In this paper, Fault diagnosis of thermal process was studied deeply based on fuzzy RBF neural network, with knowing the current situation of thermal system fault diagnosis, and analysing the key link of fault diagnosis. Fistly, Parameter pretreatment of fault diagnosis and Symptom fuzzy calculation was studied, two types of symptom, trend symptom and semantic symptom fuzzy calcuationg methods and integrated calculation were discussed. The structure and training algorithms of RBF artificial neural networks were realized, especially the grads training method. The symptom calculcation and RBF diagnose algorithm were built in FORTRAN language by modularized method. Have taken the high-pressure heater of 300MW thermal power unit as example, its fault characteristic rules were analysed, and its fault fuzzy knowledge bases were summarized. The high-pressure heater fault diagnosis model was built with diagnose algorithms, the tests of simulating fault were done under simulation supporting system, the validity of diagnose method is verified. |