The electronic system's reliability becomes the key of the normal operation of whole system; so circuit fault diagnosis has attracted more and more attention. The method based on BP neural networks is an effective approach of analog circuit fault diagnosis. In this paper, aiming at the drawbacks of fault diagnosis methods based on BP neural networks for analog circuit, a combinatorial optimization diagnosis scheme is proposed. First, the initial weights of BP neural networks are optimized by genetic algorithm( GA) to avoid local minima in the scheme, and then the BP neural networks is finely tuned with Levenberg-Marquardt ( L-M) method in the local solution space to look for the optimum solution,or approximate optimal solutions. The scheme makes good use of the mapping capabilities of BP neural networks and the global search ability of GA; it also accelerates the networks' learning speed. Experimental results show preliminarily that the scheme comprehensively improves the whole learning process approximation and generalization ability, and effectively promotes analog circuit fault diagnosis performance based on BP neural networks.%电子系统的可靠性已成为影响系统正常运行的关键,因此电路故障的诊断越来越受到重视.基于BP神经网络的诊断方法是目前实现模拟电路故障诊断的有效方法之一.文章针对已有BP神经网络模拟电路故障诊断技术的不足,提出了一种组合优化的诊断方案.该方案采用遗传算法优化确定BP神经网络的初始权值,以规避BP神经网络易陷入局部极小值的不足,然后应用L-M方法在这个局部解空间里对BP神经网络进行精调,搜索出最优解或者近似最优解.该方案发挥了BP神经网络的广泛映射能力和遗传算法的全局搜索能力,加快了网络的学习速度,综合提高了整个学习过程中的逼近能力和泛化能力,有效提升了基于BP神经网络模拟电路故障诊断的性能.
展开▼
