D21FA: Finite Element Method Linear Analysis Winter 2022

Lecture 1: Introduction to FEA

Lecturer: Shadi Mohamed

1.1 Introduction

Finite element analysis is a computer based numerical method that is used in the analysis of complex
mechanical and structural problems. It is used to obtain numerical solutions and can be applied over a wide
range of geometric functions and loading conditions. The method can be used for analysis of static problems
and dynamic problems and can be used to analyse linear systems and non-linear systems. A linear system is
a system where a linear relationship exists between the force and deflection. A non-linear system is a type
of system in which a non-linear relationship exists between, for example, force and deflection.
Finite element analysis is not restricted to mechanical and structural analysis. The method can be applied to
problems that involve heat loss, fluid flow, electric potential, etc. Due to this wide range of application, the
method is very popular. The method is used in the analysis of many two and three dimensional problems.
As finite element analysis problems become larger and more complex, they can take longer to solve. A wide
range of finite element analysis software exists today (LUSAS, ABAQUS, ANSYS).

Figure 1.1: Computer finite element analysis of an T-Beam

The method works by taking a problem and modelling it as a set of smaller elements through the use of nodes
and elements. Physical and geometrical properties are assigned to the elements and loads and displacements
are applied to the nodes. A set of equations then derives the numerical values. The best way to see finite
element analysis in action, is to see how it is applied and used to solve simple basic problems.

1-1
 dx
u
1-2 Lecture 1: Introduction to FEA
1 2

x-axis

1
1.2 Development of a1'Finite Element 2 2'

1.2.1 u1 an element
Step 1 - Creating u2

Initially, a finite element with uniformx-axis

cross sectional area and subjected to axial load will be developed. An
element can be created between the nodes, as demonstrated in Figure 1.2.

Node 1 Element Node 2

x,u

P1 L P2

L
Figure 1.2: Element

Let us consider a line element of length L with two nodes, 1 and 2, at the two ends. Each node has one
degree of freedom that is the displacement u. The problem has one dimension such that the two nodes and
u+(du/dx)dx
in fact any point on the member can only move
dx in the axial direction.

u
1.2.2 Step 2 - Nodal displacements
1 2
Now that the element has been established the nodal displacements can now be considered to obtain the
x-axis
general relationship for displacement at any point on the element.

1 1' 2 2'

u1 u2

x-axis

Figure 1.3: Axially loaded member

Node 1 Element Node 2
x,u
If a horizontal load is applied to the element, node 1 will displace to 1 and node 2 will displace to 20 . The
0

displacement of node 1 to 10 can be considered Las u1 and the displacement of node 2 to 20 can be considered
as u2 . This is illustrated in Figure 1.3. These displacements can then be written as the displacement vector,
u.  
u1
u= (1.1)
u2
The displacement along the member is required. Since there are two nodal displacements, a function with
two unknowns can be used (usually, but not always, a complete polynomial is used). Thus, the displacement
function can be written as.
u = a1 + a2 x (1.2)
Where u is the displacement at x along element 1 2, a1 and a2 are unknown constants and x is a co-ordinate
along the beam. Using the displacement function, we can determine a1 and a2 in terms of the displacements
u1 and u2 .
Lecture 1: Introduction to FEA 1-3

At node 1 we have x = 0. Substituting this into the displacement function, (1.2) we get

u1 = a1 + a2 (0) (1.3)

Therefore
u1 = a1 (1.4)
Now consider the other node. At node 2 we have x = L. Substituting into the displacement function,

u2 = a1 + a2 L (1.5)

Previously, the coefficient a1 was determined. Therefore,

u2 = u1 + a 2 L (1.6)

Rearranging it terms of the constant a2 ,

u2 − u1
a2 = (1.7)
L
With both a1 and a2 determined, they can be substituted back into the displacement function, to give
u2 − u1
u = u1 + x (1.8)
L
This can be written in the matrix form as
 
 x x
 u1
u= 1− L L (1.9)
u2
P1 P2
1.2.3 Step 3 - Strain
L
With the displacement function obtained, the strain can be determined along the member. Figure 1.4
illustrates the strain at a position along the member.

u+(du/dx)dx
dx
u

1 2

x-axis

1 Figure 1.4: Strain along the member

1' 2 2'

From Figure 1.4, the original length of a small element is denoted as dx. End 1 is displaced by u and end 2
u1 u2
is displaced by u + du
dx dx. Therefore,

x-axis du
Change in Length = u + dx − u (1.10)
dx
This simplifies to
Node 1 du Node 2
Change in Length =
Element dx (1.11)
dx x,u

L
1-4 Lecture 1: Introduction to FEA

With expressions for the change in length and the original length, an expression for the strain can be derived.
du
Change in Length dx
= = dx (1.12)
Original Length dx
This simplifies to
du
= (1.13)
dx
Previously, an expression for displacement, u, was derived to give equation 1.8. By differentiating this with
respect to x, we obtain
du u2 − u1
= (0) + (1) (1.14)
dx L
du u2 − u1
= (1.15)
dx L
Therefore,
u2 − u1
= (1.16)
L
This expression can be expressed in matrix format
 
 1 1  u1
 = −L L (1.17)
u2
This can also be obtained by differentiating equation 1.8.
The matrix
− L1 1
 
L
is the strain-displacement matrix and can be written as [B]. Now consider the vector
 
u1
u2
This is the displacement vector and can be written as u. Therefore, the expression for strain can be written
as
 = [B] u (1.18)
We are carefully dealing with a one dimensional problem, there is only one strain. Later we will be looking
at two dimensional and three dimensional situations where there can be 3 or 6 strains at any point in a body,
respectively.

1.2.4 Step 4 - Stress

With an expression for the strain, the stress in the element can now be determined. The expression to obtain
stress is given as
σ = [D]  (1.19)
[D] is the constitutive matrix and the notation is used by convention. As we are dealing with a one dimen-
sional problem, only one stress will be obtained. However, σ, can contain a number of stresses depending
on the problem. In this case, as there is only one stress, the matrix [D] only contains one term, which for a
linear elastic problem is the modulus of elasticity E.
σ=E  (1.20)
Previously, an expression for strain was derived. Substituting that into expression 1.20 yields,
σ = E [B] u (1.21)
Lecture 1: Introduction to FEA 1-5

1.2.5 Step 5 - Nodal Loads

The nodal loads now need to be related to the internal stresses. These can be included in a nodal force

P1 P2

Figure 1.5: Nodal loads

u+(du/dx)dx
vector dx  
u P1
P = (1.22)
P2
However, rather than apply just to the
1 particular loading shown above a general
2 relationship will be derived.
This can be done with the aid of virtual work. This will be of use later when we look at more general
x-axis
elements.

1 1' 2 2'
1.2.6 Step 6 - Application of Virtual Work
u1 the external forces acting on a memberu2with the internal forces in the
Virtual work can be used to relate
member.
x-axis
External Virtual Work Done = Internal Virtual Work Done (1.23)
Given a set of external applied forces, P , we can apply a set of virtual displacements, u
b. Therefore, the
external work applied can be expressed as
Node 1 Element Node 2
x,u
External Virtual Work Done = u bT P (1.24)
L
As for the internal virtual work, the internal stresses, σ, equilibrate P and the internal virtual strains, b
, are
compatible with the virtual displacement u b. From this, the internal work can be calculated using

T σ Per Unit Vol

Internal Virtual Work Done = b (1.25)

The total virtual work done

Z
Internal Virtual Work Done = T σ dV
b (1.26)
V

Now, consider the expression for strain that was derived previously.

 = [B] u (1.27)

By taking the transpose of the strain, we obtain

T = uT [B]T (1.28)

This can then be substituted into expression 1.26 to give

Z
Internal Virtual Work Done = bT [B]T σ dV
u (1.29)
V
1-6 Lecture 1: Introduction to FEA

This expression can be expanded further by using the expression for stress that was derived previously,
expression 1.21. This gives
Z
Internal Virtual Work Done = bT [B]T [D] [B] u dV
u (1.30)
V

Note, u are real displacements as they are used to obtain the real stress and u
b are virtual displacements.
Returning to the conservation of work.

External Virtual Work Done = Internal Virtual Work Done (1.31)

Z
T
b P =
u bT [B]T [D] [B] u dV
u (1.32)
V

bT is not a function of the geometry of the member. Therefore, this can be rearranged to obtain
u
Z
ubT P = ubT [B]T [D] [B] dV u (1.33)
V

Let Z
[B]T [D] [B] dV = [k] (1.34)
V
This is the stiffness matrix. Therefore
uT P = uT [k] u (1.35)
This can be simplified further to obtain the final expression

P = [k] u (1.36)

Where P is the nodal force vector, [k] is the element stiffness matrix and u is the nodal displacement vector.
As mentioned previously, the important point to note is that the derivation is completely general. No
assumptions have been made regarding the loads, element type, material properties, etc.

1.2.7 Step 7 - Stiffness Matrix

For our element, we can now derive the stiffness matrix. Let
Z
[k] = [B]T [D] [B] dV (1.37)
V

Rearranging Z
T
[k] = [B] [D][B] dV (1.38)
V
Now, consider Z
dV (1.39)
V
Expanding Z Z Z Z
dV = dxdydz (1.40)
V
But in our case, the elements have constant area throughout.
Z Z
dydz = A (1.41)
Lecture 1: Introduction to FEA 1-7

Z Z L Z L
dV = dxA = A dx = AL (1.42)
V O O
Substituting E for [D]
[k] = [B]T E[B]AL (1.43)
Now consider the strain-displacement matrix [B]. From previous discussion, it was found

[B] = − L1 L1
 

By taking the transpose of the strain-displacement matrix, we obtain

 1 
T −L
[B] = 1
L

Evaluating
− L1
 
− L1 1
 
[k] = 1 E L AL (1.44)
L
Multiplying throughout, we obtain
EA
− EA
 
P1 L L
P2
[k] = (1.45)
− EA
L
EA
L
L
1.3 Example 1

Now consider the structure shown in figure 1.6. u+(du/dx)dx

dx
u
Fixed
End 1 2 W
x-axis L

1 1' x-axis 2 2'

u1 Figure 1.6: Axially Loaded Member u2

x-axisThe Young’s modulus of the member is E and the cross sectional

Determine reactions, stresses and strains.
area is A. We will use one element.

Node 1 Element Node 2

x,u

Figure 1.7: Finite Element

This problem can be solved using

P = [k]u (1.46)
1-8 Lecture 1: Introduction to FEA

From previous, the nodal load vector is  

P1
P = (1.47)
P2
The displacement vector is  
u1
u= (1.48)
u2

The stiffness matrix is

EA
− EA
 
[k] = L L (1.49)
− EA
L
EA
L
Substituting these into equation 1.46, we obtain
  EA
− EA
  
P1 L L u1
= (1.50)
P2 − EA
L
EA
L
u2
We can now apply the boundary conditions, i.e. the known displacements and loads to expression 1.50. In
this case
u1 = 0 (Node 1 is restrained and cannot be displaced)
u2 = Unknown
P1 = Unknown
P2 = W
Therefore
EA
− EA
    
P1 L L 0
= (1.51)
W − EA
L
EA
L
u2
Multiplying the matrices, we obtain
EA
P1 = −
u2 (1.52)
L
EA
W = u2 (1.53)
L
Rearranging expression 1.53, we obtain an expression for u2
WL
u2 = (1.54)
EA
This can then be used with expression 1.52 to find the unknown nodal force.
 
EA W L
P1 = − = −W (1.55)
L EA
u2 can also be used to obtain the strain within the beam. Using expression 1.17
 
0
 = − L1 L1
 
WL (1.56)
EA
Multiplying through, the expression for strain becomes
W
= (1.57)
EA
Using the strain, the stress can be obtained.
W
σ = [D] (1.58)
EA
As mentioned previously, [D] contains only one value, E, as we are dealing with a one dimensional problem.
Therefore, the stress becomes
W
σ= (1.59)
A
Lecture 1: Introduction to FEA 1-9

1.3.1 Exact Solution

Fixed
End W
L

x-axis

Figure 1.8: Member

The stress in the member, σ, can be obtained using

F orce
σ= (1.60)
Area
The load applied is W and the cross-sectional area is A. Therefore
W
σ= (1.61)
A
The strain in the beam, , is obtained from
σ
E= (1.62)

Rearranging
σ
= (1.63)
E
Substituting the theoretical stress obtained previously, equation 1.60, into equation 1.63 to give
W
= (1.64)
EA
It is clear to see that the results obtained through the finite element method agree with the theoretical
results. However, this is not always the case!