How do you choose the best numerical scheme for DNS of turbulent flows?

Choosing the best spatial discretization scheme for Direct Numerical Simulation (DNS) of turbulent flows is critical due to the need to accurately resolve all scales of turbulence. High-fidelity schemes like spectral methods, high-order finite difference, or finite volume methods are commonly employed due to their superior accuracy and convergence properties. Spectral methods, in particular, are favored in DNS for their ability to resolve detailed wave numbers with minimal numerical dissipation and dispersion errors. The choice of scheme may depend on the specific geometry and flow configuration.

Choosing the Best Numerical Scheme for DNS of Turbulent Flows: 1. Accuracy: Select high-order schemes for precise resolution of turbulence details. 2. Stability: Ensure the scheme maintains numerical stability across different scales. 3. Computational Efficiency: Balance accuracy with available computational resources. 4. Flow Characteristics: Tailor the scheme to the specific features of the turbulent flow being studied.

2. Temporal Discretization Temporal discretization in DNS must be handled with care to ensure stability and accuracy in capturing the dynamics of turbulent flows. Typically, explicit time-stepping methods are used because of their simplicity and accuracy, but they require small time steps to satisfy stability criteria, especially in high-speed flows where the Courant-Friedrichs-Lewy (CFL) condition becomes restrictive. Implicit methods, although more computationally intensive per timestep, can allow larger time steps and may be necessary for stiff problems. The choice between explicit and implicit methods often depends on the specific requirements of the simulation, including the expected flow speeds and computational resources available.

3. Boundary Conditions The selection of appropriate boundary conditions is essential for DNS of turbulent flows as they significantly influence the accuracy and stability of the simulation. For turbulent flow simulations, boundary conditions must be defined precisely to reflect the physical situation being modeled without introducing artificial reflections or disturbances. Inlet conditions should ideally reproduce the turbulence characteristics of a fully developed flow, while outlet conditions must allow disturbances to leave the computational domain without reflecting back. Special care must also be taken with solid boundaries, where no-slip conditions are typically enforced.

Conducting a grid sensitivity analysis helps determine the optimal grid resolution. In a turbulent jet simulation, we performed grid sensitivity studies by systematically refining the grid and analysing the changes in key turbulence metrics such as Reynolds stresses and turbulent kinetic energy. This approach allowed us to identify a grid density that balanced accuracy and computational cost. Non-uniform grid spacing can enhance resolution in critical regions while conserving computational resources. In a channel flow study, we used a non-uniform grid with finer spacing near the walls where velocity gradients and turbulence intensity were highest, and coarser spacing in the core flow region.

4. Grid Generation Grid generation plays a pivotal role in DNS because the quality and resolution of the computational mesh directly impact the accuracy of the simulation. For DNS, where all scales of turbulence need to be resolved, a highly refined and possibly anisotropic grid is necessary, particularly in regions of high gradient (near walls, in boundary layers, or in the wake of objects). The grid must be fine enough to capture the smallest eddies and coherent structures of the flow, requiring careful balancing between computational expense and numerical accuracy. Adaptive mesh refinement (AMR) can be a useful approach in DNS to enhance resolution dynamically in regions of interest, thereby optimizing computational resources.

Error analysis is an integral part of choosing the best numerical scheme for DNS of turbulent flows. It involves assessing the discretization errors—both spatial and temporal—to ensure that the numerical solution converges to the true physical solution. Since DNS requires resolving all relevant scales of turbulence, the grid independence test is crucial, where results are compared across progressively refined grids to ensure that further refinement does not significantly alter the solution. Additionally, the verification of the numerical scheme through analytical solutions or experimental data for simpler problems can provide confidence in the numerical methods and their implementations before applying them to more complex turbulent flows.