Optimal control of iterative solvers in an anisotropic adaptive finite element framework for turbulent flows.
“The iterative solver is the core of a numerical simulation and generally accounts for 60-80% of the total run time. Therefore it is essential to optimize his control and convergence to increase the performances of the simulation. The focus of this PhD work is to use the information from the discretization error estimator, used for the mesh adaptation procedure, to drive the convergence criterion of the iterative solver. This leads to faster simulations without any loss in the solution accuracy.”