基于PNN神经网络的电控发动机故障诊断  

PNN-based fault diagnosis of electrically-controlled engine

作　　者：巴寅亮[1,2] 王书提 李春兰[1,2] 郭增波[3] 加克.乌云才次克 

机构地区：[1]新疆农业大学机械交通学院,新疆乌鲁木齐830052 [2]新疆农业工程装备创新设计重点实验室,新疆乌鲁木齐830052 [3]克拉玛依职业技术学院,新疆克拉玛依834000 [4]新疆交通职业技术学院,新疆乌鲁木齐830052

出　　处：《现代电子技术》2016年第20期146-148,153,共4页Modern Electronics Technique

基　　金：国家自然科学基金(51467021)

摘　　要：PNN是前馈型神经网络,具有强大的非线性模式分类能力。提出运用PNN神经网络对发动机电控系统进行故障诊断的方法,介绍了PNN神经网络及其工作原理,以伊兰特汽车发动机电控系统为研究对象,让发动机在怠速情况下,并对其进行故障设置,运用金德KT600故障诊断仪采集发动机故障数据流,利用PNN神经网络建立诊断模型,并对网络诊断模型进行验证,诊断结果完全正确,且训练速度非常快。实验结果表明PNN神经网络具有较强的泛化能力和实用价值。The probabilistic neural network （PNN） is a feedforward neural network, and has strong ability to classify the nonlinear patterns. The method of using PNN to diagnose the fault of the engine＇s electrically-controlled system is proposed. The PNN and its working principle are introduced. The electrically-controlled system of the Elantra car engine is taken as the research object. The Kinder KT600 fault diagnosis instrument is used to collect the failure data flow of the engine while the engine is kept in idle speed. The PNN is used to establish the diagnosis model. The network diagnosis model was verified. The diagnosis results are completely correct, and the training speed is very fast, which show that the PNN has good generalization ability, and has a certain practical value.

关 键 词：PNN神经网络 发动机 电控系统 故障诊断 

分 类 号：TN926-34[电子电信—通信与信息系统] TK428[电子电信—信息与通信工程]