M.

Vable Notes for finite element method: Intro to FEM 2-D

FEM in two-dimension
1
Strain energy density: U o = --- [ σ xx ε xx + σ yy ε yy + τ xy γ xy ]
2

⎧ σ xx ⎫ ⎧ ε xx ⎫
⎪ ⎪ ⎪ ⎪ 1 T
Define: { σ̃ } = ⎨ σ yy ⎬ { ε̃ } = ⎨ ε yy ⎬ U o = --- { σ˜ } { ε˜ }
⎪ ⎪ ⎪ ⎪ 2
τ
⎩ xy ⎭ γ
⎩ xy ⎭
Generalized Hooke’s law
Plane stress (All stresses with subscript z are zero)
Plane strain (All strains with subscript z are zero)
Plane Stress Plane Strain
[ ε xx + νε yy ]
σ xx = E ------------------------------- [ ( 1 – ν )ε xx + νε yy ]
2 σ xx = E ------------------------------------------------
(1 – ν ) ( 1 + ν ) ( 1 – 2ν )
[ ε yy + νε xx ] [ ( 1 – ν )ε yy + νε xx ]
σ yy = E ------------------------------- σ yy = E ------------------------------------------------
2 ( 1 + ν ) ( 1 – 2ν )
(1 – ν )
E
E τ xy = --------------------γ xy
τ xy = --------------------γ xy 2(1 + ν)
2(1 + ν)
Plane Stress Plane Strain
1 ν 0 (1 – ν) ν 0
- ν 1 ν (1 – ν)
E 0 E 0
[ Ẽ ] = ------------------ [ Ẽ ] = --------------------------------------
(1 – ν )
2
(1 – ν) ( 1 – 2ν ) ( 1 + ν ) ( 1 – 2ν )
0 0 ---------------- 0 0 --------------------
2 2
T
{ σ˜ } = [ Ẽ ] { ε˜ } [ Ẽ ] = [ Ẽ ]
2
E E ⁄ (1 – ν )
ν ν ⁄ (1 – ν)
1 T 1 ˜ T T 1 T
U o = --- { σ̃ } { ε } = --- { ε } [ Ẽ ] { ε˜ } = --- { ε˜ } [ Ẽ ] { ε˜ }
˜
2 2 2

8-1
M. Vable Notes for finite element method: Intro to FEM 2-D

Strain-Displacement
∂u ∂v ∂u ∂v
ε xx = ε yy = γ xy = +
∂x ∂y ∂y ∂x
∂
0
∂x
⎧u ⎫
{ε} = 0 ∂ ⎨ ⎬
∂y ⎩v ⎭
∂ ∂
∂y ∂x
Displacements approximating:
n n
(e) (e)
u(x) = ∑ u i f i ( x, y ) v(x) = ∑ v i f i ( x, y )
i=1 i=1
⎧ (e) ⎫
⎪ u1 ⎪
⎪ ⎪
⎪ (e) ⎪
⎪ v1 ⎪
⎪ (e)
⎪
⎪ u2 ⎪
⎪ ⎪
⎪ (e) ⎪
Define the nodal displacement vector as: { d } = ⎨ v2 ⎬
⎪ ⎪
⎪ • ⎪
⎪ ⎪
⎪ • ⎪
⎪ (e) ⎪
⎪ un ⎪
⎪ ⎪
⎪ (e) ⎪
⎩ vn ⎭

⎧u ⎫ f1 0 f2 0 … … fn 0
⎨ ⎬ = {d}
⎩ ⎭
v 0 f1 0 f2 … … 0 fn

8-2
M. Vable Notes for finite element method: Intro to FEM 2-D

∂
0
∂x
f1 0 f2 0 … … fn 0
{ε} = 0 ∂ {d} = [B]{d}
∂ y 0 f1 0 f2 … … 0 fn
∂ ∂
∂y ∂x

∂f 1 ∂f 2 ∂f n
0 0 … … 0
∂x ∂x ∂x
[B] = ∂f 1 ∂f 2 ∂f n
0 0 … … 0
∂y ∂y ∂y
∂f 1 ∂f 1 ∂f 2 ∂f 2 ∂f n ∂f n
∂y ∂x ∂y ∂x ∂y ∂x
• Matrix [B] is called strain-displacement matrix
(e) 1 T T
U o = --- { d } { B } [ Ẽ ] { B } { d }
2
(e) (e)
Strain Energy:U = ∫ Uo dV
V
(e) 1 T T 1 T (e)
U = ∫2
--
- { d } { B } [ Ẽ ] { B } { d } dV = --
2
- { d } [K ]{d}
V
(e) T
Element stiffness matrix: [ K ] = ∫ [ B ] [ E ] [ B ] dV
V
(e) T (e) e
Variation in potential energy: δΩ = { δd } ( [ K ]{d} – {R })

8-3
M. Vable Notes for finite element method: Intro to FEM 2-D

Overview of approximate methods

Some Jargon
1-D Heat Conduction:
x
2 L
d T = 0 0≤x≤L Differential Equation
2
dx
– k dT = 0 Natural Boundary Condition
dx
x=0
T = T
x=L o Essential Boundary Condition

Beam Bending
py

L
2 P
d ⎛ d v⎞
2
2⎝
⎜ EI 2 ⎟ = p y 0≤x≤L Differential Equation
dx dx ⎠
2
d ⎛ d v⎞
V = – ⎜ EI 2 ⎟ = P
dx⎝
dx ⎠ Natural Boundary Condition
2
d v
M = EI = 0
2
dx
v(0) = 0
dv Essential Boundary Condition
= 0
dx
x=0

8-4
M. Vable Notes for finite element method: Intro to FEM 2-D

Approximation of boundary value problem

Γe
Γn
Ω

Lu = f in Ω Differential Equation

D n u = g n on Γe Natural Boundary Condition

D e u = g e on Γn Essential Boundary Condition

n
φj set of approximating functions
u = ∑ cj φj set of φj is complete and independent.
j=1

n
ed = ∑ cj Lφj – f Error in Differential Equation
j=1
n
en = ∑ cj Dn φj – gn Error in Natural Boundary Condition
j=1
n
ee = ∑ cj De φj – ge Error in Essential Boundary Condition
j=1

8-5
M. Vable Notes for finite element method: Intro to FEM 2-D

Commonality and Differences in

Approximate Methods
Commonalities

• Produce a set of algebraic equations in the unknown constants cj.

• Choose φi to set one (or two) of the errors ed, ee, or en to zero
• Minimize the remaining error(s).
Differences

• Which error is set to zero

Domain Methods: ee =0 or en =0
Boundary Methods: ed=0
• Error Minimizing Process
Independence of φi

φi can be obtained from a linear combination of other φi’s in

• No
the set.
• If the set of functions φi are not independent then the equations in
the matrix will not be independent and the matrix will be singular.
Completeness of φi

• In a series sequence no term should be skipped.

• If a set is not complete then the solution may not converge for
some problems.

8-6
M. Vable Notes for finite element method: Intro to FEM 2-D

Error Minimization
Weighted Residue

(d) (e) (n)

∫ ∫ ψi ed dx dy + ∫ ψ i e e ds + ∫ ψ i e n ds = 0
Ω Γe Γn
FEM-Stiffness version: ee = 0

(d) (n)
∫ ∫ ψi ed dx dy + ∫ ψ i e n ds = 0
Ω Γn
FEM-Flexibility version: en = 0

(d) (e)
∫ ∫ ψi ed dx dy + ∫ ψ i e e ds = 0
Ω Γe
BEM: ed = 0

(n) (e)
∫ ψi e n ds + ∫ ψi e e ds = 0
Γn Γe
FEM: Discretization process is on domain of the entire body Ω
BEM: Discretization process is on the boundary of the body Γ

• In FEM stiffness matrix the equilibrium equation on stresses (differen-

tial equations) and boundary conditions on stresses (natural boundary
conditions) are approximately satisfied.
• In FEM flexibility matrix the compatibility equation on stresses (dif-
ferential equations) and boundary conditions on displacements (essen-
tial boundary conditions) are approximately satisfied.

8-7
M. Vable Notes for finite element method: Intro to FEM 2-D

Constant Strain Triangle (CST)

• Displacements are linear in x and y, resulting in constant strains.
u = a0 + a1 x + a2 y v = b0 + b1 x + b2 y
∂u ∂v ∂u ∂v
ε xx = = a1 ε yy = = b2 γ xy = + = a2 + b1
∂x ∂y ∂y ∂x
(e)
y (e)
3 ⎧ u1 ⎫
v3 (e) (e) ⎪ ⎪
(e)
u3
v2 u ( x, y ) = ∑ N i ( x, y )u i ⎪ (e) ⎪
3 i=1 ⎪ v1 ⎪
(e) 2 (e)
⎪ ⎪
v1 u2 ⎪ u( e ) ⎪
(e) ⎪ 2 ⎪
1 (e) 3 { d } = ⎨ ⎬
u1 x (
⎪v ⎪ e )
(e)
v ( x, y ) = ∑ N i ( x, y )v i ⎪ 2 ⎪
⎪ (e) ⎪
i=1 ⎪ u3 ⎪
⎪ ⎪
⎪ (e) ⎪
⎩ v3 ⎭

⎧ ε xx ⎫ y 23 0 y 31 0 y 12 0
⎪ ⎪ 1 (e) (e)
⎨ ε yy ⎬ = ------
2A
- 0 x 32 x 13 x 21 { d } = [ B ] { d }
⎪ ⎪
⎩ γ xy ⎭ x 32 y 23 x 13 y 31 x 21 y 12

x ij = x i – x j y ij = y i – y j

(e) T T
[K ] = ∫ [ B ] [ E ] [ B ] dV = [ B ] [ E ] [ B ]tA
V
3 1 N3 1
2 3
2 3
1 N2 2
1
N1 1
1

8-8
M. Vable Notes for finite element method: Intro to FEM 2-D

Pascal’s Triangle
• Used for determining a complete set of polynomial terms in two
dimensions.‘
1

x y

x2 xy y2

x3 x2 y xy2 y3

x4 x3y x2y2 xy3 y4

x 5 x4y x3y2 x2y3 xy4 y5

• Greater the degree of freedom, less stiff will be element.

• Interpolation functions are easier to develop with area
coordinates.

8-9
M. Vable Notes for finite element method: Intro to FEM 2-D

Natural Coordinates
• Coordinates which vary between 0 and 1 or -1 and 1.
• Natural coordinates and non-dimensional coordinates.
1-d Coordinates
x
x1 Node 1 Node 2
x2

⎛ x – x2 ⎞ ⎛ x – x1 ⎞
L 1 ( x ) = ⎜ -----------------⎟ L 2 ( x ) = ⎜ -----------------⎟
⎝ x 1 – x 2⎠ ⎝ x 2 – x 1⎠
Possibility 1 Possibility 2

ξ=1 ξ=−1 ξ
ξ ξ=1

Node 1 Node 2 Node 1 Node 2

L1 ( ξ ) = ( 1 – ξ ) L1 ( ξ ) = ( 1 – ξ ) ⁄ 2
L2 ( ξ ) = ξ L2 ( ξ ) = ( 1 + ξ ) ⁄ 2
2-D Triangular elements (Area Coordinates)
L=
I 0
K LJ
LK AI
L I = ------
AI A
AJ J AJ
L J = ------
AK A
AK
L K = --------
L= A
I 1
LI + LJ + LK = 1
I

8-10
M. Vable Notes for finite element method: Intro to FEM 2-D

Linear Strain Triangle

5 K

J 3
6
2

1
I

LI ( LI – 1 ⁄ 2 )
N 1 = -------------------------------- = 2L I ⎛ L I – ---⎞
1
(1)(1 ⁄ 2) ⎝ 2⎠
LI LJ
N 2 = ------------------------------ = 4L I L J
(1 ⁄ 2)(1 ⁄ 2)
Homework Problem: Write cubic interpolation functions using area coor-
dinates for nodes 1,2 and 10.
7 K
6
5
8
J 4
10
9
3

1 2
I

8-11
M. Vable Notes for finite element method: Intro to FEM 2-D

Bi-Linear Quadrilateral
u = a 0 + a 1 x + a 2 y + a 3 xy v = b 0 + b 1 x + b 2 y + b 3 xy
∂u ∂v
ε xx = = a1 + a3 y ε yy = = b2 + b3 x
∂x ∂y
∂u ∂v
γ xy = + = ( a2 + b1 ) + a3 x + b3 y
∂y ∂x

y
(e) (e)
v3 v4
(e) a a (e)
u3 u4
4
3 4 (e)
b u ( x, y ) = ∑ N i ( x, y )u i
x i=1
4
(e)
v ( x, y ) = ∑ N i ( x, y )v i
b
(e) (e)
u1 1 2 u2
i=1
(e) (e)
v1 v2
Interpolation functions in natural coordinates

ξ = x⁄a η = y⁄b
1 1
N 1 = --- ( 1 – ξ ) ( 1 – η ) N 2 = --- ( 1 + ξ ) ( 1 – η )
4 4
1 1
N 3 = --- ( 1 – ξ ) ( 1 + η ) N 4 = --- ( 1 + ξ ) ( 1 + η )
4 4
∂N i ∂N i ∂ξ 1 ∂N i ∂N i ∂N i ∂η 1 ∂N i
= = --- = = ---
∂x ∂ξ ∂x a ∂ξ ∂y ∂η ∂y b∂η

8-12
M. Vable Notes for finite element method: Intro to FEM 2-D

Other Quadrilaterals
Complete quadratic
y

a a

7 8 9
b
6 x
4 5

1 2 3

• The stiffness (row and column) related to node 5 is known at

element level and as rows and columns of other elements do not
add to it.
Quadratic element often used in practice:
y

a a

6 7 8
b
5 x
4

1 2 3

• When internal nodes are eliminated care has to be exercised to

ensure the mesh from such elements will converge.

8-13
M. Vable Notes for finite element method: Intro to FEM 2-D

Mechanical Loads
There are three types of mechanical loads
1. Concentrated Forces or Moments
• The loads must be applied at nodes when making the mesh.
• Theoretically the stresses are infinite at the point of application, hence
in the neighborhood of concentrated load a large stress gradient can be
anticipated.
2. Tractions
• Forces that act on the bounding surfaces.
S x = σ xx n x + τ xy n y S y = τ yx n x + σ yy n y
nx and ny are the direction cosines of the unit normal
• Tractions has units of force per unit area and are distributed forces.
• Usually the tractions are specified in local normal and tangential coor-
dinates.
• These distributed forces must be converted to nodal forces.
In two dimensions the bounding surface is a curve. Distributed forces can
be converted to nodal forces as was done in 1-d axial and bending prob-
L

lems. (work equivalency) Rj = ∫ p ( x )fj ( x ) dx

0
p pL/2 2pL/3
pL/2 pL/6 pL/6

Linear Lagrange Quadratic Lagrange

L

3. Body Forces
• Forces that act at each and every point on the body.
• Gravity, magnetic, inertial are some examples.
• These forces must be converted to nodal forces.
(See section 3.9)

8-14