Methods for Parallel Calculation and Integration of Smooth and Non-Smooth Multibody Systems
In recent years, both the size and the complexity of multibody systems have increased dramatically, leading to high computational times. To ensure that multibody simulations can still be used effectively in the future, methods for decreasing the computational time must be developed.
There are already several methods for this purpose. Topic of this research work are methods for the parallel calculation and numerical integration of smooth and non-smooth multibody systems. Existing methods are adapted to the special demands of multibody dynamics and new methods are developed.
Parallel methods for multibody dynamics can basically be divided in three groups: parallel co-simulation of systems of the same or different physical domain, parallel methods within in the integration schema and parallelization within the system (internal parallelization). Closely linked to the internal parallelization methods are dynamic, static and hybrid load balancing algorithms, that are individually adapted to multibody dynamics.
Besides these individual parallelization methods, combining them is one of the most promising ways for the reduction of computational time by parallelization methods. In particular, combining internal parallelization and parallelization within the integrator leads to complex requirements on the simulation environment.
Regarding the effective integration of multibody systems, integration schemas with implicit parts are taken into account. Thereby, interesting and ambitious challenges must be solved regarding the calculation of the implicit parts.
Apart from the parallelization within the integration schemas, methods must be developed to construct schemas of higher order and to adapt the time step size automatically within in smooth and non-smooth part for non-smooth multibody dynamics.