基于重复覆盖算法的模拟电路故障诊断研究

Approach of soft fault diagnosis in analog circuit based on repeated covering algorithm

针对传统构造神经网络的方式中存在的问题,基于领域覆盖算法(Neighborhood Covering Algorithm,NCA)来构建神经网络对电路进行故障诊断,运用重复覆盖算法(Repeated Covering Algorithm,RCA)对采用领域覆盖算法构建的神经网络的网络结构进行修缮。最后用某滤波电路进行仿真分析,比较两者的诊断结果。通过仿真结果可以看出,在保障准确率的前提下,重复覆盖算法能降低神经元个数,优化了网络的结构,提高神经网络的泛化能力,验证了该方法的可行性。

In view of the problem of the application of traditional neural network in analog circuit fault,a method based on neighborhood covering algorithm( NCA) is used to diagnose the fault,and repeated covering algorithm( RCA) is used to improve the deficiencies of the neural network which is built based on neighborhood covering algorithm. Finally,the well-trained neural network is used to diagnose a filter circuit faults,by comparing the differences of diagnostic rates. The experiment shows that the improved repeated covering algorithm can build the networks efficiently and the fault diagnosis is improved.