search

sandialabs / LCM

Laboratory for Computational Mechanics
Other
12 stars 7 forks source link

ACE cleanups for ACI/NH: different convergence behavior when running ACE problem w/o Erosion with Tempus vs. Piro #80

Open ikalash opened 1 year ago

ikalash commented 1 year ago

In attempting to run an erosion-free MiniErosion denudation problem w/ both Piro and Tempus, I discovered that the convergence history is different. Tempus requires much more nonlinear solves in the mechanical problem (see attached output). I am conjecturing that the initial guess used in Tempus in each time-step is not the same as the one in Piro / not correct. We may want to look into this if Tempus ends up being the method of choice for ACI/NH runs using ACE.

Despite the different convergence history, Piro and Tempus produce the same solution for both the mechanical and the thermal problems.

NOTE: even though the Piro output file has "erosion" in it, the code was hacked to turn off Erosion.

out_piro_erosion.txt out_tempus_noerosion.txt

ikalash commented 1 year ago

I forgot to mention: we should look at this after #78 has been resolved.