基于产生式人工神经网络的汽轮发电机组故障诊断专家系统研究

Research on Productive Artificial Neural Network(PANN) Based Fault Diagnosis Expert Systems for Turbogenerator

针对人工神经网络训练样本难以获取的困难，提出1种从专家系统获取训练样本的方法，使得产生式人工神经网络能够有效地逼近传统专家系统的诊断推理行为。系统已被应用于125MW汽轮发电机组的故障诊断。

A productive artificial neural net-work (PANN )based expert system is introducedin this paper,which is based on the architectureof the fault tree of the knowledge base. Amethod for acquiring the training samples forPANN is also described to ensure it to approxi-mate the diagnosis behavior of a specialist diag-nosis system-KBFDES. Because of the abili-ties of neural networks, the PANN expert sys-tem is essentially superior to the conventionalexpert system.