摘要
针对现有基于前馈式神经网络的引信故障预测算法存在自适应能力差、故障预测误差大等缺点,提出了一种基于简单递归神经网络的引信故障预测算法。该算法把经验和简单递归网络在故障预测方面的优势结合在了一起,大大缩短了网络训练时间,提高了引信的故障预测精度,而且有很强的适应能力和良好的稳定性。用此算法对引信工作参数进行预测,仿真、计算结果表明当Elman网络中隐含层神经元节点个数为50个时,引信各个工作参数的预测结果值与实测值之间的误差不到±1%,预测误差最小,而且每次的预测结果之间波动很小,证明了此算法的合理性。
In view of the existing fuze fault based on feedforward neural network prediction algorithm has poor adaptive capacity, fault prediction error and shortcoming, proposed a fuze fault simple prediction algorithm based on recurrent neural network. The algorithm combines the advantages of experience and simple recurrent network fault prediction combined together, greatly reducing the network training time and improve the accuracy of the fuze failure prediction , and has a strong adaptability and good stability. With this algorithm to forecast working parameters for fuze, simulation, calculation results indicate that when the Elman network when the number of neurons in the hidden layer nodes for 50, fuse all working parameters error between the predicted and observed values less than ±1%, prediction error minimization, and each time it fluctuated between predictions was small, prove the reasonableness of this algorithm.
出处
《电子设计工程》
2015年第9期61-63,共3页
Electronic Design Engineering
