Fault diagnosis to complex system with one method is insufficient. The characteristics of fault diagnosis based on neural networks and fuzzy logic systems and their union wers discussed. A kind of effective method to safeguard industrial production was presented: graduation fault diagnosis and alarm system. Fault diagnosis and alarm system based on industrial control nets were simulated and designed. The results show that the algorithm based on radial basis function neural network and fuzzy logic has faster training speed and better generalization ability, and it can distinguish multiroutes faults.