Please write here suggestion on how to improve the code as well as you wishlist of things you would like to see in the future. Also take a minute to write what are you interested in develop and contribute to the code.
Suggestions
"Tune" the numerical options of Saturne: we all have our best practise guidelines for RANS, LES simulations. Changing some numerical options sometimes allow to reduce the CPU time by a factor of up to 4 I noticed, without any impact on the quality of the results. The development team at EDF aknowledges that the standard precision of the algorithms could be reduced in Saturne, and in both StarCD and Fluent the "epsilons" used are much smaller. It would be interesting to test the effect of these numerical parameters on CPU time, quality of solution on a couple of simple test cases both in RANS and LES (fully dvpd wall turbulence and a flow with detachment).
Outlet BC in LES: the constant reference pressure prescribed and the Neumann on the velocity is not correct, it generates odd-even oscillations on the velocities upstream of the outlet. All people doing LES at the University use convective BC.
Lance Script: In many of the cluster tested, the compilers are not available on the nodes (due to licensing problems most of the time) therfore it is assumed that the applications are compiled on the head node and the executable sent to the queue. It would be useful to have the lance script divided in two so that the commercial compilers can be used.
Gradient reconstruction: i have tried the least squares with extended neighbors is not working in most of the cases. More over it is working, is slower then the iterative grad reconstruction.
Unstructured mesh: is it possible to use the unstructured mesh for LES? Which are the limits? Someone has tried is in some applications that is not channel flow.
Contributions
Inlet BC for LES: A divergence-free version of the SEM has been implemented. It needs to be compared to previous versions of the SEM. Additionaly, problems of pressure fluctuations observed with the previous version of the SEM should be fixed.
Hybrid RANS-LES method with a single grid and two velocities field developed in version 1.2. Needs to be adapted for version 1.3.1.
Ideas for future development
LES SGS modelling: Most LES models are relatively simple to implement. There are not many LES models available in Saturne at the moment and the WALE model for instance could be a good candidate for implementation and testing on a few flows. It has nice properties that the Smagorinsky model does not have (zero CS in laminar flows,etc.) and it simpler to code and more general than the Smagorinsky Dynamic model.
Analytical and Numerical Wall Functions: These two types of wall functions have proved to be very effective in different applications, inlcuding Buoyant flows, which is one of the topics that we need to use Saturne for, as far as KNOO is concerned. Implementation of these WFs in Saturne could be very beneficial for industires as well as academia. A few years ago, F. Wolff implemented the Analyitical WF in Saturne but probably due to lack of validations, it was removed in newer versions. We need to see if first of all, EDF is happy with us to go ahead with implementing these WFs and secondly, we need to know if there is currently any work going on at EDF to implement them or not.
Convective schemes: The availability of only CD and Up-wind is a bit restrictive. I'm looking also for other convective schemes that can be used also for LES in unstructured mesh (could be a solution to avoid the spurious oscillation found in some unstructured mesh applications).