The effects of real combined variances in components and modules of aero engines, due to production tolerances or deterioration, on the performance of an aircraft engine are analysed in a knowledge-based process. For this purpose, an aero-thermodynamic virtual evaluation process that combines physical and probabilistic models to determine the sensitivities in the local module aerodynamics and the global overall performance is developed. Therefore, an automatic process that digitises, parameterises, reconstructs and analyses the geometry automatically using the example of a real turbofan high-pressure turbine blade is developed. The influence on the local aerodynamics of the reconstructed blade is investigated via a computational fluid dynamics (CFD) simulations. The results of the high-pressure turbine (HPT) CFD as well as of a Gas-Path-Analysis for further modules, such as the com-pressors and the low-pressure turbine, are transferred into a simulation of the performance of the whole aircraft engine to evaluate the overall performance. All results are used to train, validate and test several deep learning architectures. These metamodels are utilised for a global sensitivity analysis that is able to evaluate the sensitivities and interactions. On the one hand, the results show that the aerodynamics (especially the efficiency ηHPT and capacity _mHPT)are particularly driven by the variation of the stagger angle. On the other hand, ηHPT is significantly related to exhaust gas temperature (Tt5), while specific fuel consumption (SFC) and mass flow _mHPT are related to HPC exit temperature (Tt3). However, it can be seen that the high-pressure compressor has the most significant impact on the overall performance. This novel knowledge-based approach can accurately determine the impact of component variances on overall performance and complement experience-based approaches.