Investigation of adaptive time-step strategies for high-order accurate incompressible simulations

Adaptive time-step algorithms can improve considerably the effectiveness of unsteady flow computations. Several adaptive time-step strategies are available in the literature but in all cases conservative time-step choices (small time steps) lead to a large number of time integration steps, while aggressive time-step choices (large time steps) lead to a large number of rejected time integration steps, and in both cases the efficiency and/or robustness of the adaptive strategy may be far from optimal. An appropriate adaptive strategy should instead guarantee both robustness (small-number of rejected time integration steps) and efficiency (small-number of time-integration steps for a given accuracy). In this work several adaptive time-step strategies have been adopted for the numerical solution of the unsteady incompressible Navier-Stokes and Reynolds-Averaged Navier-Stokes equations based on a high-order accurate discontinuous Galerkin space discretization. Three different classes of time integration methods have been considered, the linearly implicit Rosenbrock-type Runge-Kutta schemes [2], linearly implicit Rosenbrock-type two-step peer schemes [3] and ESDIRK schemes [2]. In oder to assess the adaptive time-step methods for both autonomous and non-autonomous (time-dependent boundary conditions) DAE systems of increasing stiffness, we will present the results obtained in the comuptation of unsteady laminar and turbulent flows around a circular cylinder at increasing Reynolds numbers ranging from Re = 100 to Re = 3900.