July 3, 2019 at 9:56 am
Gaurav Sharma
Subscriber

MicaOli 

Thanks for sharing the model. I could reproduce your problem.

As the thickness of pipe is very small compared to the length, the known limitation of handling thin features triggers the issue we are observing. I tried with a 16 GB VRAM and the problem only improves slightly:

 

There are the following two options that can be tried in this scenario:

  1. Upgrade to graphics card with more memory
  2. Reduce the problem size

Though the reduction of problem size demands a better understanding of simulation objectives, to come up with approximations that can possibly be made. As a trial, I kept only the rods at the top and removed the components beneath, and used dummy fixed supports at the cut faces. The analysis now runs well on my 16GB machine at max fidelity:

 

I guess thats the only way to handle this geometry, which otherwise doesn't appear to be most suited for Discovery Live.

Thanks,

Gaurav