Abstract: This paper introduces a vibration diagnostic system for reciprocating machine. Its main function is not only to monitor operation state but also to diagnose fault for reciprocating machine. The software system consists of two parts, one is in the diagnostic machine and the other is in the PC computer. It can perform diagnosis on spot, realize selflearning and can adapt to many types of machine. It takes main factor regressive model as character extractor and ANN model as classifier. With the combination of the two technologies, the system can perform data sampling, signal analyzing, character extracting and fault recognizing automatically. It can reduces the artificial factors in the diagnosing process, so the accuracy of the diagnosis and the level of intelligence and automation is raised.