Question for discussion
 

When I was taking an interview in a university, I was asked the following question which is hard for me to answer, so I would like to receive your opinoin, thanks.

1)In your opinion, what are the major problems/difficulties of current turbulence models? 2)What do you think are the improvements and lines of research which should be pursued in turbulence modelling and turbulence in general?

Re: Question for discussion
 

see my question on line Why turbulence models are not universal?hope it will clear all your doubts.if not feel free to post even personally.

senthil k

Re: Question for discussion
 

Whoooaaa!! A good one. Well, let me try:

Extended application of turbulence models originally developed in the 70-s and 80-s reveals a piece of information crucial for future work: there is no silver bullet. The amount of research work in the Reynolds averaged Navier-Stokes (RANS) has proven (conclusively, for me) that there will never be a RANS-type model which will be universal. However, I am also close to the conclusion that models available today are adequate for most engineering applications and fall into two main categories:

- 2-equation models, applicable to about 90 % of cases - RSTM models, for people who a-priori know that the phenomenon under consideration strongly depends on secondary flows.

Future improvements in RSTM modelling will be mainly associated with tuning and near-wall treatment; I expect no ground-breaking results.

The second area of research in the LES arena, where the foundations have been laid and it can be expected that this will be delivered to the industry in the next 5-10 years by a combination of increased computer power and better modelling. Open modelling questions (of my interest) are in the treatment of inlet and wall boundaries, as well as in the requirements on numerical algorithms posed by LES. Also, there is the issue of delaing with large amounts of data produced by LES runs in an efficient and meaningful way.

In the short term (because of limitations on computer power) we need work on the combined LES-RANS models, which will deliver improved results with only a relatively low increase in computational cost (~ a factor of 10). In the longer run, this work will be made obsolete through the increase in computer power.

From the industrial point of view, there is also a number of questions we need answered, dealing with the numerical issues of LES simulation: order of accuracy, locally refined meshes, as well as viable modelling (expected to be of wider applicability than RANS).

Comments, please!

Hrv