Department of

Mechanical & Aerospace

Engineering

Finite Element Analysis I – MECH 5605

Fall 2018
1
 Course Outline
Chapter 1: Introduction: Overview of the Finite Element Method

Chapter 2: Direct Approach

• Scalar spring elements
• 1D-problem: stress elements; heat conduction elements
• 2D-problem: truss structure; truss structure – Special cases (oblique constraints and MPC); Constant Strain Triangle (CST)

Chapter 3: Mathematical Approach

• Variational Formulations (Rayleigh-Ritz Method (minimum potential energy theory)) – Integral formulation
• Weighted Residual Method: Collocation method; Subdomain method; Least-squares method; Galerkin method

Chapter 4: Application of Mathematical Approach

• 1D elements: Axial members – Rod Elements; Beam elements; Frame elements; 1D heat conduction element
• higher order 1D elements: quadratic and cubic elements; Global, local and natural coordinates; Isoparametric elements
• 2D elements: Rectangular linear elements: Heat conduction element; Stress element
• Numerical integration (Gauss Quadrature)

Chapter 5: Special Topics: Model Order Reduction

• Model Symmetry
• Model Axisymmetry
• Model condensation
2
 2. Direct Approach
 1-D Problems: Stress
Consider the problem of axially loaded strut with linearly varying circular cross-section area
as shown in the figure. Find the deformation field under the applied loading.
 Analytical Solution:

d1

dx
L

d2
3
 2. Direct Approach
 1-D Problems: Stress

 Computational Solution: Finite Element Analysis

Solution Steps:

1. Idealize the problem and Discretize the domain

2. Assume interpolation functions that represent element behaviour

(interpolation polynomial)

3. Establish element properties through developing equations of motion for an

element (e.g. from linear elastic theory)

4. Assembly

5. Applying loads and boundary conditions

6. Solving

7. Post-processing
4
 2. Direct Approach
 1-D Problems: Stress
1. Idealize the problem and discretize the domain

• Simplified by four elements, each with uniform cross-section

• Each section has same average area of corresponding region of the original domain
• Note this introduces error due to the approximation
• More elements with smaller steps would reduce the error (sensitivity analysis)

(1)
1
(2)
2
(3)
3 x
(4)
4 (5)

5
 2. Direct Approach
 1-D Problems: Stress
Node i
2. Interpolation Function

X Node j

X
• Deformation field is denoted by
• Considering linear interpolation function where we assume that deformation of an element varies linearly
over the length of the element (1D linear element)
• Then, deflection anywhere in the element can be expressed as

• The ends of the element (nodes i and j) have ‘known’ deflection values and

Substitute into deflection equation: Solve for and

at
at ,
6
 2. Direct Approach
 1-D Problems: Stress
Node i
2. Interpolation Function

X Node j
• The deflection distribution within the element is thus Δ

• Rearrange to collect and terms

=

• Shape functions defined from terms in brackets

• In matrix form

7
 2. Direct Approach
 1-D Problems: Stress
2. Interpolation Function
• Linear Shape functions  Properties of shape functions:

• Each shape function has a value of 1 at the

corresponding node and 0 at any other node
1 at node
0 at node

(Evaluate shape functions at and to

 Notes on shape functions: verify)

• Can be used to describe variation of any variable in • The sum of all shape functions is 1 1
element (e.g. temperature or stress)
• More useful for higher order elements (quadratic) or
higher dimensional (2D, 3D) elements

8
 2. Direct Approach
 1-D Problems: Stress
3. Developing Element Equation
• Recall we want to develop a relationship in the form of:
• Assuming linear elasticity using Hooke’s law:
Average stress
Average normal strain

• Rearrange: Δ
∆
• We can then express strain in the element in terms of nodal displacements as:

• From equilibrium: 0 and P = =

• The nodal forces can then be written as: ;

1 1
• In matrix form:
1 1

• The element stiffness matrix: K= where = 9

 2. Direct Approach
 1-D Problems: Stress (1)

3. Developing Element Equation 1

(2)
• Element 1 Stiffness Matrix where 2
(3)
3 (4)
4 (5)
• Element 2 Stiffness Matrix where

• Element 3 Stiffness Matrix where

• Element 4 Stiffness Matrix where

10
 2. Direct Approach
 1-D Problems: Stress (1)

4. Assembly 1
(2)
2
(3)
Stiffness matrix for element (1):
Local 3 (4)
4 (5)
:

We use the correspondence of local and global node numbers to determine the position in the
global stiffness matrix:
Global

0 0 0
0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 0 0 0

11
 2. Direct Approach
 1-D Problems: Stress (1)

4. Assembly 1
(2)
2
(3)
Similarly, the stiffness matrix for element (2)
3 (4)
Local
4 (5)

Global

0 0 0 0 0
0 0 0
0 0 0
0 0 0 0 0
0 0 0 0 0

12
 2. Direct Approach
 1-D Problems: Stress (1)

4. Assembly 1
(2)
2
(3)
The global stiffness matrices for elements (3) and (4) are:
3 (4)
0 0 0 0 0
4 (5)
0 0 0 0 0
0 0 0
0 0 0
0 0 0 0 0

0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
0 0 0
0 0 0

13
 2. Direct Approach
 1-D Problems: Stress
 Notes on Global Stiffness Matrix:
4. Assembly
• There is one row per Degree of Freedom (DoF)

Assembly is performed by adding the global stiffness • For higher dimensions, extra rows and columns
matrices of all elements: are added to accommodate extra DoFs in the y-
and z-axes. e.g. in 3-D, and are n x 3 for a
system with n nodes. (considering only
translation DoFs)
0 0 0
0 0 • The matrix is symmetric i.e. .
0 0
0 0 • Diagonal terms are greater than 0 since material
0 0 0 properties are positive scalars i.e. a positive
force occurs in the element (tension) for an
increase in element length

14
 2. Direct Approach
 1-D Problems: Stress (1)

5. Applying Boundary Conditions 1

(2)
• The top of the bar is fixed, so 0 2
(3)
• The external load, , is applied at node 5
• Apply the boundary condition and load to the governing equation: 3 (4)
4 (5)
0 0 0
0 0 0
0 0 = 0
0 0 0
0 0 0
• Minimum number of B.C.’s required

Dimension of DOFs to be constrained

 Notes:
problem
• The stiffness matrix may no be longer symmetric after 1D 1: ux
application of B.C.’s 2D 3: ux, uy, Rz
• Sufficient B.C.’s must be applied to obtain a unique solution
3D 6: ux, uy, uz, Rx, Ry, Rz
and prevent rigid body motion
15
 2. Direct Approach
 1-D Problems: Stress
5. Applying Boundary Conditions

 Notes:

• Examples of applying boundary conditions in 2D

R=0

• If all prescribed displacements are 0 the problem has homogeneous boundary conditions

16
 2. Direct Approach
 1-D Problems: Stress
5. Applying Boundary Conditions
 Notes: Special case:
• If all prescribed displacements are 0, the problem then has homogeneous boundary conditions

• If a prescribed displacement is non-zero, e.g. , the problem then has nonhomogeneous boundary
conditions.

• To solve nonhomogeneous boundary conditions problem

1- Check which non-zero DoFs being prescribed.

2- Highlight the corresponding column(s) in the stiffness matrix which represent stiffness factors to
be multiplied by the prescribed displacement.
3- multiply the columns to their corresponding prescribed field variable and bring the whole term
to the right hand side of the equation of motion.
4- add the new term to the force vector.
5- Now nonhomogeneous DoFs are turned zeros which allows eliminating their corresponding
rows.

17
 2. Direct Approach
 1-D Problems: Stress
5. Applying Boundary Conditions
 Notes: Special case:
0 0 0
0 0 0
0 0 = 0
0 0 0
0 0 0

0 0 0 0
0 0 0
0 0 + 0 δ= 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0
0 0 = 0 0 δ
0 0 0 0
0 0 0 0 18
 2. Direct Approach
 1-D Problems: Stress
(1)
5. Applying Boundary Conditions
1
(2)
 Notes: Special case:
2
0 0 0 0 δ (3)
0 0 δ 3 (4)
0 0 = 0
0 0 4
0 (5)
0 0 0

• Solve the matrix equation:

0 0 • The reaction force, if required, can be
0 0 obtained in the post-processing step from
=
0 0
0 0 or

19
 2. Direct Approach
 1-D Problems: Stress
(1)
6. Solve the set of simultaneous equation to obtain the deflection
1
(2)

We must solve 2
(3)
′ ′
3 (4)
Where
′ is the global stiffness matrix after B.C.’s 4
′ is the vector of external (applied) loads (5)

Various methods available to solve

′ ′

• If the inverse of ′ is easily obtained, then

′ ′

20
 2. Direct Approach
 1-D Problems: Stress
(1)
7. Post-Processing
1
(2)

• Once the full displacement vector has been obtained, we can find reaction forces, 2
(3)
internal forces, strains, stresses, etc… as desired
3 (4)
4 (5)
 Reaction forces
• Occur at any node with specified displacement B.C.
• In this case at node 1
• Use complete governing equation (from Step 4)

0 0 0
0 0 0
0 0 = 0
0 0 0
0 0 0

i.e.

21
 2. Direct Approach
 1-D Problems: Stress
(1)
7. Post-Processing
1
(2)
2
 Internal forces (3)
• Force exerted on element by node 3
• In global system these cancel (sum=0) and don’t appear (4)
• Use elemental stiffness matrix e.g. 4 (5)

 Elemental strain
• Average strain is expressed as
• Shape functions defined from terms in brackets

• Since displacement is a linear function of

• Strain is constant in a linear element:

22
 2. Direct Approach
 1-D Problems: Stress
(1)
7. Post-Processing
1
(2)
2
(3)
 Elemental stress
• From Hooke’s law, stress is expressed as 3 (4)
4 (5)

Since strain is constant for a linear element, stress is also constant

Note for this simple case, by inspection we can also calculate

23
 2. Direct Approach
 1-D Problems: Stress
(1)
7. Post-Processing
1
(2)
2
(3)
 Intermediate values
• If we want the displacement (or other variable) of a specific point inside an 3 (4)
element, use the shape functions
4 (5)

We only need the initial (undeformed) coordinate of point,

Where

24
 2. Direct Approach
 1-D Problems: Stress
Notes: For complicated Geometries

• More complex geometry can be analyzed using the direct approach if we break the geometry down properly, e.g.

(Moaveni, Example 1.4)

• Balance forces at each node to obtain set of governing equations

• e.g. at node 2:

25
 2. Direct Approach
 1-D Problems: Stress
Notes: For complicated Geometries

• More complex geometry can be analyzed using the direct approach if we break the geometry down properly, e.g.

If using the direct stiffness approach, note that elemental stiffness matrix for elements
2 and 3 involve the same nodes

0 0 0 0
0 0
0 0
0 0 0 0

0 0 0 0
0 0
0 0
0 0 0 0 26