A linear aerodynamics-based preconditioner for high-fidelity aeroelastic analysis and sensitivity analysis

International audienceHigh-fidelity aeroelastic analysis and sensitivity analysis are pivotal components in gradient-based aerostructural optimization. These analyses often give rise to linear sub-problems that require iterative solvers to maintain realistic CPU/Memory consumption. However, the convergence behaviour of iterative solvers is highly influenced by the applied preconditioner. In this paper a linear aerodynamics-based preconditioner is presented with the capability of reducing the solution time of high-fidelity aeroelastic analysis and sensitivity analysis. It is constructed in Schur complement form by a panel method. Hence, it is referred to as a low-fidelity Schur complement. The proposed solution method is demonstrated on the ONERA M6 wing. The structural model of a classical wing box layout has 1812 degrees of freedom and is solved by Nastran. The high-fidelity aerodynamic model is based on the compressible Euler equations and consists of 1.11 million cells. The solver used for the Euler flow analysis and sensitivity analysis is the multi-block structured flow solver elsA. The efficiency of the preconditioner is studied for a subsonic case of Mach 0.60 and a transonic case of Mach 0.85. The application of the low-fidelity Schur complement as a preconditioner shows up to three times faster CPU times for the Mach 0.85 sensitivity analysis case