Aerodynamic shape optimization via discrete adjoint formulation using Euler equations on unstructured grids

A methodology for aerodynamic shape optimization on two-dimensional unstructured grids using Euler equations is presented. The sensitivity derivatives are obtained using the discrete adjoint formulation. The Euler equations are solved using a fully implicit, upwind, cell-vertex, median-dual finite volume scheme. Roe\u27s upwind flux-difference-splitting scheme is used to determine the inviscid fluxes. To enable discontinuities to be captured without oscillations, limiters are used at the reconstruction stage;The derivation of the accurate discretization of the flux Jacobians due to the conserved variables and the entire mesh required for the costate equation is developed and its efficient accumulation algorithm on an edge-based loop is implemented and documented. Exact linearization of Roe\u27s approximate Riemann solver is incorporated into the aerodynamic analysis as well as the sensitivity analysis. Higher-order discretization is achieved by including all distance-one and -two terms due to the reconstruction and the limiter, although the limiter is not linearized. Two-dimensional body conforming grid movement strategy and grid sensitivity are obtained by considering the grid to be a system of interconnected springs. Arbitrary airfoil geometries are obtained using an algorithm for generalized von Mises airfoils with finite trailing edges. An incremental iterative formulation is used to solve the large sparse linear systems of equations obtained from the sensitivity analysis. The discrete linear systems obtained from the equations governing the flow and those from the sensitivity analysis are solved iteratively using the preconditioned GMRES (Generalized Minimum Residual) algorithm. For the optimization process, a constrained nonlinear programming package which uses a sequential quadratic programming algorithm is used;This study presents the process of analytically obtaining the exact discrete sensitivity derivatives and computationally cost-effective algorithms to efficiently use them in a design environment. Storage issues are circumvented by developing algorithms which perform matrix-vector operations during the construction itself. To validate the unstructured shape optimization procedure, an arbitrary symmetrical airfoil is optimized in an inviscid transonic flow regime