Experimental data validation and/or grid independence
 

Dear all,

I have a very fundamental question. I have some experimental data that I would like to validate my CFD results against. Should I test my CFD solution for grid independence? Or would an adequate comparison with the experimental data be sufficient enough to call the CFD solution (with a grid not tested for independence) a representation of what's happening within the system?

In other words, should only a grid independent solution used for validation against experimental data? Or can a solution that provides adequate agreement with experimental data be deemed as final (without doing any grid independce studies)?

I would appreciate your response.

Thank you.

First, you need to know how your data were elaborated especially if they were statistically averaged. Then, you have to chose among the CFD methods which can be the suitable for your comparison.
Grid independence is a valuable method but its application depends on the CFD methodology you are using

As Filippo said, make sure you are comparing apples with apples, not only in terms of data postprocessing but also in terms of geometry. Then, a mesh convergency analysis is usually the subsequent step.

Re:
 

Thank you for your replies. I understand the part about comparing apples to apples but I am not clear about the answer. So to clarify, let us say I start with a computational grid, solve the problem using CFD, compare results with experimental data and get a match? Do I still have to proceed with a grid independence study on the CFD front or what I have is a sufficient solution?

It depends on parameter your looking for, can you let us know the flow/measurement parameters

"solve the problem using CFD" means nothing ... you have a mathematical formulation for the flow problem that is solved numerically. But you have to solve the mathematical problem that solve the type of variable that can be congruently compared to your experimental data.
Alternatively, (in principle as I don't know if your problem allows for that) you can use DNS that provide all point-wise and time dependent field. Then, you can extract all the variables you need for a congruent comparison to experiment.
Grid independence study is a task that can be useful or not, depending on your formulation