基于赋初值BP算法的模拟电路故障的诊断

BP Algorithm Given Initial Value of Analog Circuit Fault Diagnosis

提出了采用赋初值BP神经网络的方法,进行模拟电路故障的诊断;该方法分两步实现:首先对网络输入层到末级隐层之间权值的设计,能简化网络结构,缩短训练时间;其次对BP网络输出层权值和阈值的设计,以加快收敛速度、抑制局部极小、减少了学习过程的振荡现象;文中通过心电信号放大器电路实例,对其诊断方法的原理与实现进行了较深入的研究,并通过计算机仿真,模拟诊断该电路,结果证明:该方法具有鲁棒性、准确性与快速性。

A given initial value using the method of BP neural network method take analog circuit fault diagnosis. The realization of the method have two steps： First, the network input layer to the end of the class weights between the hidden layer design, to simplify the net- work structure and short the training time; second, the output layer weights and the threshold value of BP network design, to speed up the convergence rate and check local minimum, at the same time, to reduct the oscillation phenomenon of the learning process. For the example of ECG amplifier circuit diagnosis and in--depth studies , to verify the diagnosis by simulating , we conclude that the use of the method can improve the robustness of the method, accuracy and speeding.