论文部分内容阅读
针对凝汽器故障诊断问题,提出了一种基于粗糙集和证据理论相结合的故障诊断方法。利用粗糙集相对约简的不唯一性,对凝汽器故障征兆进行分类,形成不同的证据来源,既实现了证据理论对于同一事物要求有不同的证据来源的要求,又对故障征兆参数进行了降维处理,减小了网络的规模,有效缓解了由于输入参数过多给网络带来的收敛困难问题。该诊断方法将粗糙集、神经网络和证据理论有机地结合在一起,使三者优势互补,充分利用了凝汽器故障征兆的冗余、互补信息。实例证明,基于多故障诊断网络信息融合的诊断识别准确性和可靠性比基于单一故障诊断网络的诊断识别有较大的提高。
Aiming at the problem of condenser fault diagnosis, a fault diagnosis method based on rough sets and evidence theory is proposed. Using the non-uniqueness of relative reduction of rough sets, the classification of condenser fault symptoms forms different sources of evidence, which not only fulfills the requirement of evidence theory that different evidence sources are required for the same thing, but also carries out the fault symptom parameters The dimension reduction process reduces the size of the network and effectively alleviates the problem of convergence caused by excessive input parameters to the network. The diagnostic method combines rough set, neural network and evidence theory organically, so that the three are complementary in each other ’s advantages and make full use of redundant and complementary information of condenser fault symptom. The example proves that the accuracy and reliability of diagnosis identification based on multi-fault diagnosis network information fusion are greatly improved than that based on single fault diagnosis network.