摘要
水轮发电机组的故障诊断具有模糊性和耦合性,提出一种基于模糊神经网络FNN的水轮发电机组振动故障在线诊断方法。首先,对反映转子振动状态的轴心轨迹用分形维数提取其结构特征,实现图形量化,以便FNN在线识别;接着,以6种典型振动故障为研究对象,在总结了包括轴心轨迹在内4类共14种故障征兆的基础上,分析各故障征兆的模糊属性,给出它们的模糊处理;然后,建立一种六层的前向FNN映射征兆到故障间的模糊推理,并给出学习算法修正网络参数;FNN通过自学习可保证良好的在线诊断精度。实例分析结果验证了其可行性。
Considering the fuzzy and coupling characteristics of fault diagnosis for hydroelectric generating sets, this paper proposed a vibration fault on-line diagnosis method based on fuzzy neural networks（FNN）. At first, fractal dimension was applied to extracting structure feature of shaft orbit, which reflectd vibration state of rotor, to realize figure quantification for online recognition of FNN. Aiming at six typical vibration faults, including shaft orbit, four categories in all fourteen fault symptoms were summarized, their fuzzy attributes were analyzed and the corresponding fuzzifications were described. Then a sixlayer FNN was established to express fuzzy inference from symptom to fault, and its learning algorithm was provided to modify network parameter. By self-learning, FNN could guarantee a higher precision of the on-line diagnosis. The analytical results of cases verify the feasibility of the diagnosis system based on FNN.
出处
《计算机应用研究》
CSCD
北大核心
2007年第12期231-234,共4页
Application Research of Computers
基金
国家"973"重大资助项目(2002CB312200)
关键词
故障诊断
盒维数
模糊神经网络
水轮发电机组
轴心轨迹
fault diagnosis
box counting dimension
fuzzy neural networks
hydroelectric generating sets
shaft orbit
