hi~~

I've done a simulation and want to discuss the numerical accuracy. Could someone tell me how to do it? Or which book should I make reference to ? Thank you for your help!

hi

Do you have analytical results to check against? You can check the numerical results against results obtained from a known equation that can be used to calculate the same variable. This will give you an idea of the numerical accuracy.

I don't have analytical results. My project is a turbulent problem. I don't know how to check the numerical accuracy(ex: from temperature distribution, y+, mesh refinement, etc.). Thank you for your help~~

try to get the experimental or DNS data for your case, if no experimenal or DNS data is available then try to run your case for available experimental or DNS data and compare your result with them to check the accuracy of your code and then based on this you can assure the accuracy of your original case.

The question is: what do you mean by "numerical accuracy"?

If you wish to address the accuracy of you method, ie. whether it is 1st order, 2nd order or something else, simple refine you mesh and your errors should get smaller by some amount... now how do you define the error, that's a different question. Using DNS data is probably you best bet, if you do not have access to any data for the same problem you are investigating, you can (should?) compute some other case for which you can find data.

I would suggest taking a look at: Roache, PJ: "Verification and validation in computational science and engineering" May not answer you question(s) directly but will definitely help you.

-- Jarmo

about mesh refinement - run your calculations with grid twice as fine and see by how much your results change. this will tell you how optimal with the grid you are.

uh...

that's sound like "grid independent test"...