The fire data error series from accuracy test of the airborne fire control system is generally a non-stationary random process. The stability and error model are usually undiscovered. According to he error property, the auto regression model of the error data serial was built based on the accuracy test data series using the theory of time serial model. The character function of AR model and the character root of the function were introduced to evaluate the fire control system stability. The validity of the model was showed by a half physics test data of an airborne optical-electronic fire control system.