problems in elasticity by the framework method, J . Appl. Mech. 6, 169— 175 (1941). 2. R. Courant, Variational methods for the solution of problems of equilibrium and vibrations, Bulletin of American Mathematical Society, 49, 1-23 (1943). Alexander Hrennikoff Richard Courant John Argyris Finite Element Method was developed for solution of mechanical problems, and applied for the first time in aeroengineering in 1943- 1944 by John Argyris.
Δu(x)v(x)dx=∫ Ω f (x)v(x)dx v(x) – test function (arbitrary function on Ω) u(x) – trial function Weak form of boundary value problem (Now u should have only the first derivative) ∫ Ω ∇ u ∇ v dx−∫ ∂ Ω ∂u ∂n v ds=∫ Ω f vdx
v dx−∫ ∂ Ω ∂u ∂n v ds=∫ Ω f vdx Boundary conditions: Dirichlet BC: Neumann and Robin BC are specified in the weak form: u(x∈∂Ω)=u 0 v(x∈∂Ω)=0. ∫ Ω ∇ u ∇ v dx−∫ Γ R p(u−q)vds=∫ Ω f vdx ∂u ∂n =p(u−q), x∈Γ R
v dx−∫ ∂ Ω ∂u ∂n v ds=∫ Ω f vdx Discretization: 1. Meshing – dividing the simulation domain into elements (typically triangular). 2. Finite-dimensional test and trial spaces, defined over the mesh. Trial space Test space FD test space U={v∈H1(Ω):v=u 0 on Γ D } V={v∈H1(Ω):v=0 on Γ D } V h ⊂V ∫ Ω ∇ u h ∇ v h dx−∫ ∂ Ω ∂u h ∂n v h ds=∫ Ω f v h dx By testing this relation with basis functions of V h space we obtain a system of linear equations
v dx−∫ ∂ Ω ∂u ∂n v ds=∫ Ω f vdx Discretization: 1. Meshing – dividing the simulation domain into elements (typically triangular). 2. Finite-dimensional test and trial spaces, defined over the mesh. Trial space Test space FD test space U={v∈H1(Ω):v=u 0 on Γ D } V={v∈H1(Ω):v=0 on Γ D } V h ⊂V ∫ Ω ∇ u h ∇ v h dx−∫ ∂ Ω ∂u h ∂n v h ds=∫ Ω f v h dx By testing this relation with basis functions of V h space we obtain a system of linear equations Important! To find a gradient of the solution another weak problem should be solved u gr is a trial gradient function, u is the solution obtained previously ∫ Ω u gr vdx=∫ Ω ∇ uv dx
from points to triangles For a long type mesh construction was a manual operation Making best meshes since 1934 ! No thin triangles: Delaunay triangulation maximizes minimum angle in triangles. P. Cignoni, C. Montani, R. Scopigno (1998). "DeWall: A fast divide and conquer Delaunay triangulation algorithm in Ed". Computer-Aided Design. 30 (5), 333–341(1998).
(GNU GPL) mesh builder with GUI (gmsh.info) C. Geuzaine and J.-F. Remacle. Gmsh: a three-dimensional finite element mesh generator with built-in pre- and post-processing facilities. International Journal for Numerical Methods in Engineering 79(11), 1309-1331 (2009).
perfectly matched layer for the Absorption of electromagnetic waves, Journal of Computational Physics 114, 185-200 (1994). Oscillating solutions decay as 1/r(d-1)/2, d is the number of dimensions. It is slow. We need to find an absorbing boundary condition! Let's look for an absorbing layer instead of a boundary condition!
material properties of PML? σ x is the conductivity (both electric and “magnetic”, because we do not need to restrict ourselves to real materials in simulations) σ y = 0 which means than our material is an anisotropic absorber (uniaxial). PML does not influence evanescent waves exp(-γx). We can increase decay if let σ x be a complex value. ∂ ∂ x → 1 1+i σx (x) ω ∂ ∂ x
N. Wells et al., Automated Solution of Differential Equations by the Finite Element Method, Springer, (2012) aka The FeniCS book NASA Technical Paper 3485, C. J. Reddy et al., Finite element method for eigenvalue problems in electromagnetics (1994) http://www.eecs.wsu.edu/~schneidj/ufdtd/chap11.pdf