Diagonally Implicit Runge-Kutta Methods
Dear MBDyn Developers,
I was thinking about the implementation of a diagonal implicit Runge Kutta method for MBDyn. Something like "sdirk4" from the NASA report TM–2016–219173. For a prototype implementation applied to implicit ODE's see also the attached file: radau5.m One of the advantages of such a method would be better stability in case of a variable step size. So, I would like to ask you about your opinion, if this kind of method would be feasible for MBDyn. Of course I understand, that currently the number of stages is limited by the number of states stored by structural nodes. So, either I do not perform a prediction between the stages, or I have to use the "integrators" branch as a starting point. In the first case only elements using automatic differentiation would be supported.