1: \begin{abstract}

2: The finite element methods (FEM) are important techniques in engineering

3: for solving partial differential equations, but they depend heavily

4: on element shape quality for stability and good performance. In this

5: paper, we introduce the \emph{Adaptive Extended Stencil Finite Element

6: Method} (AES-FEM) as a means for overcoming this dependence on element

7: shape quality. Our method replaces the traditional basis functions

8: with a set of \emph{generalized Lagrange polynomial (GLP) basis functions},

9: which we construct using local weighted least-squares approximations.

10: The method preserves the theoretical framework of FEM, and allows

11: imposing essential boundary conditions and integrating the stiffness

12: matrix in the same way as the classical FEM. In addition, AES-FEM

13: can use higher-degree polynomial basis functions than the classical

14: FEM, while virtually preserving the sparsity pattern of the stiffness

15: matrix. We describe the formulation and implementation of AES-FEM,

16: and analyze its consistency and stability. We present numerical experiments

17: in both 2D and 3D for the Poisson equation and a time-independent

18: convection-diffusion equation. The numerical results demonstrate that

19: AES-FEM is more accurate than linear FEM, is also more efficient than

20: linear FEM in terms of error versus runtime, and enables much better

21: stability and faster convergence of iterative solvers than linear

22: FEM over poor-quality meshes.\\

23: Key Words: finite element methods; mesh quality; weighted least squares;

24: partition of unity; accuracy; stability

25: \end{abstract}

26: