Smooth DMS–FEM: A new approach to solving nearly incompressible nonlinear elasto-static problems

The DMS–FEM, which enables functional approximations with C1 or still higher inter-element continuity within an FEM-based meshing of the domain, has recently been proposed by Sunilkumar and Roy [39] and [40]. Through numerical explorations on linear elasto-static problems, the method was found to have conspicuously superior convergence characteristics as well as higher numerical stability against locking. These observations motivate the present study, which aims at extending and exploring the DMS–FEM to (geometrically) nonlinear elasto-static problems of interest in solid mechanics and assessing its numerical performance vis-a-vis the FEM. In particular, the DMS–FEM is shown to vastly outperform the FEM (presently implemented through the commercial software ANSYS®) as the former requires fewer linearization and load steps to achieve convergence. In addition, in the context of nearly incompressible nonlinear systems prone to volumetric locking and with no special numerical artefacts (e.g. stabilized or mixed weak forms) employed to arrest locking, the DMS–FEM is shown to approach the incompressibility limit much more closely and with significantly fewer iterations than the FEM. The numerical findings are suggestive of the important role that higher order (uniform) continuity of the approximated field variables play in overcoming volumetric locking and the great promise that the method holds for a range of other numerically ill-conditioned problems of interest in computational structural mechanics.

► We extend the smooth DMS–FEM to nonlinear elasticity problems of interest in solid mechanics.
► It enables higher order continuous functional approximations across inter-element boundaries.
► The method is applied without any special numerical artifacts to nearly incompressible systems.
► Stabilized or mixed FEM is conventionally used to arrest volumetric locking in such problems.
► Compared to FEM, DMS–FEM also requires fewer linearization and load steps for convergence.