CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   Will somebody help!!! Need resources on DNS (

JOHN May 27, 2002 15:19

Will somebody help!!! Need resources on DNS
Hi all; I need resources on Direct Numerical Simulation! Most of the things I am viewing on the net doesn't provide me with enough info; I need information on algorithms of DNS, example etc... how to implement it... I am new to CFD. Peace Be With You

Tom May 29, 2002 12:34

Re: Will somebody help!!! Need resources on DNS
There is probably not a source that covers everything about DNS. If you want to have some info it is a good idea to look in journals like Physics of Fluids or Journal of Fluid Mechanics, or some recent proceedings of conferences. The problem is that there exist many numerical methods for DNS: pseudo-spectral, spectral element, finite volume, finite differences, compact differences etc. Which method is the best depends on the problem.

Perhaps it is better first to formulate your problem (Is it homogeneous, wall-bounded compressible etc.) and then to look in the journals what kind of methods are used for that problem.

By the way, if you are new in CFD: DNS is not easy, it will take a lot of time to develop a code and to run it. Try to cooperate with someone.

Good luck.

JOHN May 29, 2002 13:59

Reply to Tom
Dear Tom; I would like to thank you for your advice. Finally somebody helped me. You are very right in what you have proposed. I have posed my problem very well, "DNS of incompressible, homogeneous fully developped turbulent channel flow". A rather simple problem physically, but it would be a good introduction into DNS for me. I know I'll be using either finite volume or finite difference method but I need any introduction to DNS. All the text I have read talk & talk... They just talk! I already know what they are saying. I need info about the time scales & the grid generation techniques etc... I am managing myself up till now but i am reaching some dead ends. I have searched most of the online report servers but with negative results. I found reports on special cases etc... with experimental results but nothing on the method & discretization they have used... In case you have any additional comments or any resource, a book @ least, please reply... Peace Be With You

Jongdae Kim May 29, 2002 15:52

Re: Reply to Tom
Channel flow, DNS, FFT, (lower order=2nd order) FDM/FVM etc...

If these are your interestings, why don't you borrow a book by Dr. Orlandi from university library. In the book, some source codes are explained and diskett is also included. Actually I have not used his source code in my research (LES of turbulent flow around bluff body, Dr. Orlandi explained LES schemes also in his book. Not much theory or math., which I prefer. Much better is that he just explained Navier-Stokes equations, how to discretize, details of his source codes, and run. Results of the run is explained. Simple and easy to understand).

However, I just tested one example in my PC (pentium II, 300MHz) and I got some results. His code is very fast. If you could use high-speed vector/parallel machine, it's perfect.

You can find his book in this website. I also mentioned several times on his book as the response of other questions.

Good luck.


Jonas Larsson May 29, 2002 16:12

Re: Reply to Tom
Here is a direct link to Orlandi's book in the CFD Books Guide:

Fluid Flow Phenomena: A Numerical Toolkit

Kim, if you feel like adding a short review of it that would be great - just click on "write your own review". I haven't read that book myself.

Tom May 30, 2002 04:06

Re: Reply to Tom
A general overview of DNS can be found in Annual Review of Fluid Mechnics (1998), vol. 30, pp. 539.

Information on DNS of channel flow can be found on the website

It gives some references to articles and you can download DNS results at several RE numbers. You can consider the DNS of Kim and Moser as very reliable. Ideal to validate your own computations. You should be able to find some information about how many grid points they have used and the size of the domain. However, with less grid points and a smaller domain you probably will get already reasonable results, at least for the mean velocity and for second-order moments. A general rule for the discr. in the wall-normal direction is that you need at least three grid points in the first y^+=5, further away from the wall less grid points are required.

Articles on DNS of channel flow using FVM/FD can be found in Physics of Fluids, for instance vol. 10 (1998) pp. 2608.


All times are GMT -4. The time now is 14:33.