Adjoint-based Sensitivity Analysis for External Aerodynamics

Abstract

In gradient-based optimisation, the adjoint method predominates the traditional methods in terms of computational cost. The adjoint method makes the gradient calculation independent of the number of design variables, which makes it an attractive choice when working with a high number of design variables. The DrivAer model with three con gurations is used in both RANS-based and DDES-based adjoint simulations. The convergence of the adjoint eld is assessed by monitoring di erent probes with respect to the sensitivity parameter. The comparison of the sensitivity map between RANS-based and DDES-based adjoint is investigated, then the accuracy of adjoint solver prediction is assessed within its corresponding state simulations. In the current thesis, a brief procedure description of both sensitivity-based morphing and direct morphing is presented, and their merits and drawbacks are demonstrated by a series of evaluation cases. Furthermore, the DDES-base adjoint approach is applied on a realistic car model, Volvo XC90.