Aerodynamic optimization using high-order finite-volume CFD simulations

The growth of computer power and storage capacity allowed engineers to tackle engineering design as an optimization problem. For transport aircraft, drag minimization is critical to increase range and reduce operating costs. Lift and geometric constraints are added to the optimization problem to meet payload and rigidity constraints of the aircraft. Higher order methods in CFD simulations have proved to be a valuable tool and are expected to replace current second order CFD methods in the near future; therefore, exploring the use of higher order CFD methods in aerodynamic optimization is of great research interest and is one goal of this thesis. Gradient-based optimization techniques are well known for fast convergence, but they are only local minimizers; therefore their results depend on the starting point in the design space. The gradient-independent optimization techniques can find the global minimum of an objective function but require vast computational effort; therefore, for global optimization with reasonable computational cost, a hybrid optimization strategy is needed. A new least-squares based geometry parametrization is used to describe airfoil shapes and a semi-torsional spring analogy mesh morphing tool updates the grid everywhere when the airfoil geometry changes during shape optimization. For the gradient based optimization scheme, both second and fourth order simulations have been used to compute the objective function; the adjoint approach, well known for its low computational cost, has been used for gradient computation and matches well with finite difference gradient. The gradient based optimizer have been tested for subsonic and transonic inverse design problems and for drag minimization without and with lift constraint to validate the developed optimizer. The optimization scheme used is Sequential Quadratic Programming (SQP) with the BFGS approximation of the Hessian matrix. A mesh refinement study is presented for an aerodynamically constrained drag minimization problem to show how second and fourth order optimal results behave with mesh refinement. A hybrid particle swarm / BFGS scheme has been developed for use as a global optimizer. It has been tested on a drag minimization problem with lift constraint; the hybrid scheme obtained a shock free profiles, while gradient-based optimization could not in general.Applied Science, Faculty ofMechanical Engineering, Department ofGraduat