For many companies there is a growing desire to improve the accuracy of their CFD simulations, in order to improve correlation to experiments and to allow more of their design to be undertaken in CFD.
This seminar focussed on these high-fidelity methods and in particular the role of grid design and resolution, HPC and the formulation of the approaches themselves. Through several presentations from experts representing some of the key engineering sectors (automotive, aerospace, oil & gas, nuclear), practical examples of these methods were presented, offering attendees a chance to understand and question if these methods could be applied to their simulations.
The major CFD code vendors were also represented at the seminar, allowing attendees to see the full range of options open to their company or institution.
This desire for greater accuracy has led many to look beyond traditional Reynolds Averaged Navier-Stokes (RANS) approaches for turbulence towards Large Eddy Simulation (LES). Whilst wall-resolved LES models can in general provide a much better alternative to RANS models for unsteady flows, they do so at a much higher cost, so much higher that for high-Reynolds numbers flow these costs are too great for general purpose calculations.
Hybrid RANS-LES methods offer an attractive alternative to wall-resolved LES, where RANS methods are applied in regions of the flow which are easy to predict (attached, steady flow) and LES methods in the more challenging separated regions. These methods, if correctly applied, can provide close to LES accuracy for a greatly reduced computational cost. However, the accuracy of such approaches is heavily associated with the mesh quality & resolution, numerical schemes and the availability of HPC resources. In addition, since the first use of DES (detached eddy simulation) nearly 20 years ago, numerous alternative approaches with an array of acronyms (DES, DDES, IDDES, ZDES, SAS, WMLES, XLES) have become available. For CFD users wishing to use high-fidelity methods, the initial challenge is understanding which approach is best suited to their application.
|Date||16th November 2016|
|Organisation||University of Oxford|