多级模糊神经网络在故障诊断中的应用

Application of Multi-Grade Fuzzy Neural Networks in Fault Diagnosis of Large Machinery

提出一种基于模糊神经网络的多级故障诊断系统 ,根据多级的需要 ,除了建立故障谱知识库外 ,还分别建立了频谱特征知识库、振动变化特征知识库和故障位置特征知识库 ,并对其主要功能做了比较细致的论述 .文中特别对第 1 ,3级采用的基于模糊组织径向基函数神经网络及第 2级采用的模糊逻辑神经网络的学习算法做了较全面的论述 ,同时解决了模糊神经网络随着输入变量的增加 ,模糊规则呈指数增长带来网络训练的困难 .通过试验 ,研究了该系统在某炼油厂重催化机组故障诊断中的具体应用 .

A fault diagnostic system based on multilevel fuzzy neural networks is presented. According to multilevel, besides the fault spectrum knowledge sets established, there are frequency spectrum characteristic knowledge sets, vibration change characteristic knowledge sets and fault position characteristic knowledge sets. Moreover, there is a more careful description about main functions. In this paper, there is a more comprehensive description about their learning algorithm of the first and third levels neural networks based on fuzzy structure radix function, and the second level neural networks based on fuzzy logic, the problem of fuzzy rule with exponent growth has been solved. By emulation test, the system has been investigated in material application to fault diagnosis in large machinery of some refinery.