A simple artificial intelligence system for fault diagnosis established in this paper. The system consists of knowledge base，fuzzy reasoning module，neural network module and control module. The fault pre-diagnosis is completed in the fuzzy reasoning module by use of fuzzy synthetic judgement method. The identification of the fault types is achieved in the neural network module using competitive learning algorithm. The seven models for the single and composite. faults of a cantilever beam have been investigated. The results are satisfactory.