OBD (On Board Diagnosis) has been applied to detect malfunctions in powertrains. OBD requirements have been extended to detect various failures for ensuring the vehicle emission control system being normal. That causes further costs for additional sensors and software works. Two layers diagnosis system is proposed for a passenger car gearbox system to detect changes from normal behavior.Conventional physical constraints based diagnosis is placed on the base layer. Model based diagnosis and specific symptom finding diagnosis are built on the second layer. Conventional physical constraints based diagnosis is direct and effective way to detect the failure of system if the detected signals exceed their normal ranges. However under the case of system failure with related signals still remain in normal ranges, the conventional detection measures can not work normally. Under this case, Model based diagnosis is proposed to enhance the functionality of diagnosis system. There are several methodologies to realize Model based diagnosis, the residual method was selected among them.Proposal strategy was implemented into an AMT control system. A plant model was developed for the proposed strategy. A conventional diagnosis strategy was combined with the model based diagnosis algorithm in a Simulink model. Some of states which are not measured by sensors are estimated an observer. Residuals are calculated with measured and estimated states and evaluated with a Fuzzy logic if failures occur or not.The proposed strategy was validated in time domain simulation. Response and resolution of the proposed diagnosis system were discussed and evaluated. The simulation study indicated the effectiveness of the method.