Hi all,

how does the effective viscosity varies along the wall within the laminar flow as it progresses towards transition/turbulent onset? It starts from the minimum value of 0.01mu and then increases. My question is what is the process thats used to calculate the process, or progress of this increase? Any idea?

Sheila

When you are running a simulation with a turbulence model based on an eddy-viscosity approach, the flow is assumed to be turbulent everywhere. And in general, there is no transition model included in the standard turbulence models as found in STAR-CD. Hence, you will not be able (unless you manipulate the source terms of the transport equations for k and epsilon for instance) to predict nor simulate accurately transition from laminar to turbulent with the code.

Hi 4xF,

I am actually modelling transition by manipulating the source terms for k and w. The only problem I am having now is to match the laminar part of the profile. The onset of transition from my model is further away from that predicted by STAR CD, this results in a significant laminar region. From lookin at various data I can tell that even without my modifications, the molecular viscosity increases from 0.01VISM to VISM till transition. I've tried using linear interpolation, quadratic interpolation, power law interpolation to model this increase... some of them look great in coarse mesh, but in refined mesh it does'nt work so well. The initial part of the boundary has very high fluctuations, and the extent of this somewhat seem to depend on the value of VIST in this region. Do you have nay idea what is going wrong? Or what I can try?
At the momentI am trying to prescribe bothe the VIST and the w at this region. The case is still running, dont know if it will work or diverge..

Thanks, Sheila

Are you trying to predict transition with a 2 equations model? Or do you impose the transition location and want to see how the transition zone develops by incorporating "appropriate" source terms in the transport equations for k and omega?

Hi 4xF,

I predict transition onset according to an analytical model, similar to Abu Ghanam Shah model and then transition process is modelled modifying k-w model via changing VIST by multiplying it with a scalar that varies between 0 and 1.

Sheila

But then, I do not undestand your question, since you should look at the turbulent viscosity which should be strictly zero in the laminar zone and start increasing according to some heuristic formula in the transitional zone to match the predicted value from the turbulence model after transition has happened.

Multiplying the turbulent viscosity with a factor which varies between 0 and 1 is a good idea, though. Have you ever read the very interesting articles from: a) Saville b) Erik Dick who use an additional scalar transport equation (for the so called intermittency factor) for predicting and modelling transition?

Hi 4xF,

I am indeed using an additional scalar, the so called intermittency factor that varies between 0 and 1. But, the way STAR is written, or atleast the version I am using is that you cannot use VIST=0, there is a minimum set value of something like 0.01mu. It is porbably to reduce any chance of zero problems of divergance.

So, I will reiterate my case together with my question, as I am probably not describing it properly, as often happens when you are working on something for a while and start to think people can read your mind...

I am trying to model transition in low reynolds number cases more accurately than that modelled by traditional method used in CFD. I am solving a transport equation for intermittency factor. And the way I will be affecting the k-w equation is by multiplying this scalar with VIST. The scalar is 0 at laminar flow and increases within tranisiotnal zone to the value of 1. The value of 1 corresponds to fully turbulent case.
I am also defining the onset of transition point, ie, the location from where intermittency has an effect on the k-w values. This locatin I define results in a much longer laminar region than that described by STAR. As STAR will not accpect VIST=0, I have to use some value in the laminar region which will result in STAR workin ok within thie region and matching of data with experiment (it has been done before in University of Kentucky using their own CFD package).
Looking at data generated by unmodified STAR, it seem like they use some kind of variation of VISM witihn the (very short)laminar region. I need to know what is the method that is used to prescribe this variation so I can use that to my solution.
I hope this makes sense and I hope you or someone else will be able to help me with this as I am extrmely desperate.

Thanks Sheila

One more thing, if STAR does'nt define any tranisition onset point then is it a co-incidence that the arbitrary values of VIST within the laminar region arising from the source terms of k-w varies between 0.01VISM to roughly about VISM before transition takes place...? Some how I doubt that very much.
If that is the case, then can you suggest anything that I can try to model the laminar region? Using VIST=VISM results in too thick a boundary layer, resulting in poor values of shape factor, using VIST=0.1VISM results in too thin a boundary layer... Any help would be appreciated. Thanks again, Sheila

I can hardly believe that VISM varies unless ou modify it by user subroutines. VISM being the molecular viscosity (if I remember it well). If the turbulent viscosity cannot be set to zero, this is certainly due to some kind of internal check which avoids that the users start doing stupid things -- which is quite common for the users and quite well done by STAR ;-)

Anyway, the only way to set VIST to zero is to set k to zero, whatever the value of w is, since VIST= rho*k/w. This way, you can ensure a complete laminar flow. This is the first thing you should check. But what to do with the free stream turbulence? You should then use it for your transition model.

Another possibility to get rid off the effects of the eddy viscosity in the momentum equations is to add a source term which is the opposite (i.e. -VIST/rho*div(grad(u_i)) for the respective i-component of the velocity) to have pure laminar flow and set the values of k and w in this zone to their respective values of the free-stream.

I think that should give you some possibilities to control the turbulence for your transition model. On the other hand, if you are doing such modeling, why don't you implement your own conditioned (i.e. with the use of the intermittency factor) turbulence model and use it for calculating the eddy viscosity via the VISTUR subroutine?

Hope this brings you forward...

Thank you 4xF I shall try your suggestions. The reason I did'nt use VISTUR is because when I first started modeling the scalar, I had almost no clue about STAR and subroutines (and I still dont have much). And now it just seems too much trouble as I am very short of time. Each of my initial refined cases with y+~1.5 takes 8-12 hours to run... and I am having problem with MPI as apparently there is a bug in STAR 3.15... waiting for a new license from STAR to get that sorted out... Anyhow, thank you for your help. I am looking forward to trying it out... Sheila