基于进化神经网络的模拟电路故障诊断  

Analog Circuit Fault Diagnosis Using Evolving Neural Networks

作　　者：周世官[1] 张云[2] 

机构地区：[1]宁波工程学院 电子与信息工程学院,浙江宁波315016 [2]山东省科学院 自动化所,山东济南250014

出　　处：《计算机测量与控制》2007年第8期991-993,共3页Computer Measurement &Control

基　　金：浙江省教育厅科研项目(20050014);浙江省教育科学2006年度规划研究课题(SC123);宁波市自然科学基金项目(2006A610033)

摘　　要：基于人工神经网络的智能故障诊断系统作为人工智能技术在模拟电路故障诊断领域的应用,在实践中取得了一定的成效。但由于容差和非线性特性使得模拟电路的故障诊断趋于复杂化,以及诊断系统中神经网络的拓扑结构难以确定,因此,针对上述局限,提出了模拟电路故障诊断的进化神经网络方法;文中详细的阐述了进化神经网络的构成方式,提出了把网络的结构和权值分级进化的方法,并在两级进化的过程中使用不同的适应度函数及改进的遗传算法。举例说明诊断系统的具体实现方法,仿真结果表明,在相同的精度要求下,该算法的训练时间远小于普通的进化神经网络,对模拟电路的故障诊断有一定的实际意义。Based on the artificial neural network intelligent fault diagnosis system took the artificial intelligence technology in the analog eircuit faultdiagnosis field, has obtained better result in the praetiee. But beeause the toleranee and nonlinearity in analog eireuits make its fault diagnosis complicated, as well as in the diagnosis system the neural network topology would be determined with diffieuhy, therefore, in view of above limitation, evolving neural network method is put forward in analog circuit fault diagnosis. Detailed explaining composition way of evolve neural network, proposed the network structure and value graduation evolution method, use different fitness funetion and improved genetic algorithm while two grades evolve. Example eireuits are provided to illustrate the proposed method. The simulation result indicated that, under the same preeision request, this algorithm training time far is smaller than the ordinary evolution neuralnetwork.

关 键 词：故障诊断 进化神经网络 遗传算法 模拟电路 

分 类 号：TN7[电子电信—电路与系统]