CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   To dimensionalize or not to dimensionalize !! (http://www.cfd-online.com/Forums/main/1475-dimensionalize-not-dimensionalize.html)

 tolani owosina October 30, 1999 08:11

To dimensionalize or not to dimensionalize !!

I have run the same problem using two different unit systems ( Calories - grams -cm & Joules -kg -m ) with GAMBIT as a meshing tool and FIDAP as the solver. However in spite of several checks and crosschecks I am getting very different results ! All the examples available to me are using a non-dimensional analysis approach which to me leads to complications in post processing. I would appreciate any examples (program files) of real world problems with actual metric units in any of these codes. By the way the simulation involved is an axisymmetric conjugate impinging jet problem and the problem seems to be coming from my boundary heat flux value. Geometry and material properties look okay. Also the Cal- Gram -cm unit system gives far more reasonable results.

 Sergei Chernyshenko October 30, 1999 10:15

Re: To dimensionalize or not to dimensionalize !!

>To dimensionalize or not to dimensionalize !!

No. The only one who can help you is yourself. Learn how to use non-dimensional variables, why they should be used, and how to pre- and post-process them. It is extremely easy and very important. Or quit fluid dynamics. If you see a program which uses dimensional quantities be aware that it was, most probably, written by a person entirely non-qualified in fluid dynamics.

Yours Sergei

 D.M. Lipinski October 30, 1999 12:39

Re: To dimensionalize or not to dimensionalize !!

Using any consistent set of units should result in the same result. Apart from any major errors (wrong conversion between the units), I can think of two possible reasons of your problems:

1. Maybe some constants (e.g. gravity) that are used by the program from some sets of defaults have not been converted by you. 2. Maybe in fact you dont get convergence with SI units. Some problems expressed in a particular unit system lead to arithmetic problems (because of the orders of magnitudes of the numbers and finite arithmetic). Then typically some contributions to the influence coefficients dissappear. Your calculations converge but not to the solution of your problem. I doubt that this is the problem. This can happen only sometimes when single precision is used in the code.

By the way. The last remark in the response by Sergei Chernyshenko, I find totally inappropriate. Using dimensions or non-dimensional quantities in a code does not say anything about ones qualifications in fluid dynamics. There are cases where each of the ways has advantages. A good general purpose code would function both with non-dimensional numbers and ANY CONSISTENT SET OF UNITS.

regards

DML

 Sergei Chernyshenko October 30, 1999 15:07

Re: To dimensionalize or not to dimensionalize !!

A good general purpose code would function both with

>non-dimensional

>numbers and ANY CONSISTENT SET OF UNITS.

That's it, DML, CONSISTENT SET OF UNITS is a non-dimensional one. Always can be treated like that. This is exactly what differs it from an inconsistent one. But the responsibility in using a consistent set is with the user, not the program. Code always works with non-dimensional numbers, it has to. A good code requires to use a consistent set = requires to use non-dimensional formulation. Otherwise, it provokes human errors. What happened to Tolani. One more example of the permanent flow of such examples passing before me.

But I was not talking about qualification in programming. I was talking about a qualification in fluid dynamics. And I just summarised a result of many years of my observations: any fluid dynamicist using dimensional variables is hopelessly stupid. By the way, such are quite rare, at least at the Moscow State U. and several leading UK universities I happened to be able to observe, since everyone there was taught properly. And there is no surprise, the reason for imposing the use of non-dimensional variables is given on the first pages of all good textbooks, so what we are arguing about?

Sergei

 Valdemir G. Ferreira October 30, 1999 16:20

Re: To dimensionalize or not to dimensionalize !!

Hello friends,

I strongly agree whit both Lipinski and Chemyshenko. I had a relacted problem like this one. In my case, the Reynolds number was incorrect. So, Mr. Owosina, think about it.

Bye.

 John C. Chien November 1, 1999 12:21

Re: To dimensionalize or not to dimensionalize !!

(1). I have used two general purpose commercial CFD programs. (2). It is always a headache for me to keep checking the units for each run. (3). I think, the problem is related to the philosophy used in the construction of each module. (4). The CAD geometry tends to be less sensitive to the units used, because you can always use arbitrary scale to draw the figure and the put the dimension and unit on the drawing. (5). In a drawing with a circle on it, the dimension can be 1.0diameter. It can be 1.0 inch, or 1.0 cm. (6). But in the file, the diameter is 1.0, it is not 1.0 inch or 1.0 cm. (7). When this geometry file is imported into a CFD code, it will be interpreted as 1.0 meter, if the code was written in that units. It becomes very important to check the output to see whether the diameter is still 1.0 inch, if that was the unit used to create the circle. (8). It is not easy to spot this consistency problem, unless you carefully double check all the output to see whether they are identical to the your input. (9). In addition to this problem between the geometry module and the solver, there is also another problem associated with the solver. The solver is always coded in some units, and the selection of other units is implemented through the conversion inside the code. Some codes allows you to input only a couple of different units, say the selection between the SI and the British. (10). And then there is problem with the post-processor. Some post-processor can only take SI units input. So, you will have to convert the units first before using the post-processor. (11). So, if your original geometry is in inches, and your solver is set to receive the British units, then it will be a big problem when you use a post-processor written in SI units. (12). Some codes are trying to allow the users to use mixed units in the solver, but it can only create more confusion. (13). The non-dimensional approach does have advantages , because you can select the inlet height as a reference and force the geometry to become dimensionless (relative to the inlet height). Here there is no need to worry about the wrong units. And the inlet height will automatically combined with other parameter to form dimensionless parameter such as Reynolds number. All you need to do is to pre-compute your Reynolds number as the input to the code. (14). I would say that the use of dimensionless form is easier and more reliable. (15). The problem is most commercial codes are written in real units, and the users are anxious to see the results in real units also. (16). In a way, it is a poor combination of the anxious users and the commercial codes which tends to create many unnessary errors. (17). So, if you have to use a commercial package, try to use only one system throughout the geometry definition, mesh, the solver and the post-processing steps. And double check the results, there is no guarantee that your SI units is the same as his SI units. The same is true for the British units, in many cases, 32.2 somehow just disappeared from the sight by the engineer. (18). Now you know why the Martian Probe crashed on the Mars surface. (are you sure that it was not other planet?)

 Jim November 1, 1999 12:40

Re: To dimensionalize or not to dimensionalize !!

Owosina,

Your problem is not related to dimensionalization or non-dimensionalization.

First, check all of your input, especially any boundary conditions data. Second, make sure that all of your variables are truly from the same unit system (make sure that the grams are grams and not kilograms etc.) A recent example of mixed units is the loss of the NASA probe to Mars. One item that concerns me is the fact that you are talking about calories and Joules. Are your working with reacting flows or just heat transfer ?

As for the remainder of the discussion about non-dimensionalization, what arrogance from the pompous blow-hard, Sergei. First, the use of dimensional or non-dimensional variables does not necessary reflect a person's skill in fluid dynamics. If a person says that non-dimensional variables are never needed be careful. Similarly, if a person says that non-dimensional variables are always needed be even more careful. For calorically perfect gases (i.e. gases that are not mixing or reacting) non-dimensionalization is not difficult and can be very practical. However, for reacting gases the issue of non-dimensionalization becomes cumbersome and allows for addition steps where errors may be introduced. Also, the use of non-dimensional variables can be misleading. For example, in hypersonic flight, surface ablation (ablation: the removal of material due to heating) mass flux rates are often non-dimensionalized by the mass flux at the edge of the boundary layer (density*velocity at the BL edge). However, this is not a practical non-dimensionalization since a boundary layer edge may not be well defined for hypersonic flow and may not even exist if the shock and boundary layer merge to form a fully viscous region.

For moving body or multi-disciplinary problems, such as store release, flight dynamics or fluid-structure interaction, the coupling of fluid dynamics and the six degree of freedom (6DOF) equations of motion requires that forces and moments be transferred between the simulation algorithms. The additional effort required to non-dimensionalization both the fluid dynamics and body motions is wasted since the ultimate typical result is to determine the flight path of the moving object or displacement of the surface in real world units.

 tolani owosina November 1, 1999 16:55

Re: To dimensionalize or not to dimensionalize !!

Many thanks to Messrs J.C. Chien, Jim Mon, D.M. Lipinski and V. Ferreira for their prompt response to my question on Dimensionalisation.

I have decided to run a non-dimensional simulation of the same problem for various main reasons; 1. The solver FIDAP has many of it's tutorials

and examples in non-dimensional units which

is apparently their preference.(Some are in

dimensional units too)

2. It would be useful to compare the results of

different simulations for verification purposes.

3. Unlike Mr. Cernyshenko, I don't know everything

and would like to learn.

In response to Mr. Jim Mon's question , I am simulating a boundary heat flux (heat transfer) and the solver (FIDAP) specifies the Calorie as the unit of energy when using gram/ centimetres as units of mass and length. ( Joules when using kg. and metres).

In response to Mr. Chernyshenko , I am surprised that in spite of the fact that you know it all I have not seen your name as a major contributor to any reputable publication journals etc. Also, I was amused to note that your response to Mr. Sigve Gjerstad , (sg283@eng.cam.ac.uk) who asked essentially the same question as I did was in a much less arrogant tone. Could that be because his e-mail address reflects a British institution where as you say "you know it all"? Well, it seems you don't all know everything after all! Incidentally, I have been to Moscow myself (not to study) and while I do not know where you learnt your arrogance , I find it difficult to believe that it was in a University !

T. A. Owosina

 Sergei Chernyshenko November 13, 1999 14:45

Re: To dimensionalize or not to dimensionalize !!

Dear Tolani,

Sorry for offending you. I was in a really bad temper at the moment. Sorry again.

As for dimensionalize or not, well, you learned and made a correct decision. :)

>I have not seen your name as a major
>contributor to any reputable publication journals etc.

Check JFM or Physics of Fluids :) Except I would not claim to be a major contributor. Who could?

But again I am sorry for my language and everything :(.

Good luck.

Sergei

 All times are GMT -4. The time now is 23:01.