基于改进型小波神经网络的灭火系统故障诊断

在装甲车辆灭火系统故障诊断中,小波神经网络算法能将故障诊断定位到元件级,但各元件存在容差,导致参数变化的连续性和随机性使得诊断率不高。为了提高小波神经网络算法在灭火系统中的诊断率,针对网络在运行过程中存在着收敛效果差、训练误差大及容易陷入局部极小值的缺点,网络无法继续训练和测试,提出一种以增加动量的小波自适应神经网络的改进型算法,可以使网络运行更稳定,学习速率更快。经MATLAB仿真实验表明改进后的算法诊断率远高于普通算法。

Fire-extinguishing system fault diagnosis based on improved wavelet neural network

In armored vehicle fire-extinguishing system fault diagnosis,wavelet neural network algorithm can make fauh location to the component level, however, components tolerances results in the continuity and the randomness of the parameter changes,which lead to that the diagnosis rate is not high. In order to improve the diagnosis rate of wavelet neural network algorithm for the fire-extinguishing system, for the network during operation there is poor convergence, the training error and shortcomings easy to fall into local minimum, the network is unable to continue training and testing, proposing a new algorithm of increasing the momentum of the wavelet adaptive neural network, which can make the network run more stable, faster learning rate. The MATLAB simulation experiments show that the diagnosis rate of the improved algorithm is much higher than the ordinary one.