Ensuring Optimal Convergence Rates for Mixed Formulations with Lumping-Based Static Condensation in case of Limited Internal Continuities
Please login to view abstract download link
For standard primal formulations, the convergence behavior can be deteriorated due to the presence of locking effects, e.g. shear locking, that occur especially when analyzing thin structures of a high slenderness ratio. Employing mixed formulations, discretizing further physical fields like shear forces independently counteracts these effects if the corresponding approximation orders are selected appropriately. To reduce the additional computational effort of such formulations, static condensation can be performed in order to reduce the number of degrees of freedom to a minimum. As a drawback, the computation of the inverse of a sub-matrix required within this procedure is computationally expensive. Hence, employing approximate dual basis functions for the interpolation of the additional fields and adapting the system of governing equations by a suitable weighting, the relevant matrix part becomes a diagonal-dominant approximate-identity matrix. After performing row-sum lumping, the relevant matrix is an identity matrix, and thus, no computationally expensive computation of an inverse is required for the proposed procedure. However, deteriorated convergence rates can occur for discretizations containing points of limited internal continuity. In this contribution, the analysis results for different discretizations of a challenging benchmark problem confirm the described effect and emphasize the requirement of a proper treatment. Thus, employing enhanced approximate dual basis functions instead of approximate dual basis functions within the proposed condensation procedure is investigated to prevent the observed deteriorated convergence rates entailed by an approximate dual-based lumping procedure in case of limited internal continuities. The numerical examples show that the use of enhanced approximate dual basis functions and the proposed treatment of points with limited internal continuity ensures the correct convergence rate for arbitrary discretizations.
