This is a tricky question. I don't know the software you are using, but regardless if you using a studied model in a peer reviewed software, the non-convergence means that you are using parameters in a scope where convergence do not exists or that is too slow to be measured. Review you parameters set and remembering that non-linear equation do not necessarily converge somewhere or even present unique solutions (or any solution at all) for all parameter's space. If you are using parameters from a known system or from empirical data where you know that the system presents such stationary state, then you should report to the builders of the software as a bug report.
If you made the software yourself, try to identify errors by simulating simple situations where you are sure of the answer or share the piece of the code you are troubled with.