粗糙集-神经网络与证据理论融合推理的综合故障诊断方法研究  

Synthesized Fault Diagnosis Method Based on Reasoning Integration of Rough Set-Neural Network and Evidence Theory

作　　者：杨广 吴晓平[2] 辛登松 李庆 

机构地区：[1]海军91918部队 [2]海军工程大学信息安全系

出　　处：《武汉理工大学学报（交通科学与工程版）》2013年第1期62-66,共5页Journal of Wuhan University of Technology（Transportation Science & Engineering）

基　　金：国家自然科学基金项目资助(批准号:70471031;60774029)

摘　　要：针对目前故障诊断方法处理数据量大、故障类型复杂多变的问题时存在诊断准确率不高、实时性差,利用信息融合和综合集成原理,将粗糙集、神经网络及证据理论有机的结合,提出了基于三者集成的综合故障诊断融合推理策略:通过并行神经网络结构提高局部诊断网络的诊断能力,借助粗糙集的数据预处理,将复杂的神经网络约简并剔除其中的冗余属性从而确定网络的拓扑结构,克服了网络规模过于庞大和分类速度慢的缺点;并给出了基本可信度分配的客观化方法;充分利用各种故障的冗余和互补信息,运用证据理论的组合规则对各证据进行合成及决策,可显著提高故障诊断的准确性和效率.给出的实例验证了该方法的可行性和有效性.For the reasons of low diagnosis accuracy and bad real-time capability of traditional fault diagnosis methods in handling diagnostic problems such as lots of data and various complex faults,a synthesized fault diagnosis reasoning strategy of fusing rough sets,neural network and evidence theory is presented by means of data fusion and meta-synthesis theory.Firstly,the diagnosis ability of the local diagnosis networks is advanced through parallel neural network structure;with the rough set theory,the complex neural networks are simplified by eliminating redundant properties,which overcomes networks’ shortcomings of large scale and low-rate classification and the NN structure is determined.Secondly,an impersonal means obtaining basic reliability distribution of evidence theory is given.Finally,through taking full advantages of various redundant and complementary fault information,the accuracy and efficiency of the fault diagnosis are improved obviously by using combination rule of evidence theory.A given diagnostic example proves the method is feasible and available.

关 键 词：故障诊断 信息融合 粗糙集 神经网络 证据理论 

分 类 号：U671.99[交通运输工程—船舶及航道工程] TP183[交通运输工程—船舶与海洋工程]