Abstract： Based on the research on typical fault diagnosis model, a configurable diagnosis model of induction motor was proposed to resolve the problem of complexity of network and difficulty of training. This model contains multiple sub RBF neural networks which have multiple inputs and single output, and one type of fault can be recognized by a specific sub-model. The sub-models can be any combination based on the demands of the system, and various faults can be identified. The model is trained using the samples with feature extracted, and the effectiveness of fault diagnosis model is verified through test samples. It is shown that one sub-model can be used to recognize one specific state of motor in the configured fault diagnosis model, the structure is simple, the difficulty of model training is reduced, the fault identification capability of model and flexibility of application are improved,  providing a new method for the induction motor fault diagnosis.