A consistent and efficient framework for the time integration of multibody systems with impacts and friction
Staff - Faculty of Informatics
Start date: 22 February 2017
End date: 23 February 2017
|
|||||||||||
|
|||||||||||
Abstract: | |||||||||||
In this work, we discuss time integration schemes for the dynamic simulation of nonsmooth flexible multibody systems. We develop a framework for the consistent treatment of velocity jumps, e.g. due to impacts. A non-impulsive trajectory of state-variables is improved by an impulsive correction after each time-step if necessary. This correction is automatically chosen starting from a non-impulsive base integration scheme, which discretizes the propagation within the time-step. Consistency is achieved due to the impulsive corrections on the same kinematic level as the treatment of non-impulsive constraints. This idea stems from a time-discontinuous Galerkin setting, but is generalized concerning the splitting of non-impulsive and impulsive force propagation. In this work, we compare the behaviour of four different base integration schemes in the newly developed framework as well as a classical Moreau-Jean timestepping scheme concerning selected criteria and examples from academics and industry. It turns out that the half-explicit timestepping is a very robust and the most efficient method as far as we deal with non-stiff problems. The timestepping schemes based on the generalized-alpha method, the Bathe method and the ED-alpha method become most efficient for stiff problems with spurious oscillations. In our test cases the generalized-alpha method is the most efficient base integration scheme concerning computing time, however it may get unstable in the nonlinear regime. The ED-alpha method satisfies exactly the opposed characteristics. It is the Bathe-method, which seems to be the best compromise concerning stability and efficiency in the nonlinear regime. We propose it as a base integration scheme for timestepping methods whenever stiff problems with impacts and friction have to be solved. |
|||||||||||
|
|||||||||||
Biography: | |||||||||||
|
|||||||||||
|
|||||||||||
|