
[Sponsors] 
February 10, 2000, 09:21 
Relaminarisation and wallfunctions in CFD

#1 
Guest
Posts: n/a

Hello!
I hope someone can help me to understand this whole process of relaminarisation due a high flow aceleration (in an inlet of a gap, for example). I have heard that if the relaminarisation process is to be represented then one needs to use a good lowRe model with characteristics in accelerating flow. (1) Does it mean here a lowRe Model as a wallfunction? (2) How could I modell it in a CFDCode? (3) Could anyone give me some references about this topic? Thanks in advance 

February 10, 2000, 12:06 
Re: Relaminarisation and wallfunctions in CFD

#2 
Guest
Posts: n/a

(1). You should try to get this one of the most important paper by Jones, W.P., and Launder, B.E. (1972),"The Prediction of Laminazation with a TwoEquation Model of Turbulence", International Journal of Heat and Mass Transfer, Vol 15, pp301314. (2). Anyone doing turbulence modeling must read this paper.


February 11, 2000, 04:14 
Re: Relaminarisation and wallfunctions in CFD

#3 
Guest
Posts: n/a

John,
could you give a physical explanation what mechanisms are responsible for turning a turbulent flow into laminar when accelerated ? Mark 

February 11, 2000, 04:31 
Re: Relaminarisation and wallfunctions in CFD

#4 
Guest
Posts: n/a

Thank you, John Chien for telling me about the paper. I am just starting at this area ..., and I have already gotten the impression that Launder is really the pope in turbulence! I have read a lot, but I think I have to read lots of more papers from him!!
found good the idea from Mark that you give us your opinion about this physical phenomena! 

February 11, 2000, 07:04 
Re: Relaminarisation and wallfunctions in CFD

#5 
Guest
Posts: n/a

Hi,
As far as i know, the problem of relaminarisation in ke (or other RANS turbulence modelling) is very difficult and the short answer is that you can't do it (properly)! Basically, the ke model has no mechanism for the offset of turbulence from laminar flow  look at the model: k=0 and epsilon=0 is a valdil solution, which means that if you put no turbulence in, you will never get any out whatever you do with the velocity field. Relaminarisation problem is the opposite from the above, but also not doable to my knowledge. I think John is referring to the JonesLaunder lowRe ke model paper (better known and allegedly paraphrased by Launder and Sharma)  if that is the case, the modell cannot do relaminarisation (I tried!). If there's someone out there whoknows lots about relaminarisation modelling I would be very interested to hear the developments (the JonesLaunder paper is 1972!). Hrv 

February 11, 2000, 10:49 
Re: Relaminarisation and wallfunctions in CFD

#6 
Guest
Posts: n/a

(1). No, I have not looked into this area. (2). I think, when you have traffic jam on highway, you are going to get a lot of noise from the drivers. On the other hand, if the traffic is moving at 65 mph, everybody will be very happy. And at 75 mph, you are really free, like driving in no man's land. (3). I think, if you accelerate the flow, U(i+1) will be much larger than U(i1), so, in order to satisfy the continuity you have to suck in V(j+1), and V(j1). In other word, V(j+1) and V(j1) are no longer random in the accelerating flow. Once you suck in V(j+1) and V(j1), these two components are then squeezed and merged with U(j+1). So, the distinct random motion will disappear in the process. (4). So, my free answer is: the acceleration will suck in the fluid particle from the neighboring cells to satisfy the continuity equations, then this added momentum is then mixed and merged with the main stream in the accelerated state to satisfy the momentum and energy equations. The continuous action of this accelerated field keeps sending the fluid particle through this cleaning process to realign itself in the main stream direction.


February 11, 2000, 10:59 
Re: Relaminarisation and wallfunctions in CFD

#7 
Guest
Posts: n/a

(1). Turbulence modeling is like Alice Wonderland. And B.E. Launder is the person who opened many of these doors to it. (2). What you really need is the key to the door, once you have stepped inside, it is up to you to find the answer. (3). So, what I am saying is: the answer is not written on the face of the key.


February 14, 2000, 13:23 
Re: Relaminarisation and wallfunctions in CFD

#8 
Guest
Posts: n/a

Continuing along ..John's explanation..
This type of relaminarisation is characterized by the parameter K ..= (kinematic viscosity)* (mean velocity gradient)/(mean Energy)...and occurs for vaues greater than 10^(6) or so. REF: http://www.nd.edu/~cbourass/APS99_pres_files/frame.htm Obviously there is something more to John's 'continuity  based' explanation which limits the relaminarisation process. Technicalities apart ..I was just wondering what ..the next 'improved' simplistic explanation could be.... IS it... i.) Boundary layer separation ? ii.) Strongvorticity generation at the wall ? ( and thus 'stronger bursts' even if low in frequency ) Prabhu 

February 14, 2000, 14:42 
Re: Relaminarisation and wallfunctions in CFD

#9 
Guest
Posts: n/a

"Relaminarization" is definitely not an appropriate word to define the phenomenon. A turbulent flow does not turn into a laminar one, it stays TURBULENT all way. However, the intensity of the turbulence decreases rapidly, especially in the core flow. The main reason why the "standard" turbulence models are inadequate for the accelerating flows is the fact that the turbulence becomes largely anistropic. The turbulent momentum transfer is dominant in a streamwise direction (you may think of more degress of freedom in a streamwise direction rather than crossdirection due to acceleration as the continuity eqn must be satisfied). Now recall that most turbulence models are somehow based upon isotropy of the turbulence, at least on a microscale level, i.e. assumed is there always is "5/3" low in a power spectrum. Contrary, experimental results show a lack of the inertial range when the flow acceleration is high.


February 15, 2000, 04:50 
Re: Relaminarisation and wallfunctions in CFD

#10 
Guest
Posts: n/a

Hi,
I don't think we're talking about the same thing: for me relaminarisation happens when you take a turbulent boundary layer and sqeeze it against the wall so hard that the turbulence is completely damped (to zero!). If the boundary layer carries on developing, it needs to go through transition to become turbulent again. Hrv 

February 15, 2000, 10:05 
Re: Relaminarisation and wallfunctions in CFD

#11 
Guest
Posts: n/a

Actually, if the turbulence were damped all over the BL, one would not need any "wall functions", low Re turbulence model and etc as the flow would be laminar. Unfortunately, this does not happen even when strong "relaminarization" occurs. The turbulence intensity (turbulent kinetic energy) may be deceivingly low in the core flow so that it would be hard to detect even via LDA, or PIV provided that the windtunnel is very good. The situation is quite different at the proximity of the wall. Flow stays turbulent there even though very strong favourable pressure gradient is applied. In real world there is no chance to completely damp the turbulence in the wall region. Theoretically, it's possible though. Now think of how people develop turbulence models and validate them... Sure, they inclined somehow to use experimental data, generalize them and propose something that may be good to describe the phenomena. The same applies to "relaminarization". For that reason, we see the lowRe models and any other "nonstandard" models to pull "theoretical" results on the experimental data. Coming back to "relaminarized" turbulent flow, I'd like to reiterate that the flow is NOT laminar. Otherwise, we'd be successful using steady NS equations. I'm not aware if LES does a good job for these flows, since the tubulence structure is strongly anisotropic, and I'd be glad to hear if anybody could share any experience.


February 15, 2000, 10:56 
Re: Relaminarisation and wallfunctions in CFD

#12 
Guest
Posts: n/a

This is a very interesting topic. You are certainly right that relaminarization, in the sence that all turbulent fluctuations die out, never occurs. However, the boundary layer can undergo some kind of "reverse transition" and get a "laminar type of profile" again. This typically happens in regions of very strong accelaration. Don't you agree? I've worked a lot with supersonic turbines, where the turbulent boundary layers sometimes are accelarated from Mach 1 to Mach 2 in a few millimeters distance  you clearly see how the shape of the boundary layers become laminar in this case.
When you have a very high freestream turbulence intensity you have the same problem with nomenclature  the "laminar" parts of the boundary lauyers are certainly not laminar in the sence that they do not have any turbulent fluctuations in them. Still you can see a clear transition point downstream were the boundary layer becomes turblent. To avoid confusion I always call the "laminar region" after a strong accelaration a "relaminarizing region" and I call a laminar but freestream turbulence disturbed boundary layer a "laminarizing" boundary layer. Perhaps someone could add some more insight into what differs between a "truly laminar", a "relaminarizing" and a "laminarizing" boundary layer? Can we think of them as essentially laminary? What do we neglect then? 

February 15, 2000, 12:15 
Re: Relaminarisation and wallfunctions in CFD

#13 
Guest
Posts: n/a

There is a difference between 'turbulence model (model)' and 'turbulence flow (physics)'. I hope we all agree that not a single turbulent model (math model) is good enough so far to capture all physics of turbulece flow, simply because the 'physics model' of turbulent flow is still not well established (we still don't know why turbulene). For this reason, turbulent models may not be able to model relaminarisation, just as they can not model transition. In the model, the relaminarisation has to be treated as 'laminartypeturbulence', because in real laminar flow the turbulent model is not needed at all. How can a model be used to do something in which it is not needed?
However, in the real flow, relaminarsation exists indeed. When flow losses its stability, it turns to turbulence from laminar through transition. This process can be reversed when there is a longlasting favorite pressure gradient (such as flow acceleration) and the flow regains its stability, it then turns to laminar again. I don't think there is a major difference between natureborn laminar and turbuleceborn laminar, although they are different in the process of CFD modelling. 

Thread Tools  
Display Modes  

