Non‐Linear Reduced Order Modelling of Transonic Potential Flows for Fast Aerodynamic Analysis

ABSTRACT This work presents a physics‐based reduced order modelling (ROM) framework for the efficient simulation of steady transonic potential flows around aerodynamic configurations. The approach leverages proper orthogonal decomposition and a least‐squares Petrov‐Galerkin (LSPG) projection to construct intrusive ROMs for the full potential equation. To improve accuracy in regions affected by strong gradients and shock waves, a spatially weighted LSPG formulation is introduced, yielding enhanced robustness compared to unweighted projection. The ROM training relies on a non‐linear ‐transformed Halton sampling of the parameter space, which concentrates samples in shock‐prone regimes and improves generalization without increasing offline cost. The methodology is implemented within the open‐source Kratos Multiphysics framework and validated on two benchmark configurations: the 2D NACA 0012 airfoil and the 3D ONERA M6 wing. The resulting ROMs achieve accurate reconstructions of aerodynamic fields and coefficients, with relative errors on the order of , while reducing the dimensionality of the full order models by approximately three orders of magnitude. Although the corresponding speed‐ups ( in 2D and in 3D) remain modest for the present linear subspace setting, the results highlight the potential of physics‐based intrusive ROMs as reliable surrogates for transonic flows. In the shock‐dominated regimes examined, the proposed intrusive ROM provides more accurate and physically consistent solutions than standard data‐driven surrogates. The framework provides a solid baseline for future extensions incorporating hyper‐reduction and non‐linear ROM strategies.