Aiming at the urgent needs of those manufacturing factories of hobbing machine and their consumers, who want to improve the quality of fault diagnosis in their after services, and reduce the costs of seeking for faults, a networked fault diagnosis expert system for hobbing machine tools based SPF tree (System-Phenomenon-Fault) is put forward. The operation model of this web-based system and the architecture of expert system are brought out firstly. Based on them, some key technologies are discussed, such as the methods of multi-source data acquisition, knowledge representation, the reasoning mechanism of the knowledge system, and the network architecture based on P2P(Peer-to-Peer). Above all, the implementation methods of knowledge representation and reasoning mechanism based SPF tree are the points, which include the SPF’s definition, birth, description of elements, growth of tree trunk, mathematics description, learning-self process, statistic, retrieving and et al. A prototype system based on this study is tried out in a hobbing machine company and its feasibility has been verified in the application.