Help needed with answers to CFD interview questions
Hello,
I will be appearing for a CFD Analysis job interview, I need some help with an interview question which I came across during my interview preparation: How do you know your solution is the correct one if there are no experimental solution available for comparison? I know it sounds a bit "basic". However, I got to thinking because there could be a number of ways to deal with it. One solution could be to check by reducing the grid size, and with a smaller time-step and see that the error asymptotically reaches 0. But this is a convergence validation, right? We still do not know for sure that the solution is correct. Please let me know if my understanding is correct on the above, and also what would you answer for this question if asked in an interview? |
|
Can you give me the best criterion usually checked for Tet mesh in ICEM CFD
Sent from my HTC Desire 626GPLUS dual sim using CFD Online Forum mobile app |
Quote:
|
Well, if it is an interview from a CFD company, usually it is an open discussion. No one expect a fixed answer to that. Even in academic, this question still need lots of open discussion. If you think from the side of the company, what will they expect to hear from you?
I would like to say something I know about this open question: 1 check if the solution is converged. 2 check if the results are mesh independent. 3 compare it with analytical results if you can. 4 check if the solution is physical, the variable is conservative? bounded? realizable? 5 check the parameter's plot, sometimes some parameter shows a periodic trend. 6 compare the results with other CFD code If you address these item in details, I think it might be enough for your interview. Indeed, for an open discussion, from this question, the company should expect very diverse answers. |
My answer would be: You simply can not know that a CFD solution is correct without any form of experimental validation.
|
Quote:
and this is the most pathetic problem for CFD simulation. |
Quote:
|
I would say: "All models are wrong but some are useful - There is no way to know if the a model is useful / giving correct solution without experimental data".
But as some mention, you can of course do some model experiments to gain some trust for your solution. |
All times are GMT -4. The time now is 03:43. |