摘要
建立了一种 BP 和 Hopfield 构成的主从混合网络(BPHP)。该网络利用 HP 网络的动态演化过程加速 BP 网络的收敛速度,具有记忆特性好,收敛速度快、稳定性强等特点,在与恰当的特征提取方法结合使用之后可以获得较为理想的故障诊断系统。为了验证该方法的优越性,介绍了该神经网络在航空发动机故障诊断中的应用实例,仿真结果表明该方法具有很高的分类效率,具有较好的推广应用前景。
A new BPHP hybrid network structure is presented.This network possesses virtues of good memory,fast convergence ability and strongly stable capability.Application in aeroengine fault diagnosis testifies this method is prac- tical and instructional.Results show the method owns good robustness and high precision.This method is instructional and contributive to maintenance and logistics due to its high classification effectiveness.It has extensive future in the aeroengine fault diagnosis and maintenance.
出处
《弹箭与制导学报》
CSCD
北大核心
2005年第S7期392-394,共3页
Journal of Projectiles,Rockets,Missiles and Guidance
