CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   Suction (

John Amaya July 28, 2000 16:36

Hi, everybody.

I'm trying to model a chamber where the flow is driven by the suction of a blower. Most of the commercial codes require to specify an atmospheric pressure for outlets. I specified a velocity profile at the inlet of the chamber but the numerical results didn't match the measurements. Does anybobody know how to handle this kind of problems?

I would be grateful if somebody could suggest a modeling approach.


John C. Chien July 28, 2000 16:52

Re: Suction
(1). You need to tell us more about your problem, the approach, the code and the measurement. (2). Normally, we don't get good results from running a code. So, there is nothing unusual about the poor comparison.

Alton Reich July 28, 2000 18:48

Re: Suction

There are various COMBINATIONS of BCs that will generally produce stable solutions that converge. The most common is to specify a mass flow (or velocity, it amounts to the same thing) at an inlet and a constant pressure at the outlet. This combination is stable because it allows the code to compute the velocities at the outlet (which will result in mass flow in = mass flow out), and the required delta P to push the mass through the domain (pressure at the inlet).

This is not the only possibility. You should be able to specify a constant pressure inlet, and a constant pressure outlet. Just make sure that the inlet pressure is greater than the outlet pressure :). In this case, the code will compute the mass flow rate required to balance viscous losses within the domain against the pressure drop.

You should also be able to flip the first case over. That is that you should be able to specify a constant velocity (or mass flow) at the outlet and a constant pressure at the inlet.

Ok, so what about your problem? I think that I would try specifying a velocity profile at your outlet (the inlet to the fan). This would be particulatly useful if you know not just the flow rate but the swirl velocity, too. On the other boundary specify constant pressure.

One final tip. "Inlet" and "Outlet" have more to do with how velocity and pressure are calculated than with the direction of the flow. You may get better behavoir if you treat your outlet that goes to the fan as an "Inlet" boundary.


Alton Reich July 28, 2000 18:51

Re: Suction
Quote John: "Normally, we don't get good results from running a code."

Perhaps, but if we don't run the code we get no results at all.

Alton :)

John C. Chien July 28, 2000 19:59

Re: Suction
(1). Both statements are correct. (2).If we don't run the code, we normally run the test. And no one would question the results. (3). When we don't get good results from running a code, we know that there is something wrong with the options we selected. (4). If we keep trying and changing the options, we will be able to get all the wrong results. (5). Then if we are still alive then, it is possible to figure out the right answer. (6)Design iteration by definition is to eliminate the wrong answers. If we don't get the wrong answers, the chances of getting the right answer is very small. (7). For a brand new problem, if we don't have any information about the answer, it is very difficult to know whether the answer from running a code is right or wrong. (8). Anyone who disagrees with this statement-7,, can just pick any fluid dynamic problem and run the code. There is no way he will be able to get the right answer, if he has no prior knowledge about the problem and the answer. (9). Any interested in validating this statement-7, can try to run the flow through a pipe problem. The pipe has a 90 degree turn in to y-direction after one diameter from the inlet. This is followed by a 90 degree turn into z-direction after two diameter from the first turn. It then make another 90 degree turn into the original x-direction, after another two diameter of length. It then make a final 180 degree U-turn back in negative x-direction, after two diameter of length. Each turning radius is one diameter. (10). So, the question is are you going to get the right solution at the pipe exit by running the code? There are 4 turns only in this pipe. The real world piping is much more complicated than this. Well, how many of you will be able to get the right answer? A very simple flow through a pipe with 4 turns. (laminar or turbulent)

Jim Park July 29, 2000 10:40

Re: Suction

I don't remember who said this, but it is a great truth!

No one believes the CFD results except the analyst,


Everyone believes the experimental results, except the experimentalist!

John C. Chien July 29, 2000 19:50

Re: Suction
(1). Well, two more correct statements. (2). Analyst deals with the ASSUMPTIONS, THEORIES, ALGORITHMS, OPTIMIZATION, AND INTERPRETATIONS. These have to be absolutely correct in order for the analyst to get the results. But , assumption is assumption, theory is theoroy, and even the interpretation is always subjective. (3).They have to assume that there is H2O on Mars first,in order to carry out the second step. So, if you say that the turbulence model is not really correct, then it is not acceptable to the analyst. He must assume that it is absolutely correct in the first place. (4). I think, people tend to think that measurement is just like taking the body temperature or the body weight. If the thermometer says that you have high fever, then you'd better see a doctor right away. (5). In reality, the aerodynamic measurement can only give one the static pressure and the total pressure. You don't get the velocity at all!!! And the funny thing is that after this complex flow field measurement, he then use an IDEALIZED 1-D THEORY to do the data reduction!!! SO, Actually it is very hard to measure both the speed and the direction, not to mention the turbulence components. (6). Now, we are in big trouble. To validate the 3-D cfd results, we need the test data. Unfortunately, the test data is calculated based on the 1-D inviscid theory. (7). Well, how many people have actually measured the skin friction directly (wall shear stress)? (8). I think,cfd results are based on the assumptions, while the assumptions are used to interpret the test data.

Ding Li July 31, 2000 00:36

Re: Suction
If we analyse the eigenvalues of inlet and outlet, we will find 4 positive eigenvalues in inlet(define flow into domain is positive) and 1 in outlet( 3D compressible and subsonic flow). This means we need four boundary conditions for inlet and one for outlet. We usually set back pressure in outlet and v/u = w/u = 0 in inlet and as to two other inlet boundary, you can cheese total pressure and total temperature, or mass flow and total temperature.

Good luck!


Fred Uckfield July 31, 2000 09:30

Re: Suction
"All models are wrong but some are useful"

G. E. Box

John C. Chien July 31, 2000 11:39

Re: Suction
(1). Very good statement. (2). It is like saying that in cfd, 99% of the time, we are getting the wrong answers. (3). In other words, the wrong answers and models are being used to derive the right conclusion.

Fred Uckfield July 31, 2000 12:00

Re: Suction
As long as the ends justify the means...???

A dangerous attitude (often used by nasty little dictators and religous extremists) but one that, despite itself, is fraught with success.

What interests me is how wrong can your model get before such time as the usefullness is compromised?


John C. Chien July 31, 2000 12:40

Re: Suction
(1). In real world, no one is investing in the modeling. How many companies are developing turbulence models? I would say "none". (2). Actually, we are using very old models. Most of them are over 20 years old. (3). If I say that I am helping a company which has an equivalent of over 200 years history, you might think that the situation would be different, right? Wrong! The company does not do cfd or turbulence modeling at all. But there are some limited number of commercial cfd codes licenses on the system, that's all. And all I know is ,on this side of the company, I am the only one who is actually running the commercial cfd codes.

All times are GMT -4. The time now is 09:09.