非线性模拟电路故障的BPNN诊断算法设计与实现

Algorithm and Its Realization of Fault Location for Nonlinear Analog Circuits Based on BP Neural Network

提出了用Volterra频域核作为故障特征对弱非线性定常动态网络进行故障诊断的方法,利用计算方法求出电路响应在各种常见故障状态下的Volterra级数解的各阶频域核,将其输入给BP神经网络,利用BPNN的分类功能建立故障字典,对实测的故障网络的各阶频域核进行测试样本分类来实现故障诊.通过故障诊断实例给出了各阶频域核的统一递推离散算式,并采用了改进BPNN算法及其程序实现.

A fault location method for nonlinear circuit by taking the cores of Volterra series in frequency domain as the fault characters is proposed. Via calculating the frequency cores of Volterra series in frequency domain for the response of usual fault states and feeding them into BPNN as the fault features, the fault dictionary is built. The fault location for nonlinear analog circuit is completed by BPNN classifying the Volterra frequency cores tested from the nonlinear network. The paper presents the unified recursive computing formulae for frequency cores,an improved BPNN algorithm and its program. At last, an example of this method is given.