Grid generation is the heart of numerical solution.Based on DNS which came from well known theory it is considered to suit all the flow applications (laminar and turbulent). The problem is how to make grid (mesh) that adopted to be used for most of flow applications including laminar and turbulent flows.
From a theoretical point of view, there is no problem in immagine a grid that resolves all practical flow problems in DNS, just consider that the mesh size must be as small as required to get the cell Reynolds number = 1 everywhere.
As a consequence, the practical limits are in the computational power that is required for the numerical computation (grid generation + simulation), a power that increases as the Re number increases. This is the only issue that limits the present DNS capability.
Tapan is righ, the mesh must solve every gradient and that can be ensured only by a dynamical check of the local cell Reynolds number, using the proper local velocity. The standard check on the y+ values is focused only for ensuring wall resolved conditions.
Dear Tapan K. Sengupta
in my opinion, assessing that a real DNS is performed means that a check in the whole computational mesh must ensure that in each numerical cell the convective and diffusive characteristic times have a ratio of O(1). The cell Reynolds number is a way to check this constraint, provided that the local characteristic velocity is properly based on the local velocity gradient. That is quite similar to the adoption of the sensors used in presence of shock waves (high velocity gradients) simulation.
If this proper check is done during the simulation one can have fidelity in the results and a second order accuracy in time and space produces results largely according the spectral solution.
- Badges
- Science topic
- Similar topics
- Correction