Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.
Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
Magnetic Brake example fails at higher rotation rates- Detail: NaN or Inf found when solving linear system using SOR.
Posted Sep 5, 2014, 7:38 p.m. EDT Low-Frequency Electromagnetics Version 4.3b 2 Replies
Please login with a confirmed email address before reporting spam
I'm currently modeling a situation quite like COMSOLs example of the magnetic brake as expressed in the magnetic_brake.mph file in the example forums. I've gotten my model to work at low angular rates, but whenever I attempt to increase to higher velocities, the model issues an :" - Detail: NaN or Inf found when solving linear system using SOR" error. If I increase the angular rate by small increments from where it works to the speeds where this error is issued, at some point the model simply fails to converge.
To try to understand what was going on, I decided just explore higher angular rates using the actual magnetic_brake.mph model, and to my surprise it behaves just the same way: The model is distributed with an initial angular rate of 1000 rpm (as set by the dV0 term in the Global Definitions: Parameters list), but try bumping it up to 10,000 and it fails to converge and by 100,000 it also issuses "NaN or Inf found when solving linear system using SOR".
As everyone has access to the online magnetic_brake.mph example, could someone help me understand what I have to do to get such mef models to converge when using higher velocities?
To try to understand what was going on, I decided just explore higher angular rates using the actual magnetic_brake.mph model, and to my surprise it behaves just the same way: The model is distributed with an initial angular rate of 1000 rpm (as set by the dV0 term in the Global Definitions: Parameters list), but try bumping it up to 10,000 and it fails to converge and by 100,000 it also issuses "NaN or Inf found when solving linear system using SOR".
As everyone has access to the online magnetic_brake.mph example, could someone help me understand what I have to do to get such mef models to converge when using higher velocities?
2 Replies Last Post Sep 8, 2014, 11:46 a.m. EDT
Please login with a confirmed email address before reporting spam
Posted:
10 years ago
In a similar project finer meshing extended the speed range of a model.
Cheers
Edgar
--
Edgar J. Kaiser
emPhys Physical Technology
www.emphys.com
Cheers
Edgar
--
Edgar J. Kaiser
emPhys Physical Technology
www.emphys.com
Please login with a confirmed email address before reporting spam
Posted:
10 years ago
Turning all domains to "Extremely Fine" still gives the same failure, as is easily verified by the Magnetic Brake example at 10,000 RPM, all domains to Extremely Fine.
Part of what I really don't understand is why the "Angular Rate" input should cause this to happen -- its just another value, another number in the equations, and its nowhere near being actually large or vanishingly small in terms of its magnitude for the purposes of finite element calculations.
Is there any way to manually select a different solver methodology, or even adjust solver steps so it can manage to converge on a solution that easily converged with slightly different inputs? Can I 'hint' the solver with working solutions from lower RPMs so it knows better initial conditions?
The target, again, is just to make COMSOL's own MagneticBrake example produce a converged result at 10,000 RPM.
Part of what I really don't understand is why the "Angular Rate" input should cause this to happen -- its just another value, another number in the equations, and its nowhere near being actually large or vanishingly small in terms of its magnitude for the purposes of finite element calculations.
Is there any way to manually select a different solver methodology, or even adjust solver steps so it can manage to converge on a solution that easily converged with slightly different inputs? Can I 'hint' the solver with working solutions from lower RPMs so it knows better initial conditions?
The target, again, is just to make COMSOL's own MagneticBrake example produce a converged result at 10,000 RPM.