针对热力系统故障发生的模糊性，提出将模糊辨识理论和人工神经网络相结合，建立回热系统故障诊断层次模型．该模型根据回热系统以加热器为中心的特点，分层诊断，缩小了知识库，减少了计算量，并利用了神经网络“边学习、边辨识”的优点，使故障诊断迅速、准确．经仿真实验，证明了模型辨识效果良好． Aiming at the fuzziness of thermal system fault, this paper proposes to combine the theory of fuzzy identification and artificial neural network to establish the fault diagnosis hierarchy model of regenerative system. According to the characteristics of the regenerative system with heater as the center, the model stratified diagnosis, reduced the knowledge base and reduced the computational complexity. The model made use of the advantages of “edge learning and edge identification” of neural network to make fault diagnosis quickly and accurately. The simulation experiments show that the model recognition effect is good.