Abstract: In this paper, the model-blocked of device structure and model-layered of faults diagnosed device are discussed. A new intelligent diagnostic model which combines artificial neural networks (ANN) with expert system (ES) is developed. Its knowledge base, knowledge representation, and knowledge acquisition are emphatically studied, and the diagnosing process is described. The tests are carried out in the laboratory on MK9-5 cigarette machine. The experimental results show that the new intelligent diagnostic approach has higher effectiveness and liability than the common diagnostic expert system. It is expected to be further developed and improved to be applied in practice.