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

Numerical methods and mathematical models of Elmer
Post Reply
wisnu24
Posts: 8
Joined: 13 Nov 2023, 09:36
Antispam: Yes

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

Post by wisnu24 »

Hi Guys! I'm new to this platform as well as the apps. I have been conducting a simulation of the static current of the PCB to see any drop voltage across the PCB, and I got the error of the following:
ERROR:: IterSolve: Numerical Error: System diverged over maximum tolerance.
you can see the file attached in the following link:
https://drive.google.com/file/d/1K3riP6 ... sp=sharing
can someone help me?
thankyou
Rich_B
Posts: 423
Joined: 24 Aug 2009, 20:18

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

Post by Rich_B »

Hello,

The message says that the solution did not converge. Also in your attached solver log, this message occurs multiple times:
ERROR:: ElementMetric: Degenerate 3D element: 32680
ElementMetric: Body Id: 1 DetG: 0.000000E+00
ElementMetric: Node: 1 Coord: 1.826950E-03 -1.730610E-02 0.000000E+00
ElementMetric: Node: 2 Coord: 1.626650E-03 -1.730610E-02 0.000000E+00
ElementMetric: Node: 3 Coord: 1.826950E-03 -1.730600E-02 1.800000E-05
ElementMetric: Node: 4 Coord: 1.626650E-03 -1.730600E-02 1.800000E-05
ElementMetric: Closest distance: 1-3 (8984-26835) |dCoord|: 1.800028E-05
There are some bad elements, so if the mesh can be cleaned up, then you will have a chance to obtain convergence.

Rich.
pcb-1.png
pcb-1.png (288.65 KiB) Viewed 911 times
kevinarden
Posts: 2313
Joined: 25 Jan 2019, 01:28
Antispam: Yes

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

Post by kevinarden »

I looked at this model and it is a linear 3D tet mesh, rather large, and as noted not well formed linear tets. I agree it is a mesh issue. I tried several different iterative parameters with no luck. If you could post the geometry could try some re-meshing techniques.

The linear tet is the least quality element available in FEM.
wisnu24
Posts: 8
Joined: 13 Nov 2023, 09:36
Antispam: Yes

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

Post by wisnu24 »

Hi @Rich_B and @kevinarden! Thank you for your response. what do you mean by geometry? Is it a 3D file? I generate mesh using Salome where the input file extension is STEP that I modify from my dxf PCB file, to do some extrude in onShape. I don't know that I meshed that well, I attached the 3D file with the following
link https://drive.google.com/file/d/1nNSk5c ... sp=sharing.


anyway, I just followed the tutorial on YouTube how to simulate a DC current flowing through a PCB plane. following is my reference :
https://www.youtube.com/watch?v=hNHTwpe ... 12&t=4329s

thank you.
kevinarden
Posts: 2313
Joined: 25 Jan 2019, 01:28
Antispam: Yes

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

Post by kevinarden »

I mean the quality of the mesh elements. Quality can impact convergence in a solution. The complexity of the geometry makes it very difficult to make a quality mesh. Linear tets are the worst quality element, they are used because the mesh is easy to generate. Quadratic tets are better, but then your number of nodes significantly increases. I would take on side of the geometry mesh quads, then extrude it into a brick element, if 3D is required. Doing it in 2D with quad elements would be even better.
wisnu24
Posts: 8
Joined: 13 Nov 2023, 09:36
Antispam: Yes

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

Post by wisnu24 »

can you help me to refine the mesh, I tried to generate the quad tets like you said using salome but I found difficulty about the tools for meshing. the source of using salome is quite low on the internet. I cannot find how to mesh that generate quad tets. Will be helpful if you can refine a mesh using my following step file :
https://drive.google.com/file/d/1nNSk5c ... sp=sharing
kevinarden
Posts: 2313
Joined: 25 Jan 2019, 01:28
Antispam: Yes

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

Post by kevinarden »

By quad tet, I meant quadratic tet as opposed to linear tet, in salome this is an easy conversion. Modify, change linear/parabolic order.
This can also be done in ElmerGrid using increase
ElmerGrid 8 2 Mesh_1.unv -autoclean -increase
However this did not help the elements were still degenerate (badly formed)
So I took your step file, extracted one side to make a plane surface. Meshed that surface with quad elements, then extruded those into bricks.
I suppose the solution could be 2D with just the quad elements?
Anyway the bricks worked
potential.PNG
potential.PNG (71.71 KiB) Viewed 816 times
https://github.com/mrkearden/pcbelmer
the salome file and unv file was too large for github
on my google drive is everything
https://drive.google.com/drive/folders/ ... sp=sharing
kevinarden
Posts: 2313
Joined: 25 Jan 2019, 01:28
Antispam: Yes

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

Post by kevinarden »

I added a 2dcase to the github, of course this could be extruded to 3D as well
https://github.com/mrkearden/pcbelmer
however the 2d results look the same
2dresults.PNG
2dresults.PNG (75.27 KiB) Viewed 814 times
wisnu24
Posts: 8
Joined: 13 Nov 2023, 09:36
Antispam: Yes

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

Post by wisnu24 »

thankyou kevin
Post Reply