For the condition that lower accuracy exists in complex fault diagnosis of machine unit, an artificial immune intelligent method based on immune mechanism is proposed, dimensionless immune detectors are constructed and complex fault can be detect and diagnosed. Match thresholds are used and candidate detectors from nonself-space are generated, which can effectively reduce the black hole and unclear border. Immune programming is introduced into feature construct of complex fault diagnosis in order to obtain new feature parameter. A few type immune detectors are integrated and diagnosed by using evidential theory, which can directly be applied to complex fault diagnosis. These excellent immune detectors can improve the accuracy of complex fault diagnosis, and the experiment results show that complex fault can be accurately and rapidly diagnosed.