The turbogenerator vibration faults have the character of variety. Many faults often occur synchronously. This paper introduces a diagnosing model based on parsimonious covering theory and probability. A model for turtogenerator's fault diagnosis is proposed. The availability of this method is proved by two fault diagnosis examples of turbogenerator. The results show that the model proposed can be used for multi_fault diagnosis together. It may make up shortage for some of expert systems and neural networks in some aspect. From the practice,this model has higher reliability and practicability.