F.E.a Assignment
F.E.a Assignment
Table of contents
Introduction ................................................................... 3
Theory ............................................................................ 5
Apparatus....................................................................... 5
Procedure ....................................................................... 5
Calculations.................................................................... 7
Discussion ....................................................................... 8
Conclusion.................................................................... 10
References .................................................................... 10
Appendices ................................................................... 10
2
Introduction
Saint-Venant's principle named after the French elasticity theorist Jean Claude Barré de Saint-Venant
can be stated as
"... The strains that can be made in a body by the application, to a small section of its surface, of a
system of forces statically approximate to zero force and zero couple, are of negligible magnitude at
distances which are large compared with the linear dimensions of the part."
The initial statement was published in French by Saint-Venant in 1855. Although stated verbally the
principle is familiar among mechanical engineers in this informal formulation. More present
mathematical literature gives a precise view in the context of partial differential equations. An early
participation to this came from von Mises in 1945.
The Saint-Venant's principle permits elasticians to replace complex stress distributions or weak
boundary conditions into ones that are easier to solve, as long as that boundary is geometrically short.
Quite similar to the electrostatics, where the electric field because of the i-th momentum of the load (
with 0th being the remaining charge, 1st the dipole, 2nd the quadrupole) decays as over space,
Saint-Venant's principle mentions that high order momentum of mechanical load ( momentum with
order higher than torque) decays so quick that they don’t need to be considered for regions far from
the short boundary. Therefore, the Saint-Venant's principle can be considered as a statement on the
asymptotic behavior of the Green's function by a load at point.
Loads must be dispersed over a finite area. In practice it is quite hard to find out accurate
stresses close to the point of application of a load. Engineers deal with this problem by applying
St. Venant's principle which indicates that statically equivalent systems of forces make the
same stresses and strains within a body except in the immediate region where applied loads
are present. Hence, the stresses deduced in the middle of a beam are not affected by the way
the ends are supported as long as the supporting forces and moments are statically alike to
those in the mathematical model. Local problems are dealt with by separate methods.
Aim
When examining a circular hole in a plate, the circular hole in the middle will cause a
concentration of stress. This will increase the level of the stress at specific points near the hole
to a value higher than the nominal stress in the rest of the plate. The actual aim of this
investigation/ exercise is to find the stress concentration factor at point A on the plate and to
get the longitudinal stress profile between the points A & B. It is also necessary to study the
value of the stress at point C. When carrying out this exercise two different types of elements
would need to be used and a convergence study would need to be performed with each
element.
3
Problem Specification
Geometry:
A long plate of thickness (t) and depth (D), as shown above. The plate can be suggested to be of
infinite length, thus, the deficiency in length dimension. The plate has a circular hole with a
specific diameter (d), situated in the middle of the plate precisely.
Loading:
Uniformly distributed load (w= 150 N/mm) over the ends of the plate, as shown in the figure
above.
Properties of the material:
Young’s modulus (E= 7000 N/mm²), Poisson’s ratio (ν= 0.33).
Properties of the geometry:
Thickness of the plate, t= 4 mm
Depth of the plate, D= 50 mm
Diameter of the hole, d= 20 mm
Type of elements:
2D plane stress 4 & 8 noded quadrilateral elements
4
Theory
The main concept of the St. Venant’s principle which is related to modeling the length of the
plate is to assume/ estimate a length with dimension of more than twice the depth of the plate,
i.e. L > 2D.
The formula to calculate the longitudinal stress concentration factor at point A is given below:‐
Where K= the longitudinal stress concentration factor, d= the diameter of the hole (mm), D=
the depth of the plate (mm).
The empirical formulae (hand calculation) involved in this assignment are as follows:‐
K= and =
Where K= the longitudinal stress concentration factor, = the peak longitudinal stress (N/
mm²) at point A,
= the nominal stress (N/ mm²), = uniformly distributed tensile load (N/ mm), D= the
depth of the plate (mm), t= the thickness of the plate (mm), d= the diameter of the hole (mm).
Apparatus
For this assignment, the finite element analysis software (Ansys) was used.
Procedure
Outline
This exercise required the use of engineering decision in the size of length of the plate either
side of the hole it was necessary to model. This was essential that adequate length was
modelled so that at the limits of the model the stress field had established from the disturbance
due to the hole. To determine this St.Venant’s principle was used, i.e. L > 2D. For this
assignment the length was 2.5 times more than the depth, i.e. L= 2.5 × D= 2.5 × 50= 125 mm.
With this length size, symmetry was used in order to cut down the size of the actual finite
element model. Then the part of the plate that was modelled was split up into 3 geometrical
areas, each one with four sides. The boundary conditions were stated by the use of symmetry.
Method of analyses using the Ansys software
4 Noded quadrilaterals (4 Node 42):
5
• The geometry of the shape was formed by making the key points. To do this the
preprocessor was opened, modelling was clicked on followed by create then key
points in active CS. All the values were typed inside the box for each key point.
• Then
lines
were
connecte
d to each
key
point, by
opening
the
preproc
essor,
then
modelli
ng was
clicked
on
followed
by
create
lines.
Then in
the lines menu, the straight line was chosen. All the key points were joined with lines
into 2 split rectangles/ squares. Then the left square/ rectangle was split into two
triangles.
• Next, the rectangle/ square and triangles were filled with the areas, i.e. they all turned
blue indicating they had areas. This was done again by opening the preprocessor, then
clicking on modelling and then on create and then areas, arbitrary and by lines. So
each shape was filled in separately, by clicking on each line then on apply, to fill the
shape. Then ok was selected to close the box.
• A circle was made at the origin (0, 0) for the next step. This was done by clicking on
preprocessor and then on modelling. Then on create areas, arbitrary, circle and then
solid. In the box, zeros were put for x and y and then a radius value which was half of
the diameter.
• The circle was subtracted from the rectangle, by opening the preprocessor and then
clicking on modelling, operate, Booleans, subtract and then on areas. The rectangle
was selected including the triangles and then apply was clicked on, this changed the
colour to pink. The same was done for the circle and then the ok button was selected.
• Next the element type was chosen by clicking on preprocessor, element type and
add/edit/delete. In the first box add was selected, then in the second box solid and 4
node 42 were selected. This was changed to 8 node 82 for this analysis. Then in the
element type window, the options button was chosen. In the window, K3 was changed to
plane stress w/thk and K5 to nodal stress.
6
• Then the geometric properties were defined by clicking on then preprocessor, then on
real constants and then on add/edit/delete. The add button was selected and then type
1 plane 42 was selected. Then a thickness value was added of 4.
• Next the Young’s modulus and poisson’s ratio were added. By opening the preprocessor
and then clicking on material props and then on material models. Then structural was
selected, then linear, elastic and isotropic.
• The next step was to mesh the element using the mesh tool in the preprocessor. Number
1 was typed in the window in the number of elements divisions box. For the other mesh
densities the number of element divisions were changed to 2, 4 & 8.
• After this step, the static loads/ displacements were applied to the model. Then the
pressure was applied to the model.
• Next, the problem was solved by clicking on solution, solve, current LS.
• Next, the results were plotted on a contour plot, nodal solution. The results ere listed of
the nodal solution.
• Then the nodes were plotted. including the numbers/ elements
• This procedure was repeated for other mesh densities and the 8 node 82 element type.
Results & Graphs
The results are in a spreadsheet in appendix 1 and the graphs are in appendix 2.
Calculations
Longitudinal stress concentration factor at point A for a mesh density of 8:‐
7
K= 3 – 3.13 + 3.66 ‐ 1.53
K= 3 – 1.252 + 0.5856 – 0.09792
K= 2.236
Empirical (hand calculated values) for a 4 node 42 element with mesh density 8 for the
longitudinal stress concentration factor (K):‐
= = = = 62.5 N/mm²
= 138.57
K= = = 2.217
Empirical (hand calculated values) for a 8 node 82 element with mesh density 8 for the
longitudinal stress concentration factor (K):‐
= 62.5 N/mm²
= 137.61
K= = = 2.202
Discussion
From the graph of longitudinal stress at point A against the mesh density (log n) in Appendix 2,
for the 4 node 42 curve, as the mesh density (log n) increases from 0 to 2, the longitudinal
stress at point A also increases steadily. Then from mesh density (log n) 2 to 3, the curve
flattens out. This shows that the longitudinal stress only increases slightly up to this point. This
demonstrates that as the number of elements increase, the longitudinal stress at point A is
more. Moreover, for the 8 node 82 curve, as the mesh density (log n) increases from 0 to 1, the
longitudinal stress at point A also increases higher than the 4 node 42 curve. Then from mesh
density (log n) 1 to 2 the curve slowly flattens out, i.e. the longitudinal stress at point A slowly
increases. This happens more from mesh density (log n) 2 to 3. This again shows that the
longitudinal stress at point A is high when more elements are used. Both curves converge to the
same point at mesh density (log n) 3; again this is where the element number is high (192) and
the real mesh density (n) is 8.
For the next graph in appendix 2 of longitudinal stress at point C against the mesh density (log
n) for the 4 node 42 curve, as the mesh density (log n) increases from 0 to 1, the longitudinal
stress at point C decreases slowly. Also from a mesh density (log n) from 1 to 2, the longitudinal
stress at point C decreases rapidly. Finally from mesh density (log n) 2 to 3, the curve slightly
8
flattens out, i.e. the longitudinal stress at C slowly decreases. This explains that the longitudinal
stress at point C starts off high but decreases vastly with an increasing mesh density. Therefore,
the longitudinal stress at C is low when the element number is high. Also the stresses turn
negative with an increase in mesh density and element number. Moreover, for the 8 node 82
curve, the longitudinal stress value at point C is negative at mesh density (log n) 0. Then as the
mesh density increases (log n) from 0 to 1, the stress increases steadily. Then again from a
mesh density (log n) of 1 to 3, the longitudinal stress at C increases slower. The two curves
again converge near enough to the same point. This shows that at a high element number or
mesh density the stress value is almost the same for both nodal elements.
Moreover, from the graph of longitudinal deflection of point C against mesh density (log n) for
the 4 node 42 curve as the mesh density (log n) increases from 0 to 1, the longitudinal
deflection at C increases at a high rate. From mesh density (log n) 1 to 2 the deflection at C
increases even more at a high level. Finally, from log n 2 to 3, the deflection at C increases a
little. For the 8 node 82 element curve as log n increases from 0 to 1, the longitudinal deflection
at C increases steadily higher than the value of the 4 node 42 curve. Then from a mesh density
of 1 to 3 (log n) the deflection at C increases a little then goes to a constant value. Again they
almost converge to a same value. This shows that as the element number increases/ mesh
density the longitudinal deflection at point C is high.
Next, for the graph of the percentage error in the stress values from the stress fields of the 2
elements against the mesh density for the 4 node 42 curve in appendix 2 as log n increases
from 0 to 1, the percentage error increases quite steadily. But when the mesh density increases
from 1 to 3, the percentage error decreases steadily. For the 8 node 82 curve, as the mesh
density increases from 0 to 1, the percentage error increases a little but to a lower value than
the 4 node 42 curve. Then from log n 1 to 3, the percentage error decreases by a big margin.
This shows that at a higher mesh density (4 & 8) and a higher element number (48 & 192), the
percentage error decreases rapidly. Additionally, the 8 node 82 quadrilateral element produces
more accurate results than the 4 node 42 quadrilateral element.
Additionally, for the graph of the percentage error in the stress values from the stress fields of
the 4 elements against the mesh density for the 4 node 42 curve, as the mesh density (log n)
increases from 1 to 2, the percentage error decreases slowly. As log n increases from 2 to 3 the
percentage error decreases further. Moreover, for the 8 node 82 curve, the percentage error
decreases more than the 4 node 42 curve from log n (1 to 3). This demonstrates that again at a
higher mesh density or element number the percentage error decreases vastly. Also that the 8
node 82 element is more accurate than the 4 node 42 element as there is less percentage errors
in the stress values at nodes between 4 elements.
Finally, for the last graph in appendix 2, i.e. the longitudinal stresses at mesh density 8 between
points A & B using nodally averaged values against the node number for both the 4 node 42
element and the 8 node 82 element as they are almost identical, at node 1 (A), the stress value
is high. Then from nodes 1 (A) to 9 (B), the stress values decrease by a big difference. This
shows that the average stress value between A & B decreases from A to B. So at point B (away
from the hole), the stress value is low and the more nearer to the hole the more the stress
value, whether the quadrilateral element is a 4 node 42 type or a 8 node 82 type.
From the calculations of the longitudinal stress concentration factor at A with mesh density 8,
the value is 2.236 which is close to the empirical (hand calculated) value for the 4 node 42
9
element of 2.217. This shows good agreement between the 2 values. So this shows that the
longitudinal stress values at point A are accurate and reliable. For the 8 node 82 element, the
empirical value for K= 2.202, again which is close to the original K value of 2.236. Again this
shows that the results are accurate and reliable.
Further work which could be done for this type of problem is maybe to use more different
types of mesh densities or more types of quadrilateral elements or different shaped elements.
Conclusion
The higher the mesh density or element number, the higher the longitudinal stress at point A.
Also the longitudinal stress at point C decreases for a 4 node 42 element, with an increasing
mesh density. Also for an 8 node 82 element, the stress at C increases with increasing mesh
density, but it is negative. Moreover, as the element number increases/ mesh density the
longitudinal deflection at point C is high. The more meshed the quadrilateral element, the more
accurate the results. Also the 8 node 82 element is more accurate to use for this assignment
problem as there are less percentage errors in the results. Additionally, the stress values were
accurate as the stress concentration factor value for empirical and the original value were
almost the same for both types of elements. The stress values are higher nearer to the hole and
they are lower away from the hole. Further work could be done to extend the investigation
mentioned in the discussion.
References
• Practical finite element analysis lecture/ lab book.
• http://en.wikipedia.org/wiki/Saint‐Venant%27s_principle
• Assignment handout
Appendices
These consist of nearly all the results analysis on separate pages.
10
1 123.08 1 -5.9746
2 132.84 2 -2.6483
3 137.61 3 -1.0515
8 Node
82:
Deflectio
4 Node 42: n at
Deflection at point C point C
Deflection
log n (Ux) log n Deflection (Ux)
0 0.17222 0 0.17548
1 0.17628 1 0.19185
2 0.18746 2 0.19376
3 0.19208 3 0.19389
2 66.489
70.027
59.602
66.907 15.854
Percent
age
error
3 90.668 log n (%)
93.377 1 19.345
86.269 2 15.854
11
90.452 7.881 3 7.881
2 70.528
70.34
65.373
66.295 6.213
Percent
age
error
3 92.532 log n (%)
92.507 1 14.942
90.964 2 6.213
91.095 1.709 3 1.709
12
13
14
15