基于DENN的NSGA-Ⅱ算法的串联多药室火炮内弹道性能优化  

Optimization of Interior Ballistic Performance of Series Multi-chamber Gun via DENN-based NSGA-II Algorithm

作　　者：王渤[1] 罗懿 薛涛 张小兵[1] WANG Bo;LUO Yi;XUE Tao;ZHANG Xiaobing(School of Energy and Power Engineering,Nanjing University of Science and Technology,Nanjing 210094,China)

机构地区：[1]南京理工大学能源与动力工程学院,江苏南京210094

出　　处：《弹道学报》2023年第2期20-27,共8页Journal of Ballistics

基　　金：国家自然科学基金(12072160,12202198);江苏自然科学基金(BK20220951)。

摘　　要：为了提高串联多药室火炮的内弹道性能,根据串联多药室火炮的发射特点,建立了串联多药室火炮的经典内弹道模型并开展了数值模拟,分析了影响串联药室内弹道性能的主要因素。基于此,采用传统第二代非支配排序遗传算法(NSGA-Ⅱ)对串联多药室火炮的内弹道性能进行了多目标优化,初步分析了主副药室装药质量、活塞质量、点火延迟对内弹道性能的影响。由于NSGA-Ⅱ算法在设计变量维度较高时易陷入局部收敛并导致优化无效,本文进一步考虑火药弧厚、弹丸行程长等设计变量,提出了结合机器学习神经网络(NN)和差分进化算法(DE)加速传统NSGA-Ⅱ的新型优化算法(DENN-NSGA-Ⅱ)。该方法通过让NSGA-Ⅱ算法的初始种群在NN预测的Pareto前沿附近生成,实现对串联多药室火炮内弹道性能的高效优化。数值实验结果表明:相比与传统NSGA-Ⅱ方法,DENN-NSGA-Ⅱ方法通过优化初始种群的生成位置,收敛速度更快,能够得到更理想、更全面的优化结果,实现更高效地内弹道性能优化。To improve the interior ballistic performance of series multi-chamber gun,a classical interior ballistic model of series multi-chamber guns was established and numerically simulated according to the launch features of the series multi-chamber gun.The main factors affecting the interior ballistic performance of the series multi-chamber guns were analyzed.Based on this,the traditional second-generation non-dominated sorting genetic algorithm(NSGA-II)was utilized to perform multi-objective optimization on the interior ballistic performance of the series multi-chamber guns.The charging masses of the main charge chamber and the auxiliary charge chamber,the mass of the piston,and the ignition delay on ballistic performance were analyzed preliminarily.Given that the NSGA-II algorithm tends to fall into a local convergence when the design variable dimension is overly high,resulting in an invalid optimization,the present work further considered extra design variables,such as arc thickness of propellants and projectile stroke length,and proposed a combination of machine learning neural network(NN)and differential evolution algorithm(DE),namely DENN-NSGA-II,aiming to accelerate traditional NSGA-II.The proposed method can provide a high-quality optimization of the interior ballistic performance for the series multi-chamber gun by generating the initial population used in the NSGA-II algorithm that is near the Pareto front predicted by the NN.Numerical results show that the initial population of DENN-NSGA-II generated near the predicted Pareto has a faster convergence speed than random generation of DENN,and can obtain more ideal and comprehensive optimization results and achieve more efficient interior ballistic performance optimization.

关 键 词：串联多药室火炮 内弹道优化 遗传算法 神经网络 差分进化 

分 类 号：TJ303.4[兵器科学与技术—火炮、自动武器与弹药工程]