Learning Forum

[hhalwa2]

Hello there,

I am using ANSYS Fluent 19.2 to simulate a supersonic inlet at Mach 1.8 with RANS and LES turbulence models. I am working at the stage of the mesh convergence test. So I started from 6, 9, 18, and 40 million nodes where I got converged simulations. And since the higher resolution mesh requires higher computational cost, I decided to use a data interpolation technique from a coarser mesh to a finer one to decrease this cost. The finer mesh has 60 million nodes. But the data did not interpolate properly!!. Then I tried to use a replace mesh technique, but an error pops up with "MPI Application rank 6 killed before MPI_Finalize() with status -(a bunch of numbers)". 

 

Why I can't interpolate data from a coarser mesh to a finer one for same geometry inlet? and why I can't even replace a coarser mesh by a finer? Is there anything I have to do to get the advantage of using these two techniques?

 

Thanks in advance for the help,

[Rob]

Is the problem repeatable, and do you have enough RAM (120GB-ish) to handle the case?

[hhalwa2]

Yes, it is repeatable, I am using LINUX 96 processors and 250 GB RAM. I have no problem with interpolating data from 18 to 40M nodes, but it has appeared when I tried to interpolate data from 18 and 40 to 60 M nodes. See the attached figures. The first figure is supersonic inlet Mach contour of converged simulation of 40 million nodes, while the second figure represents the supersonic inlet interpolated data Mach contour of 60 million nodes. From the later figure, you can see just a blue patch at the spike tip, which I don't know what it is implying!!

[Rob]

Cell count shouldn't make a difference. Can you check the scale of both models and also if you initialise the larger model then use interpolation. Is it the same interpolation file as you used for 18 -> 40 M cells?

[hhalwa2]

Yes, you're right. I wasn't aware of it. The reason was incompatibility in scales.

Thanks