基于集成多神经网络在模拟电路故障诊断中的研究

以进行模拟电路故障诊断为主要目的,针对单神经网络故障字典法在进行复杂电路系统故障诊断时,对多故障和多任务诊断的不足之处,讨论了基于多故障的神经网络集成技术,采用集成多神经网络来提高诊断速度和精度,提出了集成多神经网络故障字典法来解决多故障任务,对基于层次分类模型的多重结构神经网络进行了研究,给出了两种对故障定位的统一融合算法,克服了采用单神经网络多故障时学习速度慢,出现新故障的网络要重新进行学习等缺点.并给出了应用实例.

Research of Fault Diagnosis for Analog Circuit Based on Integrated Multiple Neural Networks

When diagnosing complex analog circuit, the deficiencies on diagnosing multiple fault and multiple assignment by single neural network is obvious. Aim to fault diagnosis of analog circuit, using the multiple neural networks to improve the velocity and precision on diagnosis is discussed. By research on multiple neural networks, two integrated multiple neural networks fault dictionary methods fulfill multiple fault diagnosis and their united fusion algorithms are proposed. The deficiencies of low learning velocity on diagnose multiple fault by applying single neural network and the necessity of the relearning for new fault are overcome by this schemes. Finally, example is also given.