Adjoint-based Optimization of a Hypersonic Air Intake with RANS Credibility Constraint
Please login to view abstract download link
We present an adjoint-based design strategy robust to the structural uncertainty inherent in RANS turbulence closures. The proposed approach leverages the Eigenspace Pertur- bation Method (EPM) [1] implemented in SU2 to obtain uncertainty estimates of the targeted performance. The magnitude of these estimates provides a credibility indicator concerning the numerical prediction. The adjoint-based optimization problem is then for- mulated under the constraint that this credibility indicator is below a certain threshold, confining the optimum search to design space regions in which RANS CFD predictions are deemed credible. The reference test case is an air intake in the supersonic regime [2]. Results highlight that the enforcing of the credibility constraint drives the optimization process toward a substantially different solution, see Fig. 1.
