Visit-tutorial-mir

Resources

Slides:

• Tutorial Slides (Tutorial content courtesy of Jeremy Meredith)

Data:

• torus_mat.silo from the Tutorial Data.

Set up a Filled Boundary and Mesh plot with PLIC

• open torus_mat.silo -- this is supposed to be a thin green shell surrounded by two other materials
• open the Controls -> Material Options dialog
• choose PLIC from the Algorithm list
• hit Apply
  • note the warning about only applying to new plots due to our caching of potentially expensive reconstructions.... no problem, we can just hit ReOpen to force a recalculation
• back to main window, create a Filled Boundary Plot of the variable Material, and hit Draw
  • add a Mesh plot, and zoom in a bit
  • note that some cells have the material ordering wrong, and some have guessed slopes incorrectly
  • zoom waaay in to the upper-left portion of the torus, and note that even where it's done everything correctly, it's still discontinuous

Change to the Discrete algorithm, show with and without annealing

• in Material Options, choose Discrete from the list
  • then set the annealing time to zero, hit Apply, and ReOpen
  • note in the main window the random starting conditions for mixed cells
  • increase the annealing time to 1 second, hit Apply, then ReOpen
  • note it's much better, though it may not get the thin shell correct

Change to Isovolume, note the smoothness and the holes

• in Material Options, choose Isovolume from the list, hit Apply, then ReOpen
  • note that it hasn't reconstructed the green shell -- that's actually one of those 3-material holes, though the results are nice and smooth

Change to Equi-T, note the missing shell and the wobbles

• in Material Options, choose Equi-T (Tetrahedral clipping) from the list, hit Apply, then ReOpen
  • note that it's also missing the green material, so accuracy still isn't good
  • zoom in to upper-right section of torus, and note the wobbles from the poor clipping
  • stay zoomed in for now....

Change to Equi-Z, note the lack of wobbles, but still missing shell

• in Material Options, choose Equi-Z (improved finite element Zoo clipping) from the list, hit Apply, then ReOpen
  • note the wobbles are gone, but we're still missing the green material
  • this calls for iteration!

Turn on iteration, tune the parameters

• so in Material Options, click the Iteration checkbox; leave the iterations and convergence at their default (i.e. 5 iterations, 0.4 convergence); hit Apply, then ReOpen
  • note we picked up some of the green shell, but it's not looking all that great
  • let's tune the iteration parameters:
    • lower the convergence factor to 0.3 (which means change less on each iteration; note that 0.4 is about as high as I'd recommend you go, as the higher the value, the less likely it will converge)
    • and to compensate, increase the number of iterations to 10
    • (aside: a few more iterations would still have worked well with a 0.4 C.F., but if you look at the results from 4, 5, 6, 7 iterations, you will see some oscillation around the correct answer, so a slightly lower C.F. really is better here)

Show material selection on Pseudocolor and Mesh plots

• Finally, note that material selection can be used for all sorts of analysis features, not just seeing your materials:
  • delete the Filled Boundary Plot
  • add a Pseudocolor plot of U
  • go to Controls -> Subset, or click the matching icon next to the plot
  • turn off all the materials except "2 B", and hit Apply
  • note that the plot and all queries will now operate on the selected data