7 Triangular Elements

Here a number of triangular elements are discussed. Two linear elements, the standard linear
element and the nonconforming linear element are discussed in quite some detail. The
quadratic triangular element is briefly introduced towards the end.

7.1 The Linear Triangular Element

The most basic type of triangular element is the linear element, with three nodes at the
vertices, for which the shape functions vary linearly. The shape functions for this element can
be constructed as follows: consider a triangle with vertices

( x1 , y1 ), ( x 2 , y 2 ), ( x3 , y 3 )
( x2 , y 2 )

The first shape function, N 1 ( x, y ) , is a linear function with

( x3 , y3 )
value 1 at the first node and zero at the other two, so

a1 x1  b1 y1  c1  1
( x1 , y1 ) N 1 ( x, y )  a1 x  b1 y  c1 a1 x 2  b1 y 2  c1  0 (7.1)
a1 x3  b1 y 3  c1  0

which can be solved to get

 y 2  y3 x  x3  x2  y  x2 y3  x3 y 2 
N 1 ( x, y )  (7.2)
2

where  is the area of the triangle,

2  x1  y 2  y 3   x 2  y 3  y1   x3  y1  y 2 
  x1  x 2  y 2  y 3    x 2  x3  y1  y 2 
, (7.3)
  x 2  x3  y 3  y1   x3  x1  y 2  y 3 
  x3  x1  y1  y 2    x1  x 2  y 3  y1 

185
and the other two shape functions can be determined similarly. Note that the nodes should be
numbered counterclockwise around the triangle, as is
done here, so that the area  is positive. ( x2 , y 2 )

These shape functions are also called area ( x3 , y3 ) A1 ( x, y )

coordinates, because they are the ratios of the areas A  A3
2
shown in the figure here to the total area of the
triangle: for example, N 1  A1 /  (and clearly ( x1 , y1 )
N1  N 2  N 3  1 ).

7.1.1 Local Coordinates

The local coordinates appropriate to this triangular element are

x  x1  y3  y1    y  y1 x3  x1  x  x1  y3  y1    y  y1 x3  x1 
 
x2  x1  y3  y1    y 2  y1 x3  x1  2
(7.4)
x  x  y  y1    y 2  y1 x  x1  x2  x1  y  y1    y 2  y1 x  x1 
 2 1 
x2  x1  y3  y1    y 2  y1 x3  x1  2


( x2 , y 2 )
(0,1) 3
2
3
( x3 , y3 )

1 1 2
( x1 , y1 ) (0,0) (1,0) 

global local
coordinates coordinates

186
With these definitions, the local coordinates of the three nodes ( x1 , y1 ), ( x 2 , y 2 ), ( x 3 , y 3 ) are
(0,0) , (1,0) and (0,1) .

The inverse of these relations are

4
x( , )   N i ( , ) xi
i 1
4
(7.5)
y ( , )   N i ( , ) y i
i 1

and the shape functions in terms of the local coordinates are

Shape Functions for the Linear Triangular Element:

N1  1    
N2   (7.6)
N3  

7.1.2 Transformation into Local Coordinates

For the transformation from the physical coordinate system into the local coordinate system,
one again uses the transformation matrices (6.7). Again, as in (6.9), one can write

 3 N i N i 
i 1  xi 
3
yi
i 1
 
J   Nd x (7.7)
 3 N i x 3 N i

i 1  i i1  yi 


with the sum now over three nodes, and with

 N i   x1 y1 
    1  1 0 
Nd     , x   x 2 y 2  (7.8)
 N
 i   1 0  1
    x3 y 3 

187
One also has the Jacobian of the transformation J  J  N d x and the relation between the
derivatives in local and physical coordinates (as in 6.11):

 N i   N i 
 x  
1 

 N   J    J 1 N d
 i
 
 iN
 y    

Because of the simplicity of the element, these quantities can be evaluated in closed form:

 x  x1 y 2  y1  1  y 3  y1 y1  y 2 
J 2 , J 1  , J  2 (7.9)
 x3  x1 y 3  y1  
2  x1  x3 x 2  x1 

and
 N i 
 x  1  y  y y 3  y1 y1  y 2 
 N   2 3
 x3  x 2 x1  x3 x 2  x1 
(7.10)
 i
 2 
 y 

7.1.3 Derivatives of the Shape Functions

An interesting property of the linear triangular element is that

Li
N i   ni (7.11)
2

where Li is the length of the element edge opposite node/vertex i and n i is the unit normal to
that edge. This is a constant vector over the element.
ey
For example, from the figure, with 3 n1
 

e x  n1 cos   n sin  y3  y2 ex
L1
e y  n1 sin   n cos 
x2  x3 2

1
188
where n is perpendicular to n1 , then

N 1 N
N 1  ex  1 ey
x y


1
2
 y 2  y3 e x  x3  x2 e y 
L
  1 n1
2

Once the values of p have been obtained at the three nodes, the gradient can be evaluated
through

n
p   p i N i
i 1
(7.12)
1   y 2  y3   y  y1   y  y2 
  p1    p2  3   p3  1 
2   x 3  x 2   x1  x3   x 2  x1  

or

1
p   p1 L1n1  p 2 L2 n 2  p3 L3n 3  (7.13)
2

which is a constant vector.

The gradients normal to the element edges are then p  n i , where

n1 
1
L1

 y3  y 2 e x  x2  x3 e y 
n2 
1
L2

 y1  y3 e x  x3  x1 e y  (7.14)

n3 
1
L3

 y 2  y1 e x  x1  x2 e y 

189
The gradient can also be conveniently written in terms of a shifted global coordinate system,
where the vertex 1 is position at position x1 , y1  . In terms of these coordinates,

1
 x 2 y 3
2

and
 1 1 
   0   p1 
x 2 x 2
p    p  (7.15)
 1  x3 x 1  2 
 3  p 
 y 3 x 2 y 3 x 2 y 3 y 3   3 

3
x3 , y3 

1 0,0  x2 ,0 

7.1.4 Some Integrals

The integral of a function over an element is given by

1 1

 f x, y dS  J   f ( , )dd

E 0 0
(7.16)

Some integrals involving the shape functions are {▲Problem 1}

2 1 1 
  
E N j dS  3 , j  1, 2, 3 , E N i N j dS  12 1 2 1 (7.17)
1 1 2

190
7.1.5 The Boundary Integrals

Looking now at the boundary integral, one has

1 1 1 1

 dC1  L1  d ,
1 2 0
 dC 2  L2  d  L2  d ,
2 3 0 0
 dC3  L3  d
31 0
(7.18)

where L1 is the length of the line joining nodes 1 and 2, etc.

The shape functions along the three edges are given in the following table:

N1 N2 N3
C1 1  0
C2 0  (or 1   ) 1   (or  )
C3 1  0 
Table 7.1: Shape Functions for the Linear Element along Element Edges

For natural boundary conditions of the type

p
 A, (7.19)
n

were A is a constant, one obtains the boundary vectors

1  0  1
L1   L2   L3  
C N j dC1  2 1, C N j dC 2  2 1, C N j dC3  2 0 (7.20)
1
0 2
1 3
1

7.2 Numerical Integration

The numerical integration rule for triangular regions takes the form

191
 1 1

  f ( , )dd  
N
i 1
Wi f ( i , i ) (7.21)
0 0

where the weights Wi and integration points si are given in the table below. The rule
integrates polynomials of the order r exactly using an N – point rule.

N i i Wi r
1 1 1
1 3 3 2 1
3
1
2 , ,0
1
2
1
2 , 0, 1
2
1
6 , ,
1
6
1
6 2
Table 7.2: Quadrature for triangles

7.3 The Nonconforming Linear Triangle

The non-conforming triangular element has the following shape functions:

N 1  1  2  2
N 2  1  2 (7.22)
N 3  1  2

( x1 , y1 ) (0,1) 3
2  N 1  1 
3   N   0
1 3  N 1  0   2  
  N   1 
 2    2   N 3  0
2 ( x3 , y3 ) 1
  N 3  0
( x2 , y 2 ) 3 2
1 1 
(0,0)  N1  0 
 N   0 
(1,0) 
 2  
 N 3  1 

global local
coordinates coordinates

192
If the shape functions of the standard, conforming, element are denoted by N i , then

N i  1  2N i (7.23)

The equations

3
x( , )   N i ( , ) xi
i 1
3
(7.24)
y ( , )   N i ( , ) y i
i 1

still hold only the coordinates xi , y i refer to the nodes which are now mid-way along element
edges, not at the vertices.

For this element one has

 N i   x1 y1 
    1  1 0 
Nd     2  , x   x 2 y 2  (7.25)
 N i   1 0  1  x3
   y 3 

so that N d is  2 times the N d of the standard (conforming) triangular element. Also,

 x y   3 N i N i 
 
3
xi yi
     i 1   
i 1  x 2  x1 y 2  y1 
J    N d x  2 (7.26)
 x y   3 N i 3 N
 x3  x1 y 3  y1 
 
 x
   i 1  i
i1 i yi 


which is  2 times the J matrix of the standard (conforming) triangular element.

The inverse Jacobian matrix is

193
    
 x    1  1  y 3  y1 y1  y 2  
J 1   x     (7.27)
  2  2  x1  x3 x 2  x1  
 y y 

and the Jacobian determinant is

J  J  N d x  8 s  2  (7.28)

where  s is the area of the triangle with vertices ( xi , y i ) and  is the area of the complete
triangular element.

Also, using (6.11),

 N i 
 x  1  y 2  y3 y 3  y1 y1  y 2 
 N    x 2  x1 
(7.29)
 i  2 s  x 3  x 2 x1  x3
 y 

which is a similar expression to that of the standard (conforming) triangular element.

7.3.1 Derivatives of the Shape Functions

In this case,

Li
N i  ni (7.30)


where Li is the length of he element edge associated with node i and n i is the unit normal to
the edge.

For example, from the figure, with

194
 e x  n1 cos   n sin  ey
e y  n1 sin   n cos  n1

where n is perpendicular to n1 , then 2 1 ex
 L1 / 2
y2  y3 
N 1 N 2 
N 1  ex  ey 3
x y
x3  x2

1
2 s

 y 2  y3 e x  x3  x 2 e y 
L1
 n1


Once the values of p have been obtained, the gradient can be evaluated through

n
p   p i N i
i 1
(7.31)
1   y 2  y3   y  y1   y  y2 
  p1    p2  3   p3  1 
2 s   x3  x 2   x1  x3   x 2  x1  

or

1
p  p1 L1n1  p 2 L2 n 2  p3 L3n 3  (7.32)


which is a constant vector.

The gradients normal to the element edges are then p  n i , where

n1  
2
L1

 y3  y 2 e x  x 2  x3 e y 
n2  
2
L2

 y1  y3 e x  x3  x1 e y  (7.33)

n3  
2
L3

 y 2  y1 e x  x1  x 2 e y 

195
7.3.2 The Boundary Integrals

Looking now at the boundary integral, one has

1 1 1 1

 dC1  L1  d ,
1 2 0
 dC 2  L2  d  L2  d ,
2 3 0 0
 dC3  L3  d
31 0
(7.34)

where L1 is the length of the line joining nodes 1 and 2, etc.

The shape functions along the three edges are given in the following table:

N1 N2 N3
C1 2  1  2 1  2 1
C 2 3 1 1  2  1  2
(or  1  2 ) (or 1  2 )
C 31  1  2 1 1  2
Table 7.3: Shape Functions for the Nonconforming Element along Element Edges

For natural boundary conditions of the type

p
 A,
n

were A is a constant, one obtains the boundary vectors

0  1  0 
 N j dC1  L1 0,  N j dC 2  L2 0,  N j dC 3  L3 1 (7.35)
C1 2
1 C2  3
0 C3 1
0

7.3.3 Some Integrals

The integrals of the shape functions over an element are

196
 
N
E
j dS 
3
, i  1, 2, 3 (7.36)

The shape functions have a useful orthogonality property; when integrated over an element,

0 if i  j

E N i N j dS    if i  j
(7.37)
 3

7.4 The Quadratic Triangular Element

The quadratic triangular element has midside nodes in addition to those at the vertices, and a
function is interpolated as

p  a1  a 2 x  a3 y  a 4 x 2  a5 xy  a6 y 2 (7.38)

The shape functions are


N 1  1     1  2  2 
3 N 2   2  1

N 3   2  1
(7.39)
N 4  4 1     
(0, ) 
1
5 ( , ) 1 1
N 5  4
N 6  4 1     
2 2 2
6
1 4 2
 
( 12 ,0) 

For example, taking   0 , the non-zero shape functions are

N 1  1  3  2 2 , N 2    2 2 , N 4  4  4 2 (7.40)

The one-dimensional quadratic 3-noded element is recovered by letting   (  1) / 2 (so that

the interval is now  1,1 ).

197
7.5 Problems

1. Derive the integral relations (7.17).

198