1 Line Elements

The first family of elements we will consider are one-dimensional, or line elements. Within this
group, we will look in detail at:
• Bar elements, which can only carry compressive and tensile loads.
• Beam elements, which carry compressive, tensile, and bending loads.
Line elements can experience a number of different forces:
Body forces act throughout the structure, and are often used to represent the effect of gravity,
or some other form of acceleration.
Traction forces are also distributed forces, but act only on the surface of the elements.
Point loads act only at discrete locations. Point loads are usually assumed to act directly on
a node within the structure.
Point loads are the easiest forces to model in an FE system, so we will focus primarily on them.

1.1 Bar Elements

Consider the simple two bar element system shown below:

L1 L2

F
u1 1 u2 2 u3

A2
A1
x, u

As shown, each element has only two nodes (some element types can have more). The longi-
tudinal displacement of points along the structure’s length from their initial unloaded position
is clearly a continuous function of x: u(x). Using this displacement function, it is possible to
determine the strain within the structure:
du
= (1.1)
dx
and hence also the stress:
σ = E (1.2)

3
1. Line Elements 1.1 Bar Elements

Instead of trying to find the continuous displacement function, we will attempt to find the
displacements at the nodes only: u1 , u2 and u3 .
The method we will be using to find the nodal displacements is essentially equivalent to the
Rayleigh-Ritz energy minimisation method covered in ENG3171.

Example
Consider the figure above, and assume that:

A1 = 2A2 = 2A
L1 = L2 = L
u1 = 0

Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y

4
1. Line Elements 1.1 Bar Elements

Y
Y
Y
Y

1.1.1 Nodes
Every element has a number of nodes which are used as reference points within the element. The
nodes are the only part of an element whose location and displacements are tracked. Each node
is assigned a local number, and a global number. In the figure above, there are two elements,
each with two nodes. However, two of the nodes overlap, and are only counted as a single global
node. The mapping between global and local node numbers can be summarised in a connectivity
table. Connectivity tables like this can be used for any element type.

Nodes
Elements
i j (local nodes)
1 1 2
(global nodes)
2 2 3

In the remainder of these notes, the following notation practices will be adhered to as well
as possible:

• vectors/matrices will be represented using bold symbols (u, K), or underlined when hand-
written (u, K);

• local nodes will be represented using letters (i, j, k, l, . . .);

• global node numbering will use Arabic numbers (1, 2, 3, 4, . . .).

1.1.2 Shape Functions

Shape functions are used to model the displacement field within each element. The hope is
that much more complicated global displacement fields can captured by using many simpler
displacement fields at an elemental level. One shape function is required for each node. The
simplest shape functions are linear, although much more complex forms are possible. First,
define a new intrinsic or natural coordinate η with origin at the centre of the element:

5
1. Line Elements 1.1 Bar Elements

i j

Global coordinates xi
x
xj

i j
Intrinsic coordinates
η = −1 η η = +1

The two types of coordinate are related by:

2(x − xi )
η= −1 (1.3)
xj − xi

Two shape functions are assigned to the element (one per node):
1−η
Ni (η) =
2 (1.4)
1+η
Nj (η) =
2

1 1
Nj
Ni

0.5 0.5

0 0
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
η η

These shape functions can be used to express the variation in displacement across the element:
 
  ui
u(η) = Ni ui + Nj uj = Ni Nj = Nq (1.5)
uj

u(η)

ui uj

i j

The strain in the element is given by:

du du dη
= =
dx dη dx
 
d d 2(x − xi )
= (Ni ui + Nj uj ) −1
dη dx xj − xi
−ui + uj
=
xj − xi

6
1. Line Elements 1.1 Bar Elements

If we define a new matrix called the element strain-displacement matrix :

1 1
B= [ −1 1 ] = [ −1 1 ] (1.6)
xj − xi le

then the strain can be written as:

 = Bq (1.7)
and similarly the stress within the element is:

σ = EBq (1.8)

Looking at Equations 1.5, 1.7 and 1.8, it can be seen that linear shape functions allow for a
displacement u which varies across the element, while the stress and strain are constant within
the element. In order to produce elements with varying stress and strain, higher order shape
functions would be required. For example, to produce a bar element with linearly varying strain,
quadratic (degree 2) shape functions would be required.

1.1.3 Element Stiffness Matrix

The total potential energy of the one dimensional system is given by:
X1Z XZ XZ X
χ=U +V = σA dx − ufe A dx − uTe dx − Fn un (1.9)
e
2 e e e e e n

where fe is a body force per unit volume applied to element e, Te is a traction per unit length,
and Fn is a force applied to (global) node number n. The strain energy in each element can be
expressed in terms of some of the vector forms developed in the previous section:
Z
1
Ue = σA dx
2 e
Z
1
= σ T A dx
2 e
Z
1
= (EBq)T BqA dx
2 e
Z
1
= qT BT EBqA dx
2 e

Making the substitution:

xj − xi le
dx = dη = dη
2 2
leads to:  Z 1 
1 T Ae le Ee T
Ue = q B B dη q
2 2 −1
R1
Performing the integral ( −1 dη = 2) and substituting for B from Equation 1.6 leads to:
 
1 T Ae le Ee −1  
Ue = q 2
−1 1 q
2 le 1
 
1 T Ae Ee 1 −1
= q q
2 le −1 1
1
= qT kq
2

7
1. Line Elements 1.1 Bar Elements

where the stiffness matrix is:

ui uj
Ee Ae
 
k= 1 −1 ui
(1.10)
le −1 1 uj

1.1.4 Element Force Vector

The other terms in Equation 1.9 can be assembled into vector form. Starting with the second
(body force) term of Equation 1.9:
Z Z
ufe A dx = Ae fe Ni ui + Nj uj dx
e Ze
= Ae fe Nq dx
e Z

= Ae fe qT NT dx
e
1−η
 
Z 1
le  2 
= Ae fe qT  dη
2 −1 1 + η

2
 
A l f
e e e 1
= qT
2 1

The body force vector is:  

Ae le fe 1
f= (1.11)
2 1
Following a similar approach, the traction force vector is:
 
Te le 1
T= (1.12)
2 1

The point loads will be treated next when we assemble the global stiffness matrix.

1.1.5 Global Stiffness Matrix and Force Vector

We can now express Equation 1.9 in a matrix form as:
X1 X X X
χ= qT kq − qT f − qT T − Fn un (1.13)
e
2 e e n

The simple two-element structure we have been using as an example has two elements with
two nodes each, and a total of three global nodes. We first define a vector of global displacements:
 
u1
Q =  u2 
u3

8
1. Line Elements 1.1 Bar Elements

Then, using the connectivity table, the global stiffness matrix is assembled:

u1 u2 u3 u1 u2 u3

E1 A1
" 1 −1 0 # u1
E2 A2 0
" 0 0 # u1
K= −1 1 0 u2 + 0 1 −1 u2
l1 l2
0 0 0 u3 0 −1 1 u3

u1 u2 u3
E1 A1
− E1l1A1 0
 
l1 u1
= − E1l1A1 E1 A1
l1 + l2
E2 A2
− E2l2A2  u2
E2 A2 E2 A2
0 − l2 l2 u3

The global force vector is:

         
1 0 1 0 F1
A1 l1 f1   A2 l2 f2   T1 l1   T2 l2   
F= 1 + 1 + 1 + 1 + F2 
2 2 2 2
0 1 0 1 F3

The total potential energy in the system (Equation 1.9/1.13) can now be expressed in matrix
form as:
1
χ = QT KQ − QT F (1.14)
2
From this point, we may apply the Rayleigh-Ritz method and differentiate the total potential
energy with respect to the unknown displacements. This is equivalent to taking the gradient
of the scalar function χ with respect to the unknown displacements. In the case that K is
symmetric (guaranteed for our applications), this can be found as:

∇Q χ = KQ − F = 0

or alternatively:
KQ = F (1.15)
However, this expression must be used with caution. If boundary conditions are applied (just as
we did in the first example where the fixed displacement u1 was removed from the calculation)
then some terms in the matrices above must be modified.
Another commonly used method of deriving stiffness and force matrices is the Galerkin
approach, which will not be covered here.

1.1.6 Boundary Conditions

There are two main ways of applying boundary conditions to FE problems: the elimination and
the penalty methods. Penalty methods are useful for applying complex boundary conditions
involving enforced relationships between node displacements, but will not be covered here.
Consider a three node system:
    
K11 K12 K13 u1 F1
 K21 K22 K23   u2  =  F2 
K31 K32 K33 u3 F3

Now assume that one of the three displacements is fixed to a particular value: û. Quite often û
will be set to zero to enforce an absolutely fixed condition on the node. Assuming that it is the

9
1. Line Elements 1.1 Bar Elements

first node to which the boundary condition is applied:

    
K11 K12 K13 û F1
 K21 K22 K23   u2  =  F2 
K31 K32 K33 u3 F3
    (1.16)
K12 K13   F1 − K11 û
u2
⇒  K22 K23  = F2 − K21 û 
u3
K32 K33 F3 − K31 û
The second and third rows of the equation above can be used to find the unknown displacements
u2 and u3 :     
K22 K23 u2 F2 − K21 û
=
K32 K33 u3 F3 − K31 û
   −1   (1.17)
u2 K22 K23 F2 − K21 û
⇒ =
u3 K32 K33 F3 − K31 û
The original first equation can now be used to find the reaction force F1 required to keep node
1 at its enforced displacement û:
F1 = K11 û + K12 u2 + K13 u3 (1.18)

Example
Re-solve the first example using the FE method. The figure is repeated here for convenience.

L1 L2

F
u1 1 u2 2 u3

A2
A1
x, u

Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y

10
1. Line Elements 1.1 Bar Elements

Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
/

11
1. Line Elements 1.1 Bar Elements

1.1.7 Trusses
So far we have only considered bar elements deforming in one dimension. It is reasonably
straightforward to adapt what we have already developed to a multi-bar structure (truss) in
two, or even three dimensions. Consider the bar below which has been rotated by an angle θ
from the horizontal:

vj
ũj
uj
y
ỹ j
x̃
θ ũi = ui cos θ + vi sin θ
x ũj = uj cos θ + vj sin θ

vi
ũi

ui
i

Its longitudinal nodal displacements ũi and ũj can be expressed in terms of the horizontal and
vertical components as:
ũi = ui cos θ + vi sin θ
ũj = uj cos θ + vj sin θ
or, using the traditional shorthand l = cos θ and m = sin θ:

ũi = lui + mvi

ũj = luj + mvj

We now update the displacement vector q as:

 
  ui
ũi  vi 
q̃ = q= 
ũj  uj 
vj

The relationship between these two forms can be expressed as:

q̃ = Lq (1.19)

where:  
l m 0 0
L= (1.20)
0 0 l m
This can be used to adapt the element stiffness matrix of Equation 1.10:

ũi ũj
Ee Ae
 
k̃ = 1 −1 ũi
le −1 1 ũj

12
1. Line Elements 1.1 Bar Elements

into the stiffness matrix for a bar element of a truss:

k = LT k̃L
ui vi uj vj
 l2 lm −l2 −lm  ui
Ee Ae  lm m2 −lm −m2  vi (1.21)
=
le  −l2 −lm l2 lm

uj
−lm −m2 lm m2 vj

Example
Calculate the global stiffness matrix for the two-bar truss below, and determine the deflection
of node 2:

v2 2 F
u2

A1 = A
A2 = 2A
1 2 L1 = √
L
L2 = 3L

v1 v3
u1 60◦ 30◦ u3
1 3

Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y

13
1. Line Elements 1.2 Beam Elements

Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
/

1.2 Beam Elements

Beam elements are permitted to carry shear forces and bending moments. This means that, in
addition to translational displacements of nodes, we must also account for rotations. Like bar
elements, beam elements have two nodes, but each of the nodes has two degrees of freedom.

14
1. Line Elements 1.2 Beam Elements

vi vj

vi0 vj0
i j

Here the displacement vector consists of the transverse nodal displacements v, and the beam
element slopes v 0 = ∂x
∂v
:
 
vi
 vi0 
q=  vj 


vj0
Note that we are now not considering the longitudinal displacements of the beam nodes.

1.2.1 Shape Functions

For bar elements we used linear shape functions to describe displacements as a function of
location within the element. In order to account for the variable slope within beam elements, it
is necessary to use higher order shape functions. Defining the four Hermite shape functions in
terms of the intrinsic coordinates:
1
Ni1 = (1 − η)2 (2 + η)
4
1
Ni2 = (1 − η)2 (η + 1)
4
1
Nj1 = (1 + η)2 (2 − η)
4
1
Nj2 = (1 + η)2 (η − 1)
4
with the transverse displacement:

dv dv
v(η) = Ni1 vi + Ni2 + Nj1 vj + Nj2 (1.22)
dη i dη j

The shape functions look like:

15
1. Line Elements 1.2 Beam Elements

1
0.5
Ni1

Ni2
0.5
0

0
-0.5
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
η η

0.5
1

0
Nj1

Nj2
0.5

-0.5
0

-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1

η η

Note how each shape function is designed to have either a zero or unity value or slope at each
node. The intrinsic coordinate η is related to the global coordinate x in the same way as bar
elements:
2(x − xi )
η= −1
xj − xi
meaning that again:
le
dx = dη
2
so:
dv dv dx le dv
= = (1.23)
dη dx dη 2 dx
Using this, Equation 1.22 can be re-written as:
 
vi
 le le
  vi0 
v= Ni1 2 Ni2 Nj1 2 Nj2  vj  = Nq
  (1.24)
vj0

It is also possible to find an expression for the strain through the beam thickness:


z

16
1. Line Elements 1.2 Beam Elements

The strain is given by:

z d2 v
=− = −zκ ≈ −z 2
R dx
and from Equation 1.24 we have:
v = Nq
d2 v d2 N
⇒ = q
dx2 dx2
Differentiating Equation 1.23 allows this to be re-expressed in terms of η:
d2 v 4 d2 v
=
dx2 le2 dη 2
4 d2
= 2 2 (Nq)
le dη
4 d2 N
 
= 2 q
le dη 2
where:
d2 N  3
= 2 η 3η−1 le
− 32 η 3η+1 le

dη 2 2 2 2 2

The strain through the beam thickness can therefore be represented as:

 = Bq

where:
4 d2 N
B = −z (1.25)
le2 dη 2

1.2.2 Element Stiffness Matrix

The bending strain energy stored in the beam is (from ENG3171):
Z  2 2
1 ∂ v
U= EI dx
2 ∂x2
and the potential energy in the beam is given by:
X1Z  2 2 XZ
∂ v X X
χ=U +V = EI 2
dx − pv dx − Fn v n − Mn vn0 (1.26)
e
2 e ∂x e e n n

where p is a load per unit length, Fn is a point load at location n, and Mn is a moment applied
at location n. The strain energy in each element can now be written as:
Z  2 2
1 ∂ v
U= EI dx
2 e ∂x2
Z  2 T  2 
1 T 16 d N d N
= EIq 4 2
q dx
2 e le dη dη 2
T  2 
4EI T 1 d2 N
Z 
d N
= 3 q 2
dη q
le −1 dη dη 2
9 2 3
− 49 η 2 3
 
Z 1 4η 8 η(3η − 1)le 8 η(3η + 1)le
4EI  3 η(3η − 1)le (3η − 1)2 le2 /16 − 3 η(3η − 1)le (9η 2 − 1)le2 /16 
= 3 qT  8 9 2 3
8
9 2 3
 dη q
le −1
 − 4 η − 8 η(3η − 1)le 4 η − 8 η(3η + 1)le

3 2 2 3 2 2
8 η(3η + 1)le (9η − 1)le /16 − 8 η(3η + 1)le (3η + 1) le /16

17
1. Line Elements 1.2 Beam Elements

Integrating each term in the matrix with respect to η gives:

1
U = qT kq
2
where k is the element stiffness matrix:
vi vi0 vj vj0
 12 6le −12 6le  vi
EI 6le 4le2 −6le 2le2  vi0
k= 3 (1.27)
le  −12 −6le 12 −6le

vj
6le 2le2 −6le 4le2 vj0

Note that we could also have used the standard definition for strain energy as a starting
point: Z
1
Ue = σ dV
2 e
Z
1
= qT BT EBq dA dx
2 e
Z  2 T  2 
1 2 T 16 d N d N
= Ez q 4 q dA dx
2 e le dη 2 dη 2
Z  2 T  2 
1 T 16 d N d N
= EIq 4 2
q dx
2 e le dη dη 2
with the rest of the derivation following as above.

1.2.3 Element Force Vector

We first treat the distributed load p across the beam. In the case that p is constant, its contri-
bution to the total potential energy of an element is:
ple 1
Z Z
pv dx = N dη q
e 2 −1
= f Tq
= qT f
where the contribution to the element force vector due to the distributed load is:

ple /2  vi

f = ple2 /12  vi0

(1.28)
ple /2
 
vj
−ple2 /12 vj0

The contribution of external shear forces and moments applied at nodes must simply be
placed in the correct locations in the global force vector:

 .. 
.
Fn
 
vn
F = 
Mn 0
 
  vn
..
.

18
1. Line Elements 1.2 Beam Elements

1.2.4 Global Stiffness Matrix and Force Vector

The elements of the global stiffness matrix and force vector for a beam based system are assem-
bled in exactly the same manner as the bar/truss system.

Example
Find the displacement of the centre of the first element in the two-element beam below. Note
that we are using a sign convention with v positive in the vertically upwards direction.

v
1 2 3 I1 = 2I
x 1 2 I2 = I
p

2L L

Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y

19
1. Line Elements 1.2 Beam Elements

Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
/

20