Representing turbulence by a small number of quantities, such as intensity and length scale, for example, is appropriate and efficient in many engineering situations. Resolving most of or even all turbulent motion by means of Large-Eddy Simulation (LES) or Direct Numerical Simulation (DNS), respectively, provides much more information but is computationally very demanding. Recent years have witnessed an ever-increasing availability of computer power so that the approach can now be applied by many researchers. Indeed, a minimum number of operations, determined by the grid size and the required time steps, needs to be executed to obtain sound separation of length and timescales between the smallest and the largest resolved ones. During recent years, the required performance threshold is met by more and more computer systems. Also, discretization methods and solution algorithms have improved as a result of decades of scientific activities in this field. As a consequence, meaningful DNS and LES can now be performed for more and more applications. For the same reason, a central issue of LES, subgrid-scale modelling, has become less critical today as the grid scales are further away from the resolved scales than before. Still, these methods present lots of pitfalls, and a cost-effective simulation requires optimal models. Much work has been done on improving discretization schemes, subgrid-scale models and other model contributions such as generation of in flow turbulence. On this basis, the development and application of these methods and models continues to be a very active field of research. More and more data sets from DNS nowadays provide detailed and accurate reference for improved understanding and development of physical models.
|Publisher||Springer International Publishing|
- Heat and Mass Transfer
- Numerical and Computational Physics
- Fluid Dynamics