Abstract: A new fault diagnosis method for high precision servo simulator is presented in this paper. Based on the field observation, a fault feature pattern base is built up. The ant colony optimization (ACO) is used for the optimal classification problem of the fault feature patterns, in which an index function based on the sensitivity and the specificity is defined to evaluate the performance of the diagnostic rule obtained by ACO, in order to decrease the redundant information of the fault features so that the diagnosis rules can be reduced. Diagnosis results of several types of faults of a real high precision servo simulator show that the diagnosis method based on ACO is characterized by fast convergence speed, strong robustness, high diagnosis accuracy, and reliable results.