In order to resolve the problem that the Petri Net is restricted in the process of diagnosis in the complex system, Adaptive Petri Nets (APN) is ameliorated building fault diagnosis models aimed to accurately inspection troubles with some predigested incomplete and uncertain characteristic information of the monitoring system. Rough Set (RS) is adopted to deal with the flocks of data which is discrete, and the correlative information could be disinterred as using correlative rules mining. The composite algorithm exerting the preponderance in illation and diagnoses has expansive application potential. A parameter optimization and diagnosis example of the linear induction motor (SLIM) based on the theory is presented, the availability and efficiency of the method has been proved.