Computational mechanics and its various applications have experienced a significant development over the last decade. From the numerical analysis point of view, one deals with physical laws of subsystems, e.g. deformable bodies, described by Partial Differential Equations (PDE), which are coupled by physical constraints. Their discretizations and the treatment of constraints are one of the key issues in this problem class. In particular the coupling of time-dependent problems leads to systems of Partial Differential-Algebraic Equations (PDAE’s). In my work I present a time integrator for this problem class, which combines the plausible scheme of the RATTLE algorithm with the benefit of variable numerical dissipation by the Generalized-alpha methods. The convergence proof is discussed in detail. For the discretization of the time-depending PDE´s and their algebraic constraints, this scheme is applied to the method of lines and its reversed counterpart. The latter allows to adapt the spacial mesh at each time integration step, where the step size is also variable. The challenge is the dynamic behavior of entries in the derived saddle point problems. I discuss some aspects on constraint formulations and projection techniques due to the sensitive influence of perturbation.
At the end I give some examples of systems of rigid and deformable bodies (flexible multibody systems). The simulation of pantograph with catenary dynamics demonstrates the potential of the approach.
«Computational mechanics and its various applications have experienced a significant development over the last decade. From the numerical analysis point of view, one deals with physical laws of subsystems, e.g. deformable bodies, described by Partial Differential Equations (PDE), which are coupled by physical constraints. Their discretizations and the treatment of constraints are one of the key issues in this problem class. In particular the coupling of time-dependent problems leads to systems of...
»