Comparing Multi-Element Airfoil Flow Solutions Using Multiple Solvers with Output-Based Adapted Meshes

Abstract

In this work, Reynolds-averaged Navier–Stokes (RANS) solutions for a multi-element airfoil computed with Massachusetts Institute of Technology Solution Adaptive Numerical Simulator (SANS), United States Department of Defense High Performance Computing Modernization Program CREATE^TM-AV Kestrel component Conservative Field Finite Elements (COFFE), and SU2 using output-based adapted meshes are compared with RANS solutions computed using manually generated meshes adhering to “best practices.” The adapted meshes are generated using the process implemented in SANS, which seeks to find a mesh that minimizes the error estimate of a given output functional. The manually generated meshes are the result of a systematic study of the influence of a range of meshing parameters on solution accuracy. Using the adapted meshes generally leads to a reduction of more than one order of magnitude in lift and drag error relative to the manually generated meshes of a comparable node count for all of the computational fluid dynamics solvers. Alternatively, the node counts of the adapted meshes are more than an order of magnitude smaller than the manually generated meshes for a given lift and drag error level. The improved accuracy in lift and drag computed with the adapted meshes can partially be attributed increased resolution around the stagnation points and trailing edges of the airfoil elements relative to the manually generated meshes. Notably, the adapted meshes provide accurate results despite having $y^{+}$ values generally significantly larger than one.