基于不完备模糊规则库的信息融合故障诊断方法  

Information Fusion Method for Fault Diagnosis Based on Incomplete Fuzzy Rule Bases

作　　者：徐晓滨[1,2] 吉吟东[2] 文成林[3] 

机构地区：[1]清华大学自动化系,北京100084 [2]清华大学信息技术研究院,北京100084 [3]杭州电子科技大学自动化学院,杭州310018

出　　处：《南京航空航天大学学报》2011年第B07期55-59,共5页Journal of Nanjing University of Aeronautics & Astronautics

基　　金：国家自然科学基金(61004070;60934009;61034006;60874105)资助项目;中国博士后基金(20100470353)资助项目;国家科技支撑计划(2009BAG12A08)资助项目

摘　　要：在基于模糊推理的故障诊断专家系统中,规则库的不完备常常会引起系统的某些输入产生的输出不确定。针对此问题,将模糊推理与证据理论相结合提出了一种新的融合诊断方法。首先根据证据的随机集表示及随机集扩展准则将模糊输入映射到输出,并聚合生成故障命题的基本概率赋值(BPA),以度量由规则库不完备和输入模糊性引起的输出不确定。然后利用Dempster组合规则将多个不完备规则库提供的BPA融合,用融合结果判决故障,并以电机转子故障诊断为例,验证了所提方法可以有效地提高故障确诊率。In the fuzzy-reasoning expert system of fault diagnosis, the fuzzy rule base is often incomplete, thus leading to uncertain outputs for some possible inputs. A fusion fault diagnosis algorithm is presented based on the fuzzy rule reasoning and the evidence theory. Firstly, using the random set formalism of evidence and the extension principle of random set, fuzzy inputs are mapped to outputs, then integrated to generate basic probability assignment （BPA） for measuring the uncertainties of outputs caused by the incompletion of system and the fuzziness of inputs. Then, the Dempster combination rule is used to fuse BPA coming from different incomplete rule bases, and faults can be decided by fusion results. Finally, fault diagnosis tests on machine rotor show that the method can effectively improve diagnostic rates.

关 键 词：故障诊断 模糊推理 信息融合 证据理论 随机集理论 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP206[自动化与计算机技术—控制科学与工程]