The Statistical Energy Analysis (SEA) method has been shown to be a powerful predictive tool of structural-acoustic design. However, there are many difficulties when the method is applied to an engine structure. It is because 1) the engine structure is so complicated that both the internal and coupling loss factors have to be determined in situ by measurements and 2) the general matrix inversion to calculate the loss factors, with the power injection method, can not simply be used as it produces many negative and unreliable results. The problems are attacked with several new approaches, which include using the equivalent mass, the selective geometric averaging, the consistency and reciprocity relationships, the back estimation technique and the coupling damping. Finally, an experimental validation is done. It shows that the obtained SEA model generally can predict the vibration tendency of the engine structure when damping and isolation treatments are applied. It is also concluded that the drive points should be selected in places as far as possible from the coupling lines and one or two drive points for each subsystem may be enough in engineering practice.