I’m currently trying to simulate a 3D printed hard drive enclosure with a SAS expander backplane. Because of the restrictions imposed by the backplane, spacing between the disks is limited and I expect them to need heavy cooling.
After struggling to get a decent mesh for a while, I think I might finally have one that’s OK enough to sim. However, when I started the run, it failed on iteration 1 complaining of continuity time step errors.
Anyone have any advice for how I can resolve this?
I think as a general advice on these kinds of situations, the best approach should be to simplify your CAD model so that the mesh can get to a reasonable size. If the hard drive is in an open environment, maybe applying for instance an external flow volume should give you more stable results than applying the boundary condition directly on the model.
Looking at the meshing parameters from your last successful meshing run, the max non-orthogonality is 89.57 which will probably lead to the divergence of your simulation (for more information, take a look here).
If reducing complexity is not an option, at any rate I don’t think you should be able to run the simulation on a Community account, due to it being very computationally consuming. Try looking at the following tutorial as an example:
However, as one last comment, I think something that might help reducing the computational demand of your analysis is to toggle the Compressible option to “off”:
My advice is to start with a simplified geometry—it doesn’t need to exactly match your real geometry. Consider whether bolts, nuts, and other intricate features are truly necessary for a CHT study like this (in most cases, they aren’t). Often, it’s better to set aside the CAD model and create a simplified 3D representation from scratch.
First, test the models you’re using, as well as your workflow and meshing approach. Once you’re confident in those, you can move on to a more complex geometry.
I was following that tutorial which mentioned to turn compressible on. Is this not needed for a “close enough” simulation?
Regarding simplifications, I was able to greatly reduce the size of the mesh by deleting the plastic pieces that I don’t particularly care about. I manually generated the flow volume in CAD using a boolean subtract with all components of interest, then deleted everything except for that flow volume and the drives / SAS expander chip.
All in all, I’m really concerned about the ability to cool the drives + the ability to cool the expander. This geometry is essential to that, but I don’t care about the temperatures of the plastic, so I was able to remove those pieces. It’s still around 15.3M cells though, so I will need to do something else to bring down the sizing.
Additionally, nonOrthogonality is still high even with the changes - it seems that the boundary layers are hotspots for this - would adding or removing a boundary layer help?
Thanks! I did try to simplify it more since the original post, but I think you’re right I need to just set aside the model and make a simplified 3D model for this purpose
15.3M of cells its a huge mesh for such a simple (if you succeed simplifying) model. 2nd advise: Never use the automatic mesher, do use the user controlled, hex-dominant mesher, and refine the mesh were the temperature gradients are expected to be large.
I thought so too, but I can’t seem to get it down to something more reasonable using the standard mesher. Unfortunately hex-domininant isn’t an option for CHTv2 simulations, so I have been trying to work within that restriction.
I might just do a simple incompressible flow simulation to see where my airflow ideally would be going and then adjust from there