Does adaptive mesh refinement always give better result?
 

I am a bit struggling with whether or not to use AMR in steady state subsonic simulation. I do CFD for road car everyday and I use AMR for few years.

I usually set 2 level mesh refinement and perform 2~3 times mesh adaption for each simulation. For some cases, after the second mesh adaption, the solution got back to where before AMR started. Between the first and the last adaption, the solution fluctuates.

I have observed that the fluctuation is cased by the newly refined mesh changes flow field around the car and in the next refinement, the new adapted mesh changes the entire flow field again, so on and so forth. Sometimes it makes the solution cannot convergent because of that.

Is there any guideline of how many times and how frequently the adaption need to be in order to get a reasonable result? In my case, or other bluff body steady state simulation, if after all the AMR, the solution is very close to the solution without AMR, then should I not use it at all but using only a finer mesh region to capture the wake of the car?

Thank you and looking forward to your reply

What about the criteria you used for adapting the grid?

Thanks for your reply. I am now using FloEFD and it does not allow user to set the criteria. However I am moving to Openfoam and I am looking to use AMR, so could you advise any criteria to me?

Sent from my LG-H818 using CFD Online Forum mobile app

The key is to guide the grid to adapt according to the gradients of the variables that more characterize the flow problem

Yes, sure but sustently you need more than one refinement before getting an accurate result. Have many time and how long between each period of refinement you set?

Sent from my LG-H818 using CFD Online Forum mobile app

Or you can do away with the problem entirely...

How long? long enough for the solution to converge before doing any mesh refinement. Then there is no longer any question.