CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   NPARC or WIND (

Matt Mason August 8, 2000 15:11

I am new at using WIND 3.0 (NPARC).

I would like some advice on how to determine if a CFD analysis can be trusted or considered good.

I have a couple of year's experience using FEA programs for structural analysis. In using FEA (Algor) I would use the convergence of the Von Mises stress as the mesh density was increased. Is this a good criterion for CFD?

I do not have a luxury of using a wind tunnel to validate my model and CFD runs, so I have to rely on WIND and some common sense.


John C. Chien August 8, 2000 15:57

(1). Well, from the point of view of design, it is nice to have some experience in using structure and stress analysis codes, in addition to the cfd analysis experience. (2). But the problem is, fluid dynamic problem is completely different from the solid mechanics problem. It is many order of magnitude more difficult than the solid mechanics problem. And in most cases, we don't know whether there is a solution at all. (3). But if you are dealing with the kind of problem which has been studied by the code developer, in this case the AEDC and NASA, then, as long as you stay within the boundary of the problem definition, you should be all right. That is if they were able to obtain the solution consistently in the past for a class of problem, then, you can trust the solution. (4). The benchmark test cases are designed to serve that purpose. (5). For new problems and applications, you are on your own. But still, yopu can follow the same process to validate the solution and do the parametric study. (6). In solid mechanics, you can find a solution, in fluid mechanics, that's not always the case. (7). Common problems which have been studied for a long time are, flow over a wing, jet flow, flow through a nozzle. For these problems, you should be able to get some useful results. For more complex problems, one needs to have the test data as a reference point. Otherwise, any solution is possible. (8). Remember that general purpose codes are nothing but libraries of subroutines, it is your responsibility to model the geometry, the mesh, the turbulence, the solution. How you do it will determine whether you will be able to find the solution or not. It is similar to the computer games,except that cfd problems normally have millions of degree of freedom, which is much larger than the unknowns in a computer game. (9). By the way, there is no replacement for a good testing. In laminar flow, it is possible to predict the solution, in turbulent flow, the best one can do is simulation.

Matt Mason August 8, 2000 16:13

Thank you for the response.

I am studying the fluid flow over a wing-body configuration for the determination of lift and drag basically. I am also interested in internal flow such as an intake duct for an engine (jet or recip.).

Other areas of interest are studying winglets and high lift devices (STOL configurations).

I have looked at the validation example from NPARC, but I am also interested in other examples of wings, and etc.

John C. Chien August 8, 2000 16:36

(1). I would say that as long as you start with simple geometry, everything should be all right. (2). So, the complexity of the configuration will determine whether you are going to have headache or not. (3). You can take a look at the AIAA journals to see what kind of problems have been studied successfully by using cfd. And cfd on PC is not the right platform for complex configuration anyway.

Ed Blosch August 8, 2000 18:51

Hi Matt, Northrop-Grumman has a very competent CFD group in California. They had a paper invited to Reno 2000 based on reputation for accurate drag predictions. If they don't contact you first, you may wish to track them down. I'll send you an address off-line. Best wishes

Bilal Bukhari August 9, 2000 14:40

The CFD analysis can be trusted if you get your residual decay by three to four order of magnitude, provided you have modeled the physical problem correctly, i mean applied the bd.cds. correctly. Then you view the problem in your post processor and can see if your solution make some sense or not. If it make some sense and your residual decay by Three to four order of magnitude you can trust the results.


John C. Chien August 9, 2000 15:04

(1). Assuming that the solution is mesh independent. (2). But in real 3-D flow, this is a tough problem to handle on PC. The solution is always a function of the mesh used in 3-D.

Dr. H.K.Narahari August 29, 2000 07:51

All of the points mentioned above (viz ., correct BCs, stable and converging scheme, 3-4 order fall in residue and mesh independence) are 'necessary conditions' only and unfortunately are not 'sufficient' in the classical math sense. Unfortunately therefore we can not assert that the solutions obtained thus are absolutely correct. A general process followed by most researchers, when they develop a new code is to validate it against known experimental data ( forward / backward facing step, cavity, M6 wing etc) and hope the results on their config is OK. We need to do a lot of sanity & consistency checks before we accept the results on face value .... BTW, if CFD was straight forward we would all have no future :)

John C. Chien August 30, 2000 21:00

(1). I like the last comment. (2). I think, most 3-D fluid dynamic problems are largely un-solved. (3). There is no short cut and those who dedicate their time doing systematic analysis of fluid dynamic problems, will be rewarded eventually. (4). Don't feel bad, if you are not recognized at your work place. (if people are still using flow over flat plate and flow through pipe loss data to estimate 3-D complex duct loss, then, I guess, the cfd experience can only be learned. It is personal experience and you can't find it in the code, user's guide, or references.

All times are GMT -4. The time now is 06:15.