
[Sponsors] 
July 6, 2001, 20:49 
help

#1 
Guest
Posts: n/a

Hello everyone, I met the hint " reversed flow in 76 faces on pressureoutlet 3"in my calculation.
Does anybody know what it means and why there is so many faces (76)? (in my model there isn't so many faces) Why did it happen? thanks 

July 7, 2001, 05:48 
Re: help

#2 
Guest
Posts: n/a

(1). Obviously, you have reverse flow at the exit, and that is not a good sign. So, you are getting the warning. (2). Well, how many faces do you have at the exit?(faces of the cells)


July 7, 2001, 10:57 
Re: help

#3 
Guest
Posts: n/a

I have eight pipe faces at the exit.I set all of these faces' B.C as one pressureoutlet3. Am I right?
Thanks a lot lucy 

July 7, 2001, 13:37 
Re: help

#4 
Guest
Posts: n/a

(1). Why not display your mesh and take a look at the cells at the exit. Do you mean that there are only eight cell faces at the exit?


July 7, 2001, 14:00 
Re: help

#5 
Guest
Posts: n/a

I have found the reverse flow velocity vector and the cell faces at the exit,but I still wonder why it happened? Could you help me? thanks


July 7, 2001, 15:38 
Re: help

#6 
Guest
Posts: n/a

(1). I think, it is between you and the vendor's code. (2). I mean, it happens only when you are trying to solve your problem with the vendor's code. (3). So, it is a good idea to have a good discussion with the vendor of the code. right? At least, it is related to your problem definition, I mean the boundary conditions.


July 7, 2001, 16:17 
Re: help

#7 
Guest
Posts: n/a

Thank you very much. best regards lucy


July 9, 2001, 02:15 
Re: help

#8 
Guest
Posts: n/a

I don't think it is necessarily the vendor's code. Your geometry and mesh density determines if you will get reversed flow at the exit. It is also not necessarily incorrect if you get reversed flow with a pressure outlet boundary condition. It one's again depends upon you geometry and mesh density unless you have made an error somewhere else.


July 9, 2001, 03:42 
Re: help

#9 
Guest
Posts: n/a

(1). Based on my experience with Fluent and CFXTASCflow codes, it is possible to get reverse flow at the exit and even at the inlet "during the iteration", which is a consequence of the algorithm used and possibly partly due to the initial guess of flow field. (2). If one gets reverse flow at the exit when the solution is converged, then it is a good idea to change or modify the exit condition or location.


July 10, 2001, 16:50 
Re: help

#10 
Guest
Posts: n/a

How about changing your pressure outlet to an outflow???? In this way you eliminate one problem which probably caused by wrong initial guess at pressure outlet BC... In this way, you can determine if your grid is the main culprit of the problem...


July 11, 2001, 02:14 
Re: help

#11 
Guest
Posts: n/a

(1). Outflow would be easier, but you need to make the exit like the outflow, that is to extend the exit location further downstream. (like a longer duct or pipe attached) (2). I would say that when there is a reverse flow at exit, it is a good idea to move the exit location further downstream. (this applies to the inlet too, move the inlet location further upstream to avoid reverse flow at the inlet. regardless of the code you are using.)


July 11, 2001, 02:43 
Re: help

#12 
Guest
Posts: n/a

But what if you are modelling a real geometry (it is built and experimentally tested) and you get reversed flow in your CFD model. You can't move the boundariesWhat do you do now?


July 11, 2001, 03:27 
Re: help

#13 
Guest
Posts: n/a

I agree with John
reverse flow at BCs is never a good thing : You can't be sure of what goes in and out of the domain in this case and this generally lead to a lack of accuracy. I extend the domain in most case, even if this extension doesn't exist in the true geometry. Of course, in use the real outlet or inlet position for post processing like pressure drop, ... Best regards Alain 

July 11, 2001, 05:06 
Re: help

#14 
Guest
Posts: n/a

(1)."I extend the domain in most case, even if this extension doesn't exist in the true geometry." This is an excellant statement about the nature of the CFD. (2). In the definition of CFD, the numerical analysis and mathematical modeling in fluid mechanics, it requires the use of analysis and modeling. (3). Analysis and modeling simply say that in CFD the geometry model, the physical model , and the turbulence model do not have to be "exactly the same as the real one". (4). A 3D problem can be analyzed as a modeled 1D problem. (5). The experience of the researcher in that particular problem will be the key to the most appropriate way to model it. (5). So, in the analysis and modeling, the goemetry can be changed (different from the real geometry) as long as the result is useful for the purpose of analysis.


July 11, 2001, 05:22 
Re: help

#15 
Guest
Posts: n/a

I agree, but hopefully we are moving in the direction where we can simply model the geometry and conditions 'as it is'.


July 11, 2001, 13:55 
Re: help

#16 
Guest
Posts: n/a

Hi...
I totally agree with all of you... I do always extend my geometry to such a point that back flow would not affect the analysis of my real objective... To Livo... it would be too idealistic at the moment to model the whole set of problem with a given geometry in one shot.... normally it takes some dissection and assumption to achieve a more realistic model... 

July 27, 2001, 17:49 
Re: help

#17 
Guest
Posts: n/a

The real problem here is that the assumption of constant pressure at the boundary location is incorrect. Obviously the pressure is not constant at the boundary location or you would not be getting recirculation (assuming the solution is converged, grid independence, correct models, ...).
Modelling the geometry/problem "as it is" requires detailed knowledge of the flow conditions at the boundaries (assuming there is a reasonable way to input these conditions in the code). If you don't have this knowledge, sometimes you have to make comprises, 1) by artificially extending the domain or, 2) by placing the boundaries at a location where you do have detailed knowledge of the flow. Since 2) is often unmanageable (in the limit you are modeling an infinite domain with infinite detail!) the first choice is often the only choice. Just my 2 cents 

February 28, 2011, 01:45 
residence time disrbution

#18 
New Member
eswar
Join Date: Feb 2011
Location: india
Posts: 2
Rep Power: 0 
hi sir/mam my self eswar, doing M.Tech chemical.
my question is to find the residence distribution of tracer sample using pulse input in gambit. does any one could help me please. my project is studies on static mixing. my tracer sample is (CONGO RED DYE). my experiment procedure is 1) inlet fluid water ( continuous flow) 2) inlet tracer sample (pulse injection) 3) i want to find the concentration of tracer sample at regular time intervals. please help me thanks in advance 

Thread Tools  
Display Modes  

