Abstract: The authors proposed a model for an expert system based on artificial neural networks. The neural networks are used as the knowledge base for the expert system. In the model, the production rules are standardized into binary rules, each rule containing only two antecedents and one conclusion. In order to store the rules in the neural networks, the binary rules are coded by a method presented in the model. A modified back propagation (BP) algorithm is used to train the neural network. They also designed a new inference mechanism based on the neural networks for the proposed model, and built a prototype system using the model for animal recognition. The results indicate the validity and rationality of the model.