Greetings,
Is there a general rule, in what situations to use different constraints handlers? And please do not tell me that solution is Trial and Error: use constraint handler that is good for your certain situation . I know only that I can not use Transformation method when retained node is not constrained in any other constraint.
Second question is a recommended number of iterations for each of the alghoritm solutions. I know that for Newton-Rhapson is 6-10, for Modified Newton goes to 1000 or 2000, but i don't know for the rest.
When we use transformation method, we manipulate the displacement matrix and simply put zeros in the restrained DOFs. Solver removes the associated rows and columns to these DOFs from stiffness matrix. This makes the stiffness matrix smaller and enhances the computation speed. However, if boundaries are in some ways connected to each other, removing the associated rows and columns can affect the stability of solution. In this case, the solution fails in very first steps of computation.
When we use penalty method, we manipulate the stiffness matrix and simply put large numbers in the restrained DOFs. Solver does not change the stiffness matrix, and will take more time for solution. The large numbers in stiffness matrix can affect the results if they are too large or too small.
In general, I will always at first use transformation method and if my model failed in the first steps (i.e. great chance of having boundary problems), I will change to penalty method.
About your second question, it really depends on your problem. I will not use more than 50 iterations for each of these algorithms. Instead, I will try to reduce my time-step to converge within 50 iterations.
. I know only that I can not use Transformation method when retained node is not constrained in any other constraint.