【摘要】 提出了一个基于模糊隶属度和规则的分类层次诊断模型。针对该模型,首先以汽轮机通流部分故障为对象,讨论了层次分类的方法,根据结构和故障分解的原则建立了故障诊断树;其次根据热力参数的实际情况,选择模糊隶属度函数并确定隶属度函数的算法;最后综合这两种方法的优点,设计了故障节点的知识组织结构,把每个故障节点的知识库分成工况参数、初始证据源、证据模式、神经网络信息、模糊规则库、索引知识等6个部分。该模型既减少了故障判断的搜索数量,又把诊断所需的各种模糊不确定的知识用模糊神经网络的权重来表示,知识的获取通过模糊神经网络的训练进行,解决了知识获取的"瓶颈"问题。经过实际故障诊断验证,该模型对于通流部分故障诊断具有很好的适用性。图2表4参2。更多还原

【Abstract】 A hierarchic classification diagnosis model, based on fuzzy membership grade and rules is being proposed. First of all, the hierarchic classification method is discussed by taking steam turbine blading faults as the object concerned, and a fault diagnosis tree, based on decomposition of structure and fault events, is set up. Next, in accordance with actual conditions given by the thermal parameters, a fuzzy membership grade function is chosen and its algorithm fixed. Finally the advantages of these two methods are synthesized in the design of the fault nodes’ organizational structure of knowledge with the knowledge repertory of each fault node divided into different compartments for storing behavior parameters, original evidences, proof patterns, neural network information, fuzzy rule bank and index knowledge. This model, which not only saves searching work for fault decisions, but also expresses fuzzy and uncertain knowledge by the fuzzy neural network’s weight and acquires knowledge through the fuzzy neural network’s training, thereby solving the "bottleneck" of knowledge acquisition. Practical probation of fault diagnosing capability shows that the model suits the purpose of blading fault diagnosis quite well. Figs2, tables4 and refs2.更多还原