基于补偿模糊神经网络的核电百万千瓦级汽轮发电机的故障诊断

Fault diagnosis for the million-kilowatt-class nuclear power turbo-generator based on compensatory fuzzy neural network

在独特的、自行设计的多功能核电百万千瓦级汽轮发电机试验台上,人工模拟了多种故障,选择反映其故障状态的不同频段参数组成特征向量,利用补偿模糊神经网络(CFNN)系统进行故障诊断分析。由于网络引入了补偿模糊神经元,使其容错率更高,系统更稳定。仿真实验证明该模型在智能诊断中具有收敛速度快,诊断精度高,而且适应性强等优点。

In this article,the different frequency band parameters and measures are selected to form the characteristic vector by making use of the compensation fuzzy neural network（CFNN）system to simulate and diagnose many kinds of faults on the unique multifunction million-kilowatt-class nuclear power turbo-generator test-bed.Because the compensation fuzzy neural ceil is introduced into the network,it can lead to the network fault-tolerant rate to be higher and the system to be more stable.The simulating experiment has proved that this model has many merits,such as： the convergence speed is quick,the diagnosis precision is high,and compatibility is strong and so on.