Page 1 of 1

big computational cost

Posted: Wed May 10, 2023 10:31 am
by giom
Hello everyone!

I modelled an infill masonry wall in 3d because I want to study the out-of-plane behavior. In particular, I used 2d elements in xz plane. I set a pushover analysis with displacement control to understand the in-plane behavior before. However, the computational cost in term of time is very big! In particular, in the best of cases after 1 hour of analysis the process is still at 50% because it needs many iterations for each increment.

Are there any options in the analysis setting that allow to reduce the iterations? Or other solutions to make the analysis faster?

I attach the model file to show you my setting.

Thank you very much!

Giovanni

Re: big computational cost

Posted: Thu May 11, 2023 8:35 am
by STKO Team
Try this:
S1A_pushover_damage_original.zip
(777.4 KiB) Downloaded 159 times
22 mins for the whole analysis:
PH_1.png
PH_1.png (58.59 KiB) Viewed 1016 times
  • Use the smallest number of fibers that keep the section symmetric (an hexagon in this case) You don't need a fine mesh for rebars.
    FIB.png
    FIB.png (14.2 KiB) Viewed 1016 times
  • Use disp-beam column, you don't need the extra effort of a force-beam column for rebars with that very fine mesh
  • In the analysis use 30 max iterations (and put the double in the maxDim of Krylov, because with adaptive time steps we allow 2*maxIter, so in this way, Krylov updates the tangent matrix only once per time step)
    Also, use a larger tolerance. you are modeling in mm, and an absolute error norm of 1.0e-5 mm is very hard to reach when you have so many nodes
    AN.png
    AN.png (13.18 KiB) Viewed 1016 times