基于AFPN的汽车故障诊断研究

作者：金永夫1，郝平1*，张华波1，袁华炜2 日期：2010-06-28/span> 浏览：2804 查看PDF文档

基于AFPN的汽车故障诊断研究
金永夫1，郝平1*，张华波1，袁华炜2
(1.浙江工业大学 信息工程学院，浙江 杭州 310014; 2.绍兴文理学院 计算机学院,浙江 绍兴 312000)

摘要：针对汽车故障诊断专家系统的知识具有动态性和不确定性的特点.提出了一种基于自适应模糊Petri网（AFPN）的汽车故障诊断方法。该方法首先根据模糊产生式规则建立相应的模糊Petri网模型，然后利用BP算法，通过样本数据对模糊Petri网中的权值等进行反复的学习训练，最后利用训练得到的参数和故障征兆发生概率对故障进行诊断。这种方法使知识库能动态更新，并可以避免依靠人工经验设置带来的不确定性。诊断结果表明，和传统故障树方法相比，该方法具有很多优点。
关键词：自适应模糊Petri网；故障诊断；知识表示；知识学习；汽车
中图分类号：TP206+.3文献标识码：B文章编号：1001-4551(2010)05-0090-05

Research on auto fault diagnosis based on AFPN

JIN Yong-fu1, HAO Ping1, ZHANG Hua-bo1， YUN Hua-wei2
(1. College of Information Engineering, Zhejiang University of Technology, Hangzhou 310014, China;
2. Deportment of computer, Shaoxing University, Shaoxing 312000, China)

Abstract: Aiming that knowledge in the automobiles fault diagnosis expert system is dynamic and uncertain, a method based on the adaptive fuzzy Petri nets（AFPN） was proposed to solve the problem. In this method, at first a fuzzy Petri net was set up by the fuzzy production rule. Then the parameters of the fuzzy Petri net were trained by BP learning algorithm. At last, the fault origin can be found through the trained parameters and fault inferences. Using this method, the knowledge can be updated dynamically, and the uncertainty coming from people’s setting can be avoided as well. The diagnosis result indicates that this method has many more merits than the fault tree diagnosis method.
Key words: adaptive fuzzy Petri nets(AFPN); fault diagnosis; knowledge representation; knowledge learning; automobiles