This master thesis is a principal study of how model based diagnosis with dynamic models can be applied to the turbo-jet engine RM12. A model based diagnosis system can be used to detect failures automatically. Automatic fault detection is of interest since it simplifies the reperation and increases the safety. A principal study of model based diagnosis applied to RM12 has therefore been performed. The study was performed by linearizing a nonlinear engine model of RM12 in an operating point utilizing a perturbation technique. Then the linear model was reduced with two different methods, reduction via balanced realization and reduction with physical interpretation of states using modal analysis. The diagnosis system was designed from the reduced models with a method called the Minimal Polynomial Basis method. Simulation of faults were performed with the linearization of the engine model. The principal study shows that model based diagnosis with dynamic models applied to RM12 isolates small sensor and actuator faults. The study also shows that a small and well conditioned design model gives the best performing diagnosis system.