PALMO: An OVERFLOW Machine Learning Airfoil Performance Database

The characterization of airfoil performance across a range of Mach numbers, Reynolds number, and angles of attack remains a key aspect for a vast array of mid-fidelity aerospace analysis and design methods. The computation of aerodynamic loads on rotors and wings often relies on previously computed airfoil performance tables. Although the mid-fidelity tools rely on these tables to reduce their computational cost, the process of creating accurate lookup tables is often a computationally intensive, and time-consuming task. Due to the still high computational cost of high-order CFD solvers, airfoil performance tables at proximal but mismatched Reynolds and Mach numbers are often used in conceptual design. These tables are sometimes created either with existing experimental data, which are limited by the test Mach and Reynolds number, or using a lower-fidelity approach such as XFOIL or MSES, Refs. [1-2]. Past studies by Patt and Youngren, Ref. [3], and Cornelius and Schmitz, Ref. [4], document both the need for higher refinement implementations of C81 tables and the improvements obtained by using them, highlighted when analyzing rotors with multiple airfoils, large changes in radial chord distribution, and variable-speed (varying RPM) operation, Ref. [4]