Documente Academic
Documente Profesional
Documente Cultură
Esther Koo
Chen Shen
Brian Chantrupon
Esther
Brian
Chen
efk28
cs829
bc566
Sources: None
10 hours
12 hours
12 hours
2) One quarter of the plate was taken, so the new geometry became width = 0.15 and
length = 0.5 . The quarter-plate was meshed using rectangular elements.
MeshGeometry.m
% Choose a meshing method
% "1" for mesh directly generated from Matlab "BoxGrid_2D" function
% "2" for mesh generated from Ansys and loaded from "loadFromGridFile"
% function
LoadChoice = 1;
% Plot instruction
PlotInstructions.plot_mesh
PlotInstructions.plot_node
= 'no';
= 'no';
PlotInstructions.plot_boundary = 'no';
PlotInstructions.plot_contour
PlotInstructions.plot_vector
PlotInstructions.plot_deformed
PlotInstructions.plot_fringes
nnpe = 4;
=
=
=
=
%
%
%
%
%
'yes';
'yes';
'yes';
'yes';
nsd = 2;
% number of spatial dimensions
if LoadChoice==1 % use LoadChoice =1 only when the geometry is a
% rectangular
% Set this information for each problem
xl
xr
yb
yt
=
=
=
=
0.0;
0.15;
0.0;
0.5;
nx = 12;
ny = 12;
%
%
%
%
For symmetry boundary conditions, the second surface (left) had no horizontal displacement
and no tangential traction. The bottom surface was broken up into two surfaces: the first
surface (bottom) and the fifth surface (crack). The first surface had symmetry conditions with
no vertical displacement and no tangential traction. The fifth surface was treated as a free
surface (no tangential or normal tractions). The only natural boundary condition entered in
InputData.m was = 30 at the third surface (top). Since the plate was thin, the plane
stress condition was used.
InputData.m
% Define the essential BCs
% boundStruct.SurfEssV = [];
boundStruct.SurfEssV = [2 1 0
1 2 0];
%
%
%
%
The crack was located at the last third of the bottom surface, so boundStruct was redefined
in MeshGeometry.m by giving the last third of the elements and associated nodes of the
bottom surface to the fifth surface and the remaining elements and nodes of the bottom
surface to the first surface.
MeshGeometry.m
% Use the provided box grid generator
[nCoords,elCon,boundStruct]=BoxGrid_2D(nsd,xl,xr,yb,yt,nnpe,nx,ny);
BoundaryNodes(1).Nodes = [1:2/3*nx+1];
2
2
2 ).
The Von Mises stress for a planar plate was defined as2 (
+
+ 2
The Von Mises
PostProcessor.m
% Calculate Von Mises stress
von_mises = sqrt(3/2*(sigma_xx.^2 + sigma_yy.^2 + 2*sigma_xy.^2));
% Plot stresses
if strcmp(PlotInstructions.plot_contour,'yes')
% Here are two options for plotting stress. Try MakePatchPlot and
% PlotContour to see which you prefer.
%
range=max(max(abs(sig)));
%
MakePatchPlot(sigma_xx,range,meshStruct, 'xx stress ');
%
MakePatchPlot(sigma_yy,range,meshStruct, 'yy stress ');
%
MakePatchPlot(sigma_xy,range,meshStruct, 'xy stress ');
plotcontour(meshStruct,sigma_xx, 'xx stress ',15,'area','on');
plotcontour(meshStruct,sigma_yy, 'yy stress ',15,'area','on');
plotcontour(meshStruct,sigma_xy, 'xy stress ',15,'area','on');
plotcontour(meshStruct,von_mises, 'Von Mises Stress ',15,'area','on');
end
Because there was a cusp between the first and fifth surfaces, stress at that node (and
therefore maximum stress) would go to infinity as the number of elements in the mesh
increased. The y displacement of the rightmost node (the free end of the crack in the quarterplate and center of the crack in the whole plate) was used to determine convergence of the
solution as seen in Table 1.
Table 1. Convergence values
Number of Elements in x-direction
12
30
90
120
240
360
Displacement (m)
1.3373e-04
1.5071e-04
1.5758e-04
1.5839e-04
1.5959e-04
1.5998e-04
2x2x2
4x4x4
8x8x8
16x16x8
64x64x8
Max xx at
x=2m (1D)
[MPa]
Max xx at
x=2m (2D)
[MPa]
Max xx at
x=2m (3D)
[MPa]
Max xx at
x=0m (1D)
[MPa]
Max xx at
x=0m (2D)
[MPa]
Max xx at
x=0m (3D)
[MPa]
2.3176
2.3176
2.3176
2.3176
2.3176
2.3863
2.3304
2.3095
2.3165
2.3176
2.292
2.3465
2.3145
2.3166
2.3176
4.6352
4.6352
4.6352
4.6352
4.6352
4.7451
4.7426
4.7727
4.9234
5.9295
4.853
4.8387
4.9219
5.308
6.9953
stress at x = 2m for all three element types are the same up to five significant digits
and thus agree very well.
c) Displacement converges faster than stress when refining the mesh because stress is a
second derivative of displacement and thus takes longer to reach an exact solution, if
it is able to reach a solution at all.
d) For 1D, 2D, and 3D elements, the displacement results display very little effects from
stress singularities because they are mainly concentrated on the fixed end. As seen in
the results, all displacement values ultimately converge at three significant digits as
the mesh is refined. Stress is calculated as force divided by area. When a beam is
loaded at one end, the four sharp corners of the fixed end of the beam and the location
of the point source can have an infinitely small area. As the mesh is refined, the area
becomes smaller, resulting in a value of stress that becomes larger. The maximum
stress at the fixed end of the beam modeled with 1D elements does not exhibit stress
singularities because the stress values stay the same as the mesh is refined. However
for 2D elements, when the mesh is refined from 16x16x8 to 64x64x8, the maximum
stress value suddenly increases by 18.5%. A similar trend is shown in 3D elements
but at an even earlier point during mesh refinement: as the mesh changes from 8x8x8
to 16x16x8, the maximum stress value increases by 7.55%; and when the mesh
changes from 16x16x8 to 64x64x8, the stress value increases by 27.4%.
Unfortunately for 2D and 3D elements, a refined mesh leads to stress singularities and
lowers the likelihood of convergence at the fixed end of the beam. The extent of
stress singularities can be seen when comparing stress values between all element
types at a refinement level of 64x64x8. 2D elements lead to a 24.5% increase in
stress value compared to 1D elements, while 3D elements lead to a 40.5% increase in
stress value from 1D elements.
e) In order to determine displacements, 3D elements provide the most accurate result.
However efficiency is decreased because at least 7803 degrees of freedom are
required to obtain a converged solution. When stresses are determined, 1D elements
are sufficient to obtain the stress within the beam and at the free end of the beam.
The stress value obtained with 1D, 2D, and 3D elements is ultimately the same,
except that 1D elements reach the solution earlier with a coarser mesh. As a result,
1D elements for calculating the stress at the free end of the beam and within the beam
are accurate and efficient. During calculation of the stress at the fixed end of the
beam, 1D elements do not exhibit stress singularities as the mesh is refined.
Although efficiency is maintained, accuracy decreases. As a result, 2D elements and
3D elements provide more detail of stress behavior at the fixed end of the beam
because the refined mesh allows for a more accurate stress value.
4)
a) The bracket was meshed in a 2D modeller as seen in Figure 1.
c) The response surfaces for (10-H,R1), (10-H,R2), and (W,H) are shown in the below.
Though the response surface for (R2, T) was requested, it is not shown because T was
made to not be a varying parameter which is what was advised by the Professor.
Figure 4. 3D response surface plot of the H and R1 parameter vs. the maximum von Mises
stress
Figure 5. 3D response surface plot of the H and R2 parameter vs. the maximum von Mises
stress
Figure 6. 3D response surface plot of the W and H parameters vs. the maximum von Mises
stress
d) Using the response surfaces shown in Figure 4, Figure 5, and Figure 6, Ansys will
find candidate points below the maximum stress constraint that is input into the
optimization window. Figure 7 shows the candidate points generated by ANSYS
using the 3D response surfaces. The candidate point is chosen that has been verified
to be below the stress limit of 100 kPa and has minimized the volume the most. In this
case, Candidate Point 1 is the best solution, with a 10-H value of 5.7525m (meaning H
is 4.2475m), R1 value of 2.0827, R2 value of 2.4956, and 30-W value of 4.949m
(meaning W is 25.051m). The total volume of the bracket using these parameters is
50.424 m3, and the complete geometric configuration is detailed in Figure 8. Figure 9
shows the von Mises equivalent stress plot for the optimal configuration.
Figure 7. Candidate points and their verified values as generated by ANSYS optimization for the
support bracket
Figure 8. The optimized support bracket according to Candidate Point 1 parameters. The
dimensions are shown estimated to 1 significant figure.
Figure 9. Von Mises equivalent stress plot for the optimal configuration