Mesh quality and simulation time

[jingansong]

Hi everyone,

I'm running a simulation, whose mesh size is big and the element number is small. But it still takes a lot of time. Could it be due to the poor mesh quality of some regions? I didn't request data output very frequently.

 

[burningDiesel]

Implicit or Explicit?
In both cases the answer can be "yes".
Anyway, what is "big", "small" and "a lot of time" and "poor mesh quality" in your case?

 

[jingansong]

Hi sdebock,

Thank you! It's explicit, the number of element is about several thousand, and the input file is only about 1M. But it will take one week according to what is done even on the cluster using 12 CPUs. I had some other simulations on the cluster before, the number of element was several hundred thousand, and input file was about 100M, but it took about one week. So obviously the currently simulation has problem. I think I need to check the mesh.

 

[burningDiesel]

In explicit, it's not only the mesh. You have to consider the stable time increment (which among other things, also depends on the mesh quality). But it is mainly influenced by the (smallest) element size, and the material properties.
If you have, even if it is only one, element with high stiffness but small dimensions, it can slow down the entire simulation.
Things you can do:

1. print output EDT to get an idea of the min/max and average stable time increment per element
2. look into mass scaling and time scaling, there is an entire chapter on it in the users manual.