基于退火BP神经网络的模拟电路故障诊断方法

针对BP神经网络在模拟电路故障诊断中存在的固有缺陷,提出模拟退火(simulated annealing,SA)算法与BP算法相结合的神经网络故障诊断方法,发挥SA算法全局寻优的特点来优化BP神经网络的学习过程,避免网络训练收敛速度慢和陷入局部极小;应用MATLAB编写程序对一典型模拟电路的仿真故障样本进行训练,在其它初始条件都相同的情况下,基于提出的算法神经网络比传统BP算法神经网络训练收敛速度快且诊断精度高,是一种新的有效的模拟电路故障诊断方法。

Fault Diagnosis of Analog Circuit Based on Simulated Annealing BP Neural Network

To avoid the limitation of the BP neural network in fault diagnosis of analog circuit,a neural network fault diagnosis combined with simulated annealing and BP algorithm is presented.In the diagnosis the method of global optimization is used in BP network learning to avoid the slow convergence and prevent it from falling into local minimum.A MATLAB program is written to train a typical analog circuit simulated fault samples.On the same initial conditions,the new diagnostic method is fast convergence and mor...