You are on page 1of 11

Report 2

Structural Optimization of 2D structures.


Continuum optimization.

41561, Structural Optimization

TECHNICAL UNIVERSITY OF DENMARK

Authors:

Group 06

Felipe Olivos | s220050

Rafael Parrado | s213153

Date:

May 12th , 2023


REPORT 2: Structural Optimization of 2D structures. Continuum optimization.
41561 Structural Optimization | Group 06
incluir cálculo de deforma-
ción para cierta fuerza en
Table of contents
rango elástico
1 Continuum optimization with top88 . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Continuum optimization with Comsol . . . . . . . . . . . . . . . . . . . . . . . . 5


2.1 Numerical modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Experimental test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

A XXXX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

Contributions
This work has been developed in equal parts by both authors. Therefore, no clear distinction
among the different parts is presented here, as both have contributed to the code generation,
case study and peer-check. This allowed both of us to learn throughout the process.
After the main body of the report, appendices are included, where further comments and
pictures of the numerical modelling and experimental testing are presented, which are not part
of the main body of the report.

Abbreviations
CS Continuation Strategy SO Shape Optimization
DD Design Domain TO Topology Optimization
FR Filtering Radius Vf Volume fraction
Msz Mesh Size VM Von Mises
Rev Reversible Loads

Summary
This report is drawn up within the framework of the course 41561, Structural Optimization,
and constitutes the second of two reports, where topology optimization of two-dimensional
continuum structures is addressed. Based upon the design explored in Report 1 [1], the Matlab
script Top88.m, which performs topology optimization of a continuum in 88 lines, is employed
for further investigation. The results are analyzed, and comparisons are carried out.
In addition, a second structure is examined using the commercial software Comsol Multiphysics.
The study begins by conducting topology optimization to obtain a design that maximizes stiff-
ness. Then, the obtained design is adjusted with the application of shape optimization aiming
to minimize VonMises stresses. Finally, the designs are 3D printed, and tested in the laboratory.
y comparamos los resultados
experimentales con los teóri-
May 12th , 2023 Page 1 of 10
REPORT 2: Structural Optimization of 2D structures. Continuum optimization.
41561 Structural Optimization | Group 06

1 Continuum optimization with top88


Compliance optimization of a 2D building’s facade has been performed in [1] using truss el-
ements. In this study, topology optimization is performed adopting the same structure as a
continuum instead. The Matlab script Top88.m [2] is employed for the analysis. The width
and height of the building are H = 200 m and B = 40 m, which yields an aspect ratio of 5. A
lateral unit point load is applied on the top left edge of the structure.
To begin the study, a sensitivity analysis is performed by varying relevant parameters as mesh
size (Msz), filtering radius (FR) and volume fraction (Vf). These are summarized in tab. 1.1,
and the final designs illustrated in fig. 1.1.
Table 1.1: Parametric study. Parameters and compliance.
Parameter FR-A. FR-B. FR-C. Msz-A. Msz-B. Vf-A. Vf-B.
nelx 32 32 32 16 40 32 32
nely 160 160 160 80 200 160 160
Vf 0.40 0.40 0.40 0.40 0.40 0.30 0.50
Penalty 3 3 3 3 3 3 3
FR 1.5 2.0 3.0 1.5 1.5 1.5 1.5
Compliance 360.14 367.28 385.75 380.13 355.89 466.18 297.76

B
Fw

(a) Geom. (b) FR-A. (c) FR-C. (d) Msz-A. (e) Msz-B. (f) Vf-A. (g) Vf-B.
Figure 1.1: Filtered densities of the optimized designs. Note that not all the cases are comparable compli
pair-wise. Comparisons are made below. For space reasons, not all the cases are shown. ance

Firstly, it can be observed that the increase in the FR leads to cleaner but more compliant
structures (cases FR-A,B,C). Similarly, refining the mesh shows a similar trend, resulting in
more complex but less compliant structures (cases Msz-A,B). Lastly, the volume fraction is
modified to indicate, as expected, that the lower the Vf, the more lightweight the achieved
structure is, though sacrificing stiffness (cases Vf-A,B). this last case is not comparable with
the other two (we should mention it)
Case FR-A is now adopted for the application of the so-called continuation strategy. Fixed
penalty factors and filtering radii have been used in the cases above. However, a continuation
strategy is often applied where these parameters are, after certain iterations, modified. For
illustration purposes, this is only applied here for the penalty factors, which is initialized at 1.5
increasing every 30 iterations in steps of 0.25 (See case CS-B). The resulting optimized designs
eliminar, ya que se cita la fig.
are illustrated in fig. 1.2, and the convergence plots in fig. 1.4. 1.2 mas adelante
Interestingly, the application of this strategy yields simpler structures, but again, sacrificing
stiffness. This can be explained by the use of an initial relatively low penalty factor, being then
gradually increased to higher values, responsible for reducing the performance. The application

May 12th , 2023 Page 2 of 10


quizá está al revés, compliance baja (levemente), por lo que la es-
tructura es más rígida.
podemos mencionar que la mejora en
compliance es mínima, pero que es de-
seable obtener estos casos ya que no
tiene grandes costos computacionales.
REPORT 2: Structural Optimization of 2D structures. Continuum optimization.
41561 Structural Optimization | Group 06

in order to study changes on the


of this technique to FR is not explored further here. optimize design.
So far, a single point load towards the left edge of the building has been considered. Yet,
lateral loads can be reversible. Thereby, a slight modification is done in the code to address
optimization with two load cases. In particular, the objective function is transformed to the
weighted average of the compliances sum, C = w1 C1 + w2 C2 , where wi is a scalar (weight)
and Ci is the compliance due to load i. The analysis is performed for two sets of weights, case
Rev-A (w1 = w2 = 0.50) and Case Rev-B (w1 = 0.70, w2 = 0.30). The designs and convergence
plots are presented in fig. 1.3 and fig. 1.5, respectievly.
deberiamos
poner donde
y en qué di-
rección fue-
ron aplica-
das las car-
gas

(a) CS-A (b) CS-B (a) Rev-A, set 1. (b) Rev-B, set 2.
Figure 1.2: Continuation strategy, designs. Figure 1.3: Reversible loads, designs.
900 0.42 600 0.42
CS-A, c=360.1 Rev-A, c1 =374.2
CS-B, c=357.3 Rev-A, c2 =374.8
800 550 Rev-B, c1 =373.3
CS-A, Vf
Rev-B, c2 =389.1
CS-B, Vf
700 Rev-A, Vf
500 Rev-B, Vf
Comp.

Comp.
Vf

Vf
600 0.40 0.40
450
500
400
400

0.38 350 0.38


0 50 100 150 200 0 50 100 150 200 fig
Iter. Iter. 1.3.a
Figure 1.4: Continuation strategy, convergence. Figure 1.5: Reversible loads, convergence.
It can be observed that, for equal weights, the same compliance is achieved. Two load bearing
mechanisms are formed, mainly in the form of arches connected by shear elements. However, for
the second set of weights, it is apparent that the algorithm prioritizes minimizing compliance
for the left load, thereby leading to a more efficient structure for that load (observe the upper
node moving towards the left edge).
Finally, in figures, designs from [1] are replicated. Though depicting similar trends, the designs
are not equal. Continuum optimization allows for some moment transfer, as elements are not
trusses. Depending on the application, both methods can prove to be certainly useful for
optimization purposes. recuerdo que Federico habló de la energía de deformación en es-
te punto, pero en el contexto de top. opt. es lo mismo que mi-
nimizar complience, deberíamos hacer notar este punto?

May 12th , 2023 quizá debamos mencionar también que si bien se parecen, los Page 3 of 10
elementos estructurales trabajan de manera muy distinta, ya que
para la versión de optimización continua hay transferencia de
momentos en casi todos los elementos.
REPORT 2: Structural Optimization of 2D structures. Continuum optimization.
41561 Structural Optimization | Group 06

RPA to try to squeeze in comparison with report 1, but not exceed 2 pg.
The stresses and sensitivity (or strain energy) plots are not presented here for the sake of space,
yet they provide insightful information to be further assessed.
Opino que deberíamos ponerlos en el anexo.
(1.5 pages) Matemáticamente, es lo mismo optimizar mi-
XXX nimizando la energia de deformacion que la
complience. La única diferencia podría ser
• Introduction to the problem. Brief description. en cómo se calcula la tensión
• Parameter effects study (filtering radius, mesh, volume fraction). Comparison of the per-
formance, and (convergence plots). Include here a picture with the problem description.
• Continuation strategy. (a) With penalty factor (b) with filtering radius. Convergence
plots. Comparison of performance.
• Reversible loads. Introduction to the problem and the solution implementation. Different
applied weights, topology designs, convergence plots. Solutions not implemented (max
compliance).
• comparison with report 1. hago la comparación?

May 12th , 2023 Page 4 of 10


Seguro q es acero? Si es así puede ser una explica-
ción de la diferencia de desempeño con la realidad
REPORT 2: Structural Optimization of 2D structures. Continuum optimization.
41561 Structural Optimization | Group 06
continuum
2 Continuum optimization with Comsol
2.1 Numerical modelling
A 2D plane-stress cantilever subjected to a vertical force F = 100 kN upon the tip is optimized.
Steel material is adopted (E = 205 GPa, ν = 0.30). The case study description can be found
in [3]. The discretization consists of triangular mesh and quadratic serendipity elements. First,
topology optimization (TO) is performed in the linear-elastic regime aiming at minimizing
compliance. Thus, the normalized strain energy is used as the objective function. Similarly
to sec.1, different mesh sizes and filtering radii are tested. In addition, projection using the
hyperbolic tangent function is performed. The parameters used for the final design are presented
in tab. 2.1. The analysis is performed, and the projected density of the final design (β = 8) is
shown in fig. 2.1.
Table 2.1: Topology optimization in Comsol. Parameters.
Vol [mL] Vf [-] Msz Max/Min [mm] FR β psimp
56 0.4917 3 / 1.5 0.3Msz [1,2,4,8] 3 centrar
(DD)
It can be observed that the specified design domain might be restricting the optimal design to
be achieved. An improved design can be then obtained if the DD is extended (See app. XX).
Based upon the 0.5 threshold of the filtered density, shape optimization (SO) is carried out.
The objective is to minimize VonMises (VM) stresses across the structure. Specifically, a stress
concentration is observed in the cantilevered part. Hence, the average p−norm of the normalized
VM stress is used as an objective function, constraining the Vf specified above. Again, different
p-values are tested, but the outcomes are shown for p = 10. The moving boundaries are allowed
to move dmax =Msz. The resulting structural shape is depicted in fig. 2.1.

Stress concentrations Smoothed edge over the


previous
top opt
design

Es justo comparar ambos


si no parten de la misma
estructura inicial?

Figure 2.1: Optimized designs. TO, Projected density (β = 8) (left), SO, Optimized shape (right).

The contour used as an initial design for SO is now adopted to run a FE simulation in order to
compare TO and SO designs. However, in this process, a slight loss of material occurs. Thus,
SO is performed with a new volume constraint (55 mL). The VM stresses are plotted on the
deformed shape in fig. 2.2. (see smoothed zone
in fig.2.1 right)
As anticipated, SO design proves to improve performance in terms of stresses, with a reduction
of more than 50%. This high value can be explained given that the highest stresses are found in
the cantilevered part. Its depth is increased after SO, resulting in a higher moment of inertia.
Nevertheless, unexpectedly, lower displacements are encountered in SO design. Although this
is not usual, as SO tends to compromise stiffness in an effort to reduce stress concentrations, in
this specific case, the deflections are greatly affected by the slender cantilever, whose stiffness

May 12th , 2023 smaller? Page 5 of 10


REPORT 2: Structural Optimization of 2D structures. Continuum optimization.
41561 Structural Optimization | Group 06

is increased after SO implementation (increased moment of inertia).

Max. displacement Max. displacement


15.2 mm 12.2 mm

Max stress Max stress


1.45E4 MPa 6.41E3 MPa

Figure 2.2: Comparison of the optimized designs. Vol.= 56 mL. TO (Left), SO (Right).
i.e., a perfect fixed boundary vs a clamped boundary
2.2 Experimental test with external force introduction to the samples,
respectively.

Samples of the TO and SO designs are 3D-printed and sub-


jected to destructive testing (settings in tab. 2.2). The sam-
ples are positioned on the testing device, manually aligned,
and clamped to the base plate, as depicted in fig. 2.3. Al-
ready, a difference with the FE model is observed, as the
boundary conditions defer from the applied in the lab.
Loading device
Table 2.2: Settings for 3D-printing.
2mm Shell Thk. Wall Thk. Density Pattern Weight TO/SO
Clamping plate & bolts
xx mm xx mm 20% Triangular xx / xx g
Figure 2.3: Testing device.
A force-controlled loading is carried out until failure or reaching a displacement of XXX mm.
The tested specimens and the load-displacement curves are presented in fig and fig, respectively.
tested specimens
load disp curve 19gr/ 16gr
comments on the results, differences with FE modelling, reasons.

May 12th , 2023 Page 6 of 10


REPORT 2: Structural Optimization of 2D structures. Continuum optimization.
41561 Structural Optimization | Group 06

(2.5-3 pages)
XXX
• Introduction to the problem. Brief description. Boundary conditions, loads, materials.
• Topology optimization. Settings in Comsol, objective function, constraints, design do-
main.
• Shape optimization. Settings in Comsol, objective function, constraints.
• Parameter effects study (filtering radius, mesh). Comparison of the performance, and
(convergence plots).
• Comparison of performance of final designs. Include projection.
• Experimental testing. 3D printing settings (shell and wall thickness, volume fraction,
infill pattern, weight) and brief description, pictures of the specimens maybe not needed,
sketch or picture with the setup. Plot results (load displacement curve, stiffness).
• Analysis of results. Leave some discussion for later. Observe the stiffness, the ductility,
the strength, failure forms.
• Leave maybe the why for the conclusions? Namely: 3D printing quality, temperature
effects, infill, pattern, imperfections, straightness, out of plumbness, loading device, flat-
ness, stress concentrations, clamping and prestressing on the base. Something about the
sound when failing and cracking.
• Some of the pictures can be left for the appendix e.g. failure?

May 12th , 2023 Page 7 of 10


REPORT 2: Structural Optimization of 2D structures. Continuum optimization.
41561 Structural Optimization | Group 06

3 Discussion
(0.5 pages)


May 12th , 2023 Page 8 of 10


REPORT 2: Structural Optimization of 2D structures. Continuum optimization.
41561 Structural Optimization | Group 06

Appendix A
XXXX
XXX

May 12th , 2023 Page 9 of 10


REPORT 2: Structural Optimization of 2D structures. Continuum optimization.
41561 Structural Optimization | Group 06

References
[1] 41561, Structural Optimization. Report 1: Structural Optimization of a 2D structure. DTU
Civil and Mechanical Engineering. Technical University of Denmark. 2023.
[2] Erik Andreassen andothers. “Efficient topology optimization in MATLAB using 88 lines
of code”. inStructural and Multidisciplinary Optimization: 43 (2011), pages 1–16.
[3] 41561, Structural Optimization. Final Project, description. DTU Civil and Mechanical
Engineering. Technical University of Denmark., 2023.

May 12th , 2023 Page 10 of 10

You might also like