基于混合型专家系统的重型机床故障诊断  

Mixed-expert-system-based fault diagnosis for heavy machine

作　　者：王远航[1] 邓超[1] 吴军[2] 熊尧[1] 

机构地区：[1]华中科技大学数字制造装备与技术国家重点实验室,湖北武汉430074 [2]华中科技大学船舶与海洋工程学院,湖北武汉430074

出　　处：《计算机集成制造系统》2010年第10期2139-2147,共9页Computer Integrated Manufacturing Systems

基　　金：国家自然科学基金资助项目(50975109);国家863计划资助项目(2009AA043403)~~

摘　　要：考虑了重型机床结构和故障的复杂度,将重型机床故障分为三个等级,引入规则诊断体和案例诊断体的概念,在产品结构树的基础上演化出结构故障树的知识组织方式,分别用规则诊断体间的映射方法和框架表示法描述故障关联和故障案例,并提出了数据库组织结构。根据重型机床规则特征,通过BP神经网络训练规则诊断体实现基于规则推理的诊断,同时提出基于模糊隶属度的案例诊断体CBR诊断方法,通过最近相邻算法实现未知案例的匹配检索,介绍了案例重用和调整,针对案例库管理,用基于TC相似度的K-近邻算法实现新案例入库,并介绍了原始样本案例库"分类关联—聚类验证—调整入库"的构建过程。最后,进行了原型系统的开发。Considering the complexity of heavy machine structure and fault,the faults of heavy machine was divided into three levels,and the concepts of＂rule diagnosis body＂and＂case diagnosis body＂were introduced.The knowledge organization mode of＂structure fault tree＂was evolved based on the product structure tree.And then mapping between rule diagnosis bodies and frame representation were used to describe fault association and fault case respectively.Also,the organization of database was put forward.According to rules characteristics of heavy machine,the rule-based diagnosis was implemented by Back Propagation（BP）nural network trained by the information of rule diagnosis bodies.Besides,the case-based reasoning diagnosis based on case diagnosis body converted to fuzzy membership was proposed as well,in which matching retrieval of to-be-solved case was performed through the nearest neighborhood algorithm.Case reuse and case adjustment were introduced.Aiming at case-base management,new cases were stored by Knearest-neighbor algorithm based on TC similarity.The construction process of original sample cases was introduced through＂classification of associations-verification by clustering-storing after adjustment＂.Finally,prototype system was developed.

关 键 词：重型机床 故障诊断 专家系统 故障树分析 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程] TP206.3[自动化与计算机技术—控制科学与工程]