Assignment 4

AE4ASM003 Linear Modelling (incl. F.E.M)

Lennarth Cohen (5002591)
Delft University of Technology
 1
Part 1: Triangular element

The questions posed in this chapter will be solved using a triangular membrane element
with four nodes. This element is given in Figure 1.1, the coordinates of the nodes (in
Cartesian coordinate system) are given in the figure. Further information that is needed
to solve the questions will be introduced during solving if necessary.

Figure 1.1: Triangular membrane element

1.1. Derive the shape functions for this given triangular

membrane element
The shape functions will be derived in the Cartesian coordinate system. Furthermore,
the displacement function is assumed to be given by Equation 1.1.

u = c1 + c2 · x + c3 · y + c4 · x · y (1.1)

The shape functions belonging to the horizontal displacements can be obtained by sub-
stituting the nodal coordinates into the assumed displacement function. The following
system is obtained if this is done.

ui = c1 + 2 · w · c2
uj = c1 − 4 · w · c2
uk = c1 + 3 · h · c3
um = c1 − w · c2 + h · c3 − w · h · c4

1
1.1. Derive the shape functions for this given triangular membrane element 2

This system of equations can be rewritten to a matrix/vector form, which is given by the
following.

    
ui 1 2·w 0 0 c1
 uj  1 −4 · w 0  
0   c2 
 = 
 uk   1 0 3·h 0   c3 
um 1 −w h −w · h c4

This matrix equation is of the form ⃗u = A · ⃗c, where ⃗u the displacement vector, A is the
matrix with the nodal coordinates, and ⃗c the vector with the coefficients of the assumed
displacement function (in horizontal direction).

Since the coefficients of the assumed displacement function is what is to be obtained,

the A matrix should be inverted to obtain the expression for the vector with these as-
sumed displacement equation coefficients.

   −1  
c1 1 2·w 0 0 ui
c2  1 −4 · w 0 0   uj 
 
 = 
 c3   1 0 3·h 0   uk 
c4 1 −w h −w · h um

Inverting this matrix leads to the following coefficients.

   
c1 6 · (2 · ui + uj )
 c2   3·(ui −uj ) 
 = 1 · 
c3  18  
w (1.2)
 − 2·(2·ui +uh j −3·uk ) 
c4 5·ui +7·uj +6·uk −18·um
w·h

The shape functions of the displacement function in horizontal displacement direction

can now be obtained by rewriting these coefficients, together with the assumed dis-
placement function in horizontal direction, in the form given in Equation 1.3.
 
ui
[ ]  uj 
u(e) = Ni Nj Nk Nm  
 uk  (1.3)
um

When the now readily obtained coefficients are substituted into the assumed (horizontal)
displacement equation, the system below is obtained.

[
1 3 · (ui − uj ) 2 · (2 · ui + uj − 3 · uk )
u (e)
= 6 · (2 · ui + uj ) + ·x− ·y
18 w h
]
5 · ui + 7 · uj + 6 · uk − 18 · um
+ ·x·y
w·h
1.1. Derive the shape functions for this given triangular membrane element 3

Rewriting this equation isolating the factors depending on ui , uj , uk and um . The follow-
ing expression is obtained.

[( ) ( )
1 3·x 4·y 5·x·y 3·x 2·y 7·x·y
u (e)
= 12 + − + · ui + 6 − − + · uj +
18 w h w·h w h w·h
( ) ( ) ]
6·y 6·x·y 18 · x · y
+ · uk + − · um
h w·h w·h

The expression above is of the form given in Equation 1.3. This means the shape
functions for horizontal displacements can be isolated from this expression. These
shape functions in horizontal direction are then found to be given by the expressions
below.

( )
1 3·x 4·y 5·x·y 2 x 2·y 5·x·y
Ni = · 12 + − + = + − +
18 w h w·h 3 6 · w 9 · h 18 · w · h
( )
1 3·x 2·y 7·x·y 1 x y 7·x·y
Nj = · 6− − + = − − +
18 w h w·h 3 6 · w 9 · h 18 · w · h
( )
1 6·y 6·x·y y x·y
Nk = · + = +
18 h w·h 3·h 3·w·h
x·y
Nm =−
w·h

Using the same displacement function for the displacement in vertical direction, the
vertical displacement function can be written in the form given in Equation 1.4.

v (e) = c5 + c6 · x + c7 · y + c8 · x · y (1.4)

Inserting the (Cartesian) nodal coordinates, and rewriting to the matrix/vector form leads
to the following system.

    
ui 1 2·w 0 0 c5
 uj  1 −4 · w 0 0   c6 
 =  
 uk   1 0 3·h 0   c7 
um 1 −w h −w · h c8

This matrix is similar to the matrix equation that was solved for the horizontal displace-
ment system. This (vertical) matrix/vector equation will thus lead to the same shape
functions as in the horizontal displacement system. Thus, both the horizontal and ver-
tical shape functions are fully described by the following shape functions.
1.2. Derive the expression for the strain-displacement matrix B 4

2 x 2·y 5·x·y
Ni = + − +
3 6 · w 9 · h 18 · w · h
1 x y 7·x·y
Nj = − − +
3 6 · w 9 · h 18 · w · h
y x·y
Nk = +
3·h 3·w·h
x·y
Nm = −
w·h

1.2. Derive the expression for the strain-displacement ma-

trix B
The strain-displacement matrix is the matrix B, as it occurs in Equation 1.5. This ma-
trix relates the displacement vector to the strain vector. The strain vector is given by
Equation 1.6. Furthermore, the displacement vector is given by Equation 1.7.

⃗ϵ = B · ⃗u (1.5)

   ∂u 
ϵxx ∂x
⃗ϵ =  ϵyy  =  ∂y 
∂v
(1.6)
∂u ∂v
γxy ∂y
+ ∂x
 
uix
 uiy 
 
 ujx 
 
 ujy 
⃗u =  
 ukx  (1.7)
 
 uky 
 
umx 
umy

The strain-displacement matrix is then obtained by substituting the expressions for the
displacement for u(e) (horizontal) and v (e) (vertical) directions of the element. This leads
to the following expressions.

 
∂
∂x
(Ni · uix + Nj · ujx + Nk · ukx + Nm · umx )
⃗ϵ =  · uiy + Nj · ujy + Nk · uky + Nm · umy ) 
∂
∂y
(Ni
∂
∂y
(Ni · uix + Nj · ujx + Nk · ukx + Nm · umx ) + ∂x (Ni · uiy + Nj · ujy + Nk · uky + Nm · umy )
∂

This expression for strain can be written as the strain-displacement matrix B, and the
displacement vector ⃗u. This system is given by Equation 1.8, where the different βindex
and δindex with index=i, j, k & m, are respectively the partial derivatives with respect to
x and y of the shape functions Ni , Nj , Nk , and Nm . These expressions are also given
below.
1.3. Explain why solving the integral is not as straight forward as in class 5

 
βi 0 βj 0 βk 0 βm 0
⃗ϵ =  0 δi 0 δj 0 δk 0 δm  · ⃗u (1.8)
δi βi δj βj δk βk δm βm

∂Ni ∂Nj ∂Nk ∂Nm

βi = βj = βk = βm =
∂x ∂x ∂x ∂x
∂Ni ∂Nj ∂Nk ∂Nm
δi = δj = δk = δm =
∂y ∂y ∂y ∂y

Working out the derivatives leads to the expressions for β, and δ as given below.

1 5·y −1 7·y y y
βi = + βj = + βk = βm = −
6 · w 18 · w · h 6 · w 18 · w · h 3·w·h w·h
−2 5·x −1 7·x 1 x x
δi = + δj = + δk = + δm = −
9 · h 18 · w · h 9 · h 18 · w · h 3·h 3·w·h w·h

This leads to the final strain-displacement matrix, as given by Equation 1.9. With the
β’s and δ’s as expressed above.
 
βi 0 βj 0 βk 0 βm 0
B =  0 δi 0 δj 0 δk 0 δm  (1.9)
δi βi δj βj δk βk δm βm

1.3. Explain why solving the integral is not as straight

forward as in class
The integral that is to be solved is the integral given by Equation 1.10. This equation
is not as straight forward to solve as in class, since the shape functions are a function
of x, y and xy. This means that the shape functions are more complex than the in-
class case (quadratic of order, not linear any more). This leads to the derivatives of the
shape functions in the matrix B being non-constant, thus making solving the integral
more difficult. A solution to this problem is using a higher-order integration method to
solve this integral, which is given by Equation 1.10.
∫
B T DB dV (1.10)
V

The integration method that could be used to solve this integral is, for example, the
in-class mentioned method of Gaussian quadrature.

1.4. Calculate the force vector F for this given load case
The loading on the triangular membrane is given by a point load on node k, and a
distributed load of q on the horizontal segment from node j (x=-4w) to x=0. In this
1.4. Calculate the force vector F for this given load case 6

section the force vector will be calculated, this is done by splitting this section into
two subsections. One subsection calculating the force vector contribution of the point
load, and one subsection calculating the force vector contribution of the distributed
load. These force contributions are obtained by using the principle of external work
on an element, which is given by Equation 1.11. This equation will be re-written in the
following subsections to fit the type of loading (point loading or distributed loading).
∫
Wp(e) = (u)T (P ) dA (1.11)
A

1.4.1. Point load force vector contribution

For a point load, Equation 1.11 can be rewritten to Equation 1.12. The point load on
node k gives contributions in Fkx , and Fky directions only. Using Equation 1.13 the
force vector contribution of the point loading on node k can be obtained. This point
load contribution to the force vector will later be combined with the distributed load
contribution to the force vector to obtain the total force vector of the element.

(e)
Wpoint = ui · Fix + vi · Fiy + uj · Fjx + vj · Fjy + uk · Fkx + vk · Fky + um · Fmx + vm · Fmy (1.12)

 
0
 0 
 
 0 
 
∂Wpoint
(e)
 0 
(Fpoint ) (e)
= 
=  (1.13)
∂u −P · sin(θ) 
 
 P · cos(θ) 
 
 0 
0

1.4.2. Distributed load force vector contribution

For a distributed load, Equation 1.11 can be rewritten to Equation 1.14. It is important to
note that in this derivation it is assumed that the load q is given in units N /mm. Further-
more, note that now in the equation (Equation 1.14) (u)T is replaced by (N )T (ucomponents )T ,
this replacement comes from Equation 1.3.
∫
(e)
Wdistri. = (N )T (ucomponents )T (P ) dydx (1.14)
A

This integral can be further rewritten to obtain the following.

1.4. Calculate the force vector F for this given load case 7

∫ 0 [ ]
(e) ∂W (e) 0
Fdistri. = = (N |y=0 ) (ucomponents )
T T
dx
∂ucomponents −4·w −q
∫ 0
(e) [ ]T
Fdistri. = 0 −q · Ni (y = 0) 0 −q · Nj (y = 0) 0 0 0 0 dx
−4·w
∫ 0
(e) [ ]T
Fdistri. = −q · 0 Ni (y = 0) 0 ·Nj (y = 0) 0 0 0 0 dx
−4·w
∫ 0
(e) [ 2 ]T
Fdistri. = −q · x
0 3 + 6·w 0 13 − 6·w
x
0 0 0 0 dx
−4·w
(e) [ −4·q·w −8·q·w
]T
Fdistri. = 0 3
0 3
0 0 0 0

1.4.3. Total force vector F for this load case

The total force vector of the element is obtained by taking the sum of the derived force
vectors for the point- and distributed loads. The calculation of the total force vector F
is given in Equation 1.15. Here the value for θ, which is 30°, is filled in.

     
0 0 0
 0   −4·q·w   −4·q·w 
   3   3 
 0   0   0 
     
 0   −8·q·w   −8·q·w 
=   3   
(e) (e)
F (e) = Fpoint + Fdistri. 3
 −P  +  0  =  −P  (1.15)
√ 2    √ 2 
 3·P   0   3·P 
 2     2 
 0   0   0 
0 0 0
 2
Part 2: Panel under tension

A panel with dimensions as indicated in Figure 2.1 is tested by hanging many small
weights from it (more than schematically shown). This leads to a distribution of 257.5
kg/m on the bottom horizontal edge of the structure. Moreover, the top 50 mm of the
panel is clamped. The thickness of the panel is 2 mm, and the panel is made out of
Aluminium with a Young’s modulus of 70 GPa and a Poisson ratio of 0.3.

Figure 2.1: Panel under tensile load

2.1. Give an example of what this test can represent in

real life
Looking at the test case, it can be seen that in this case not only the edge is clamped,
but the top 50 by 400 mm is clamped. This has an influence on the behaviour of the
structure to the distributed load which is applied on the bottom of the structure.

The clamped area can be interpreted as a section which is for example welded or
joined/linked to its surroundings. This welding/joining is not perfectly on the boundary,
but always exceeds the interface between panel and surrounding slightly, which can
be seen on Figure 2.2. This welding/joining makes the boundary condition application
solely on the edge invalid, and clamping (U1=U2=U3=UR1=UR2=UR3=0) an area/sec-
tion therefore is a better approximation of reality.

8
2.2. Discuss the approach to idealise this panel and test setup 9

Figure 2.2: Welding of structure

Taking into account the entire panel with the applied loading, not only the boundary
condition aspect, the structure/test case could for example represent a panel which is
part of an aircraft fuselage. During flight, the aircraft fuselage panels experience tensile
and bending due to the differences in air pressure in and outside the aircraft. This test
case could represent this loading of the thin panels, since the clamped section closely
mimics the way that panels are linked in a fuselage structure.
The assumptions made are that the panel is flat (t«w and h), the panel is uniform in
material behaviour and that the loading is uniform along the edge of the panel. The
panel behaviour is often not as optimal as assumed in the simulation, due to manufac-
turing not being able to be perfect, leading to manufacturing imperfections. And the
material will always be slightly anisotropic. Lastly, the loading of a panel is rarely con-
stant in fuselage panels. The influence of these assumptions is that the simulations will
fail to model these imperfections are they are impossible to predict and thus analysed.
Therefore, the analyses can be used as an indication of magnitudes, however, safety
factors and possible experimental testing might be used to omit the risks involved with
improper modelling of the structure.

2.2. Discuss the approach to idealise this panel and test

setup
The structure dimension in thickness direction (2 mm) is very small compared to the
other dimensions (width= 400 mm, and height = 650 mm). This leads to the conclusion
that the structure can be modelled as a 3D shell part. Furthermore, the loading can be
modelled as a shell edge load which is placed normal to the bottom horizontal edge.
Since the dimensions in Abaqus were inserted in mm, I needed to convert the 257.5
kg/m to N /mm, which is done as shown below.

257.5 kg/m = 257.5 · 9.81 N /m = 2526.075 N /m = 2.526075 N /mm

It should be noted that the loading is downwards in the y-direction, so the load should be
inserted negatively in the shell edge loading menu when applying loads to the Abaqus
2.2. Discuss the approach to idealise this panel and test setup 10

model. Lastly, the clamped section on the top 400 mm by 50 mm means that in this
section the boundary condition is U1=U2=U3=UR1=UR2=UR3=0.

Looking at the structure, it can also be seen that there is one plane of symmetry (in
y-axis). This plane can be used to reduce the amount of numerical computations, by
cutting the structure through half vertically and applying the symmetry boundary condi-
tions and loads. The boundary conditions and loads that will need to be applied are as
follows:

• Boundary conditions: The boundary conditions that need to be applied along

the cut axis of symmetry are U1=U3=UR2=0. This means that the displacements
in the x-axis and z-axis remain 0 in the axis of symmetry, and that the rotation
perpendicular to the axis of symmetry are continuous when comparing the half-
structure to the full structure.
• Loads: Since a distributed load is applied vertically to the bottom horizontal edge,
the distributed load that remains on the cut structure is a load per unit length that
is equal to the load per unit length on the full structure.
Using all this information, the structure can be modelled as shown in Figure 2.3 in
Abaqus. This (simplified) panel model where the y-axis symmetry is used will be used
in the subsequent sections where it can be used. Later it will be seen that for option
II in Section 2.4 the half-model can not be used for option II, since the structure is not
symmetric in the y-axis any more.

Figure 2.3: Abaqus model of panel

2.3. Model the panel as a flat panel, discuss results and effects of idealisation 11

2.3. Model the panel as a flat panel, discuss results and

effects of idealisation
The discussion on the loading and boundary conditions above in Section 2.2 discussed
the most important aspects of the Abaqus model, which together with the part 2 de-
scription at the beginning of this chapter complete the data that is needed to model
this panel in Abaqus. The panel is modelled using Linear quadrilateral elements (type
S4R). This type of element ensures accuracy and computational efficiency while being
used in thin to moderately thick shell structures. Modelling the panel in Abaqus results
into the stresses as shown in Figure 2.9, and displacement in U1 and U2 as shown in
Figure 2.6.

Figure 2.4: U1 Figure 2.5: U2

Figure 2.6: Displacement plots Abaqus

2.3. Model the panel as a flat panel, discuss results and effects of idealisation 12

Figure 2.7: S11 Figure 2.8: S22

Figure 2.9: Stress plots Abaqus

The mesh cells for the clamped top section are 20 mm (width) by 25 mm (height) to be
as close as possible to the 20 mm by 20 mm size which was asked for in the assign-
ment description. The remaining structure was divided into 20 mm by 20 mm mesh
cells. This meshing means the discretization of the panel, which leads to less accurate
solutions of the problem (simulation settings). Depending on both the mesh sizing, and
the loading and boundary conditions, the accuracy of the model to the reality can be
increased or decreased as desired.

This panel is modelled as a panel with a certain Poisson ratio ν, and a clamped section
as described above, together with the uniform distributed load, the panel will not sim-
ply deform vertically to elongate. The panel will show a necking behaviour due to the
decomposition of the stresses (and thus displacements). Looking at the displacement
behaviour, this indeed shows that the structure ’necks’. This is visible by looking at
the U1 displacements, the structure is symmetric thus the other half of the structure
would be obtained by mirroring this figure in the y-axis and then the loads would be
negative instead of positive (see Appendix A for full structure simulation results). This
change in sign indicates that in the x-direction, the sides move towards the symmetry
axis. Looking at U2 further proves that the panel is necking (again, see Appendix A
for full structure simulation). This plot namely shows that the panel is moving vertically
downward, this is a result of the distributed load being in vertical downward direction.

Looking at the stresses in the panel, the S11 (σxx ), S22 (σyy ), and S12 (τxy ) that follow
from the displacements qualitatively show the same behaviour that we would expect
from the simulation case as well. Furthermore, it can be noted that the stresses are a
direct result of the obtained displacements, and vice versa. For this simulation case, it
was easier to identify the physical phenomenon of necking (what we expected) in the
2.4. Two options to reduce the displacement 13

displacement results, however.

The results of these simulations are very accurate for the exact case which was simu-
lated. However, the validity of the results of a simulation compared to a ’real life case’ is
always questionable. This is due to the imposing of the loads and boundary conditions.
When a boundary condition or load is not closely mimicking reality, the deviation be-
tween real life- and simulation results might be very off. Therefore, it is important to use
the boundary conditions and loads that best represent the real life case, but obtaining
these exact is very, if not, close to impossible. Furthermore, if the loading and bound-
ary conditions accurately mimic reality, the meshing is another factor that influences
the accuracy of the simulation. The trustworthiness of the simulations are thus highly
dependent on several factors.

2.4. Two options to reduce the displacement

Two options were proposed to reduce the displacement in the simulated panel. These
options are option I: doubling the thickness of the middle 150 mm by 600 mm section.
And option II: adding two L-stiffners on top of the original panel. Figure 2.10 shows the
two options, left; option I, right; option II: L-stringers on blue vertical lines.

Figure 2.10: Two options to reduce thickness

2.4.1. Option I: Doubling the thickness

For option I, the middle section is double the thickness of the original panel. This means
that the thickness of this middle part is 4 mm instead of 2 mm. This doubling of the thick-
ness is done by generating a back panel and a master panel and tie-ing them together
with a tie constraint. Since the panel is again symmetric in the y-axis. The master panel
is 200 mm by 650 mm. and the back panel is 75 mm by 600 mm (since the clamped
section is not taken extra thickness in back panel. The U1, U2, U3, S11, S22, S33, and
S12 results of this simulation can be found in Appendix A. The most important values
for reducing the displacement in the panel are the magnitude of the displacement, and
the U1, U2, and U3. Therefore, only these plots will be used to compare the options.
2.4. Two options to reduce the displacement 14

For this option I, the assumptions that are made are that the extra thickness is added
on one side of the already existing panel. This has an influence on the way the struc-
ture bends, as the panel is not symmetrical when looking at the cross-section in the
yz-plane. This assumption of adding the extra thickness on one side of the panel is
made since this is the most logical way adding thickness is done in the manufacturing
process. Note that the resulting structural behaviour of adding the extra thickness on
one side of the panel can have both a positive or a negative effect on the panel, de-
pending on the application/load case of the panel.

Looking at the simulation results for the magnitude of the displacement and U1, Fig-
ure 2.13, and the U2 and U3 plots are obtained, Figure 2.16. After the results of option
II, both configuration will be discussed in terms of Abaqus results.

Figure 2.11: Magnitude of U option I Figure 2.12: U1 option I

Figure 2.13: Umag and U1 option I

2.4. Two options to reduce the displacement 15

Figure 2.14: U2 option I Figure 2.15: U3 option I

Figure 2.16: U2 and U3 option I

2.4.2. Option II: Adding two L-stiffners

The second option consist of the original panel, onto which two L-stiffners of the ge-
ometry in Figure 2.10 are added in the vertical direction on the two vertical blue lines
indicated in Figure 2.10. For simulating this case in Abaqus, it is important to note that
the panel is not symmetric any more, thus the symmetry axis can not be used to reduce
the simulation size of the panel. The stiffeners make this panel asymmetric, thus the
full panel, including stiffeners, needs to be simulated. Figure 2.17 shows the way the
stiffeners are mounted on the panel. This is done by aligning the leftmost vertical part
of each stiffener with the blue lines that were discussed before.

Figure 2.17: Stiffener on panel

The stiffness as added to the original panel in option I is rather different from the way
that the stiffness is added to the panel in option II in terms of the added geometry. To
find out the effects on the original panel, option II simulation results will be used. The
magnitude of the displacement, and the U1, U2 and U3 displacements, are given in
Figure 2.20 and Figure 2.23.
2.4. Two options to reduce the displacement 16

Figure 2.18: Magnitude of U option II Figure 2.19: U1 option II

Figure 2.20: Umag and U1 option II

Figure 2.21: U2 option II Figure 2.22: U3 option II

Figure 2.23: U2 and U3 option II

The results of the original panel are given below, magnitude of displacement, and the
U1, U2, and U3 displacements, Figure 2.26 and Figure 2.29. The Abaqus results of
these three options can be summarized as follows.
2.4. Two options to reduce the displacement 17

Figure 2.24: Magnitude of U original panel Figure 2.25: U1 original panel

Figure 2.26: Umag and U1 original panel

Figure 2.27: U2 original panel Figure 2.28: U3 original panel

Figure 2.29: U2 and U3 original panel

2.5. What would need to be changed to the model in the new case? 18

Table 2.1: Abaqus panel results

Umag,max U 1max U 2max U 3max volume

option
mm mm mm mm mm3
original 0.0108 0.0011 -0.011 0 5.2 · 105
I 0.009 0.0013 -0.0089 0 7.0 · 105
II 0.1578 0.0016 -0.0099 0.158 7.0 · 105

Looking at the simulation results of the displacements of Umag , U1, U2, U3, and the
volume of the options. It can be noted that for both options I and II the volume is the
same, thus indicating that for the same material, the mass is also equal. Furthermore,
the U3 in option II is suddenly non-zero, which has a significant effect on the panel
behaviour, moreover, the Umag,max is significantly higher than the other options, which
is not desirable.

Looking at the properties of the different options, option II is significantly worse than
option I. Not only in terms of performance, but also the manufacturing of the panel is
more difficult, since the production of the L-stiffners is relatively time-consuming and
labour-intensive compared to adding a panel on top of the original panel. Thus, I would
prefer option I over option II.

2.5. What would need to be changed to the model in the

new case?
If the bottom was first clamped, and then loaded with the same weight, the following
has to change in Abaqus.
• The clamped section on the top has to be removed, since this is not valid any
more. This previously clamped section is now free to move.
• A new boundary condition has to be applied to the bottom, this bottom is now
clamped instead of the top of the panel. For partition the section that you want
to clamp later. Then this section can be selected in the boundary condition (BC)
menu to clamp, which means that an encastre (U1=U2=U3=UR1=UR2=UR3=0)
boundary condition needs to be applied.
• The distributed load remains the same, as the panel is said to be loaded with the
same weight. This load needs to be applied at the bottom horizontal edge again.
• Depending on the option that is to be simulated, also the location of the added
thickness and stiffeners needs to be changed, since these need to overlap the
section that is not clamped. This means in option I and II, respectively, the added
panel or stiffeners need to be translated vertically.
The implications these changes have on the results is that the stress concentrations
will be high near the clamped bottom. This is due to the fact that the region resists the
applied load while this region is clamped. Furthermore, the displacements in the sec-
tion above the ’new’ clamped section will displace more, the further from the clamped
section. This is due to the free-edge on top and the clamped section, close to the bot-
tom now.
 A

A.1. Original panel results

Figure A.1: U1 Figure A.2: U2

Figure A.3: Displacement plots Abaqus (original panel)

Figure A.4: U3

19
A.1. Original panel results 20

Figure A.5: S11 Figure A.6: S22

Figure A.7: Stress plots S11 & S22 Abaqus (original panel)

Figure A.8: S33 Figure A.9: S12

Figure A.10: Stress plots S33 & S12 Abaqus (original panel)
A.2. Option I: Doubling the thickness of the mid-section 21

A.2. Option I: Doubling the thickness of the mid-section

Figure A.11: U1 option I Figure A.12: U2 option I

Figure A.13: Displacement plots Abaqus (option I)

Figure A.14: U3 option I

A.2. Option I: Doubling the thickness of the mid-section 22

Figure A.15: S11 option I Figure A.16: S22 option I

Figure A.17: Stress plots S11 & S22 Abaqus (option I panel)

Figure A.18: S33 option I Figure A.19: S12 option I

Figure A.20: Stress plots S33 & S12 Abaqus (option I panel)
A.3. Option II: Adding two L-stiffeners 23

A.3. Option II: Adding two L-stiffeners

Figure A.21: U1 option II Figure A.22: U2 option II

Figure A.23: Displacement plots Abaqus (option II)

Figure A.24: U3 option II

A.3. Option II: Adding two L-stiffeners 24

Figure A.25: S11 option II Figure A.26: S22 option II

Figure A.27: Stress plots S11 & S22 Abaqus (option II panel)

Figure A.28: S33 option II Figure A.29: S12 option II

Figure A.30: Stress plots S33 & S12 Abaqus (option II panel)