基于仿真数据的神经网络故障诊断方法  

Neural network method for faults diagnosis based on simulation data

作　　者：赵导[1] 齐晓慧[1] 田庆民[1] 任红伟[2] 甄云卉[1] 

机构地区：[1]军械工程学院光学与电子工程系,河北石家庄050003 [2]军械工程学院训练部,河北石家庄050003

出　　处：《计算机工程与设计》2010年第9期2020-2022,共3页Computer Engineering and Design

基　　金："十一五"装备预先研究基金项目(51325050203)

摘　　要：针对某些模拟电路的历史故障信息,专家知识及其诊断经验难以获取的状况,提出了一种基于仿真数据的神经网络故障诊断方法。通过使用PSpice电路仿真软件模拟实际电路,生成训练样本训练神经网络,从而建立了电路的输出响应与电路中元件实际值之间的映射。以电路的输出响应和技术指标为判断依据,诊断电路的当前状态、定位故障元件及其偏差。最后以带通滤波器电路为例,对整个过程进行了仿真试验,验证了方法的可行性。For some analog circuits, the information of faults history and expert knowledge is not easy to be obtained. Considering this status, a neural network method for faults diagnosis based on simulation data is proposed. PSpice is used to simulate real circuit and generate training samples to train the neural network. Thus, the output of the circuit is mapped the actual value of components in the circuit. Making the output and the index of the circuit as judgement standard to diagnosis the present state of circuit, locating fault component and its deviation. A band-pass filter circuit is taken as the example and the feasibility of method is verified.

关 键 词：故障诊断 模拟电路 电路仿真 神经网络 诊断方法 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]