 2016, 11(4): 51-58. ZHANG Z H, NIU C, QIAN H F, et al. Impact experiment of six-layer pyramidal lattices sandwich panels subjected to near field underwater explosion[J]. Chinese Journal of Ship Research, 2016, 11(4): 51-58.
[4]	WANG H, CHENG Y S, LIU J, et al. The Fluid-Solid Interaction Dynamics between Underwater Explosion Bubble and Corrugated Sandwich Plate[J]. Shock and Vibration, 2016, 2016(3): 1-21.
[5]	BATRA R C, HASSAN N M. Response of Fiber Reinforced Composites to Underwater Explosive Loads[J]. Composites Part B Engineering, 2007, 38(4): 448-468. doi: 10.1016/j.compositesb.2006.09.001
[6]	LEBLANC J, SHUKLA A. Dynamic Response of Curved Composite Panels to Underwater Explosive Loading: Experimental and Computational Comparisons[J]. Composite Structures, 2011, 93(11): 3072-3081. doi: 10.1016/j.compstruct.2011.04.017
[7]	LEBLANC J, SHUKLA A. Response of Polyurea-coated Flat Composite Plates to Underwater Explosive Loading[J]. Journal of Composite Materials, 2015, 49(8): 965-980. doi: 10.1177/0021998314528263
[8]	LUL, MENGX H, MAOZ P, et al. DeepXDE: a deep learning library for solving differential equations[J]. SIAM Review, 2021, 63(1): 208-228. doi: 10.1137/19M1274067
[9]	冯麟涵, 杨俊杰, 焦立启. 基于RBF神经网络的船舶冲击谱速度数据挖掘与预报[J]. 振动与冲击, 2022, 41(13): 189-194. FENG L H, YANG J J, JIAO L Q. Data mining and prediction of ship shock spectral velocity based on RBF neural network[J]. Journal of Vibration and Shock, 2022, 41(13): 189-194.
[10]	ZHOU Y, MENG S, LOU Y, et al. Physics-Informed Deep Learning-Based Real-Time Structural Response Prediction Method[J]. Engineering, 2024, 35(4): 140-157. doi: 10.1016/j.eng.2023.08.011
[11]	高明贺, 石成英, 王游. 冲击响应谱分析方法研究[J]. 科技视界, 2012(28): 117-194. GAO M H, SHI C Y, WANG Y. Research on the Analysis Method of Shock Response Spectrum[J]. Science & Technology Vision, 2012(28): 117-194.
[12]	施广宏, 石成英, 韩华锋. 系统部件对冲击载荷的响应分析[J]. 电子产品可靠性与环境试验, 2010(4): 24-26. XIE G H, SHI C Y, HAN H F. Analysis for Shock Response of Parts[J]. Electronic Product Reliability and Environmental Testing, 2010(4): 24-26.
[13]	谢浩, 冯麟涵, 吴静波, 李晓文, 郭君. 舰船冲击谱若干计算方法比较研究[J]. 噪声与振动控制, 2017, 37 (4): 115-120. XIE H, FENG L H, WU J B, et al. Comparative Study on Several Calculation Methods for Ship’s Shock Spectra[J]. Noise and Vibration Control, 2017, 37 (4): 115-120.
[14]	SMALLWOOD D. Improved recursive formula for calculating shock response spectra[J]. 1980.
[15]	武国宁, 胡汇丰, 于萌萌. 深度学习中的正则化方法研究[J]. 计算机科学与应用, 2020, 10(6): 1224-1233. doi: 10.12677/CSA.2020.106126 WU G N, HU H F, YU M M. Regularization Methods in Deep Learning[J]. Computer Science and Application, 2020, 10(6): 1224-1233. doi: 10.12677/CSA.2020.106126