Skip to main content
11 events
when toggle format what by license comment
Aug 29, 2020 at 9:00 history tweeted twitter.com/StackMma/status/1299632902616031234
Aug 29, 2020 at 3:59 comment added xzczd Just tested in v12.0.1, the ndsz warning no longer pops up, at least in this specific example.
Aug 29, 2020 at 3:59 history edited xzczd CC BY-SA 4.0
added 84 characters in body
Aug 29, 2020 at 3:36 history edited xzczd
edited tags
Jan 20, 2019 at 19:51 vote accept Woody
Jan 18, 2019 at 16:11 answer added xzczd timeline score: 7
Jan 18, 2019 at 15:38 comment added Woody @xzczd Sorry, should say "increase t0 to 3.8595/f". I've edited the question.
Jan 18, 2019 at 15:37 history edited Woody CC BY-SA 4.0
edited body
Jan 18, 2019 at 15:25 comment added Woody @HenrikSchumacher Sorry, I should have clarified. Instead of using r = 0, which I realize will introduce a singularity, I use a small value. In this case, 10^-6, to avoid any infinite expressions. Is that what you are refering to? Also, from my understanding FEM does not work for spatial derivatives with an order higher than 2. Mine has spatial derivatives up to 3.
Jan 18, 2019 at 15:12 comment added Henrik Schumacher You introduced the singularities yourself by using polar coordinates: While the actual PDE on the disk may have smooth solutions, the polar coordinates pulls this solution back to something with a singularity for $r= 0$ because the differential of the polar coordinate mapping is singular for $r=0$. So any discrete method for solving the PDE will exhibit large errors because the fundamental assumptions in all convergence proofs for discrete methods exploit that the solution of the continuous problem is smooth. I'd suggest to apply FEM (by using NDSolve) directly on the original disk.
Jan 18, 2019 at 14:31 history asked Woody CC BY-SA 4.0