I would recommend saving the iteration count from get_iter() to your logfile and having an if-clause within you programs major loop checking if get_iter() = 0. The solver is happy thinking that the convergence is good enough. That is the reason why the residual error stays constant after it, because the simulation does nothing. On the EpotBiCGSTAB issue, you state that the solver takes 0 minor iterations after the fourth major iteration round. Simple thing first: On the Gauss-Seidel solver (EpotGSSolver), the relaxation parameter is set by function set_w(): If you believe it to be suspicious, report using the Report Phish button in Outlook or send to Derek and others, Subject: Re: Continuation Simulation: ConvergenceĬAUTION: This email originated from outside of Thermo Fisher Scientific. Iteration Number Iteration Error Solver IterationsĞmittance Error I am seeing over a large parameter space that the emittance converges nicely with the electric potential solver. I wrote in an emittance error metric that finds |e-e*|/|e*| (where e is the emittance) with an initial value at the first iteration equal to one. BiCGSTAB), or under-relaxation factor (with the GS solver). I see no sensitivity in the final result with accuracy target, solver method (GS vs. Thank you for the thoughtful response! I took some time to explore the space and your intuition that the solver error hurestic was not well suited for specific geometries appears to be correct.
0 kommentar(er)
