Professional Documents
Culture Documents
The next element we develop is the 4--node (2--D) rectangle element for plane--
stress and plane--strain analyses.
4 3
1 2
x
a
115
DRAFT
u(x, y) = a1 + a2 x + a3 y + a 4 xy
v(x, y) = a5 + a6 x + a7 y + a 8 xy
1
Recall Pascal’s triangle y
x
x2 xy y2
x3 x2 y x y2 y3
116
DRAFT
4
u(x, y) = N i(x, y) u i = N iu i
i=1
4
v(x, y) = N i(x, y) v i = N iv i
i=1
v4 v3
u4 u3
4 3
v1 v2
1 2
u1 u2
If we use the polynomial shown earlier, and substitute for x and y at all corner
nodes, we can write the nodal displacements in terms of the a values as u = C a, and
inverting C to get finally the shape functions as shown below,
117
DRAFT
x − y + xy
N1 = 1 − a
b ab 1 3
x − xy
N2 = a
ab 3
4
1
xy
N3 =
ab 3
z
4
y xy
N4 = − y
b ab
118
DRAFT
These shape functions are the 2--D Lagrange polynomials of order 1. Note each N i is
non--dimensional
Properties:
4
Ni = 1
i=1
4
∂N
∂x
i
=0
i=1
4
∂N
∂y
i
=0
i=1
Suppose we know the temperature at each node of the element, we can use these
some interpolating functions to compute the temperature at a point (x,y) in the element
4
T(x, y) = Ni(x, y) Ti = N iT i
i=1
Parametric Coordinates
In preparation for later development of isoparametric elements, let’s introduce
parametric coordinates to describe the element geometry and formulation.
x and η = y .
The change of coordinate then takes the following form: ξ = a
b
Thus, x = ξ a, and y = η b.
119
DRAFT
Note that x = f(ξ) and not x = f(ξ, η), and similarly y = f(η) and not y = f(ξ, η),
4
u(ξ, η) = Ni(ξ, η) ui = N iu i
i=1
4
v(ξ, η) = N i(ξ, η) v i = N iv i
i=1
If you examine the following figure and use the element edge equations not adja-
cent to the node, you will get the shape functions in terms of the isoparametric coordi-
nate system.
η
N 1 = (1 − η)(1 −
1−η=0
4 3
1 1−ξ=0
ξ=0
1 η=0 2
ξ
120
DRAFT
N 1(ξ, η) = (1 − ξ) (1 − η)
N 2(ξ, η) = ξ (1 − η)
N 1(ξ, η) = ξ η
N 1(ξ, η) = η (1 − ξ)
4
Ni = 1
i=1
4
∂N
∂ξ
i
=0
i=1
4
∂N
∂η
i
=0
i=1
Strain Computation
∂u = ∂x ∂u + ∂y ∂u
∂ξ ∂ξ ∂x ∂ξ ∂y
∂u = ∂x ∂u + ∂y ∂u
∂η ∂η ∂x ∂η ∂y
or
121
DRAFT
∂u ∂x ∂y ∂u
∂ξ ∂ξ ∂ξ ∂x
=
∂u ∂x ∂y ∂u
∂η ∂η ∂η ∂y
Treating these as a set of 2 simultaneous equations and solve for ∂u and ∂u. We
∂x ∂y
can solve when det J > 0.
∂x ∂y
∂ξ ∂ξ a 0
J =
∂x ∂y
∂η ∂η 0 b
1
a 0
−1
[J ] =
0 1
b
Thus,
∂x
1 ∂u = 1
Á x = ∂u = a
∂ξ a ∂N i
∂ξ i
u
i=1
122
DRAFT
∂v ∂x ∂y ∂v
∂ξ ∂ξ ∂ξ ∂x
=
∂v ∂x ∂y ∂v
∂η ∂η ∂η ∂y
[J]
4
Á y = ∂v = 1 ∂v = 1
∂y b ∂η b
∂N i
∂η i
v
i=1
4 4
γ xy = ∂u + ∂v = 1
∂y ∂x b
∂N i
∂η i
u + 1
a ∂N i
∂ξ i
v
i=1 i=1
Summarizing,
123
DRAFT
Áx 1 ∂N 1 0 1 ∂N 2 0 1 ∂N 3 0 1 ∂N 4 0 u1
a ∂ξ a ∂ξ a ∂ξ a ∂ξ
v1
Áy = 0 1 ∂N 1 0 1 ∂N 2 0 1 ∂N 3 0 1 ∂N 4 u2
b ∂η b ∂η b ∂η b ∂η
v2 =
γ xy 1 ∂N 1 1 ∂N 1 1 ∂N 2 1 ∂N 2 1 ∂N 3 1 ∂N 3 1 ∂N 4 1 ∂N 4 u3
b ∂η a ∂ξ b ∂η a ∂ξ b ∂η a ∂ξ b ∂η a ∂ξ
v3
u4
v4
u(x, y) = a1 + a2 x + a3 y + a 4 xy
v(x, y) = a5 + a6 x + a7 y + a 8 xy
and switching to parametric coordinate systems does not change the order of the
polynomial. Let’s examine the strain variation,
γ xy = ∂u + ∂v = a 3 + a4 x + a6 + a8 y linear in x,
∂y ∂x
124
DRAFT
This element because the variation of Á x is linear in y, behaves much better for
bending problems.
Stiffness Computation
[K ]8×8 = [B ] T
8×3 [ D ] 3×3 [B ]3×8 dV e
Ve
x=a y=b
K
8×8
=t B(x, y) T
8×3
D
3×3
B(x, y)
3×8
dx dy
x=0 y=0
Change (x,y) to (ξ , η)
x=a ξ y=a η
dx = a dξ dy = a dη
Integration limits:
Substituting,
125
DRAFT
1 1
K
8×8
In other words,
the implication is that once you calculate K in symbolic form in terms of B and D,
the computations will be very efficient. The D matrix for plane stress is
1 ν 0
E ν 1 0
[D ] =
(1 − ν 2)
(1 − ν)
0 0
2
1−ν ν 0
[D ] = E ν 1−ν 0
(1 + ν)(1 − 2ν)
(1 − 2ν)
0 0
2
126
DRAFT
127
DRAFT
a a
4 7 3
b
6 PSR − 8
8
b
1 5 2
a1
u(x, y) = 1 x y x 2 xy y 2 x 2y xy 2
a2
a3
a4
a5
a6
a7
a8
Similarly,
128
DRAFT
a9
V(x, y) = 1 x y x 2 xy y 2 x 2y xy 2
a 10
a 11
a 12
a 13
a 14
a 15
a 16
u1 a1
u2 a2
u3 a3
u4 a4
C
u5
= a5
u6 a6
u7 a7
u8 a8
u~r = [C ] ~a
−1 ~
a = [C ]
~
ur
129
DRAFT
u(x, y) = N{a }
{u e} = [C ]{a }
−1
u(x, y) = N[ C ] {u e}
1 −a −b a2 ab b2 − a 2b − a 2b a 2b
1 a −b a2 − ab b 2 − a 2b a 2b a 2b
1 a b a2 ab b2 a 2b a 2b a 2b
[C ] =
1 −a b a2 − ab b 2 a 2b − a 2b a 2b
1 0 −b 0 0 b2 0 0 0
1 a 0 a2 0 0 0 0 0
1 0 b 0 0 b2 0 0 0
1 −a 0 a2 0 0 0 0 0
1 0 0 0 0 0 0 0 0
Inverting [C ] using ,
−1
[ N ] = N[C ]
130
DRAFT
In explicit form,
xy(a − x)(b − y)
N1 =
4a 2b 2
− xy(a + x)(b − y)
N2 =
4a 2b 2
xy(a + x)(b + y)
N3 =
4a 2b 2
− xy(a − x)(b + y)
N4 =
4a 2b 2
− (a 2 − x 2)y(b − y)
N5 =
2a 2b 2
x(a + x)(b 2 − y 2)
N6 =
2a 2b 2
(a 2 − x 2)y(b + y)
N7 =
2a 2b 2
− x(a − x)(b 2 − y 2)
N8 =
2a 2b 2
(a 2 − x 2)(b 2 − y 2)
N9 =
a 2b 2
131