As a current STAR user I like to ask a question on conjugate heat transfer.

I work on CHT problems in complex geometries. Using the standard k-e model I learned that the results are more or less a good guess but not a real trustworthy result. Sometimes I try a low Reynolds k-e model but I do not like it due to its bad convergence. y+ mainly is in the correct range (standard k-e, low Re k-e)

Recently I got a copy of CFX's validation report concerning Heat Transfer Predictions using advanced two equation turbulence models. The report covers results for several test cases comparing different turbulence models.

It seems that CFX's k-e model with scalable wall functions is a pragmatic step forward leading to better results compared to the standard k-e model.

Their SST model with adaptive wall treatment seems to be even better.

Searching this site I found out that Fluent has a SST model which may lead to questionable results (due to its implementation ?). Star has no SST model.

Does anyone can comment on that ?

We have tested both Fluent's and CFX's SST models on a 2D turbine heat transfer test-case and we obtained very similar results with both codes. The main difference we saw was that Fluent's SST implementation had a slightly bigger overshot in the reattachment point (the blade had a partially separated pressure side). My feeling is that the SST implementation in CFX is a bit more robust and mature (not surpricing since Menter works for CFX and they have had this model for a long time). The heat-transfer results with the SST model were good compared to other low-Re two equation models, but not sensational. Note that the test we made were with a low-Re grid with y+ well below 1.

About the problems you read about with Fluent's SST model - you have probably seen my comments about strange results in 3D. Note that this was with an early Fluent 6 version and I haven't investigated it any further, so don't draw too many conclusions based on that - I was just asking if someone else had had similar problems and noone responded so it might just be a misstake I made, I dunno.

Do these low-Re models employ wall functions?

I get confused as some people refer to low-Re in terms of near wall low Re effects, others in terms of bulk flow low Re nos.

No they don't. SST is basically a combination of k-Omega at the walls (and its viscinity) and k-Epsilon in the far field. As Jonas pointed out it was developped by Florian Menter who is heading turbulence development and reserach at CFX, and one of its goal is to alleviate the heavy cost of the low-k-e model (y+ ~ 0.1) in the aero and mechnical industries for example. In essences it combines the advances of k-Omega at the walls whilst eliminating its flaws away from the wall by switching to a standard k-epsilon and ensures a conservation of the stresses at the junction if I am correct. You should be able to get a paper on Menter's model on the CFX site.

Hi

Just to let you know that Star will have SST model in v 3.2 which is due in a couple of mouths.

Regards

Staruser

will it have SST with automatic wall treatment (automatic local switching between wall function and low Reynolds formulation dependent on the local y+) ?

According to presentations at the last UGM Star 3.2 will feature 'automatic wall treatment'. The presentation of updates for 3.2 will probably be available on adapco-online soon.

Funny. I guess they would have to since CFX has had this for about a year and a half now.

Neale

I might sound strange, but I am using star-cd version 3.15, and sst model is already available.. check the supplementary notes.