ENGG302/MECH452

Introduction to Finite
Element Analysis

Lecture 2

Dr Peter L Green

2018
Recap
• Finite Element is powerful and has many applications.

• It could ‘boost’ your final year projects?

• It is an approximation method:

All models are wrong but some are useful.

• You will learn how to use Abaqus (coursework).

• You will understand what Abaqus is doing (exam).

• Limitation 1: The Finite Element Method is computationally costly.

Content
• Symmetry – a simple trick to reduce the cost of your model.

• The ‘assemble system equations’ part of the Finite Element analysis.

• A generic expression for 1D truss elements.

• How can software can automatically develop stiffness matrices?

 Symmetry
• Plate with, hole at its centre, tensile test (pressure 𝑃).

• Our Finite Element model predicts the maximum stress in the

plate (for Assessment 1c !).
𝑦

𝑃 𝑃
Symmetry
• 2 lines of symmetry in the geometry and loading conditions.

𝑥
Symmetry
• We only have to model a quarter of the plate (‘the quarter plate
model’).

𝑥
Symmetry
• As the plate deforms, points along this line will not deform in the
y-direction

𝑥
Symmetry
• As the plate deforms, points along this line will not deform in the
x-direction

𝑥
 Symmetry
• Here is a schematic of our quarter plate model. This is an example of
planar symmetry.

• Quarter of the number of elements – a large saving in computational cost!

𝑦

𝑃
Symmetry
• An example of cyclic symmetry:

• Similar to planar symmetry

but in a cylindrical coordinate
system.

• Examples: spline fittings,

propellers…
 𝑧
Symmetry
• An example of axial symmetry:

• Distribution of the variable of

interest (stress, for example), is
constant around the
circumference of the part

• Very common – Abaqus includes

axisymmetric elements for this
reason

• Loading conditions, material

properties and boundary
conditions must be symmetric with
respect to the axis of revolution 𝑟
 𝑧
Symmetry
• Examples:

• Pressure vessels
• Cylindrical containers
• Pistons
• Rotating equipment (cement mills,
for example)
• Rocket casing
• Cooling towers
• And more!

𝑟
Symmetry
• Repetitive symmetry.

• These are less common…

 Symmetry
• Be careful!

• Include loading and boundary conditions when considering symmetry:

𝐹
Not a line of
symmetry
Symmetry

𝐹 𝐹/2
1D Truss Elements
• So, what is Finite Element actually doing?
• We will look at 1D truss elements first.
• You use these in Assessment 1.

𝑥 Element

Nodes
1D Truss Elements
Assumptions:

1. Axial dimension is very large compared to the cross section.

2. Load acts uniformly over the cross section.

3. Stress and strain are uniform over the cross section.

𝑥
1D Truss Elements
Assumptions:

4. Nodes can only move in the 𝑥 direction.

5. Element self-weight is neglected.

6. Material properties are constant.

𝑥
1D Truss Elements

• Degrees of freedom: the number of coordinates needed to

define the position of each node. In this case, each node has
1 degree of freedom only.

𝑥
1D Truss Elements
• Step 1: discretise the geometry into nodes and elements.

• After selecting the element type and properties (1D truss in this
case) we must then assemble the system equations. For a static
stress analysis:

𝑭 = 𝑲𝒖

𝐹1 𝑢1
𝑭 = 𝐹2 , 𝒖 = 𝑢2
⋮ ⋮
 1D Truss Elements
• Step 1: discretise the geometry into nodes and elements.

• After selecting the element type and properties (1D truss in this
case) we must then assemble the system equations. For a static
stress analysis:

𝑭 = 𝑲𝒖

𝐹1 𝑢1
External force 𝑭 = 𝐹2 , 𝒖 = 𝑢2
applied to each ⋮ ⋮
node

Position of each
node (in 1 direction
for our example)
1D Truss Elements
• Step 1: discretise the geometry into nodes and elements.

• After selecting the element type and properties (1D truss in this
case) we must then assemble the system equations. For a static
stress analysis:

𝑭 = 𝑲𝒖 Global stiffness
matrix. This is what
we want to find.
𝐹1 𝑢1
𝑭 = 𝐹2 , 𝒖 = 𝑢2
⋮ ⋮
1D Truss Elements
• Example: the stepped-bar problem
𝐿1 𝐿2

𝐹
𝐸2 , 𝐴2
𝐸1 , 𝐴1

• 2 cylindrical bars, each with constant cross-sectional area and

Young’s modulus, subject to a tensile force, 𝐹.

• We will discretise (or mesh) our problem using 1D truss elements.

 1D Truss Elements
𝐿1 𝐿2

𝐹
𝐸2 , 𝐴2
𝐸1 , 𝐴1
Discretise into a general Finite Element model with 2, 1D truss elements:
𝐹1 𝐹2 𝐹3

𝑢1 𝑢2 𝑢3

𝐹1 𝑘11 𝑘12 𝑘13 𝑢1

𝐹2 = 𝑘21 𝑘22 𝑘23 𝑢2
𝐹3 𝑘31 𝑘32 𝑘33 𝑢3
1D Truss Elements
• We will look at a rule you can use to find the stiffness matrix.
Crucially, it is programmable.
𝐿

𝐹1 𝐹2
𝑢1 𝑢2

• Internal forces:

𝐹𝑖𝑛𝑡 𝐹𝑖𝑛𝑡
𝐹1 𝐹2
1D Truss Elements
𝐹𝑖𝑛𝑡
• For node 1: 𝐹1

• Truss elements assume constant cross sectional area, 𝐴, and

constant Young’s modulus, 𝐸.
• Say 𝜎 is the stress in the element then:

𝐸𝐴
𝐹𝑖𝑛𝑡 = 𝜎𝐴 = 𝜖𝐸𝐴 = 𝑢2 − 𝑢1
𝐿
1D Truss Elements
𝐹𝑖𝑛𝑡
• For node 1: 𝐹1

• Truss elements assume constant cross sectional area, 𝐴, and

constant Young’s modulus, 𝐸.
• Say 𝜎 is the stress in the element then:

𝐸𝐴
𝐹𝑖𝑛𝑡 = 𝜎𝐴 = 𝜖𝐸𝐴 = 𝑢2 − 𝑢1
𝐿

Strain – difference in length

divided by the original length of
the element
1D Truss Elements
𝐹𝑖𝑛𝑡
• For node 1: 𝐹1

• Truss elements assume constant cross sectional area, 𝐴, and

constant Young’s modulus, 𝐸.
• Say 𝜎 is the stress in the element then:

𝐸𝐴
𝐹𝑖𝑛𝑡 = 𝜎𝐴 = 𝜖𝐸𝐴 = 𝑢2 − 𝑢1
𝐿

• Finally, 𝐹1 = −𝐹𝑖𝑛𝑡 (equilibrium) so:

𝐸𝐴
𝐹1 = (𝑢1 − 𝑢2 )
𝐿
1D Truss Elements
𝐹𝑖𝑛𝑡
• For node 2: 𝐹2

𝐸𝐴
𝐹𝑖𝑛𝑡 = 𝜎𝐴 = 𝜖𝐸𝐴 = 𝑢2 − 𝑢1
𝐿

• 𝐹2 = 𝐹𝑖𝑛𝑡 (equilibrium) so:

𝐸𝐴
𝐹2 = (𝑢2 − 𝑢1 )
𝐿
1D Truss Elements
• We have:
𝐸𝐴
𝐹1 = (𝑢1 − 𝑢2 )
𝐿
𝐸𝐴
𝐹2 = (𝑢2 − 𝑢1 )
𝐿

• In matrix form:

𝐹1 𝐸𝐴 1 −1 𝑢1
= 𝑢2
𝐹2 𝐿 −1 1

• This is the general stiffness matrix of a 1D truss element.

Knowing this makes it easier to derive the stiffness matrices of
complex systems. We will illustrate this with the stepped-bar
example…
 Stepped Bar Problem
𝐿1 𝐿2

𝐹
𝐸2 , 𝐴2
𝐸1 , 𝐴1
Discretise into a general Finite Element model with 2, 1D truss elements:
𝐹1 𝐹2 𝐹3

𝑢1 𝑢2 𝑢3
Stepped Bar Problem

𝐹1 𝐹2 𝐹3

𝑢1 𝑢2 𝑢3

𝐹1 ? ? ? 𝑢1
𝐹2 = ? ? ? 𝑢2
𝐹3 ? ? ? 𝑢3
 Stepped Bar Problem

𝐹1 𝐹2 𝐹3

𝑢1 𝑢2 𝑢3

First, we only consider the resistive forces that arise because of element 1.
This allows us to make use of the general expression we derived earlier…

𝐹1 ? ? ? 𝑢1
𝐹2 = ? ? ? 𝑢2
𝐹3 ? ? ? 𝑢3
Stepped Bar Problem

𝐹1 𝐹2 𝐹3

𝑢1 𝑢2 𝑢3

𝐹1 𝐸1 𝐴1 1 −1 𝑢1
= 𝑢2
𝐹2 𝐿1 −1 1

𝐹1 ? ? ? 𝑢1
𝐹2 = ? ? ? 𝑢2
𝐹3 ? ? ? 𝑢3
Stepped Bar Problem

𝐹1 𝐹2 𝐹3

𝑢1 𝑢2 𝑢3

𝐹1 𝐸1 𝐴1 1 −1 𝑢1
= 𝑢2
𝐹2 𝐿1 −1 1

𝐸1 𝐴1 𝐸1 𝐴1
𝐹1 − ? 𝑢1
𝐿1 𝐿1
𝐹2 = 𝐸1 𝐴1 𝐸1 𝐴1 𝑢2
− ?
𝐹3 𝐿1 𝐿1 𝑢3
? ? ?
 Stepped Bar Problem

𝐹1 𝐹2 𝐹3

𝑢1 𝑢2 𝑢3

Now we consider the resistive forces that arise because of element 2. Note
that node 2 connects elements 1 and 2. It must therefore be subject to the
resistive forces of both elements…

𝐸1 𝐴1 𝐸1 𝐴1
𝐹1 − ? 𝑢1
𝐿1 𝐿1
𝐹2 = 𝐸1 𝐴1 𝐸1 𝐴1 𝑢2
− ?
𝐹3 𝐿1 𝐿1 𝑢3
? ? ?
Stepped Bar Problem

𝐹1 𝐹2 𝐹3

𝑢1 𝑢2 𝑢3

𝐹2 𝐸2 𝐴2 1 −1 𝑢2
= 𝑢3
𝐹3 𝐿2 −1 1

𝐸1 𝐴1 𝐸1 𝐴1
𝐹1 − ? 𝑢1
𝐿1 𝐿1
𝐹2 = 𝐸1 𝐴1 𝐸1 𝐴1 𝑢2
− ?
𝐹3 𝐿1 𝐿1 𝑢3
? ? ?
Stepped Bar Problem

𝐹1 𝐹2 𝐹3

𝑢1 𝑢2 𝑢3

𝐹2 𝐸2 𝐴2 1 −1 𝑢2
= 𝑢3
𝐹3 𝐿2 −1 1

𝐸1 𝐴1 𝐸1 𝐴1
− ?
𝐹1 𝐿1 𝐿1 𝑢1
𝐸1 𝐴1 𝐸1 𝐴1 𝐸2 𝐴2 𝐸2 𝐴2
𝐹2 = − + − 𝑢2
𝐿1 𝐿1 𝐿2 𝐿2
𝐹3 𝐸2 𝐴2 𝐸2 𝐴2 𝑢3
? −
𝐿2 𝐿2
 Stepped Bar Problem

𝐹1 𝐹2 𝐹3

𝑢1 𝑢2 𝑢3

We have now accounted for all the internal forces in the system, so the
remaining elements of the stiffness matrix must be equal to 0:

𝐸1 𝐴1 𝐸1 𝐴1
− 0
𝐹1 𝐿1 𝐿1 𝑢1
𝐸1 𝐴1 𝐸1 𝐴1 𝐸2 𝐴2 𝐸2 𝐴2
𝐹2 = − + − 𝑢2
𝐿1 𝐿1 𝐿2 𝐿2
𝐹3 𝐸2 𝐴2 𝐸2 𝐴2 𝑢3
0 −
𝐿2 𝐿2
 Stepped Bar Problem
• The nice thing about creating stiffness matrices in this way is that it is
easily programable.

• We select 1D truss elements – the software then ‘looks up’ the generic
stiffness matrix for these elements.

• When we discretise the geometry the software generates a connectivity

table, telling us how the elements are connected together.

• In our example, it would tell us that element 1 is connected to element 2,

by node 2.

• Once we enter the material properties, it can automatically create the

stiffness matrix.
 Rigid Element Problem
𝑒2 𝐹3
Element number
𝑢3
𝐹1 𝑒1 𝐹2
𝑢2
𝑢1
𝐹4
𝑒3
Rigid element
𝑢4

All elements have the same

properties (𝐴, 𝐸, 𝐿)
 𝐹2 𝐸𝐴 1 −1 𝑢2
Rigid Element Problem = 𝑢3
𝐹3 𝐿 −1 1

𝑒2 𝐹3

𝑢3
𝐹1 𝑒1 𝐹2
𝑢2
𝑢1
𝐹4
𝑒3

𝐹1 𝐸𝐴 1 −1 𝑢1 𝑢4
= 𝑢2
𝐹2 𝐿 −1 1
𝐹2 𝐸𝐴 1 −1 𝑢2
= 𝑢4
𝐹4 𝐿 −1 1
Rigid Element Problem

𝐹1 𝑢1
𝐹2 𝐸𝐴 𝑢2
𝑭 = 𝑲𝒖 → = 𝑢3
𝐹3 𝐿
𝐹4 𝑢4

𝐹1 𝐸𝐴 1 −1 𝑢1
=
𝐹2 𝐿 −1 1 𝑢2
𝐹2 𝐸𝐴 1 −1 𝑢2
=
𝐹3 𝐿 −1 1 𝑢3
𝐹2 𝐸𝐴 1 −1 𝑢2
=
𝐹4 𝐿 −1 1 𝑢4
Rigid Element Problem

𝐹1 1 −1 𝑢1
𝐹2 𝐸𝐴 −1 1 𝑢2
𝑭 = 𝑲𝒖 → = 𝑢3
𝐹3 𝐿
𝐹4 𝑢4

𝐹1 𝐸𝐴 1 −1 𝑢1
=
𝐹2 𝐿 −1 1 𝑢2
𝐹2 𝐸𝐴 1 −1 𝑢2
=
𝐹3 𝐿 −1 1 𝑢3
𝐹2 𝐸𝐴 1 −1 𝑢2
=
𝐹4 𝐿 −1 1 𝑢4
Rigid Element Problem

𝐹1 1 −1 𝑢1
𝐹2 𝐸𝐴 −1 2 −1 𝑢2
𝑭 = 𝑲𝒖 → = 𝑢3
𝐹3 𝐿 −1 1
𝐹4 𝑢4

𝐹1 𝐸𝐴 1 −1 𝑢1
=
𝐹2 𝐿 −1 1 𝑢2
𝐹2 𝐸𝐴 1 −1 𝑢2
=
𝐹3 𝐿 −1 1 𝑢3
𝐹2 𝐸𝐴 1 −1 𝑢2
=
𝐹4 𝐿 −1 1 𝑢4
Rigid Element Problem

𝐹1 1 −1 𝑢1
𝐹2 𝐸𝐴 −1 3 −1 −1 𝑢2
𝑭 = 𝑲𝒖 → = 𝑢3
𝐹3 𝐿 −1 1
𝐹4 −1 1 𝑢4

𝐹1 𝐸𝐴 1 −1 𝑢1
=
𝐹2 𝐿 −1 1 𝑢2
𝐹2 𝐸𝐴 1 −1 𝑢2
=
𝐹3 𝐿 −1 1 𝑢3
𝐹2 𝐸𝐴 1 −1 𝑢2
=
𝐹4 𝐿 −1 1 𝑢4
Rigid Element Problem

𝐹1 1 −1 0 0 𝑢1
𝐹2 𝐸𝐴 −1 3 −1 −1 𝑢2
𝑭 = 𝑲𝒖 → = 𝑢3
𝐹3 𝐿 0 −1 1 0
𝐹4 0 −1 0 1 𝑢4

𝐹1 𝐸𝐴 1 −1 𝑢1
=
𝐹2 𝐿 −1 1 𝑢2
𝐹2 𝐸𝐴 1 −1 𝑢2
=
𝐹3 𝐿 −1 1 𝑢3
𝐹2 𝐸𝐴 1 −1 𝑢2
=
𝐹4 𝐿 −1 1 𝑢4
 Conclusions

• Exploiting symmetry can help reduce computational cost.

• Each element type has generic expressions that are associated with it.

• Combining this with material properties and a connectivity table, it is

then possible to construct the stiffness matrix for linear elastic problems.

• Note: once defined, the stiffness matrix remains constant in the entire
analysis.

• For our linear static analysis the element properties (𝐸, 𝐴, 𝐿) will not
change, even if this is unrealistic!