CoilSolver problems with boundary layers and parallel processing

Numerical methods and mathematical models of Elmer
Post Reply
didzisB
Posts: 1
Joined: 21 Aug 2023, 15:45
Antispam: Yes

CoilSolver problems with boundary layers and parallel processing

Post by didzisB »

Hi!

I am trying to solve a case (following this article https://hal.science/hal-01336357/document) where three AC coils are generating volume force in a metal volume. Initially I am not interested about the generated liquid metal velocity and would like to only calculate the electromagnetic part of the problem.

I have previously calculated similar cases with "MagnetoDynamics" solver by simply defining current density as a volume force. However, it is not so easy to define volume force if coils are not axially symmetrical. So for this task I decided to use CircuitBuilder and CoilSolver and followed the provided tutorial case here: https://github.com/ElmerCSC/elmer-elmag ... d/harmonic

The tutorial case runs fine with single process as well as in parallel. However, there were couple of problems with my case.
First, my case did not converge with a mesh that has boundary layers.

ERROR:: IterSolve: Numerical Error: System diverged over maximum tolerance.

Later I learned that the case converges if I use mesh without boundary layers. However, there was another problem. Turns out there is a segmentation fault error if I use parallel processing.

Backtrace for this error:

Program received signal SIGSEGV: Segmentation fault - invalid memory reference.


I used

Code: Select all

ElmerGrid 2 2 meshElmer -metis 6
to make partitions.

Does CoilSolver support mesh with prisms?
Does someone know how to fix these problems?


Didzis
Attachments
TravelingACfieldWithBoundaryLayers.zip
Not working with boundary layers
(674.38 KiB) Downloaded 44 times
TravelingACfieldWorking.zip
Working case w/o boundary layers
(626.46 KiB) Downloaded 41 times
Post Reply