Engine subsystem models are not commonly used in design optimization studies as it is computationally expensive to solve these models for a large number of iterations. To reduce the computational cost of such optimizations, a novel multi-fidelity Kriging-based optimization approach is proposed that uses shell FEMs to provide a low-fidelity response and solid FEMs to provide a high-fidelity response. This marks the first time that shell and solid models are used together in a multi-fidelity surrogate modelling approach. The shell FEMs are generated from medial surfaces that are extracted from solid component geometries in a semi-automatic manner. This approach is applied to a case study for optimizing the intercasing subsystem from the CRESCENDO whole engine model. The results show that the optimum design found by the multi-fidelity Kriging approach was on par with the optimum design found by a single-fidelity Kriging approach using only solid FEMs which is more than twice as expensive to run. The shell and solid FEMs were also shown to be well-correlated such that optimization studies employing only the shell FEMs by themselves could generate designs that are feasible with respect to the design constraints imposed on the solid model.