Vous êtes sur la page 1sur 13

DEPARTMENT OF MECHANICAL ENGINEERING

ALL Sections must be completed using Block Capitals or typed

CODE AND TITLE OF COURSEWORK


Course code:
MECHM003

Title:
FEA Assignment 2

STUDENT NAME:
Noureddine El Saidi
DEGREE AND YEAR:
MEng Mechanical Engineering 4th Year
LAB GROUP:
N/A
DATE OF LAB. SESSION:
N/A
DATE COURSEWORK DUE FOR SUBMISSION:
12/2/2016
ACTUAL DATE OF SUBMISSION:
12/2/2016
LECTURERS NAME:
Dr M Moazen
PERSONAL TUTORS NAME:
Dr Frome
RECEIVED DATE AND INITIALS:

I confirm that this is all my own work (if submitted electronically, submission will
be taken as confirmation that this is your own work, and will also act as student
signature)

Signed:

University College London, Torrington Place LONDON WC1E 7JE

Finite Element Methods for Stress Analysis (Ver. 2015/16)

A2 - 1

MECHGM05/GR05/M003: Advanced Computer Applications in Engineering


Finite Element Analysis - Coursework 2- Mark Sheet
Name:
Course:
Grade:
Introduction
[5 marks]

Comments:

University College London

Materials and
Methods
[15 marks]

Results

Discussion

[50 marks]

[15 marks]

Overall
presentation
[15 marks]

Total mark
[100 marks]

Implementation of Finite Element Analysis Equations and


Investigation of Axially Loaded Thin Plate
Noureddine El Saidi
UCL Meng Mechanical Engineering
noureddine.saidi.12@ucl.ac.uk
Word Count: XXXX
February 7, 2016

Abstract

MECHM003: FEA Assignment 2

Noureddine El Saidi

Contents
1 Introduction

2 Materials and Method

2.1

FE Solver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.2

Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.3

Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3 Results
3.1

Element Stiffness Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.1.1

3-node Element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.1.2

6-noded Element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.2

Standard Plate Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.3

Parametric Study Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4 Discussion

10

References

11

Page 2

MECHM003: FEA Assignment 2

Noureddine El Saidi

Introduction

In this study, a fully automated two-dimensional finite element solver in MATLAB was developed using three node linear and six node quadratic triangular elements, which was then used to
investigate the strain within a small thin plate exposed to a distributed force in the axial direction.

Figure 1: Diagram of the plate analysed in this study


The aim of the study was twofold, to first develop an understanding into the Finite Element
equations, and secondly to investigate the mechanical response of a plate to an axial load.

2
2.1

Materials and Method


FE Solver

A MATLAB script was developed which automatically solves the finite element equations and
begins by reading an input file containing material properties, nodes, elements, nodal boundary
conditions and nodal forces. The first step is the generation of the element stiffness matrix.
To optimise the solution, the shape function derivatives for the linear and quadratic triangular
elements are embedded in the code, and used the values .
Z Z
Ke = t



BT DB dxdy ' t BT DB

(1)

Where t is the plate thickness, h is the weighing, D is the compliance matrix for plane strain shown
below.

1
E

D=
1 2
0

1
0

0
1
2 (1 2)

(2)

Where E is the stiffness modulus and the Poisson ratio.


The B matrix, which considers the shape function is calculated as follows:
B = B1 B2 B3

(3)
Page 3

MECHM003: FEA Assignment 2

Where:

Noureddine El Saidi

1 0 0 0
B1 = 0 0 0 1
0 1 1 1
1

1
J11 J12
0
0
J 1 J 1
0
0
21
22

B2 =
1
1
0
0
J11
J12
1
1
0
0
J21
J22

N1
n
0
N
0
L1
L1
n
N1
0
N
0
L2
L2

B3 =
N
N
1
n
0

0
L1
L1
N1
Nn
0

0
L2
L2

(4)

(5)

(6)

And the Jacobian J calculated from multiplication of the


A single Gaussian integration point was used for the linear triangle at the centroid of the element.
Three integration points were used in the case of the 6 node element, which were found at L = 16
and L = 32 interchangeably, with equal weighting of 61 .
The global stiffness matrix was assembled by transforming the local nodes within each elements
stiffness matrix into the global nodes, and then positioning them in the appropriate place.
Boundary conditions were applied by removing the restricted degrees of freedom from the global
stiffness matrix.
As a pressure force was applied, the nodal forces were calculated using the work equivalent theorem,
such that the resultant force was split 1:1 and 1:4:1 for 3-node elements and 6-nde elements
respectively.
The nodal deflections can then be calculated using the equation below, which was subsequently
solved by Gaussian elimination.
K = F
(7)
The strain at a Gaussian integration point was calculated using:
Ue = Be e

(8)

This was extrapolated to the element corners using the standardised method of bilinear extrapolation using the shape function, as presented in [1][2] and shown below:
e = Ng

(9)

The shape function can be found in the appendix.


The Principal strain was calculated using the methodology found in [3]. From this, the von-Mises
strain was calculated as below [2]:
q
1
1
(10)
v =
(1 2 )2 + 21 + 22
21+
Where 1 and 2 are the principal strains.
The strain contour plot was plotted using bi-linear interpolation within each element using Matlabs
patch function.
Page 4

MECHM003: FEA Assignment 2

2.2

Noureddine El Saidi

Simulations

A number of different simulations were run to investigate the mechanical response of the plate,
The dimensions of the plate are shown in figure ?? but are shown below for convenience.
E = 100GP a
= 0.3
t = 0.002m
The first analysis was calculation of the stiffness matrix for:
1. Linear 3-node element
2. Quadratic 6-node element
The second analysis was a finite element analysis on the plate to calculate the displacement and
strain, as demonstrated in figure ??, using:
1. Two linear 3-node elements
2. Two quadratic 6-node elements
3. Four linear 3-node elements
4. Four quadratic 6-node elements
The final analysis was a parametric study to investigate the effect of:
1. Youngs modulus
2. Loading conditions
On the mechanical response of the plate, using a mesh comprised of four quadatic 6-node elements.

2.3

Validation

It is important to validate the method to ascertain its ability to model the scenario correctly and
produce valid results.
Ideally, an accurate mathematical representation of the problem would be created, however in
the absence of the capability for this a simplified stress strain approach is taken to calculate the
deflection expected directly from the definition of the Youngs modulus.
=

L
E

(11)

Which results in a deflection of 4 106 m for a uniformly stressed rod. The end deflection of the
plate measured was calculated to be
Therefore, further validation was necessary which was done by comparing the results with the
existing validated and established finite element package ANSYS. The problem was recreated in
Page 5

MECHM003: FEA Assignment 2

Noureddine El Saidi

ANSYS with both 4 elements and 96 elements, and produced identical results with those generated
by the developed MATLAB code. Therefore, it can be concluded that the method is valid.
The results are included in Appendix 2.

3
3.1

Results
Element Stiffness Matrices

Figure 2: Image of 3 node and 6 node element, and node numberings for this part of the excercise,
showing the integration points (crosses) used for the 6 node element

3.1.1

3-node Element
1.3370
0.7143
0.8242
108 0.3846
0.5128
0.3297

K3n =

3.1.2

0.7143
1.7537
0.3297
0.2885
0.3846
1.4652

0.8242
0.3297
0.8242
0
0
0.3297

0.3846
0.2885
0
0.2885
0.3846
0

0.5128
0.3846
0
0.3846
0.5128
0

0.3297
1.4652
0.3297
0
0
1.4652

6-noded Element
1.34
0.71
0.27
0.13
0.17
8 0.11
10 1.10
0.51
0
0
0.68
0.44

K6n =

0.71
1.75
0.11
0.10
0.13
0.49
0.44
0.38
0
0
0.51
1.95

0.27
0.11
0.82
0
0
0.11
1.10
0.44
0
0.44
0
0

0.13
0.10
0
0.29
0.13
0
0.51
0.38
0.51
0
0
0

0.17
0.13
0
0.13
0.51
0
0
0
0
0.51
0.68
0.51

0.11
0.49
0.11
0
0
1.47
0
0
0.44
0
0.44
1.95

1.10
0.44
1.10
0.51
0
0
3.57
0.95
1.37
0.95
0
0.95

0.51
0.38
0.44
0.38
0
0
0.95
4.68
0.95
3.91
0.95
0

0
0
0
0.51
0
0.44
1.37
0.95
3.57
0.95
2.20
0.95

0
0
0.44
0
0.51
0
0.95
3.91
0.95
4.68
0.95
0.77

0.68
0.51
0
0
0.68
0.44
0
0.95
2.20
0.95
3.57
0.95

0.44
1.95
0
0
0.51
1.95
0.95
0

0.95
0.77
0.95
4.68

Page 6

MECHM003: FEA Assignment 2

3.2

Noureddine El Saidi

Standard Plate Results

Figure 3: Plot of the different meshes constructed for task 2, and consquent deflections in each
case
Figure 4: Results of maximum and minimum displacements for investigation of
Max Deflection Min Deflection Max vonMisses Strain
[m]
[nm]
[micro strain]
3 node 2 element 4
0.09
90
3 node 4 element 4
0.09
90
6 node 2 element 5.62
2.36
271
6 node 4 element 6.27
2.12
357

different meshes
Min vonMisses Strain
[micro strain]
90
90
40
51.7

Page 7

MECHM003: FEA Assignment 2

3.3

Noureddine El Saidi

Parametric Study Results

Figure 5: Strain contour plot for the resultant strains from task 3(i) in which the Youngs modulus
of the plate material was altered

Page 8

MECHM003: FEA Assignment 2

Noureddine El Saidi

Figure 6: Strain contour plot for the resultant strains from task 3(ii) in which the orientation of
the applied load of the plate material was altered

Page 9

MECHM003: FEA Assignment 2

Noureddine El Saidi

Discussion

In this study, a finite element analysis was conducted on a thin plate subjected to an axial force.
First, the element stiffness matrix was calculated for the three node and six-node elements. It was
found that at coincident nodes the stiffness value can be seen, as seen along the diagonal. This
was expected as the mechanical response at those points explicitly is the same.
The second analysis compared the difference in using 3-node and 6-node elements. The obtained
results were different, and the resultant deformations and element strains varied. The first observation, which can be seen from figure ??, is a difference in the resultant shape, with both 3-node
element produced straight edge deformation, and displacement in the bottom edge purely horizontal. On the contrary, for both 6-node elements there was bottom edge displacement and the edges
were not straight. Furthermore, at the left edge, this intuitively curved in in the 6-node elements.
This phenomena is known as shear locking, in which there is not a sufficient number of elements
to show bending.
Another difference was the left edge, which was deformed for the 6 node cases, but straight for
the 3 node case. Intuitively, the edge will deform due to the boundary conditions which prevent
horizontal displacement of either points and tension is applied therefore the edge will curve. This
was not present on the 3 node cases, because there were only 2 nodes forming that edge both of
which were contained, therefore any deflection was impossible. This would be fixed by increasing
the number of the elements, and at a sufficiently high resolution the results should be identical.
By inspecting the displacement and there is a significant difference between the two results. The
first is that the strain was uniform in all elements for the 4 node case. This is due to shear
locking, as the plate is unable to deform laterally (as per Poissons ratio), therefore the strain
is uniform through out. Consequently, both 2 element and 4 element cases produced the same
results. Furthermore the deflection matches that derived in section 2.3, as that basic theory was
for uniform stress.
The 6 node element on the other hand was able to accommodate such deflections and therefore
the strain was not uniform. The deflection strain also increased when 4 elements were used.
The final study comprised of changing the Youngs modulus of the material, and then the orientation of the applied force. Changing the Youngs resulted in a change in both the deflection and
strain, which was invesrely proportional to the change in Youngs modulus. This was expected, as
the linear elastic simulation was conducted, therefore the response is fully Hookean, which states
the deflection is directly proportional to the load, where the constant of proportionality is the
Youngs modulus.
Changing the angle of the pressure resulted in a curvature in the the beam, which was expected
as there was a vertical component of the force. There was symmetry due to the roller support.
The location of the maximum strain changed depending on the direction of the load. When this
was pointed up, the maximum stress was in the bottom left corner, and when pointed down the
maximum stress was in the top left corner. The change is due to a change in deflection of the
beam, and given the boundary conditions the edge in tension will have the greater strain as was
seen. The minimum stress was located on the opposite corner as expected.
The model produced was extremely limited, in that it cannot be used to produce meaningful
conclusions. This is because of a small number of elements were used which used. The produced
Page 10

MECHM003: FEA Assignment 2

Noureddine El Saidi

is however valid, and matches that employed by popular FEA programs such as ANSYS, and as
such with a greater number of elements can be used to conduct finite element analysis. As a proof,
the a larger mesh was generated and compared to the output from ANSYS, which shows identical
result successful demonstrating that this method can be expanded to engineering problems.
Improvements - Appendix

References
[1] R. Cook, D. Malkus, and M. Plesha, Concepts and Applications of Finite Element Analysis.
New York, US: John Wiley and Sons, 1974.
[2] ANSYS inc, ANSYS Mechanical APDL Theory Reference.
2013.

Pennsylvania, US: ANSYS inc,

[3] S. Timoshenko and J. Gere, Mechanics of Materials, 3rd ed. London, UK: Chapman and Hall,
1991.

Page 11

Vous aimerez peut-être aussi