As one can see from Abaqus Manual (http://130.149.89.49:2080/v6.7/books/usb/default.htm?startat=pt03ch07s02abo09.html#usb-anl-aconvergeintro), the default residual error for the increment convergence is 0.005. In my case, the anaysis does not converge using this, but increasing it, it does. Using 0.01 or 0.015 as error of residuals can be considered as a good approssimation? I think it is because it is only 2 or 3 times the Abaqus default one. What is your opinion?
Many thanks in advance for your help.
Don't do this, unless you know exactly what you are doing and how the increased tolerance affects results.
If your analysis converges with higher residual, this usually means that you need more Newton iterations. You can use the solver controls. Try Analysis=discontinuous first; if this does not work, try to change other solver controls. This way is better because it only affects how you get a solution, not what is accepted as solution.
Dear Martin, I will try to modify firstly solver control parameters. As you see, I am not experienced in Abaqus; many thanks for your help.
Dear @Martin, I am performing a direct cyclic analysis (low-fatigue) and it seems that for this kind of analysis it is not possible to edit the solver control parameters.
Do you have suggestions?
Never used that, but the manual says that you can define the max number of newton iterations directly; so I would set this to a large number (like 1000).
It seems that you can also change the time increment or change CETOL. I would think that all of these options are better than to change the residual tolerance.
- Badges
- Science topic