Point sensor makes the simulation diverge?

Hello,

I added a point sensor to my simulation and it does not converge anymore (the convergence in the task manager goes up)? What might be the reason and how it could be fixed?

Thanks.

Hello Saya, can you please try to attach the log file from the solver? (it is a file ending in _Input.log in the Results folder of your project).

This would contain information about the type of simulation, simulation settings, grid settings, hardware settings, warning messages, error messages, etc... all of which are necessary to be able to provide you with a helpful answer.

Hello,
There is no file ending in _input.log, but there is files ending just _input and _input.h5? Do they work?

Your file explorer is probably hiding some file extensions (there is no file ending with _input, it must be _Input.log and the Explorer is trying to be clever....).

Files ending _input are text file. How I can attach it here?

Yes, that's correct: the log files are text files with a .log extension.
To attach it here, click "Reply", then drag & drop the Input.log file into the compose message box.

I did it. But there an error pops up: "ERROR
You do not have enough privileges for this action."

can you try again now?

There is still the same error.

This post is deleted!

It is possible that both simulations are performing equally well (or equally bad...). One major difference is that the "Convergence" that you see in the task manager is computed differently: in one case it includes the edge sensors at the source and the overall field, in the other case it includes the edge sensors, the overall field, AND the point sensor. The latter is a bit "stricter" and it may detect divergence sooner (or rather "non-convergence", since the simulation has not actually diverged).

my suggestions:

  • try running for more periods and/or use automatic convergence
  • upgrade to Sim4Life 4.4

The first file is the tutorial "3.1.7 Generic MRI System and Lead Pass" which it converges. The second one is exactly the same as the first one, I just added a point sensor at the origin and it makes the simulation not converge.

I checked the plane wave simulation, tutorial " 3.3.4 Heating from a Standard Active Implantable Medical Device" with line sensor which is attached.

427eb6a5-2266-47f8-9ef2-b541d6981478_Input.log

b03661b0-9635-4ff8-93fa-7b9e220f6b89_Input.log

And it diverges too.

How I can use automatic convergence?

None of the simulations you have shown so far truly "diverge". They all finished successfully, as indicated at the end of those log files.
Note that this does not mean that the simulations "converged", nor that the results are correct.

In principle, simulations stop when one of the following happens:

  • the excitation signal was sustained for the desired amount of time (you can see what your current settings are by clicking Source Signal button in the ribbon)
  • the simulation diverges (the numerical scheme becomes unstable and values go to infinity)
  • the solver returns an error (because of a license issue, incompatible hardware, no more space on hard drive, out of RAM, bug, etc...)
  • the solver detects that the simulation converged

Now the question is, what do you think happened in your examples?

Actually in the simulation I sent you, the convergence in the task manager goes up instead of going down. You mean if that happens, it does not mean the simulation does not converge?

Convergence, as plotted in the task manager, is computed as a rate of change of the quantities measured by the sensors:

  • when that number goes up, it means the simulation has not converged (yet?)
  • when it goes towards infinity exponentially fast, the simulation has diverged
  • when it goes below a given threshold (-50dB is the default setting), the simulation is said to have converged

Note (once again), that "convergence" depends on the sensors that have been set in the simulations, so adding a point sensor affects the "convergence" value reported in the task manager even if nothing else changes. For example, the rate-of-change of the field values reported by a point sensor placed far away from a source might go up when the traveling wave generated by the source reaches the point sensor.

To answer your question specifically: I don't know, because I can't see what you are doing. But try running the simulation for a longer time, then you'll know whether the simulation converges or diverges.

Thanks for your help.

Log in to reply