One of the main problems associated with design of fault detection (FD) strategy is availability of first generation engine. To solve this problem a methodology based on a virtual engine platform is proposed in this paper. This approach allows designing an FD algorithm in early stages of engine development. The application of this methodology is illustrated on a modern turbocharged gasoline engine by investigating the effect of a leakage in the exhaust manifold. Experimental results show good ability of the virtual engine platform to predict the effect of the leakage fault on the engine performance. Moreover unexpected results of exhaust manifold leakage effects are presented which are very useful for designing a leak detection strategy.