PolyDPG : a discontinuous Petrov-Galerkin methodology for polytopal meshes with applications to elasticity




Mora Paz, Jaime David

Journal Title

Journal ISSN

Volume Title



Over the last two decades, the computational mechanics community has witnessed a growing interest in the development of discretization of boundary value problems (BVPs) on meshes of polygons and polyhedra (i.e., polytopes) of a more general kind than the typical element shapes used in finite element (FE) methods. Among many problems addressed with polytopal discretization techniques, several authors have focused on simulating large deformation of elastomers with rigid inclusions. These works deal with the nonlinear elasticity equations in 2D and hyperelastic materials. Their results have demonstrated a larger capacity for stretching with polygonal meshes, in comparison with standard meshes (solving with a similar number of degrees of freedom). This is one of the scenarios where polytopal numerical methods have been claimed to deliver superior performance to the standard FE techniques. A specific practical problem, with similarities to the mentioned result, is a recent study by Brown and Long on the deformation under compressive loads of a silicone elastomer (the matrix material) with glass micro-balloons as inclusions, a type of composite material known as an elastomeric syntactic foam. The problem has been marked as a simulation challenge by Sandia National Laboratories (SNL). The positive experience with polygonal meshes in 2D suggests that, in 3D, methods that support polyhedral elements may turn out to be well-suited for this specific application.

The Discontinuous Petrov-Galerkin (DPG) FE methodology has proven to be a variationally flexible discretization method, crafted with automatic discrete stability via the use of optimal test functions, and naturally coupled with adaptivity. DPG has been recently introduced into the family of polygonal methods by Vaziri, Fuentes, Mora and Demkowicz, who have labeled their proposed methodology as PolyDPG. In 2D, the extension of DPG to polygonal elements has been enabled by the ultraweak variational formulation and broken test spaces. With respect to the ultraweak formulation, PolyDPG is a conforming method in 2D and, in special cases, in 3D as well. For general polyhedral meshes, PolyDPG is actually a non-conforming discretization. Given the special properties of DPG and the ultraweak variational formulation, we believe that PolyDPG can develop into a tool capable of solving difficult linear and nonlinear BVPs.

In this dissertation we have engaged in three main goals, all of which will ultimately take us closer to a solution of the presented challenge problem through PolyDPG.

First, we analyze the mathematics of PolyDPG as a non-conforming DPG method for linear BVPs in 3D, which helps in understanding its convergence. As a second goal, we develop the computational tools that are required for the implementation of the three-dimensional version of PolyDPG as a geometrically flexible, multiphysics and high order approximation FE method. A series of numerical experiments for the Poisson equation and the linear elasticity equations illustrate the convergence behavior of PolyDPG for different types of meshes and approximation orders. The work of our third goal, which is embodied in Part II of this dissertation, has as its starting point, the study of the theory related to hyperelastic materials and the nonlinear problem of finite elasticity. Next, we cover the derivation of a novel ultraweak formulation of the corresponding linearized problem. Once we have a discrete formulation suitable for PolyDPG, we propose a model problem that resembles the elastomeric syntactic foam problem. By means of a nonlinear solver, we perform a set of numerical experiments to assess the performance of PolyDPG and the new formulation. The results provide evidence that PolyDPG can indeed be a useful tool to carry out simulations that successfully capture large deformations in hyperelastic materials in 3D.


LCSH Subject Headings