摘要
依据空战态势、空战效能以及目标战役价值,采用威胁指数法建立了空战目标威胁评估模型。在威胁评估的基础上,研究了空战中基于Elman神经网络的目标威胁排序方法。考虑到Elman神经网络的学习性能和收敛性,采用附加动量项、自适应改变各参数学习率以及重置算法改进网络权值的学习算法。算例结果表明,采用Elman神经网络对空战目标进行威胁排序的方法是有效的,且改进的学习算法提高了网络的学习效率,有效地抑制了局部极小值的出现。该方法有利于提高火控系统的智能化水平。
According to air combat situation, efficiency and target battle value, a target threat estiination model is established based on threat index method. On the basis of threat evaluation, a target threat sequencing method based on Elman neural network is studied. Considering the learning capability and convergence performance of Elman neural network, the network weight learning method is improved by adopting additional momentum item,self-adaptive learning rate of the parameters and early reset algorithm. The example results showed that the threat sequencing method based on Elman neural network is effective; the improved learning method can improve the learning rate of neural network as well as restrains the local minimum effectively. The introduced method is helpful for improving the intelligence level of fire control system.
出处
《电光与控制》
北大核心
2008年第8期1-4,13,共5页
Electronics Optics & Control
基金
航空基金(05C52007)
空装预研基金资助
