基于BP神经网络的爆炸用激波管峰值压力预测方法

陈梓薇; 王仲琦; 曾令辉

doi:10.11883/bzycj-2023-0187

基于BP神经网络的爆炸用激波管峰值压力预测方法

doi: 10.11883/bzycj-2023-0187

北京理工大学爆炸科学与技术国家重点实验室，北京 100081

基金项目: 国家重点研发计划 (2021YFC3001204)

详细信息

作者简介:
陈梓薇（1998－　），女，硕士研究生，czw_bit@163.com

通讯作者:
王仲琦（1972－　），男，博士，副教授，czqwang@bit.edu.cn

中图分类号: O383
计量
- 文章访问数: 126
- HTML全文浏览量: 86
- PDF下载量: 84
- 被引次数: 0
出版历程
- 收稿日期: 2023-05-18
- 修回日期: 2024-03-15
- 网络出版日期: 2024-03-26
- 刊出日期: 2024-05-08

A method for predicting peak pressure in an explosion shock tube based on BP neural network

State Key Laboratory of Explosive Science and Technology, Beijing Institute of Technology, Beijing 100081, China

摘要

摘要: 针对爆炸用激波管缺乏相应的经验公式和数值模拟时效性差的问题，同时为了快速得到激波管内的峰值压力，建立预测爆炸用激波管试验段峰值压力的四层反向传播（back propagation，BP）神经网络。采用数值模拟方法计算激波管试验段峰值压力，计算结果与激波管爆炸试验结果进行对比，平均相对误差为2.69%。证明激波管数值模型的准确性后，将数值模拟得到的195组激波管测得的峰值压力作为输出层，激波管驱动段TNT的药量、药柱的长径比以及爆炸比例距离作为神经网络的输入层。为了加快神经网络迭代速度和提高预测精度，使用自适应矩估计（adaptive moment estimation，ADAM）算法作为神经网络误差梯度下降的优化算法。结果表明，训练好的神经网络得到的预测结果与模拟值基本吻合，预测结果与数值模拟结果的平均相对误差为3.26%。BP神经网络模型能够反映激波管爆炸的峰值压力与影响因素之间的映射关系，采用BP 神经网络模型计算时比数值模拟节约了大量运算时间。
- BP神经网络 /
- 激波管 /
- 峰值压力 /
- 自适应矩估计
Abstract: In response to the problems of the lack of corresponding empirical formulas and the poor timeliness of simulation for the explosive shock tube, and to quickly obtain the peak pressure of the shock tube used in explosions, a four-layer back propagation (BP) neural network was established to predict the peak pressure in the experimental section of the shock tube. After verifying the grid independence, numerical simulation was used to calculate the peak pressure of the test section of the shock tube, and the simulation data were compared with the experimental data of the shock tube explosion, and the average relative error is 2.49%. After proving the accuracy of the numerical simulation values, the 195 sets of peak pressure obtained from the numerical simulation in the shock tube test section were used as the output layer, and the TNT dosage in the shock tube driving section, aspect ratio of the charge column, and explosion proportional distance were used as the input layer for BP neural network training. To speed up the neural network iterations and increase the prediction accuracy, Adam's algorithm was used as an optimization algorithm for neural network error gradient descent. The results show that the predicted results obtained through the trained neural network are basically consistent with the simulated values, and the average relative error between the predicted results and the numerical values is 3.26%. In contrast to the evaluation metrics obtained using multiple regression analysis (mean absolute error (MAE) of 480 and coefficient of determination (R²) of 0.58), the four-layer BP neural network obtains a MAE of 25.4 and an R² of 0.99 for the validation set. The BP neural network model can reflect the mapping relationship between the peak pressure of the shock tube explosion and the influencing factors, and improve several times compared with the time required for numerical simulation, so it has the value of practical engineering applications.
- BP neural network /
- shock tube /
- peak pressure /
- adaptive moment estimation

HTML全文

图 1 爆炸用激波管示意图

Figure 1. Schematic diagram of the explosion shock tube

下载: 全尺寸图片幻灯片

图 2 激波管数值模型

Figure 2. Simulation model of the shock tube

下载: 全尺寸图片幻灯片

图 3 峰值压力与网格独立性验证的比较

Figure 3. Comparison of peak pressure and grid independence verification

下载: 全尺寸图片幻灯片

图 4 激波管试验装置示意图及测点布设图

Figure 4. Schematic diagram of the shock tube and the location of the measurement points of the shock tube

下载: 全尺寸图片幻灯片

图 5 不同工况下的压力时程曲线

Figure 5. Pressure-time curves for different cases

下载: 全尺寸图片幻灯片

图 6 预测激波管峰值压力的神经网络拓扑结构示意图

Figure 6. Schematic diagram of neural network topology for predicting peak pressure in a shock tube

下载: 全尺寸图片幻灯片

图 7 预测峰值压力与数值峰值压力对比图

Figure 7. Comparison of predicted and simulated peak pressure

下载: 全尺寸图片幻灯片

图 8 BP神经网络预测拟合图

Figure 8. BP neural network prediction fitting diagram

下载: 全尺寸图片幻灯片

表 1 材料参数

Table 1. Material parameters

材料	密度/(kg·m⁻³)	定压比热/(J·kg⁻¹·K⁻¹)	抗拉强度/MPa
4340钢	7830	477	818
TNT	1630
空气	1.225	717.6

下载: 导出CSV

表 2 部分激波管爆炸数值模拟数据

Table 2. The partial data of shock tube explosion by numerical simulation

药量/g	长径比	爆炸比例距离/(mm·g^−1/3)	峰值压力模拟值/kPa
10.4	2.00	3394.74	147.84
70.1	0.50	1185.72	182.17
70.1	4.00	899.78	848.38
166.2	0.50	343.76	997.61
186.9	1.33	505.42	902.55
249.3	0.75	1110.63	238.39
280.5	0.25	456.77	467.08
473.2	1.00	499.19	1352.11
584.3	0.90	387.55	1298.94

下载: 导出CSV

表 3 不同工况下测点测量的峰值超压

Table 3. Different cases of peak overpressure obtained from measurement points

测点	峰值压力/MPa
测点	5 g TNT	10 g TNT	20 g TNT	25 g TNT	30 g TNT
1	0.250	0.430	0.740	1.040	0.857
2	0.225	0.358	0.534	0.597	0.588
3	0.189	0.220	0.389	0.417	0.442
4	0.088	0.136	0.145	0.179	0.192
5	0.081	0.131	0.172	0.202	0.209

下载: 导出CSV

表 4 工况2中测点的试验数据与数值模拟结果对比

Table 4. Comparison of experimental data and numerical simulation results at measurement points in the case 2

测点	测点距原点距离/m	峰值压力/MPa
测点	测点距原点距离/m	试验	数值模拟	相对误差/%
1	1.889	0.430	0.416	3.32
2	2.772	0.358	0.346	3.33
3	4.389	0.220	0.214	2.50
4	5.000	0.136	0.141	−3.64
5	7.962	0.131	0.130	0.67

下载: 导出CSV

表 5 压力峰值试验数据与预测结果对比

Table 5. Comparison of peak pressure between test date and predicted result

TNT药量/g	测点离原点位置/m	峰值压力/kPa
TNT药量/g	测点离原点位置/m	试验	预测	相对误差/%
5	2.772	227	247.73	4.72
10	2.772	358	349.51	−2.37
10	6.481	140	134.74	−3.75
20	1.889	740	763.84	3.22
20	3.555	410	416.60	1.61
25	1.889	1040	1076.29	3.49
30	2.772	588	594.28	1.06
30	9.443	186	177.91	−4.34

下载: 导出CSV

表 6 预测评价指标

Table 6. Prediction evaluation indicators

预测模型	MAE	RMSE	MAPE%	R²
BP神经网络	25.4	43.3	3.47	0.99
多元线性回归	480	597.7	85.8	0.58

下载: 导出CSV

参考文献(18)

[1]	徐春光, 白晓征, 刘瑜, 等. 爆炸激波管管口稀疏波对试验段的影响 [J]. 国防科技大学学报, 2011(4): 1–5. DOI: 10.3969/j.issn.1001-2486.2011.04.001. DOI: 10.3969/j.issn.1001-2486.2011.04.001. XU C G, BAI X Z, LIU Y, et al. Research on the influence of rarefaction wave to the experimental section in blast shock tube [J]. Journal of National University of Defense Technology, 2011(4): 1–5. DOI: 10.3969/j.issn.1001-2486.2011.04.001. DOI: 10.3969/j.issn.1001-2486.2011.04.001.
[2]	张军, 黄含军, 王军评, 等. 炸药驱动式爆炸管的载荷计算 [J]. 装备环境工程, 2021, 18(5): 21–27. DOI: 10.7643/issn.1672-9242.2021.05.004. ZHANG J, HUANG H J, WANG J P, et al. Simulation on the blast load inside the explosively drived shock tube [J]. Equipment Environment Engineering, 2021, 18(5): 21–27. DOI: 10.7643/issn.1672-9242.2021.05.004.
[3]	崔云霄, 王万鹏, 王雷元, 等. 压缩气体驱动大型激波管内部流场的数值模拟 [C]// 2014年中国计算力学大会论文集. 2014: 528–534. CUI Y X, WANG W P, WANG L Y, et al. Numerical simulation of Flow-Field in shockwave tube driven by condensed gas [C]//Proceedings of the 2014 Chinese Conference on Computational Mechanics, 2014: 528–534.
[4]	ISMAIL A, EZZELDIN M, EL-DAKHAKHNI W, et al. Blast load simulation using conical shock tube systems [J]. International Journal of Protective Structures, 2020, 11(2): 135–158. DOI: 10.1177/2041419619858098.
[5]	刘瑞朝, 任辉启, 徐翔云. 大型爆炸激波管数值模拟 [J]. 防护工程, 2009, 31(1): 31–35. LIU R C, REN H Q, XU X Y. Numerical simulation of large blast simulators [J]. Protective Engineering, 2009, 31(1): 31–35.
[6]	谢全民. 爆破振动信号分形维数相关性研究 [J]. 振动与冲击, 2021, 40(22): 48–51, 59. DOI: 10.13465/j.cnki.jvs.2021.22.007. XIE Q M. Correlation study on the fractal dimension of blasting vibration signals [J]. Journal of Vibration and Shock, 2021, 40(22): 48–51, 59. DOI: 10.13465/j.cnki.jvs.2021.22.007.
[7]	XU Y, HUANG Y M, MA G W. A beetle antennae search improved BP neural network model for predicting multi-factor-based gas explosion pressures [J]. Journal of Loss Prevention in the Process Industries, 2020, 65: 104117. DOI: 10.1016/j.jlp.2020.104117.
[8]	袁格侠, 刘宏昭, 钱学梅, 等. 求解超高压筒形容器爆破压力的神经网络方法 [J]. 兵器材料科学与工程, 2010, 33(2): 31–34. DOI: 10.3969/j.issn.1004-244X.2010.02.009. YUAN G X, LIU H Z, QIAN X M, et al. ANN-based prediction of bursting pressure under ultra-high pressure for cylindrical vessel [J]. Ordnance Material Science and Engineering, 2010, 33(2): 31–34. DOI: 10.3969/j.issn.1004-244X.2010.02.009.
[9]	唐泽斯, 郭进, 王金贵, 等. 基于人工神经网络的气体泄爆最大超压预测研究 [J]. 中国安全生产科学技术, 2020, 16(4): 56–62. DOI: 10.11731/j.issn.1673-193x.2020.04.009. TANG Z S, GUO J, WANG J G, et al. Study on prediction of maximum overpressure in gas explosion venting based on artificial neural network [J]. Journal of Safety Science and Technology, 2020, 16(4): 56–62. DOI: 10.11731/j.issn.1673-193x.2020.04.009.
[10]	李江涛. 矿井瓦斯爆炸超压值分布的神经网络预测方法 [J]. 煤矿安全, 2013, 44(2): 157–160. DOI: 10.13347/j.cnki.mkaq.2013.02.012. LI J T. Overpressure distribution prediction of mine gas explosion based on artificial neural networks [J]. Safety in Coal Mines, 2013, 44(2): 157–160. DOI: 10.13347/j.cnki.mkaq.2013.02.012.
[11]	严国建, 周明安, 余轮, 等. 空气中爆炸冲击波超压峰值的预测 [J]. 采矿技术, 2011, 11(5): 89–90, 112. DOI: 10.3969/j.issn.1671-2900.2011.05.035. YAN G J, ZHOU M A, YU L, et al. Prediction of the peak overpressure of explosive shock waves in air [J]. Mining Technology, 2011, 11(5): 89–90, 112. DOI: 10.3969/j.issn.1671-2900.2011.05.035.
[12]	施建俊, 李庆亚, 张琪, 等. 基于Matlab和BP神经网络的爆破振动预测系统 [J]. 爆炸与冲击, 2017, 37(6): 1087–1092. DOI: 10.11883/1001-1455(2017)06-1087-06. SHI J J, LI Q Y, ZHANG Q, et al. Forecast system for blasting vibration velocity peak based on Matlab and BP neural network [J]. Explosion and Shock Waves, 2017, 37(6): 1087–1092. DOI: 10.11883/1001-1455(2017)06-1087-06.
[13]	GUO Q P, YANG S J, WANG Y C, et al. Prediction research for blasting peak particle velocity based on random GA-BP network group [J]. Arabian Journal of Geosciences, 2022, 15(15): 1351. DOI: 10.1007/s12517-022-10615-3.
[14]	郭璇, 马思远, 郭一帆, 等. 基于BP神经网络的围岩介质爆炸峰值压力预测 [J]. 振动与冲击, 2019, 38(3): 199–206. DOI: 10.13465/j.cnki.jvs.2019.03.028. GUO X, MA S Y, GUO Y F, et al. Blast peak pressure prediction for surrounding rock medium based on BP neural network method [J]. Journal of Vibration and Shock, 2019, 38(3): 199–206. DOI: 10.13465/j.cnki.jvs.2019.03.028.
[15]	陈俊霖, 侯麒麟, 宁文学, 等. 基于BP神经网络的火箭弹射击效率计算方法 [J]. 火力与指挥控制, 2022, 47(6): 87–92. DOI: 10.3969/j.issn.1002-0640.2022.06.014. CHEN J L, HOU Q L, NING W X, et al. Calculation method of rocket firing efficiency based on BP neural network [J]. Fire Control & Command Control, 2022, 47(6): 87–92. DOI: 10.3969/j.issn.1002-0640.2022.06.014.
[16]	焦李成, 杨淑媛, 刘芳, 等. 神经网络七十年: 回顾与展望 [J]. 计算机学报, 2016, 39(8): 1697–1716. DOI: 10.11897/SP.J.1016.2016.01697. JIAO L C, YANG S Y, LIU F, et al. Seventy years beyond neural network: retrospect and prospect [J]. Chinese Journal of Computers, 2016, 39(8): 1697–1716. DOI: 10.11897/SP.J.1016.2016.01697.
[17]	段宝福, 张猛, 李俊猛. 逐孔起爆震动参数预报的BP神经网络模型 [J]. 爆炸与冲击, 2010, 30(4): 401–406. DOI: 10.11883/1001-1455(2010)04-0401-06. DUAN B F, ZHANG M, LI J M. A BP neural network model for forecasting of vibration parameters from hole by hole detonation [J]. Explosion and Shock Waves, 2010, 30(4): 401–406. DOI: 10.11883/1001-1455(2010)04-0401-06.
[18]	范勇, 裴勇, 杨广栋, 等. 基于改进PSO-BP神经网络的爆破振动速度峰值预测 [J]. 振动与冲击, 2022, 41(16): 194–203, 302. DOI: 10.13465/j.cnki.jvs.2022.16.025. FAN Y, PEI Y, YANG G D, et al. Prediction of blasting vibration velocity peak based on an improved PSO-BP neural network [J]. Journal of Vibration and Shock, 2022, 41(16): 194–203, 302. DOI: 10.13465/j.cnki.jvs.2022.16.025.