模型与数据混合驱动的代理模型构建方法研究

doi:10.16182/j.issn1004731x.joss.22-1263

系统仿真学报 ›› 2024, Vol. 36 ›› Issue (3): 756-769.doi: 10.16182/j.issn1004731x.joss.22-1263

模型与数据混合驱动的代理模型构建方法研究

安靖¹^,²^,³(), 司光亚³(), 曾妙婷¹

^1.国防大学联合勤务学院，北京 100858
^2.国防大学研究生院，北京 100091
^3.国防大学联合作战学院，北京 100091

收稿日期:2022-10-21 修回日期:2022-11-28 出版日期:2024-03-15 发布日期:2024-03-14
通讯作者: 司光亚 E-mail:anj21_2000@sina.com;sgy863@sina.com
第一作者简介:安靖(1981-)，女，副教授，博士，研究方向为军事运筹学、战争设计系统工程。E-mail：anj21_2000@sina.com
基金资助:
全军军事类研究生资助课题(JY2020B031)

Construction of Surrogate Model Driven by Model and Data

An Jing¹^,²^,³(), Si Guangya³(), Zeng Miaoting¹

^1.Joint Logistics College, National Defense University, Beijing 100858, China
^2.Graduate School, National Defense University, Beijing 100091, China
^3.Joint Operations College, National Defense University, Beijing 100091, China

Received:2022-10-21 Revised:2022-11-28 Online:2024-03-15 Published:2024-03-14
Contact: Si Guangya E-mail:anj21_2000@sina.com;sgy863@sina.com

摘要/Abstract

摘要：

以某作战样式下的立体投送行动为研究对象，针对模拟仿真推演计算因子过多、计算资源开销过大、普通解析模型计算精度不足等问题，提出了一种模型与数据混合驱动的代理模型构建方法，支撑作战行动的研究。基于军事理论构建了包含武器装备、作战力量等在内的含待优化系数的立体投送解析模型组，并依托自主研发的“代理模型可视化平台”，实现前述解析模型组的复合和参数设置；通过模拟仿真系统实施推演，采集高可信度仿真数据；以高可信度仿真数据为样本，采用多目标遗传优化算法NSGA-II对解析模型中的待定关键系数进行优化，获得能够兼顾计算精度和计算速度的立体投送代理模型。经对比实验结果表明，构建的立体投送代理模型计算结果与高可信度仿真推演相比，战损最大相对误差不超过6.5%，而计算速度提升了150倍。

关键词: 立体投送, 仿真推演, 解析模型, 代理模型, 多目标优化算法, 可解释性

Abstract:

By taking the three-dimensional projection action in a certain combat style as the research object, a surrogate model construction method driven by model and data is proposed to support the operational action research, so as to solve the problem that the calculation factors are too much during simulated deduction; the calculation resource cost is too large, and the calculation accuracy of the general analytical model is insufficient. Firstly, an analytical model group of three-dimensional projections with coefficients to be optimized is constructed based on military theory, including weapons and equipment, forces, etc. In addition, the composition and parameter setting of the above-mentioned analytical model group are realized by the self-developed "visualization platform of surrogate model". The simulation system is used to implement deduction and collect high-credibility simulation data. Finally, by taking high-credibility simulation data as samples, the multi-objective genetic optimization algorithm NSGA-II is used to optimize the coefficients to be determined in the analytical model, and then a surrogate model of three-dimensional projection that considers both the calculation accuracy and speed is obtained. The experimental results show that the maximum relative error of operational loss of the constructed surrogate model of three-dimensional projection is less than 6.5%, and the calculation speed is 150 times faster than that of high-credibility simulation deduction.

Key words: three-dimensional projection, simulation deduction, analytical model, surrogate model, multi-objective optimization algorithm, interpretability

中图分类号:

TP 391.9

安靖,司光亚,曾妙婷 . 模型与数据混合驱动的代理模型构建方法研究[J]. 系统仿真学报, 2024, 36(3): 756-769.

An Jing,Si Guangya,Zeng Miaoting . Construction of Surrogate Model Driven by Model and Data[J]. Journal of System Simulation, 2024, 36(3): 756-769.

图/表 12

图1

图2

图3

表1

代理模型的待优化参数

平台名称	待定参数	参数序号
某型运输直升机	对地导弹基础命中率基础闪避率对地导弹伤害值	5~7
某型两栖登陆舰	舰炮伤害值舰炮基础命中率对空导弹基础命中率生命值基础闪避率	8~12
$⋮$
总计		76

表1

图4

图5

表2

图6

图7

图8

表3

图9

参考文献 13

1	杨镜宇, 司光亚, 胡晓峰. 信息化战争体系对抗探索性仿真分析方法研究[J]. 系统仿真学报, 2005, 17(6): 1469-1472, 1496.
	Yang Jingyu, Si Guangya, Hu Xiaofeng. Study on Simulation-based Exploratory Analysis Method of Information Warfare System of System (SoS) Encounter[J]. Journal of System Simulation, 2005, 17(6): 1469-1472, 1496.
2	Hunt K. A New Paradigm for Developing and Maintaining Scenarios for Distributed Simulation Systems[C]//1999 Fall Simulation Interoperability Workshop Papers. [S.l.]: [s.n.], 1999: 99S-SIW-179.
3	Johnson T H. Mission Space Model Development, Reuse and the Conceptual Models of the Mission Space Toolset[C]//In Proceedings of the 1998 Spring Simulation Interoperability Workshop. [S.l.]: [s.n.], 1998: 98S-SIW-155.
4	Koziel Slawomir, Leifsson Leifur. Surrogate-based Modeling and Optimization: Applications in Engineering[M]. New York, NY: Springer New York, 2013.
5	Krige D G. A Statistical Approach to Some Basic Mine Valuation Problems on the Witwatersrand[J]. Journal of the Southern African Institute of Mining and Metallurgy, 1951, 52(6): 119-139.
6	韩忠华. Kriging模型及代理优化算法研究进展[J]. 航空学报, 2016, 37(11): 3197-3225.
	Han Zhonghua. Kriging Surrogate Model and Its Application to Design Optimization: A Review of Recent Progress[J]. Acta Aeronautica et Astronautica Sinica, 2016, 37(11): 3197-3225.
7	Schulman J, Wolski F, Dhariwal P, et al. Proximal Policy Optimization Algorithms[EB/OL]. (2017-08-28) [2022-09-20]. .
8	Mnih V, Adrià Puigdomènech Badia, Mirza M, et al. Asynchronous Methods for Deep Reinforcement Learning[C]//Proceedings of the 33rd International Conference on International Conference on Machine Learning. Cambridge: JMLR, 2016: 1928-1937.
9	安靖, 司光亚, 张雷. 基于深度强化学习的立体投送策略优化方法研究[J]. 系统仿真学报, 2024, 36(1): 39-49.
	An Jing, Si Guangya, Zhang Lei. Research on the Strategy Optimization Method of Multi-dimension Projection Based on Deep Reinforcement Learning[J]. Journal of System Simulation, 2024, 36(1): 39-49.
10	蔡文杰. 基于替代模型的飞行器稳健设计软件研究与开发[D]. 绵阳: 西南科技大学, 2020.
	Cai Wenjie. Research and Development of Robust Design Software for Aircraft Based on Alternative Model[D]. Mianyang: Southwest University of Science and Technology, 2020.
11	李记超. 离散伴随和数据驱动的气动优化设计方法研究[D]. 西安: 西北工业大学, 2019.
	Li Jichao. Aerodynamic Shape Optimization Methods with Discrete Adjoint and Data-driven Techniques[D]. Xi'an: Northwestern Polytechnical University, 2019.
12	Deb K, Pratap A, Agarwal S, et al. A fast and Elitist Multiobjective Genetic Algorithm: NSGA-II[J]. IEEE Transactions on Evolutionary Computation, 2002, 6(2): 182-197.
13	韩忠华, 许晨舟, 乔建领, 等. 基于代理模型的高效全局气动优化设计方法研究进展[J]. 航空学报, 2020, 41(5): 25-65.
	Han Zhonghua, Xu Chenzhou, Qiao Jianling, et al. Recent Progress of Efficient Global Aerodynamic Shape Optimization Using Surrogate-based Approach[J]. Acta Aeronautica et Astronautica Sinica, 2020, 41(5): 25-65.

序号	参数	值
1	仿真时间序列T/h	6
2	离散时间步n	6，12，24，36，48，60
3	初始种群规模	200
4	最大进化代数	1 000
5	交叉概率	0.9
6	变异概率	0.03

模型	推演时间
代理模型	12 s
CMO	30 min(模拟速度=max)

[1]	安靖, 司光亚, 张雷. 基于深度强化学习的立体投送策略优化方法研究[J]. 系统仿真学报, 2024, 36(1): 39-49.
[2]	王国陈. 空中联合投送任务规划仿真与效能评估系统[J]. 系统仿真学报, 2022, 34(11): 2497-2506.
[3]	黄晓冬, 谢孔树, 李妮, 燕雪峰, 赵亚丽. 面向体系对抗的仿真支撑平台研究及应用[J]. 系统仿真学报, 2021, 33(8): 1914-1926.
[4]	章佳君, 钱晓超, 周金鹏, 王宣灵, 陆志沣. 基于仿真的无人机空袭沙特油田事件探索性分析[J]. 系统仿真学报, 2020, 32(6): 1195-1200.
[5]	陆湛文, 程新功, 张永峰. 基于共识粒子群的全局优化求解方法[J]. 系统仿真学报, 2020, 32(10): 1936-1942.
[6]	彭本红, 王圆缘, 屠羽. 政府规制视角下EPR契约设计及仿真分析[J]. 系统仿真学报, 2019, 31(9): 1811-1818.
[7]	刘博, 尚晓兵, 晁涛, 马萍, 杨明. 基于PCE的飞行器不确定性量化方法[J]. 系统仿真学报, 2018, 30(12): 4555-4562.
[8]	王宏伟, 朱子元, 石晶林, 苏泳涛, 石红梅, 刘智国. 基于代理模型的处理器结构设计空间探索算法[J]. 系统仿真学报, 2017, 29(5): 1077-1085.
[9]	王波, HaechangGea, 白俊强, 张玉东, 宫建, 张卫民. 考虑已知不确定性的随机Kriging建模方法[J]. 系统仿真学报, 2016, 28(6): 1261-1272.

模型与数据混合驱动的代理模型构建方法研究

Construction of Surrogate Model Driven by Model and Data

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 13

相关文章 9

编辑推荐

Metrics

本文评价