基于遗传神经网络的简易制导炸弹有控弹道设计  

Design of Controllable Trajectory of Simple Guided-bomb Based on Neural Network and Genetic Algorithm

作　　者：桑保华[1] 薛晓中[1] 

机构地区：[1]南京理工大学动力工程学院,南京210094

出　　处：《弹道学报》2007年第1期12-16,共5页Journal of Ballistics

摘　　要：结合简易制导炸弹的方案弹道实时寻优技术和控制律,提出了一种以遗传算法和神经网络为基础的有控弹道设计方法.该方法针对炸弹的控制系统构成,在内回路中引入极小邻域零收敛鲁棒稳定准则以确保内回路BP神经网格稳定,在外回路的设计中,采用附加时域指标集的适应度函数方法以跟踪阶跃响应.其中,外回路的BP神经网络的结构及初始权值采用并型遗传算法进行优化设计.在全弹道设计中,结合智能降阶动态解耦思想,采用变增益控制方法较好地实现了炸弹的有控飞行,仿真结果良好.By combining project trajectory real-time optimization of simple guided-bomb with control law, a design method about its controllable trajectory based on genetic algorithm and neural network was presented. Aiming at the control system of simple guided-bomb, a robust stabilization theory of minimum adjacent region and zero convergence was excerpted in inner loop to ensure its back propagation （BP） neural network stabilization. In outer loop, an adaptability function appended with index aggregate of time-domain was introduced to track the step response. Parallel genetic algorithm was applied to optimization of BP neural network and its initial weights. In the design of the whole trajectory, varying gains and intelligent step-reduced dynamic decoupling were employed to realize controllable flight of simple-guided bomb, and the simulation results prove well.

关 键 词：制导炸弹 神经网络 遗传算法 有控飞行 有控弹道 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置] V249[自动化与计算机技术—控制科学与工程]