Abstract: An expert system based on neural networks for predicting and judging the state of blast furnace is developed by applying the method of combining neural networks with traditional expert system. The system with high hit ratio has strong adaptability and good self-learnning function.