CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   Questions on direct numerical simulation (

feng_ji_ren April 29, 2003 08:28

Questions on direct numerical simulation
Hello everyone!

I am a graduate student.

I want to do the project:Three-dimensional direct numerical simulation on the wake of circular cylinder. (Reynolds number will be about 190~525)

It will be solved by my own program making by computer language(c language).There are many kinds of mathematical methods about this problem such as finite element,finite difference,LBG,spetral element,and so on.As I am a beginner,I don't know how to choose one of them.Because you know the time of everyone is limited.

So I want to ask which one is the best in the DNS. By the way,is it worth doing it (my project)?Sometimes I am doubted.This will be one part of my thesis .It is my fist step.

Thanks all of you for helping.

Good luck! You can contact me with my

XX_x April 29, 2003 11:41

Re: Questions on direct numerical simulation

Before that, remember that fluctuations are needed to do LES and DNS calculations. Either starting from the experiment or by any computational method (e.g. Pao spectrum or CDI, etc) The other thing is, circular cylinder. Are you talking about swirls in 3D? If you don't have experimental data available and have to create your own initial condition, then you will need time for this task. You'll see. If that is a MS project then... I tell you, forget it, you won't have time without REAL help from MANY people if you are a beginner. Are you starting your project of doing/creating your own CFD-code from zero? How many years are you talking about?

feng_ji_ren April 30, 2003 04:00

Re: Questions on direct numerical simulation
First,thanks for your responds and suggestions. In fact,I am a doctor student.Please do not laugh at me. I have 3 more years to do it.Unfortunately,I have to creat my own cfd code from zero.Because I have not the good base to go on.I am not talking about swirl in 3D .It is the wake behind the cicular cylinder(3D)

By the way,what is Pao spectrum or CDI?and your email please.

Thanks again!!

XX_x May 5, 2003 10:10

Some issues
I'm not laugh on you, but make sure you have the proper help, I mean, real supervision because there are a lot of "supervisors" outhere that they only make contact with the students from time to time.

On your question: Pao spectrum or CDI?!! There are several method to create the fluctuations for the inital condition. Pao spectrum is one of them based on FFT. which means that you have to create the fluctuations in a random way that suits the E(K) VS k spectrum. Very complicated if you are dealing with unstructured grid. CDI stands for core dynamics instabilites and is based on the creation of fluctuations using instabilities as it is explained in books, papers on trasnition from laminar to turbulence. I think you should take a course on DNS and or LES or even better, find someone in your University with such experience and work with him. Honestly The P Sagaut book won't give you too much help on this initial conditions and I haven't seen noone explaining neither Pao nor CDI explicitly, which means you are on your own, man. Find real help. A cfd code would take 1 year to create it if you are a good programmer with real knowledge of numerical methods. Another year to find why thinks don't work as they should, find some buggs (others will remain years after, until other discovers them) and with luck you will get few month to do your task. Just as an example. I have seen people working 100% in just implementing their models and some year after , someone discover that the program had some buggs making his/her results .... you know. On the other hand a DNS is not so difficult task comparing with modelling stuff. A CFD code is not only solving eqs. by discretization, you will have to create interfaces between the grid generator and solver, the post processing, etc etc etc. I suggest you to devide your code in Initializer, Solver and Post Processing, but there are more ways to do it I think you should talk with your supervisor.

I really wish you good luck. You need it

All times are GMT -4. The time now is 06:24.