基于M-ANFIS-PNN的目标威胁评估模型

doi:10.12305/j.issn.1001-506X.2022.10.19

摘要/Abstract

摘要：

目标威胁评估的目的是根据目标的属性和状态信息对目标的威胁程度进行定量估计, 为后续作战决策提供辅助支持。现有威胁评估模型大多依赖于数值信息, 难以有效处理包含定性、定量数据的目标特征信息。基于此, 提出一种改进的自适应模糊神经推理系统模型。在自适应模糊神经推理系统的基础上, 引入前件影响矩阵和后件影响矩阵对定性数据进行处理, 使得定量、定性数据的影响同时作用于模糊规则的前件参数和后件参数; 为了进一步提高模型的输出精度, 将自适应模糊神经推理系统的输出层替换为多项式神经网络; 通过基于Gower距离的近邻传播聚类算法对改进模型进行结构辨识, 确定模糊规则的初始参数。仿真实例验证了所提方法的有效性与可行性, 与其他混合属性数据建模方法相比, 所提方法具有较高的预测精度, 可为作战指挥决策提供有效的辅助支持。

关键词: 威胁评估, 自适应模糊神经推理系统, 多项式神经网络, 混合属性, 近邻传播聚类算法

Abstract:

The purpose of target threat assessment is to quantitatively estimate the threat level of the target based on the target's attributes and status information and provide auxiliary support for operational decision-making. Existing threat assessment models mostly rely on numerical information, and it is difficult to effectively process target feature information containing qualitative and quantitative data. Based on this, this paper proposes an improved adaptive network based fuzzy inference system model. On the basis of the adaptive network based fuzzy inference system, the antecedent influence matrix and the consequent influence matrix are introduced to process the qualitative data, so that the influence of the quantitative and qualitative data acts on the antecedent parameters and the consequent parameters of the fuzzy rules at the same time. In order to further improve the output accuracy of the model, the output layer of the adaptive network based fuzzy inference system is replaced with a polynomial neural network. The structure of the improved model is identified by the affinity propagation clustering algorithm based on Gower distance, and the initial parameters of the fuzzy rules are determined. Simulation examples verify the effectiveness and feasibility of the proposed method. Compared with other mixed attribute data modeling methods, the proposed method has a higher prediction accuracy and can provide effective auxiliary support for battle command decision-making.

Key words: threat assessment, adaptive network based fuzzy inference system, polynomial neural network, mixed attribute, affinity propagation clustering algorithm

中图分类号:

TP391

于博文, 于琳, 吕明, 张捷. 基于M-ANFIS-PNN的目标威胁评估模型[J]. 系统工程与电子技术, 2022, 44(10): 3155-3163.

Bowen YU, Lin YU, Ming LYU, Jie ZHANG. Target threat assessment model based on M-ANFIS-PNN[J]. Systems Engineering and Electronics, 2022, 44(10): 3155-3163.

图/表 10

图1

图2

表1

表2

图3

表3

表4

图4

表5

表6

参考文献 30

1	靳崇, 孙娟, 王永佳, 等. 基于直觉模糊TOPSIS和变权VIKOR的防空目标威胁综合评估[J]. 系统工程与电子技术, 2022, 44 (1): 172- 180.
	JIN C , SUN J , WANG Y J , et al. Threat comprehensive assessment for air defense targets based on intuitionistic fuzzy TOPSIS and variable weight VIKOR[J]. Systems Engineering and Electronics, 2022, 44 (1): 172- 180.
2	QU C W, HE Y. A method of threat assessment using multiple attribute decision making[C]//Proc. of the 6th International Confe-rence on Signal Processing, 2002, 2: 1091-1095.
3	卢盈齐, 范成礼, 付强, 等. 基于改进IFRS相似度和信息熵的反导作战目标威胁评估[J]. 系统工程与电子技术, 2022, 44 (4): 1230- 1238.
	LU Y Q , FAN C L , FU Q , et al. Missile defense target threat assessment based on improved similarity measure and information entropy of intuitionistic fuzzy rough sets[J]. Systems Engineering and Electronics, 2022, 44 (4): 1230- 1238.
4	赵烨南, 杜伟伟, 陈铁健, 等. 基于集对分析的坦克多目标威胁评估方法[J]. 火力与指挥控制, 2020, 45 (6): 108- 112.
	ZHAO Y N , DU W W , CHEN T J , et al. Multi-target threat assessment method of tank based on set pair analysis[J]. Fire Control & Command Control, 2020, 45 (6): 108- 112.
5	柴慧敏, 张勇, 李欣粤, 等. 基于深度学习的空中目标威胁评估方法[J]. 系统仿真学报, 2022, 34 (7): 1459- 1467. doi: 10.16182/j.issn1004731x.joss.21-0080
	CHAI H M , ZHANG Y , LI X Y , et al. Aerial target threat assessment method based on deep learning[J]. Journal of System Simulation, 2022, 34 (7): 1459- 1467. doi: 10.16182/j.issn1004731x.joss.21-0080
6	孔尚萍, 张海瑞, 廖选平, 等. 基于AHP与熵权法的空中目标威胁评估方法[J]. 战术导弹技术, 2018, (1): 79- 84. doi: 10.16358/j.issn.1009-1300.2018.01.14
	KONG S P , ZHANG H R , LIAO X P , et al. Aerial targets threat assessment based on AHP and entropy weight method[J]. Tactical Missile Technology, 2018, (1): 79- 84. doi: 10.16358/j.issn.1009-1300.2018.01.14
7	LUO R N , HUANG S C , ZHAO Y , et al. Threat assessment method of low altitude slow small (LSS) targets based on information entropy and AHP[J]. Entropy, 2021, 23 (10): 1292. doi: 10.3390/e23101292
8	任耀军, 袁修久, 黄林, 等. 毕达哥拉斯三角犹豫模糊Muirhead平均算子的目标威胁评估应用[J]. 电光与控制, 2021, 28 (4): 16- 20.16-20, 63 doi: 10.3969/j.issn.1671-637X.2021.04.004
	REN Y J , YUAN X J , HUANG L , et al. Pythagorean hesitant triangular fuzzy muirhead mean operator and its application in target threat assessment[J]. Electronics Optics & Control, 2021, 28 (4): 16- 20.16-20, 63 doi: 10.3969/j.issn.1671-637X.2021.04.004
9	赵若静, 杨风暴, 吉琳娜. 基于可能性理论扩展灰关联的目标威胁评估[J]. 指挥信息系统与技术, 2021, 12 (3): 23- 29.
	ZHAO R J , YANG F B , JI L N . Target threat assessment based on extended grey correlation with possibility theory[J]. Command Information System and Technology, 2021, 12 (3): 23- 29.
10	JANG J S R . ANFIS: adaptive-network-based fuzzy inference system[J]. IEEE Trans.on Systems, Man, and Cybernetics, 1993, 23 (3): 665- 685. doi: 10.1109/21.256541
11	LIU M , DONG M Y , WU C . A new ANFIS for parameter prediction with numeric and categorical inputs[J]. IEEE Trans.on Automation Science and Engineering, 2010, 7 (3): 645- 653.
12	ARMAGHANI D J , HARANDIZADEH H , MOMENI E . Load carrying capacity assessment of thin-walled foundations: an ANFIS-PNN model optimized by genetic algorithm[J]. Engineering with Computers, 2021, doi: 10.1007/s00366-21-01380-0
13	FREY B J , DUECK D . Clustering by passing messages between data points[J]. Science, 2007, 315 (5814): 972- 976.
14	MADANDOUST R , BUNGEY J H , GHAVIDEL R . Prediction of the concrete compressive strength by means of core testing using GMDH-type neural network and ANFIS models[J]. Computational Materials Science, 2012, 51 (1): 261- 272.
15	HARANDIZADEH H , ARMAGHANI D J , ASTERIS P G , et al. TBM performance prediction developing a hybrid ANFIS-PNN predictive model optimized by imperialism competitive algorithm[J]. Neural Computing and Applications, 2021, 33 (23): 16149- 16179.
16	JIA X D , DI Y , FENG J S , et al. Adaptive virtual metrology for semiconductor chemical mechanical planarization process using GMDH-type polynomial neural networks[J]. Journal of Process Control, 2018, 62, 44- 54.
17	HUANG W , OH S K , PEDRYCZ W . Fuzzy reinforced polynomial neural networks constructed with the aid of PNN architecture and fuzzy hybrid predictor based on nonlinear function[J]. Neurocomputing, 2021, 458, 454- 467.
18	ARMAGHANI D J , HARANDIZADEH H , MOMENI E , et al. An optimized system of GMDH-ANFIS predictive model by ICA for estimating pile bearing capacity[J]. Artificial Intelligence Review, 2022, 55 (3): 2313- 2350.
19	HARANDIZADEH H , TOUFIGH M M , TOUFIGH V . Application of improved ANFIS approaches to estimate bearing capacity of piles[J]. Soft Computing, 2019, 23 (19): 9537- 9549.
20	GAN G J , NG M K P . Subspace clustering using affinity pro-pagation[J]. Pattern Recognition, 2015, 48 (4): 1455- 1464.
21	HU Q , ZHANG Q , SI X S , et al. Intelligent fault diagnosis approach based on composite multi-scale dimensionless indicators and affinity propagation clustering[J]. IEEE Sensors Journal, 2020, 20 (19): 11439- 11453.
22	AHMAD A , KHAN S S . Survey of state-of-the-art mixed data clustering algorithms[J]. IEEE Access, 2019, 7, 31883- 31902.
23	ZHANG K , GU X S . An affinity propagation clustering algorithm for mixed numeric and categorical datasets[J]. Mathematical Problems in Engineering, 2014, 486075.
24	BEKTAS A, SCHUMANN R. How to optimize Gower distance weights for the k-medoids clustering algorithm to obtain mobility profiles of the Swiss population[C]//Proc. of the 6th Swiss Conference on Data Science, 2019: 51-56.
25	FOSS A H , MARKATOU M , RAY B . Distance metrics and clustering methods for mixed type data[J]. International Statistical Review, 2019, 87 (1): 80- 109.
26	AKAY O , YUKSEL G . Hierarchical clustering of mixed variable panel data based on new distance[J]. Communications in Statistics-Simulation and Computation, 2019, 50 (6): 1695- 1710.
27	CHENG Y M , JIA H . Categorical and numerical attribute data clustering based on a unified similarity metric without knowing cluster number[J]. Pattern Recognition, 2013, 46 (8): 2228- 2238.
28	QUE X , JIANG S Y , YANG J Y , et al. A similarity measurement with entropy-based weighting for clustering mixed numerical and categorical datasets[J]. Algorithms, 2021, 14 (6): 184.
29	HARANDIZADEH H , ARMAGHANI D J . Prediction of air-overpressure induced by blasting using an ANFIS-PNN model optimized by GA[J]. Applied Soft Computing, 2021, 99, 106904.
30	HARANDIZADEH H , JAHED A D , KHARI M . A new development of ANFIS-GMDH optimized by PSO to predict pile bearing capacity based on experimental datasets[J]. Engineering with Computers, 2021, 37 (1): 685- 700.

C₁	C₂	SE
C₁₁	C₂₁	[100 100]
C₁₂	C₂₁	[010 100]
C₁₃	C₂₁	[001 100]
C₁₁	C₂₂	[100 010]
C₁₂	C₂₂	[010 010]
C₁₃	C₂₂	[001 010]
C₁₁	C₂₃	[100 001]
C₁₂	C₂₃	[010 001]
C₁₃	C₂₃	[001 001]

编号	特征属性	取值范围及属性值	特征类别
1	D/m	[500, 6 000]	定量数值
2	V/(km/h)	[0, 150]	定量数值
3	CA/(°)	[-180, 180]	定量数值
4	H/m	[0, 2 000]	定量数值
5	FR/km	[0, 50]	定量数值
6	AC/(m/s²)	^{[1, 8]}	定量数值
7	C	武装直升机C1, 主战坦克C2, 自行火炮C3, 步兵战车C4, 装甲作战指挥车C5, 导弹发射车C6	定性分类
8	A	一级攻击状态A1, 二级攻击状态A2, 三级攻击状态A3	定性分类
9	DS	未被毁伤D1, 轻度毁伤D2, 中度毁伤D3, 重度毁伤D4, 报废D5	定性分类
10	CC	特强CC1, 强CC2, 较强CC3, 一般CC4, 较弱CC5, 弱CC6, 特弱CC7	定性分类

编号	D/m	V/(km/h)	CA/(°)	H/m	FR/km	AC/(m/s²)	C	A	DS	CC	R
1	2 800	60	10	0	3	3.5	C5	A3	DS1	CC1	0.994 1
2	1 500	10	15	5	3.2	2.5	C2	A1	DS1	CC6	0.943 1
3	1 000	0	5	10	3.2	2	C2	A1	DS1	CC6	0.969 9
4	3 500	60	120	5	3.5	2.5	C2	A3	DS2	CC6	0.321 4
5	2 500	30	120	8	4	2	C2	A3	DS2	CC6	0.392 7
6	2 500	60	5	11	2	2	C2	A2	DS1	CC6	0.727 6
7	1 000	5	5	100	3	1	C1	A1	DS1	CC3	0.979 4
8	3 000	60	50	400	5	1	C1	A2	DS1	CC3	0.854 7
9	5 200	150	5	800	3	1.5	C1	A3	DS2	CC4	0.546 2
10	6 000	130	150	300	3	0.8	C1	A3	DS3	CC4	0.154 3
11	4 200	100	30	300	6	1.5	C1	A3	DS2	CC4	0.347 7
12	1 300	0	5	20	10	0.5	C3	A1	DS1	CC6	0.914 2
13	1 500	5	5	2	8	0.4	C3	A1	DS2	CC6	0.905 8
14	3 800	60	20	6	8	0.4	C3	A2	DS2	CC6	0.432 9
15	2 000	0	30	21	10	0.4	C3	A2	DS5	CC6	0.119 1
16	3 500	60	160	5	10	0.5	C3	A3	DS3	CC6	0.136 7
17	1 800	18	40	5	3	3	C4	A2	DS2	CC3	0.721 9
18	1 500	0	15	5	3	3	C4	A2	DS5	CC3	0.125 6

实验数据	模型	性能评价指标
实验数据	模型	R	MSE	RMSE	E-Mean	E-StD
训练数据	C	0.913	0.697	0.835	0.099	0.836
	M	0.921	0.511	0.714	0.073	0.715
	M+PNN	0.955	0.371	0.609	0.053	0.611
数据阶段	C	0.862	0.874	0.935	0.124	0.936
	M	0.871	0.741	0.861	0.105	0.862
	M+PNN	0.918	0.537	0.733	0.076	0.734
全部数据	C	0.887	0.785	0.885	0.112	0.886
	M	0.896	0.626	0.787	0.089	0.788
	M+PNN	0.936	0.454	0.673	0.064	0.674

组号	样本数量	聚类数目		RMSE
组号	样本数量	AP	EGAP	AP	EGAP
1	800	35	12	0.656	0.646
2	800	40	14	0.633	0.641
3	800	35	14	0.699	0.689
4	800	38	15	0.644	0.631
5	800	38	14	0.683	0.706
6	800	31	16	0.662	0.653
7	800	33	13	0.717	0.718
8	800	30	14	0.685	0.673
平均值	800	35	14	0.672	0.669