Prediction of Projectile Muzzle Velocity Based on Neural Network Algorithm Combined with Clustering Association Rules

doi:10.12382/bgxb.2021.0687

Abstract

Abstract:

In view of the situation that reconstruction is needed for the muzzle velocity of a projectile measured by muzzle velocity radar in the range test, the data of two radars used in the test at the same time are fused to establish a neural network model, and the data of one radar is used to predict the data that needs to be reconstructed by the other radar. Because the prediction accuracy of the prediction model depends on the quality of the model, and the model quality depends on the quality of the sample data, we first use cluster analysis and association rules to mine high-quality samples from a large number of historical test data, and then establish a neural network for prediction. The experimental results show that: compared with support vector regression machine, the prediction accuracy of the combined algorithm constructed by clustering analysis association rules and neural network is higher, the error of predicting similar historical data is far less than 1‰, and the accuracy of predicting data significantly different from historical data is also more reasonable. The prediction results in the two cases show that the combined algorithm not only ensures the prediction accuracy, but also has certain robustness, and can be used as the prediction model of projectile muzzle velocity.

Key words: velocity initial, cluster analysis, association rules, neural network, combined algorithm

CLC Number:

TJ38

TIAN Ke. Prediction of Projectile Muzzle Velocity Based on Neural Network Algorithm Combined with Clustering Association Rules[J]. Acta Armamentarii, 2023, 44(2): 452-461.

Figures/Tables 23

Fig.1 3-layer BP nurual network topology

Fig.2 Site layout of two test radars

Fig.3 Left positive and rear positive distribution curves of Radar A

Fig.4 Variation curves of initial velocity of Radar A and Radar B

Fig.5 Absolute error and error standard variation curves

Fig.6 Variation curve of sum of squares within groups under different numbers of categories

Fig.7 Variation curve of sum of squares between groups under different numbers of categories

Fig.8 Schematic diagram of classification category

Table 1 Results of Apriori algorithm for association rule mining

左侧规则	右侧规则	支持度	置信度	提升度	样本数量
{A4}	{好}	0.2661290	0.8684211	1.066180	33
{A2}	{好}	0.3145161	1.0000000	1.227723	39

Fig.9 Frequency distribution of all association rules in Table 1

Fig.10 Visibility graph of correlation coefficient between left positive, back positive, initial velocity of Radar A and initial velocity of Radar B in combined samples

Fig.11 Variation curve of the residual sum of squares of combined sample neural network with the number of iterations

Table 2 Weights and thresholds from input layer to hidden layer of combined sample neural network

神经元	输入1	输入2	输入3	阈值
1	28.1432000	-6.1436770	8.7983940	44.7162829
2	14.5432400	-6.4585750	8.6515020	4.0440836
3	18.5108700	3.5425780	1.7351020	-3.0232857
4	20.6268800	3.0062510	2.6938990	-3.8635498
5	-22.2357100	1.1870900	13.6750550	0.3672449
6	19.0408800	-15.0458720	17.6933350	8.8356581

Table 3 Weights from hidden layer to output layer of combined sample neural network

神经元	1	2	3	4	5	6
输出层	0.33721577	0.01783879	37.37517928	37.14605252	-18.66516705	0.21881953

Fig.12 Prediction value of projectile muzzle velocity measured by radar B in combined samples by BP neural network

Table 4 Measured values and neural network predicted values of the 1st ~ 5th rounds of Radar B

弹序	实测值	神经网络预测值	SVR预测值
1	92.24	92.26248	93.79947
2	91.61	91.56277	93.84397
3	91.95	91.84691	93.81447
4	91.94	91.93622	93.80780
5	91.48	91.27619	93.88130

Table 5 Measured values and neural network predicted values of the 58th ~ 62nd rounds of Radar B

弹序	实测值	神经网络预测值	SVR预测值
58	93.59	93.70422	93.70042
59	93.47	93.58785	93.58933
60	93.38	93.48330	93.49280
61	93.30	93.36240	93.38630
62	93.47	93.48330	93.49280

Table 6 Measured values and neural network predicted values of the 66th~70th rounds of Radar B

弹序	实测值	神经网络预测值	SVR预测值
66	92.90	92.87567	93.09812
67	92.87	92.86936	93.09310
68	92.87	92.86309	93.08815
69	92.82	92.82619	93.06020
70	92.97	92.92084	93.13554

Table 7 Measured values and neural network predicted values of the 104th~108th rounds of Radar B

弹序	实测值	神经网络预测值	SVR预测值
104	92.53	92.59104	92.71547
105	92.78	92.84000	92.84771
106	92.95	93.03365	92.98245
107	92.59	92.69054	92.76176
108	92.82	92.89735	92.88517

Fig.13 Initial velocity variation curve of v1 Radar A and Radar B

Fig.14 Initial velocity variation curve of v2 Radar A and Radar B

Table 8 Measured values of v1 Radar B and predicted values of BP neural network

弹序	实测值	神经网络预测值	SVR预测值
1	85.30	85.65732	94.20604
2	85.20	85.66622	94.20604
3	85.39	85.65430	94.20604
4	85.38	85.66515	94.20604
5	85.41	85.65885	94.20604

Table 9 Measured values of v2 Radar B and predicted values of BP neural network

弹序	实测值	神经网络预测值	SVR预测值
1	84.80	84.31650	94.20603
2	84.86	84.31782	94.20604
3	84.82	84.31676	94.20603
4	84.86	84.31808	94.20603
5	84.86	84.31782	94.20603

References 27

[1]	GJB2973A—2008,火炮内弹道试验方法[S]. 北京: 中国人民解放军总装备部, 2008.
	GJB2973A—2008. Test method for interior ballistics of artillery[S]. Beijing: General Armament Department of the Chinese people's Liberation Army, 2008. (in Chinese)
[2]	梁承栋, 张贤椿, 王军. 炮口初速预测方法综述[J]. 兵工自动化, 2020, 39(8):10-15.
	LIANG C D, ZHANG X C, WANG J. Summary of muzzle velocity prediction methods[J]. Ordnance Industry Automation, 2020, 39(8):10-15. (in Chinese)
[3]	田珂, 陈铎. 基于组合模型预测火炮初速的研究与应用[J]. 火炮发射与控制学报, 2021, 42(1):30-35.
	TIAN K, CHEN D. Research and application of predicting gun muzzle velocity based on combined model[J]. Journal of Gun Launch and Control, 2021, 42(1):30-35. (in Chinese)
[4]	解维河, 孙卫国, 孙东彦. 舰炮首发弹丸初速预测新方法[J]. 兵器装备工程学报, 2019, 40(8):18-20.
	XIE W H, SUN W G, SUN D Y. A new method for predicting the muzzle velocity of the first projectile of naval gun[J]. Journal of Ordnance and Equipment Engineering, 2019, 40(8):18-20. (in Chinese)
[5]	解维河, 孙卫国, 孙东彦. 弹丸初速预测误差分析及模型改进研究[J]. 指挥控制与仿真, 2019, 41(3):113-115. doi: 10.3969/j.issn.1673-3819.2019.03.024
	XIE W H, SUN W G, SUN D Y. Error analysis and model improvement of projectile muzzle velocity prediction[J]. Command Control and Simulation, 2019, 41(3):113-115. (in Chinese)
[6]	张良均, 云伟标, 王路, 等. R语言数据分析与挖掘实战[M]. 北京: 机械工业出版社, 2018:318.
	ZHANG L J, YUN W B, WANG L, et al. R language data analysis and mining[M]. Beijing: China Machine Press, 2018:318. (in Chinese)
[7]	艾云昊, 杨超宇, 李慧宗. 基于聚类的关联规则后处理算法研究[J]. 黑龙江工业学院学报, 2020, 20(9):126-131.
	AI Y H, YANG C Y, LI H Z. Research on post-processing algorithm of association rules based on clustering[J]. Journal of Heilongjiang University of Technology(Comprehensive Edition), 2020, 20(9):126-131. (in Chinese)
[8]	赵露. 基于聚类分析的网络安全数据特征可视化融合研究[J]. 长春工程学院学报, 2020, 21(2):94-97.
	ZHAO L. Research on visual fusion of network security data features based on cluster analysis[J]. Journal of Changchun Institute of Engineering, 2020, 21(2):94-97. (in Chinese)
[9]	王青松, 姜富山, 李菲. 大数据环境下基于关联规则的多标签学习算法[J]. 计算机科学, 2020, 47(5):90-95. doi: 10.11896/jsjkx.190300150
	WANG Q S, JIANG F S, LI F. Multi label learning algorithm based on association rules in big data environment[J]. Computer Science, 2020, 47(5):90-95. (in Chinese) doi: 10.1063/1.31588 URL
[10]	田建勇, 石林江. 融合布尔矩阵和项目特性的关联规则挖掘算法[J]. 控制工程, 2020, 27(6):1004-1011.
	TIAN J Y, SHI L J. Association rule mining algorithm combining Boolean matrix and item characteristics[J]. Control Engineering, 2020, 27(6):1004-1011. (in Chinese)
[11]	杨岚雁, 靳敏, 张迎春, 等. 一种基于关联规则的MLKNN多标签分类算法[J]. 计算机工程与科学, 2020, 42(7):1309-1317.
	YANG L Y, JIN M, ZHANG Y C, et al. MLKNN multi label classification algorithm based on association rules[J]. Computer Engineering and Science, 2020, 42(7):1309-1317. (in Chinese)
[12]	DING S F, SU C Y, YU J Z. An optimizing BP neural network algorithm based on genetic algorithm[J]. Artificial Intelligence Review, 2011, 36(2):153-162. doi: 10.1007/s10462-011-9208-z URL
[13]	LONG J Q, LAN F C, CHEN J Q. Mechanical properties prediction of the mechanical clinching joints based on genetic algorithm and BP neural network[J]. Chinese Journal of Mechanical Engineering, 2009, 22(1):36-41. doi: 10.3901/CJME.2009.01.036 URL
[14]	高玉波, 张伟, 李达诚, 等. 基于反向传播神经网络的陶瓷损伤参数反演分析[J]. 兵工学报, 2018, 39(1):146-152. doi: 10.3969/j.issn.1000-1093.2018.01.016
	GAO Y B, ZHANG W, LI D C, et al. Inverse analysis of ceramic damage parameters based on back propagation neural network[J]. Acta Armamentarii, 2018, 39(1):146-152. (in Chinese)
[15]	王国辉, 张宝栋, 李向荣. 基于遗传算法的BP神经网络火炮身管烧蚀磨损量预测[J]. 火炮发射与控制学报 2019, 40(4):6-10,15.
	WANG G H, ZHANG B D, LI X R. Prediction of erosion wear of gun barrel based on BP neural network based on genetic algorithm[J]. Journal of Gun Launch and Control, 2019, 40(4):6-10,15. (in Chinese)
[16]	齐浩淳, 黄大鹏, 魏久南, 等. 基于BP神经网络的高寒山地弹药消耗需求分析[J]. 兵器装备工程学报, 2016, 37(6):97-101.
	QI H C, HUANG D P, WEI J N, et al. Analysis of ammunition consumption demand in alpine mountain based on BP neural network[J]. Journal of Ordnance Equipment Engineering, 2016, 37(6):97-101. (in Chinese)
[17]	田珂. 基于组合模型高精度预测弹丸径向速度的方法[J]. 弹道学报, 2022, 34(3):49-57.
	TIAN K. Amethod for predicting projectile radial velocity with high accuracy based on combination model[J]. Journal of Ballistics, 2022, 34(3):49-57. (in Chinese)
[18]	刘畅, 肖斌, 蒋铁军, 等. 一种改进K均值的小样本聚类算法[J]. 兵器装备工程学报, 2021, 42(增刊1):266-270.
	LIU C, XIAO B, JIANG T J, et al. An improved k-means small sample clustering algorithm[J]. Journal of Ordnance Equipment Engineering, 2021, 42(S1):266-270. (in Chinese)
[19]	薛震, 孙玉林. R语言统计分析与机器学习[M]. 北京: 中国水利水电出版社, 2020:196-362.
	XUE Z, SUN Y L. Statistical analysis of R language and machine learning[M]. Beijing: China Water Resources and Hydropower Press, 2020:196-362. (in Chinese)
[20]	廖孟柯, 樊冰, 李忠政, 等. 基于改进Apriori算法的配电网设备退役信息挖掘[J]. 科学技术与工程, 2021, 21(24):10381-10386.
	LIAO M K, FAN B, LI Z Z, et al. Distribution network equipment retirement information mining based on improved Apriori algorithm[J]. Science,Technology and Engineering, 2021, 21(24):10381-10386. (in Chinese)
[21]	邱明月. 基于Apriori关联规则的草原火灾预警及对策研究[J]. 科技经济导报, 2019, 27(26):106-108.
	QIU M Y. Research on grassland fire early warning andcountermeasures based on apriori association rules[J]. Science and Technology Economic Guide, 2019, 27(26):106-108. (in Chinese)
[22]	纪文璐, 王海龙, 苏贵斌, 等. 基于关联规则算法的推荐方法研究综述[J]. 计算机工程与应用, 2020, 56(22):33-41. doi: 10.3778/j.issn.1002-8331.2006-0158
	JI W L, WANG H L, SU G B, et al. Review on recommendation methods based on association rule algorithm[J]. Computer Engineering and Application, 2020, 56(22):33-41. (in Chinese)
[23]	于哲峰, 婿建宇, 罗跃, 等. 高超声速飞行器烧蚀后退量时序提取及基于神经网络的预测[J]. 兵工学报, 2021, 42(6):1230-1237. doi: 10.3969/j.issn.1000-1093.2021.06.013
	YU Z F, XU J Y, LUO Y, et al. Time series extraction of ablation retreat of hypersonic vehicle and prediction based on neural network[J]. Acta Armamentarii, 2021, 42(6):1230-1237. (in Chinese)
[24]	高昂, 段渭军, 李立欣, 等. 基于神经网络的无人机云服务质量控制方法研究[J]. 兵工学报, 2018, 39(9):1762-1771. doi: 10.3969/j.issn.1000-1093.2018.09.013
	GAO A, DUAN W J, LI L X, et al. Research on service quality control method of UAV cloud based on Neural Network[J]. Acta Armamentarii, 2018, 39(9):1762-1771. (in Chinese) doi: 10.3969/j.issn.1000-1093.2018.09.013
[25]	何耀民, 何华锋, 徐永壮, 等. 基于反向传播神经网络的海杂波参数估计[J]. 兵工学报, 2019, 40(12):2473-2481. doi: 10.3969/j.issn.1000-1093.2019.12.011
	HE Y M, HE H F, XU Y Z, et al. Sea clutter parameter estimation based on back propagation nurual network[J]. Acta Armamentarii, 2019, 40(12):2473-2481. (in Chinese)
[26]	王成, 吴岩, 杨廷飞. 利用改进单分类支持向量机提升舰船尾流目标的检测准确率[J]. 兵工学报, 2020, 41(9):1887-1893. doi: 10.3969/j.issn.1000-1093.2020.09.021
	WANG C, WU Y, YANG T F. Improved single classification support vector machine is used to improve the detection accuracy of ship wake target[J]. Acta Armamentarii, 2020, 41(9):1887-1893.(in chinese)
[27]	孙丽娜, 王应海, 黄永红, 等. 基于贝叶斯推断LSSVM的枪管寿命建模与预测[J]. 火炮发射与控制学报, 2018, 39(4):91-95.
	SUN L N, WANG Y H, HUANG Y H, et al. Modeling and prediction of barrel life based on Bayesian inference LSSVM[J]. Journal of Gun Launch and Control, 2018, 39(4):91-95. (in Chinese)