The proposed methodology employs geometric design parameters as inputs for training surrogate models, establishing relationships between shape characteristics and aerodynamic pressure fields. This achieves computational efficiency while maintaining physical consistency across different configurations. The methodology is tested on the AIRFRANS database, comprising high-fidelity CFD simulations of aerofoil shapes under various flow conditions. Subsequently, the framework is extended and validated using the Blendednet Aircraft dataset, demonstrating application to complex 3D configurations and establishing generalisation of the proposed method. Results demonstrate significant reduction in computational time as expected but importantly maintaining accuracy suitable for preliminary design applications. The geometric parameter-based approach exhibits strong predictive performance across unseen configurations, suggesting viability as a surrogate modeling technique for aerodynamic analysis in early-stage design optimisation workflows as already demonstrated in work by Du et al. This work contributes to the growing field of physics informed machine learning in aerospace applications, providing aerodynamic analysis tools that preserve engineering insight through interpretable geometric parameterisation