A standard state model verifying fault diagnosis method is proposed to overcome the problem in which it is hard to acquire enough samples for intelligent fault diagnosis. Firstly, the motion parameters of the system are acquired by standard state test. Then the test data is used as the standard, the simulation model of the conveyor is verified by changing the selected uncertain variables. Finally, a large number of samples are generated from the simulation model for training the neural network model so as to get a diagnostic machine. This method is used in chain circulating conveyor's fault diagnosis, and the diagnosis results are very effective.