CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   computational requirements to write own cfd prog! (http://www.cfd-online.com/Forums/main/1346-computational-requirements-write-own-cfd-prog.html)

yogi September 29, 1999 22:42

computational requirements to write own cfd prog!
 
hi guys.. well.. i had some doubts in mind before i start writing my own cfd program.. 1: how logical/helpful is it to write our own cfd code ( in my case its 'turbulence modellling over a bluff body at the free surface')instead on using softwares like FLUENT AND FIDAP? 2: assuming i start writing my own code.. what computational requirements would i need? a workstation or a PC hardware requirements. 3: as for the question on FLUENT capability i would like to ask how rigorous is the turbulence modelling in FLUENT?considering the fact that i am concentrating on the turbulence effects at the free surface this is quite important for my study.

hey .. would like to thank in advance for all ur replies and suggestions.. bye and regards yogi

John C. Chien September 30, 1999 00:31

Re: computational requirements to write own cfd prog!
 
(1). Assuming that you have the governing equations all figured out, the missing part will be the Reynolds Stresses terms. (2). The turbulence modeling deals with the closure of the Reynolds stresses. In other words, it must be modelled in terms of existing variables, such that you have enough equations for the unknowns. (3). There are several levels of turbulence modeling ranging from the Reynolds stress model to eddy viscosity models (algebraic, 1-equation, 2-equation). (4). The next step to do is to convert the governing equations into algebraic equations so that you can find a method to obtain the algebraic solutions. (5). As long as you have the background in numerical methods for partial differential equations and the Fortran programming skill, the coding part should be fairly straight forward. (the complex geometry is a separate issue) (6). The free surface is a special feature of your problem, but there are methods to handle this type of problems. (7). As a users of commercial codes, I don't care what's inside the code. I can always device various test cases to find out the information I need. That shouldn't be a difficult task. This is the only reliable way to convince yourself what's inside the commercial code.(assuming that you have free access to the code.)

yogi September 30, 1999 00:56

Re: computational requirements to write own cfd prog!
 
hey thanks.. for the comprehensive reply.. but i have a few more queries.. 1: as u mentioned that having a background of numerical methods and fortran with theoritical background of turbulence the code can be written.. but what about the complex geometry? is there a way in which i do my mesh generation in one software and then use that mesh file into my partial D E solver which i code myself and then take the output and link it to another visualistion software? 2: u mentioned about the different methods to handle surface methods..could u elaborate more on that? 3: and also reading the posted messages on the 'PC or WORKSTATION' ..i am apprehensive about the accuracy of my code that i would be writing coz of the hardware limitations. thanks again in advance.. regards bye yogi

Yogesh Talekar September 30, 1999 03:01

Re: computational requirements to write own cfd prog!
 
Is is better to write your own codes.. You can modify them easily to fit to the situations you are handling!

Also its much better to have a Workstation if you are handling turbulance.

If your problem is in 2_D or 3-D you MUST do it on worksation. Because solving even a simple 2-D problem with full N-S equations on a PC takes LIFE-TIME of a human-being! In case u have restricted access to a workstation develop your code on a PC and try to run it for few iterations of PC itself. When your are confident abt. it you can transfer it to Worksation.

Best of luck.

bye

YOGESH

Shigunov September 30, 1999 03:49

Re: computational requirements to write own cfd prog!
 
I should suppose that it is quite not necessary to write own code (and to elaborate own method, naturally) for applying it to only one problem, particularly for complex geometry.

Best regards

yogi September 30, 1999 09:46

Re: computational requirements to write own cfd prog!
 
hey yogesh thanks.. a lot but i still have one doubt unanswered .. 1: is it possible to generate the mesh in one software and then solve the PDE s using the written code and then shift the output into another visualisation software.. is this the way CFD programs are written by mundane ppl like me?? thanks a lot again. bye regards yogi(yogesh!!!)

Karl Edquist September 30, 1999 10:28

Re: computational requirements to write own cfd prog!
 
I am doing this with my own CFD code now. I use Gridgen to create the mesh, run the problem with my code, and then view the results using FAST. FAST is a freely available visualization software package for SGI workstations from NASA Ames. You must, if course, have access to an SGI machine

Karl Edquist September 30, 1999 10:29

Re: computational requirements to write own cfd prog!
 
I am doing this with my own CFD code now. I use Gridgen to create the mesh, run the problem with my code, and then view the results using FAST. FAST is a freely available visualization software package for SGI workstations from NASA Ames. You must, if course, have access to an SGI machine. There may be visualization software for other workstations/PC's available to the public.

Larry Clark September 30, 1999 16:53

Re: computational requirements to write own cfd prog!
 
I don't think you will have any problems running CFD on a PC as long as you do everything double precision on 32 bit machines. I found intermittent failures when I attempted to use single precision. The floating point processors on the Intel machines suffer little by using double precision. This can be more of a problem on some of the 32 bit unix machines; they will slow down more. best wishes.


All times are GMT -4. The time now is 08:05.