I would like to know if there is more information regarding the following warning message when running a simulation:
WARNING: Newton did not reach convergence
It seems that this warning message does not prevent the simulation from arriving at a solution. Also, per previous posts regarding this, I have tried removing the warning by increasing spacecraft capacitance, decreasing simulation dt, and increasing the number of iterations in the Poisson solver.
However, this has only prevented the warning for the beginning of my simulation and after some time it gives me the warning every other timestep. Again, the simulation is still able to progress with this warning - so what is it telling me exactly? How might my results be effected by it? I am not sure of this warning’s significance… any advice would be much appreciated, thank you!
while running a SPIS simulation I receive exactly the same warning message. Were you able to find a solution? I tried to change the spacecraft capacitance and Poisson solver parameters too, but the warning appears at every simulation time step. Do you think the results of the simulation could be reliable?
Any information would be valuable.
I haven’t found a way to stop the warning entirely, but the issue seems to be related to the mesh quality. In my cases, the results are still comparable to some theoretical results I have, but I cannot say for certain how reliable. I do notice that the potential of my spacecraft is not “smoothly” plotted with time due to this error.
It is occurring for me in regions where I have relatively large changes in mesh conditions (i.e. a region with a mesh resolution of ~1e-1 containing a sub-region with resolution ~1e-2). In my case, I have a very small spacecraft in a relatively large domain and so the mesh resolution is on completely different scales. I have been able to reduce the frequency of this warning by including sub-domains into my computational domain. I add smaller volumes within my main domain which have an intermediate mesh resolution between the large and small scales of my simulation domain and spacecraft. I’m not sure if there is a better way to do transitional mesh regions in the gmesh versions utilized by SPIS.
If you like, I can send you and example of such geometry to see how I’ve done it.
Hi Tyler,
thank you so much for your response. I don’t think in my case the mesh is an issue, since I have the same resolution throughout the whole domain. However, I noticed that this warning appears only when I use the ECSS global parameters, while with NASA parameters I have no problems. Did you also notice?
I am using custom global parameters in my simulation, but I may look into the differences between the ECSS and NASA parameters. Still not sure what is causing the warning.
Hi Anna, Tyler. I think your main issue can be mesh quality, at least I know Tyler was struggling with this, I would mesh in gmsh v4 (and above) outside of spis, and importing the (version 2) mesh into the spis project, skipping the geometry editor step entirely.
The warning just shows that the computation ran for the max defined iterations without dropping below the convergence tolerance value. From experience, an unstable or just very slow simulation can be expected. Your simulation results (e.g. potential or current vs time should show if the result really is reliable)