Expert system as a computer program can imitate human experts to diagnose fault of rotating machinery quickly and accurately, which makes diagnose automatically and intelligently and offers a guarantee for device operation in security, stability, and high-quality. The performance of diagnosis system will be degraded if the organization and management of fault diagnosis knowledge of rotating machinery that is more and complex are illogical. For this reason, the fault knowledge is divided according to the different parts and stored in databases in the form of production rules. As a result, the querying, modifying, adding, deleting of knowledge can be easily realized by the basic technology of database. Finally, an expert system based on forward-backward hybrid reasoning for general fault diagnosis of main parts of rotating machinery, such as rotor, roll bearing, gear and so on, has been designed, the rotor fault are verified and diagnosis result are accurate. The system is reasonable in structure, rich in information, easy in management and maintenance of data base.