Lecture

13: 2D Problems using CST

AML706 Finite Element Methods

 Two-dimensional Problems using

Constant Strain Triangles
•  To formulate 2D problems we will follow similar steps as in
the case of 1D FE modeling.
•  Here the displacements, tracJon, distributed body force are
funcJons of the posiJon given by (x,y)
•  The displacement vector is given by u=[u v]T
•  Where u and v are x and y components of u
•  Stresses and strains are :
T T
σ = [σ x , σ y , τ xy ] ε = [ε x , ε y , γ xy ]
•  The body force and tracJon vectors are:
f = [ fx , fy ]T T = [Tx ,Ty ]T
 FE Modeling of 2D Problem

•  For elemental volume calculaJon, we have

V = tdA
•  Where t is the thickness normal to element (along z direcJon)
•  Units of body force : force per unit volume and that of
tracJon is force per unit area.
•  The strain displacement relaJon is
T
⎡ ∂u ∂v ⎛ ∂u ∂v ⎞⎤
ε = ⎢ , , ⎜ + ⎟⎥
⎣ ∂x ∂y ⎝ ∂y ∂x ⎠⎦
•  The stress strain relaJon is σ = D
ε where D is the material
matrix (plane stress)
 2D FE Modeling


•  Two dimensional problems discussed here have two degrees
of freedom at each node as follows
•  The global displacement vector is y
Q2i
T
Q = [Q1 , Q2 , Q3 ....QN ]
x
where N is the total number of dof i Q2i-1

•  The given 2D domain is triangulated first and the

triangulaJon data is stored as nodal coordinates and
connecJvity informaJon.
•  The coordinates are stored in an NNx2 matrix where
NN is total number of nodes in the triangulaJon.
 Triangulated 2D domains


•  Here we show two examples
of triangular meshes. First one
straight edges which is tria
-ngulated evenly. The other
domain with curved boundary

 2D Element ConnecJvity


•  Considering an element, the connecJvity of a typical element
in the triangular mesh is shown here
q6

3 q5
•  The element displacement vector
is shown here as v q4
(x,y) u q3
q = [q1, q2 , q3, q4, q5, q6 ]T
2
q2
•  A typical connecJvity table is given as
1 q1

1 2 3
Element no./ local nodes The element displacement
1 1 2 3 vector q can be extracted
2 3 2 6 from global Q using
3 2 6 8 connecJvity table
: : : :
N
 The Local and Global Connect


•  The nodal coordinates shown here as (x1,y1),(x2,y2) and (x3,y3)
have a global correspondence through the connecJvity table
•  The local represent-
aJon of nodal T
coordinates and k = L k 'L
degrees of freedom
is a way of clearly
represenJng the
element characteri-
sJcs.
 Constant Strain Triangle


•  Recall the shape funcJons used to interpolate the nodal
displacements in 1D problems. Here in 2D problems we
determine the displacements inside an element from the 3
nodal displacements using linear shape funcJons.

•  By inspecJon we see that N1+N2+N3=1. This means that they

are not linearly independent.
 2D Shape funcJons


•  The plot of third funcJon is given here
•  Note that the funcJon N1 is 1 at
node 1 and 0 at nodes 2 and 3.
•  Similarly N2 and N3 also have
their values as 1 at nodes 2 and 3
respecJvely and 0 at other nodes.
•  RepresenJng the independent
shape funcJons by ξ and η, we have
N1 = ξ N 2 = η and N 3 =1− ξ − η
•  Here ξ and η are called the natural coordinates.
 2D Shape funcJons


•  Now we look at the analogy of 1D and 2D shape funcJons
•  In 1D case, x-coordinate mapped onto ξ –coordinate and
shape funcJon is a funcJon of ξ . Here in 2D problems, x, y
coordinates map onto ξ , η and the shape funcJon is a
funcJon of ξ and η.
•  The shape funcJons are now represented by area
coordinates. A point P(x,y) divides the triangle into three
areas A1, A2 and A3 as shown. Now express
the shape funcJons as area raJos.
A1 A2 A3
N1 + N 2 + N 3 = + + =1
A A A
•  For every point inside element N1+N2+N3=1
 Isoparametric RepresentaJon


•  Now displacements inside the element are expressed in terms
of the nodal values of the unknown displacement field.
u = N1 (ξ , η )q1 + N 2 (ξ , η )q3 + N 3 (ξ , η )q5
v = N1 (ξ , η )q2 + N 2 (ξ , η )q4 + N 3 (ξ , η )q6
•  Now using the definiJons of shape funcJons
N1 = ξ N 2 = η and N 3 =1− ξ − η
•  We write the displacements u,v as
u = (q1 − q5 )ξ + (q3 − q5 )η + q5
v = (q2 − q6 )ξ + (q4 − q6 )η + q6
•  As in 1D we represent N as a matrix now.
 Isoparametric RepresentaJon


•  As in 1D we represent N as a matrix now.
⎡ N 0 N 0 N 0 ⎤
1 2 3
N =⎢ ⎥
⎢⎣ 0 N1 0 N 2 0 N 3 ⎥⎦
•  The displacements in matrix form {u} = [ N ] {q}
•  We know that for an isoparametric representaJon the
coordinates inside the triangular element also can be
represented in terms of the nodal coordinates as:
x = N1 (ξ , η )x1 + N 2 (ξ , η )x2 + N 3 (ξ , η )x3

y = N1 (ξ , η )y1 + N 2 (ξ , η )y2 + N 3 (ξ , η )y3

•  OR x = (x1 − x3 )ξ + (x2 − x3 )η + x3

y = (y1 − y3 )ξ + (y2 − y3 )η + y3
 Isoparametric RepresentaJon


•  The (x,y) coordinates inside the elements are wrihen as
x = (x1 − x3 )ξ + (x2 − x3 )η + x3

y = (y − y )ξ + (y − y )η + y3
1 3 2 3

= ( xi − x )j )
By changing the notaJon as xij

x = x13ξ + x23η + x3
y = y13ξ + y23η + y3