Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
http://www.iaeme.com/IJMET/index.asp
148
editor@iaeme.com
1. INTRODUCTION
Finite element Analysis (FEA) plays a very important role in the overall mechanical
design process. It has been applied successfully to almost all kinds of problems and
the complexity of problem ranges from static to dynamics and multyphysics
problems. Today much of the work in CAE (Computer Aided engineering) is done
with the FEA softwares such as ANSYS, NASTRAN, ABAQUS etc. Stress
concentration in machine elements [1,4,5 ] has been of much academic focus in the
textbooks and the data on how much is the rise in the stress level as compared to the
nominal stress at the discontinuity is available for some standard configurations in the
literature [2,3,6,7,8,9] . It is our experience that the industrial Finite Element Analysis
user is least aware of the standard data and most of the times the computations are
done by blind use o the selection of an element and good mesh quality. This paper
presents a comparison of the stress concentration in a rectangular plate due to a
circular hole and we compare here several elements and different types of meshes. We
notice that it is very much beneficial for a practical finite element to be aware of this
and then they can predict these effects which are of utmost importance in all industry
verticals such as automobile, aerospace , pressure vessel, process piping to name a
few. Stress concentration is not discussed in depth in any of the Finite Element
textbooks [10, 11, 12] in academics and hence it is important for an user on how the
element choice and mesh quality affects the results. We believe that the CAE
Community will benefit from the present detailed study when it comes to solving
practical problems.
2. THE PROBLEM
The problem considered here is a plate of length 100 mm and width 40 mm and with a
thick ness is subjected to a tension of 900N and this force is applied as uniformly
distributed force. The thickness of the plate in the present problem is taken larger so
that both the solid and shell elements of various types can be compared. The
performance of solids is not discussed much in the literature and this is the reason we
have taken a configuration which can be useful to a variety of problems in Finite
Element Analysis. Fig.1 represents the loading and the dimensions.
B
100mm
40mm
= 20mm
B
http://www.iaeme.com/IJMET/index.asp
149
editor@iaeme.com
=2.25MPa
=4.5MPa
=9.714MPa
Figure 2
= 2.25 MPa
The nominal stress at the section B-B due to the hole cut out is given by
=
= 4.5 MPa
The stress at the point c and d which face the maximum stress levels is given by
=Kt *
http://www.iaeme.com/IJMET/index.asp 150
editor@iaeme.com
Thus the exact value of the maximum stress at the hole at point C should be
9.714 MPa.
The representation of the stress levels in the plate at different sections and points
is shown in Fig 2.
http://www.iaeme.com/IJMET/index.asp
151
editor@iaeme.com
Mesh refinement study was carried out and the results of which are presented in
Figures Fig. Nos. 5 and 6.
http://www.iaeme.com/IJMET/index.asp 152
editor@iaeme.com
http://www.iaeme.com/IJMET/index.asp
153
editor@iaeme.com
http://www.iaeme.com/IJMET/index.asp 154
editor@iaeme.com
Various types of refined meshes are shown in the figures given below. These are
dependent on the various algorithms used for mesh splitting. We have found that in
one case it leads to a polyhedral type of mesh used very much in Computational fluid
dynamics nowadays. Although it is having lots of advantages with respect to
computational time by using a Finite Volume Method, we have to keep in mind that
such kind of meshes are not used in Structural mechanics. We also wish to keep in
mind that there are some studies going on in Structual mechanics on showing that
Finite Volume Method can be used for structural mechanics but this is limited only to
one dimensions and on the other hand Finite element methods have been widely used
http://www.iaeme.com/IJMET/index.asp
155
editor@iaeme.com
in the area of Computational Fluid dynamics. The mesh and the stress plots of these
refinements are shown in Fig Nos. 11 to 14.
http://www.iaeme.com/IJMET/index.asp 156
editor@iaeme.com
Figure 15 Stress plot of Another refinement of TRIA mesh Looking like a Polygon
http://www.iaeme.com/IJMET/index.asp
157
editor@iaeme.com
http://www.iaeme.com/IJMET/index.asp 158
editor@iaeme.com
http://www.iaeme.com/IJMET/index.asp
159
editor@iaeme.com
http://www.iaeme.com/IJMET/index.asp 160
editor@iaeme.com
The tetrahdraons are of two type. The four noded ones (TETRA4) are first order
accurate and the ten noded ones (TETRA10) are second order accurate. It is usual to
use second order tetrahedrons as the first order ones are quite stiff in practice. That is
they give very less displacement as compared to the exact displacement. Most of the
automeshing in structural mechanics is done with second order tetras whereas for
fluids most of the meshing is done using first order accurate tetras as the equations
involve only first order derivatives.
The coarse and fine meshes for the first order tetrahedral elements are shown in
Fig. Nos. 27, 29 and Fig. 31 shows the coarse mesh for a second order accurate
tetrahedral element. The corresponding stress plots are given in Fig. Nos. 28, 30 and
Fig.No. 32.
http://www.iaeme.com/IJMET/index.asp
161
editor@iaeme.com
http://www.iaeme.com/IJMET/index.asp 162
editor@iaeme.com
The Pentas are 6 noded elements commonly called as wedge or prism elements
and are not used commonly in structral mechanics as they are very stiff. However
http://www.iaeme.com/IJMET/index.asp
163
editor@iaeme.com
they have to be used as filler elements as complex geometries with hexahedral meshes
present lots of problems at the interfaces .
http://www.iaeme.com/IJMET/index.asp 164
editor@iaeme.com
Sr.
No.
Element
Displacement
obtained by FEA
solution in mm
QUAD4
Coarse Mesh
0.149e-2
6.03
QUAD4
Fine Mesh
0.161e-2
7.50
QUAD4
Coarse Ruled
Mesh
0.1e-2
5.81
QUAD4
Fine Ruled
Mesh
0.11e-2
7.58
QUAD4
Coarse
Automesh
0.1e-2
5.10
QUAD4
Fine Automesh
0.11e-2
6.40
QUAD8
Coarse Mesh
0.165e-2
6.04
QUAD8
Fine Mesh
0.175e-2
7.47
TRIA3
Coarse Mesh
0.142e-2
7.29
10
TRIA3
Fine Mesh
0.142e-2
7.69
11
QUAD4
obtained
from TRIA3
Fine Mesh
polygonal type
0.157e-2
8.20
12
TRIA6
Coarse Mesh
0.159e-2
6.91
13
TRIA6
Fine Mesh
0.163e-2
8.03
14
HEXA8
Coarse Mesh
0.154e-2
5.90
15
HEXA8
Fine Mesh
0.152e-2
7.37
16
TETRA4
Coarse Mesh
0.148e-2
7.68
17
TETRA4
Fine Mesh
0.142e-2
9.11
18
TETRA10
Coarse Mesh
0.162e-2
7.65
19
PENTA6
Coarse Mesh
0.176e-2
6.98
http://www.iaeme.com/IJMET/index.asp
165
editor@iaeme.com
The major observations of the finite element analysis based on the table given above
are as follows:
1.
2.
3.
4.
5.
6.
7.
The ruled mesh of quadrilaterals is unable to predict the stresses due to worst mesh
quality. The value of stress precdicted is much smaller than the exact value of 9.714
MPa. Such a kind of mesh should not be used for stress analysis. If a cfd result is
available from such a mesh then that result should be appropriately interpolated
before it can be used for structural analysis. Even though stress improves on a finer
mesh, displacement is not and hence this comment.
If one increases the order of the element from QUAD4 to QUAD8 then displacement
improves but the stress does not. This means that whether it is QUAD4 or QUAD8,
reasonable stress representation is obtained only on a fine mesh with a better
performance obtained by QUAD8 as compared to QUAD4.
The automesh results are also of the worst quality. Typically there is a blind practice
followed by the industry user to use an automesh and make it fine but as one can see
from the table, this belief is proven totally wrong. On a fine mesh, one doesnt get the
required level of accuracy and still the stress is represented with a very less value.
The TRIA3 performs much better as compared to QUAD4. However the industry and
academics opinion about this element is that it is too stiff. This is true only for
bending. The present problem is of pure axial stress and a good performance of the
element is seen for such problems. The polygonal looking type of mesh performs
much better as compared to conventional refinement and this is not noted in the
literature of finite elements.
Increasing the order of the element from TRIA 3 to TRIA6 doesnt necessarily
improve the stress. CTRIA 6 performs better on fine mesh with results comparable to
fine mesh of TRIA3.
HEXA8 performance improves on a fine mesh as compared to coarse mesh. Industry
belief is that hexahedral mesh is alsways better than the tetrahedral one is not
necessarily true. If one can use the automesh option using TETRA10 then results are
much accurate to the exact solution. Further for practical problems the time taken to
generate a hexahedal of tetesh is an order of magnitude greater than the tetrahedral
mesh. TETRA4 is better for stress analysis but displacement is predicted on the
conservative side and further cannot represent bending properly . So in our view a
fast algorithm for automesh with TETRA10 elements is much better as compared to
HEXA8.
The PENTA element does not represent the bending and is too stiff for frequency
calculations which is shown in the reference [13] and hence is not to be used for
practical finite element analyses where the same kind of mesh is to be used further for
vibration. For static stress analysis, results are much lesser than TETRA4 coarse and
TETRA10 coarse mesh. This justifies that this element is to be used only as a filler
element wherever necessary and should not be used in general.
6. CONCLUSION
We have presented a comparison of several meshes and element types in this paper
for analyzing a standard problem of stress concentration in a plate. One can see that
there is lot of variation of the results and if the user expects that he will get closer to
the exact value by using a fine mesh then it is not true. Using a fine mesh is also not
practical always. Thus the practical finite elementer has to keep in mind two things 1.
appropriate choice of element and 2. Desired mesh quality with lesser of distorted
elements. It is the authors observation sometimes that people simply fall in love with
their meshes and then dont worry on the accuracy. This can be misleading. e.g. the
ruled mesh look fantastic but has severly distorted quadrilaterals and such types of
http://www.iaeme.com/IJMET/index.asp 166
editor@iaeme.com
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
http://www.iaeme.com/IJMET/index.asp
167
editor@iaeme.com