改进BP神经网络在航空弹药预测中的应用

对航空弹药消耗的准确预测是未来战争取胜的一个重要因素,然而战场复杂多变,传统的方法难以进行及时准确地预测.采用了BP神经网络对单个目标航空弹药进行预测,针对常规BP神经网络收敛速度慢、存在所谓"局部最小值"等缺陷,提出了具有全局收敛性的Fletcher-Reeves共轭梯度算法对常规BP网络进行改进,并将改进后的BP网络应用于单个目标航空弹药预测仿真试验中.结果表明,改进BP网络能克服局部极值、快速提高网络收敛速度,并能较为准确地预测航空弹药的需求量.

Application of Improved BP Neural Network to Prediction of Aviation Ammunition Consumption

Prediction of aviation ammunition consumption is very important in the future war. To get a accurate result,back propagation (BP) neural network is proposed to do the prediction. However, the conventional BP neural network is of two major defects which are slow convergence rate and "local minimum" problem. There-fore, the Fletcher-Reeves method which has global convergence is proposed to improve the conventional BP neu-ral network. The BP network hidden layers and their nodes are computed by experimental formulas. Then the im-proved BP network is applied to predict consumption of aviation ammunition against single target. The simulation results show that improved BP network which is quicker and more effective can predict the consumption of avia-tion ammunition accurately.