Professional Documents
Culture Documents
Fulltext03 PDF
Fulltext03 PDF
Integrating Design
Optimization in the
Development Process using
Simulation Driven Design
Linköping University
SE-581 83 Linköping, Sweden
013-28 10 00, www.liu.se
Abstract
This master thesis has been executed at Scania CV AB in Södertälje, Sweden.
Scania is a manufacturer of heavy transport solutions, an industry which is chang-
ing rapidly in order to meet stricter regulations, ensuring a sustainable future.
Continuous product improvements and new technologies are required to increase
performance and to meet markets requirements. By implementing design optimiza-
tion in the design process it enables the potential of supporting design exploration,
which is beneficial when products with high performance are developed.
The purpose was to show the potential of design optimization supported by
simulation driven design as a tool in the development process. To examine an alter-
native way of working for design engineers, elaborating more competitive products
in terms of economical and performance aspects. Furthermore, to minimize time
and iterations between divisions by developing better initial concept proposals.
The alternative working method was developed iteratively in parallel with a case
study. The case study was a suction strainer and were used for method improve-
ments and validation, as well as decision basis for the included sub-steps.
The working method for implementing design optimization and simulation
driven design ended up with a procedure consisted of three main phases, con-
cept generation, detail design and verification. In the concept generation phase
topology optimization was used, which turned out to be a beneficial method to
find optimized solutions with few inputs. The detail design phase consisted of a
parameterized CAD model of the concept which then was shape optimized. The
shape optimization enabled design exploration of the concept which generated
valuable findings to the product development. Lastly the optimized design was
verified with more thorough methods, in this case verification with FE-experts.
The working method was tested and verified on the case study component,
this resulted in valuable knowledge for future designs for similar components. The
optimized component resulted in a performance increase where the weight was
decrease by 54% compared with a reference product.
i
Acknowledgments
This thesis is the final assignment for master’s studies at Linköping University.
The work has been conducted at Scania CV AB during the spring 2019, covering
20 weeks of work equal to 30 credits. The thesis was challenging but has given us
valuable knowledge both in an academic and an industry perspective.
We would like to thank Scania and everyone involved in our thesis for making
this possible. Additional thanks to our supervisor, Johan Salomon, for the help-
fulness and engagement to make our time inspiring and pleasant. We are thankful
to perform our work at the NMBO department, the positive attitude and interest-
ing discussions has been worthwhile. Another thanks to Mikael Tellner and Tina
Louka for valuable insights within the subject of optimization and simulation.
We are also thankful for the feedback and support from our supervisor at
Linköping University, Johan Persson. And finally, we would like to show our
gratitude to our opponent, Mattias Andersson, for providing valuable feedback
and dedication of time and effort throughout the thesis.
iii
Nomenclature
Abbreviation Meaning
CAD Computer-Aided Design
CATIA V5 Computer-Aided Three-Dimensional Interactive Application V5
DOE Design of Experiments
FE Finite Element
FEM Finite Element Method
GAS Generative Assembly Structure Analysis
HEEDS Hierarchical Evolutionary Engineering Design System
KBE Knowledge Based Engineering
MDO Multidisciplinary Design Optimization
NMBO Base engine lubrication system
PD Product Development
SDD Simulation-Driven Design
TO Topology Optimization
v
Contents
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Purpose and Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Research questions . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Deliverables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.5 Delimitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Theoretical framework 5
2.1 Product development process . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 Design Process . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.2 Design Paradox . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 CAD-modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.1 Parametrization . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.2 Knowledge based engineering . . . . . . . . . . . . . . . . . 9
2.2.3 CAD model robustness and flexibility . . . . . . . . . . . . 9
2.3 Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3.1 Multidisciplinary design optimization . . . . . . . . . . . . 11
2.3.2 Topology optimization . . . . . . . . . . . . . . . . . . . . . 12
2.4 Simulation driven design . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4.1 Finite element method . . . . . . . . . . . . . . . . . . . . . 13
2.4.2 Frequency response . . . . . . . . . . . . . . . . . . . . . . . 14
2.5 Sand casting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3 Thesis methodology 19
3.1 Pre-study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.1.1 Literature study . . . . . . . . . . . . . . . . . . . . . . . . 20
3.1.2 Study of current working method and case study . . . . . . 20
3.2 Development of working procedure . . . . . . . . . . . . . . . . . . 20
3.2.1 Method development . . . . . . . . . . . . . . . . . . . . . . 21
3.2.2 Case study validation of working process . . . . . . . . . . . 21
vii
4.2 The case study component- Suction strainer . . . . . . . . . . . . . 25
4.2.1 Description of the component . . . . . . . . . . . . . . . . . 25
4.2.2 Requirements on the suction strainer . . . . . . . . . . . . . 26
5 Results 29
5.1 Developed method results . . . . . . . . . . . . . . . . . . . . . . . 29
5.1.1 Phase 1: Start-up . . . . . . . . . . . . . . . . . . . . . . . 30
5.1.2 Phase 2: Design requirements and concept generation . . . 30
5.1.3 Phase 3: Detail design and design exploration . . . . . . . . 30
5.1.4 Phase 4: Design verification and final decision . . . . . . . . 31
5.2 Case study results . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.2.1 Case study-Phase 1 . . . . . . . . . . . . . . . . . . . . . . 31
5.2.2 Case study-Phase 2 . . . . . . . . . . . . . . . . . . . . . . 32
5.2.3 Case study-Phase 3 . . . . . . . . . . . . . . . . . . . . . . 33
5.2.4 Case study-Phase 4 . . . . . . . . . . . . . . . . . . . . . . 41
5.2.5 Case study improvements . . . . . . . . . . . . . . . . . . . 42
6 Discussion 45
6.1 Methodology discussion . . . . . . . . . . . . . . . . . . . . . . . . 45
6.2 Result discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
6.2.1 Developed working method . . . . . . . . . . . . . . . . . . 46
6.2.2 Case study discussion . . . . . . . . . . . . . . . . . . . . . 47
7 Conclusion 51
7.1 Research question 1 . . . . . . . . . . . . . . . . . . . . . . . . . . 51
7.2 Research question 2 . . . . . . . . . . . . . . . . . . . . . . . . . . 52
7.3 Research question 3 . . . . . . . . . . . . . . . . . . . . . . . . . . 53
7.4 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
Bibliography 55
List of Figures
2.1 Product development process . . . . . . . . . . . . . . . . . . . . . 5
2.2 The design process paradox . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Morphological transformation levels . . . . . . . . . . . . . . . . . 8
2.4 Topological transformation levels . . . . . . . . . . . . . . . . . . . 8
2.5 Benefits with implementing KBE in the design process . . . . . . . 9
2.6 LHS example, two dimensional with four sample points . . . . . . 11
2.7 Illustration of a Pareto front . . . . . . . . . . . . . . . . . . . . . 12
2.8 Illustration of a topology optimized beam . . . . . . . . . . . . . . 13
2.9 Meshed triangular elements in a rectangular domain . . . . . . . . 14
2.10 Representation of the effect of junction and creation of hot spot . . 16
2.11 Design requirements for sand casting . . . . . . . . . . . . . . . . . 17
Introduction
1.1 Background
The most critical factor in profit-driven enterprises is the ability to develop suc-
cessful products with an economic success, where the development process is one
of the fundamental parts [2]. Since the product development process requires
man-hours, the total development cost of products can be reduced by changing
companies working methods. Scania’s development process starts with an initial
CAD model from a design engineer followed by analysis from FE-experts. Based
on the simulation results the design engineer makes changes and send the CAD
model to the simulation engineers iteratively until the product meets the require-
ments. Then the component needs to be physically tested and approved from an
assembly and purchasing point of view until the final design gets approved [3].
This development process requires many changes to the CAD model and therefore
time due to demands from the different disciplines. Since the design process is
time consuming there is short amount of time for increasing the performance of
the components even further.
The complexity of products are increasing successively and by implementing
design optimization at an early stage in the design process, the support for the
1
2 Introduction
decision making increases. The design engineer can deliver a better initial pro-
posal for the specific product, this induces fever iterations between the different
disciplines. The procedure of design optimization can to a high extent explore re-
lationships between the various properties of a product. By using an optimization
algorithm together with simulation driven design, the exploration of the design
space increases and simultaneously a product with higher performance can be
found and created [4].
1.4 Deliverables
During the thesis the following deliverables will be performed gradually:
• Investigate the requirements on the case study.
• Develop a working method for how design engineers can work with design
optimization.
• Apply the method and develop an optimized suction strainer.
• Evaluate the developed design suggestion of the suction strainer.
• Evaluate the working method where design optimization has been imple-
mented.
1.5 Delimitations 3
1.5 Delimitations
The thesis work is performed from a design engineers’ point of view. Since the main
focus is on how design optimization can be used to provide a more competitive final
product. The provided working method will focus on how the design engineering
should work to minimize the required iterations between different disciplines.
The developed working method will be applied on only one case study, but
the method should be applicable on similar products. The development of the
case study component will stop at FE-verification, which means no prototypes or
physical testing will be performed.
For this thesis Altair Inspire will be used for topology optimization, CATIA
V5 will be used for CAD-modelling, CATIA Frequency Analysis will be used for
calculations and Heeds will be used as the optimization software. These software’s
will be investigated mainly because Scania is using them today which makes it
easier to implement the findings in their development process.
4 Introduction
Chapter 2
Theoretical framework
This chapter describes the theory and literature studies relevant for the thesis and
the studied field. The chapter covers areas such as product development process,
CAD-modelling, optimization, simulation driven design and relevant manufactur-
ing methods.
5
6 Theoretical framework
The product development process starts with the planning phase which defines
the opportunities and evaluates the market objectives and potential technologies.
When ending the planning phase, the mission statement of the product should be
defined, containing the business goals, target market and limitations [5].
After the planning phase ends the concept development phase starts. The
crucial task in this phase is to identify the needs and generate several concepts
with desired performance. Followed by selecting one or a few concepts for further
development and testing.
When one concept has been chosen the process continuous into the system-level
design where the architecture of the product is determined. The product concept
is decomposed into several subsystems and parts, were the crucial components are
preliminary designed. The assembly scheme for production should also be defined
in this phase.
The following phase is detail design, where each part is fully specified in terms
of material, geometries and tolerances. The manufacturing preparation of the
components should also be completed in this phase and production cost should be
definite.
The next step is testing and refinement phase were the product performance is
evaluated by simulations or physical testing making sure it will fulfil the require-
ments before entering the final phase which is production ramp up.
In the production and ramp up phase the assembly staff are educated and if
there are any production related issues left they should be solved before the full
scale production [5].
2.2 CAD-modelling
In order to work more efficiently with product development there are several tools
which can be used. This chapter describes how to work smarter with CAD-
modelling, integrating information by knowledge based engineering (KBE) and
how to make models more usable and flexible using parametrization.
2.2.1 Parametrization
One way of making the product development process more time efficient is to
develop reusable CAD-models. This can be done by controlling the CAD-model
with parameters which are non-geometric features. The basic idea is to be able to
reuse the CAD model by modifying its geometry with minimal effort [9]. Making
it possible to generate several versions in for example a product family [10].
There are various levels of parametrization which are divided into morpho-
logical and topological transformations. Morphological transformations represent
changes in shape while topological transformations involve positioning of objects
and features within a CAD-model [11].
As figure 2.3 shows, the morphological transformation contains four levels.
Fixed object is the lowest level which corresponds to a model without the ability
to change shape. The parameterized object level represents models with possibility
to change shape with the help of parameters, lacking relations between the various
parameters. The equation-based relation level involves dependencies between pa-
rameters. The top level of morphological transformation is script-based relations
describing relations with programming [11].
8 Theoretical framework
Figure 2.5: Benefits with Implementing KBE in the design process [14].
2.3 Optimization
Optimization can be explained as finding the best solution among several feasible
ones. The feasible solutions are all solutions not violating any constraints. Ob-
jective function is the function describing the desired property which can either
be minimized or maximized. Where the objective function can represent vari-
ous properties such as weight, efficiency and manufacturing cost. The constraints
are properties expressed with functions which have limits not to be exceeded, for
example maximum allowed stress or minimum flow performance [16].
One way of describing the mathematical formulation of an optimization prob-
lem can be expressed as below. Where f(x) is one or several objective functions,
g(x) and h(x) are constraint functions and the vector x represent the design vari-
ables [16].
Objective function(s):
f(x)k k=1,2,3,..., K
Subjected to:
g(x)i ≤ 0 i=1,2,3,..., m <n
h(x)j =0 j=1,2,3,..., r <n
where
x1
x2
x= .
..
xn
2.3 Optimization 11
Within optimization experiments are changes of inputs by a given rule and iden-
tifying the correlating output. Design of experience (DOE) is the umbrella term
for techniques to efficiently guide selection of experiments [17]. Latin Hypercube
sampling is one DOE technique which is inspired by a mathematical combination
namely the Latin Square, where an N x N matrix is filled with N objects so they
cover each row and column in the matrix, see figure 2.6. This technique can be
adapted to cover and explore the design space within optimization [18].
gether with a set of constraints called boundary conditions. There exist three main
categories of boundary value problems, equilibrium and steady-state, eigenvalue
problems and transient problems. An equilibrium problem is often solid mechanic
problems where the displacement or stress distribution is defined. In an eigenvalue
problem the natural frequencies are calculated. The type of transient problems
are time-dependent and are used for instance when there is interest of finding the
response of time-varying forces [27].
According to Liu and Quek [28], the procedure of using the finite element
method consist primarily of four steps:
1. Modelling of the geometry
2. Meshing of the geometry
3. Define material property
4. Specification of boundary and loading conditions.
In engineering design the geometry is created in a CAD-software. However,
the ”real” CAD-model created by the designer is often very complex and need to
be simplified in order to perform a good analysis. The model can then be meshed
were the geometry is divided into small pieces. It is important to have the right
coarse of the mesh, providing an accurate result while decreasing the simulation
time. The type of material and which load conditions the geometry is subjected
to must be defined before the analysis is performed [28].
• Long transport sections for the melt must have a suitable wall thickness.
• Seek after a straight parting line for the component.
• Design it to be easy to pour and have into account where the ricer will be
located.
Figure 2.10: Representation of the effect of junction and creation of hot spot [31].
Thesis methodology
This chapter describes the methodology and working procedure adapted in this
thesis. The work was divided into two main parts, first a pre-study was performed
and based on those findings the main work and result generation were conducted.
The general workflow outline for the thesis is presented in figure 3.1.
The first phase, pre-study, included a literature study in the investigated field,
as well as a current situation analysis of Scanias working method to get knowledge
and find improvement potentials. In the thesis a developed method was tested
and improved with a case study component. The requirements on the case study
was in this stage defined and documented.
The second part of the work was the primary result generation, this process was
accomplished in an iterative manner, where the working method was developed
in parallel with the case study validation. Finally, the thesis result was presented
with a final developed working method as well as a finalized case study.
3.1 Pre-study
To provide a knowledge base for this thesis, studies of relevant literature in the
field and former thesis work was realized. A study of the current working method
19
20 Thesis methodology
This chapter describes the current product development method at Scania and a
pre-study of the investigated case study component. The information presented
in this chapter has been collected from internal sources at Scania and published
reports in the field.
23
24 Current situation analysis
Figure 4.2: Design engineers role the in product development process [34].
The suction strainer is low volume production item and is therefore manufac-
tured using sand casting. The material used is aluminium and has the properties
shown in table 4.1.
Table 4.1: Material properties for the suction strainer.
Properties Value
Material EN AC-43100 SF
Density 2770 kg/m3
Young’s modulus 71 GPa
Poisson ratio 0.33
Yield strength 150 MPa
Fatigue limit 45 MPa
as oil pan, balancing shafts, engine block etc. The design was also limited by the
fixed position of the mounting points, oil pick-up location and oil-stick container.
Since the component is placed inside an engine the loading conditions it must
withstand are vibrations during runtime. The vibrations induce fatigue on the
component and frequency calculations had to be performed to find stress concen-
trations and fatigue limits. The engine has a specific frequency spectrum which
defines the acceleration excitation which is acting on the component. These values
have been measured by physical testing for Scanias various engines. Generally, the
engines does not exceed a frequency range of 300 Hz, see figure 4.4. An investi-
gation of the eigenfrequency of the suction strainer must be performed. If the
natural frequency is below 300 Hz the component can start to oscillate and get
damaged. By using this limit the probability of oscillations for the component de-
creases significantly. The component must also have a adequate stiffness to ensure
a low deflection.
Figure 4.4: The acceptance criteria for reduce the risk of oscillations.
28 Current situation analysis
Chapter 5
Results
This chapter presents the gathered results from the method development and the
results from the newly developed case study component. Followed by performance
comparison between the developed and the original component.
29
30 Results
in some manner, making sure the simulations are not performed in vain with a
concept not manufacturable.
The CAD model should then be analysed with relevant simulation models, in
either the used CAD software or in another software for simulations. By setting up
the simulation framework with relevant load cases and supports the optimization
software can then explore the design space by varying the parameters in the sim-
ulation model. Resulting in various design suggestions to be compared in terms
of the desired performance. One of the design suggestions will be chosen and
analysed further for verification.
The disciplines involved for this project was primarily design engineers and cal-
culation engineers with competence in the field of vibrations- and frequency cal-
culations. The reason why those disciplines were involved was because of the
components operating environment. If this project was to develop a totally new
32 Results
concept the discipline of CFD should be involved as well, but because of perfor-
mance satisfaction for the already existing fluid pipes in the design those was not
involved.
In this phase it was decided that the design engineer should do most of the
development work and the calculation engineer should in the end verify the pro-
posed concept. The calculations performed had a level of difficulty which in some
extent was executable for a design engineer to give a good result estimation.
suitable partition plane for the component, the optimization algorithm took the
mould removal into account.
The optimization problem was then defined with the input to maximize the
stiffness as the objective. The problem was constrained by a minimum allowed
eigenfrequency of 300 Hz, as well as a minimum thickness of 5 mm, see chapter
4.2.2.
The result from the optimization run can be seen in figure 5.3. The material
placement, in orange, had a structure with several branches from the mounting
points to the suction pipe. The oil-stick cup was connected to one of the mounting
holes. The optimized solution was used as a baseline for the parametric CAD-
modelling in CATIA, the model was inserted into CATIA for simplifying the re-
creation of the optimized solution.
parameters were enabled to vary the elliptic cross sections of the braces at various
positions. The angles of where the braces were connected to the fixed geome-
tries and the ratio of where brace 2 connected to brace 1 was also varied with a
parameter. The result of the concept is illustrated in figure 5.4.
5.2 Case study results 35
Figure 5.4: The first CATIA developed concept of the suction strainer.
Minimize:
Mass [kg]
Subjected to:
Eigenfrequency > 300 Hz
Fatigue limit, σ f < 6 MPa
Geometry interference check = OK
The first CATIA analysis was a geometry check and consisted of a Clash Anal-
ysis, secondly a Frequency Case and three Harmonic Dynamic Response Cases was
calculated. The Clash Analysis was performed to make sure the suggested designs
from Heeds did not result in interference with the surrounding components in the
engine. If the Clash analysis found an interference between the suction strainer and
the nearby components, the design was considered to be failed and the Frequency
Analysis and the Harmonic Dynamic Response Cases were not computed.
The Frequency Case calculated the eigenmodes for each design. In the Fre-
quency Case the mounting holes were restrained with fixed displacement and
rotation. The Frequency Case was then used as the reference for the restraint
5.2 Case study results 37
Figure 5.6: The frequency and excitation amplitude in X,Y and Z-direction.
Optimization results
The parametric CAD model was then optimized using Heeds as the solver and CA-
TIA analysis for evaluating designs, one example of a feasible design from Heeds
can be seen in figure 5.4. The concept did fulfil the constraints regarding mini-
mum allowed eigenfrequency for the three modes, it also fulfilled the constraints
regarding fatigue limit when exciting the component according to the excitation
spectra in X-, Y- and Z-direction.
However, this concept did not fulfil the requirements regarding manufactura-
bility for sand casting since the cross sections varied too much which generates
hot spots. Nor was the concept cost efficient since it would have required several
partition planes for the braces which results in a more expensive manufacturing
process of the mould. From the Heeds optimization the conclusion was drawn
that the cross sections of the braces were large in order to reach the minimum
allowed eigenfrequency and not for the fatigue limit since no design violated the
constraint of 6 MPa. Which implies the concept could be improved to a structure
using less material while still fulfilling the eigenfrequency constraints. The deci-
sion was made to redo phase 3 from the beginning, to find a concept fulfilling the
requirements for the manufacturing method.
38 Results
Figure 5.7: The second CATIA developed concept of the suction strainer
The parameters used in the second concept can be seen in figure 5.8. For
example the material thicknesses were controlled individually for the geometries
to the left and right. The rib thicknesses placed underneath the left and right
geometry was also changed individually. The parameter values for the design is
summarized in table 5.1.
5.2 Case study results 39
To make sure the CAD-model had parameters enabling a wide range of con-
figurations with few errors, a robustness and flexibility analysis was conducted.
The robustness for the model was calculated using equation 2.1. The analysis
was performed on 100 different design configurations in the design space using the
Latin Hypercube design of experiments technique. Of these 100 designs five failed
which implies a robustness of 95%.
Response Value
Mass 0.69 Kg
Frequency Mode 1 316 Hz
Frequency Mode 2 400 Hz
Frequency Mode 3 527 Hz
Von Mises Stress X-direction 4.6 MPa
Von Mises Stress Y-direction 5.6 MPa
Von Mises Stress Z-direction 2.3 MPa
Rib Height 15.36 mm
Material Thickness R 3.28 mm
Material Thickness L 3 mm
Rib Thickness Right 5.04 mm
Rib Thickness MP12 3 mm
Rib Thickness Above 3 mm
Rib Thickness Left 4 mm
Hole Size L 100 mm
Hole Size R L 50 mm
Hole Size R U 50 mm
A Pearson correlation matrix of the two most critical responses was created,
which which was the first eigenmode and the mass. The figure 5.10 illustrates the
parameters which the responses mentioned were most sensitive to. One important
note from the correlation matrix was the first eigenmodes large sensitive to the
mass of the suction strainer. Which is a reason why the shape optimization in
Heeds reached a lower limit of acceptable mass for not violating the eigenfrequency
constraint. The frequency constraint was most sensitive to be the rib height, since
it affects the moment of inertia the most. The mass was most sensitive to the
material thickness of the left geometry and its corresponding hole size.
Response Value
Frequency Mode 1 [Hz] 314
Frequency Mode 2 [Hz] 416
Frequency Mode 3 [Hz] 508
RMS Mises Stress X-Direction [MPa] 5.5
RMS Mises Stress Y-Direction [MPa] 9.7
RMS Mises Stress Z-Direction [MPa] 6.6
Figure 5.11: Comparison between the developed and original suction strainer.
Table 5.3: Comparison of the original suction strainer and the developed concept.
Discussion
This chapter discuss the outcome from the thesis and is divided into two parts.
First the methodology is discussed, which describes how it may has affected the
outcome. In the second part the main results are discussed.
45
46 Discussion
ever not official with forms etc. The interviews would cover what the development
engineers at Scania thinks about the current working method. Which would be a
good base when developing the suggested working method and avoiding flaws in
the current method.
be performed, this would however not be time efficient given the time accessible
during the thesis.
Topology optimization
The usage of topology optimization in the product development process turned
out to be a good concept generator. The topology optimized result gave valuable
information in a relatively easy and time-effective way. The method gave inspira-
tion to the finalized shape optimized concept, which indicates a product could be
developed in a time-effective way and desirable results could still be found.
However, the optimization method requires a correct input in order to reach the
desirable results. The boundary input for the case study was relatively easy, since
the case study was an improvement of an existing component the functionalities
and boundaries were already defined. If this was a totally new concept with
few limitations the topology optimization would be more complex, this would
require the designer to first define the functionalities, surrounding components
and fixed geometries before a material boundary was developed and optimized.
The definition of the input can also restrict the exploration of possible designs,
if the product has high extent of freedom several concepts proposals needs to be
developed with different boundary input data. Since the case study was pre-defined
one concept was enough.
The authors had no previous experience of Inspire and topology optimization,
the process of developing a working concept was done iteratively. If previous
knowledge existed about the process and program the outcome may had been
somewhat differently, more experience would give a better understanding of how
the optimization work.
The case study was only subjected to vibrations which generated a more com-
plex problem. In order for the topology optimization to place material load cases
had to be defined. The solution to this was to approximate loads in the non-fixed
geometries. Depending of this approximation the result outcome may differ.
When a design engineer has gained experience the usage of the topology opti-
mization can be usable in some cases and less usable in others. If the load case is
relatively straight forward and the component is simple as a bracket for example.
Then an experienced engineer can explore and evaluate a component faster with-
out topology optimization, by using CATIA analysis and simply cutting material
and see where material is needed. While if a component has a complex geome-
try and load case, the topology optimization will perform results an experience
engineer has difficulties to generate.
The design paradox, from chapter 2.1.2, implies changes becomes more trou-
blesome and costly when project knowledge increases as the project proceeds.
48 Discussion
When connecting these statements with topology optimization one can say by set-
ting up the topology optimization problem does require substantial knowledge of
the product. When working with product development projects knowledge of the
component already exists, since Scania uses their method called lesson learning de-
scribed in chapter 4.1.1 to avoid similar problems and transfer knowledge between
employees. This results in that design engineers should be able to find key infor-
mation when starting the project making big changes with topology optimization
acceptable and cost efficient.
Shape optimization
Overall, the shape optimization did not give a large benefit for the specific case
study after the topology optimization. The main reason why was because of the
narrow design space from start due to surrounding components. The topology
optimization gave a very large impact of the result by suggesting an optimized
geometry based on the design space and load case. The shape optimization would
have been more beneficial if the design freedom was larger, giving the optimization
algorithm more to work with. The case study had the objective of minimizing the
weight, which resulted in a small improvement from topology optimization to the
shape optimization result. When optimizing a component with multiple objectives
counteracting each other the shape optimization is of much greater interest. Since
it gives the user the possibility to evaluate the different possibilities and trade-offs
for various disciplines such as CFD, FEM etc.
In order to successfully carry out shape optimization, a robust and flexible
parametric CAD model is required. In order to allow the optimization algorithm
to explore a large design space and generate relevant results. Creating a para-
metric CAD model with the properties mentioned can be time consuming if the
engineer lacks experience or if the component is complex to make parametric. For
example, if the component has geometric tolerances prohibiting changes in certain
dimensions or if the product consists of complex shapes.
Measuring the robustness of the parametric CAD model before carrying out
the shape optimization, is a good verification making sure the CAD model is
suitable for shape optimization. Note, the level of robustness is a measurement
the user must be somewhat cautious about. Robustness changes with the design
space, which means a parametric CAD model with poor flexibility, can have a high
robustness if the design space is small. It is therefore better to increase the design
space and if the robustness decreases solve the flaws with the CAD model instead
of decreasing the design space.
When developing the CAD model the manufacturability should be considered,
which will lead to abnormalities from the geometry suggested from the topology
optimization. This was a good learning from the authors who for the first concept
followed the topology optimization too strictly, which lead to a concept which was
not suitable for sand casting. Therefor, the authors defined what requirements the
component needed to be manufacturable. Before a new parametric CAD model
was developed. The extra iteration was not between disciplines but within one
discipline, which would have led to an increase in the duration if it would appear
later in the design process.
6.2 Result discussion 49
For the shape optimization more parameters could have been used to find
designs with even better performance. The trade-off between perfecting the results
and the time required applies for every project covering optimization. Since the
case study only had the objective to minimize the weight, the importance of each
parameter was rather obvious. This might not be the case if for instance a problem
with multiple objectives or a complex geometry is considered.
The optimization problem for the suction strainer was relatively straight for-
ward, however it requires skill and experience to define optimization problems
correct. By making the optimization simplified in the beginning, visualization of
the optimization framework gets easier which might increase the understanding.
The need of experience increases with the complexity of the optimization problem.
If several disciplines are involved the time duration will increase, making it more
inconvenient when several disciplines need to work in parallel.
The frequency response analysis required the excitation accelerations acting on
the suction strainer which was accessible at Scania. If this would not be the case,
simplifications could be made by approximating constant acceleration. However,
this would give a worse approximation.
The selected design from the optimization in Heeds did not have performance
values at the various constraint limits. The reasoning behind the selection was be-
cause there will always be imperfections in the manufacturing and the simulations,
leading to deviation from the performance in reality. Which might result in a final
product with lower performance than the constraint limits. Therefore, selection
of designs not too close to the constraint values are a good safety consideration.
Design verification
The validation phase is fundamental to ensure achievement of product require-
ments. The kind of validation and extension will vary depending on type of prod-
uct and uncertainties regarding the operation environment. The case study com-
ponent was verified with a more accurate frequency calculation model compared
with the one used in the optimization framework. The reason this verification
was enough for this component was the previous experience of similar products.
Physical testing was neglected due to the scope, but it would verify the component
even further.
The verification of the design was performed by the authors, the reason being
if an FE-expert was going to perform the calculations it would have taken several
weeks, since the thesis was not prioritized. Therefore, an FE-expert consulted and
directed the calculations with correct method and procedure. The final decision
was done by the FE-expert but based on the generated results from the authors.
This would not affect the trustworthiness of the results.
Depending on the product the calculation model used in the optimization
framework can vary, the type of verification and calculations model need to be
defined in the start-up phase of the project. If the calculation model is accurate
enough and experience exist of similar components, the verification step could be
excluded.
50 Discussion
Chapter 7
Conclusion
This chapter summarizes the conclusions from this master thesis, covering the
goals, purposes and answering the research questions stated in the introduction
chapter. The goal with the thesis was to develop a working process including design
optimization and simulation driven design to implement at Scania. The method
was tested on a case study, ensuring the most valuable steps was included to
successfully carry out design optimization. The purpose of the thesis was to show
the potential with design optimization in the development process and how it can
minimize iterations among disciplines. This was achieved by showing a new design
with a weight reduction of 53,5% compared to the original design with minimum
amount of discipline iterations. The weight reduction results in an increase of the
components performance and since the weight correlates directly to the material
volume needed it also results in an improvement regarding economical aspects.
In order to meet a sustainable future continuous development of each com-
ponent to maximize performance and minimize weight is needed to minimize the
global emissions. By reducing the weight for one component does not make a
substantial impact, but for each weight optimized component the sum of emission
savings increases. When the weight of the vehicle is decreased the load carrying
capacity increases, which enables companies to reduce the number of journeys
needed to move same amount of freight [1].
Design optimization has the main advantage of making sure a product has the
desired properties. A more traditional design process involves several iterations
where designers and FE-experts collaborates until the product fulfils all the re-
quirements. Since this process can be time consuming, the product tends to be
good enough but not optimized. This is the subject were the design optimization
excels, since it enables automatic design exploration which improves the perfor-
51
52 Conclusion
mance while also minimizing the time-consuming iterations between the disci-
plines.
Using the proposed working method and following the instructions step by step
will in the long run reduce the development time and generate more competitive
products. Initially there will be a threshold when developing parametric CAD
models, using topology optimization in Inspire and shape optimization in Heeds.
However, when the user has used the method for one product the gained experience
will be applicable on the next product. Since the design engineers at Scania
already uses FEM modules for simplified calculations in CATIA, this part will not
be different when using the design optimization. The last step in the method is
verification with more computationally expensive simulations by FE-experts, this
verification step are also used at Scania in their development method today. The
developed working method is therefore adaptable to Scanias current process, it will
minimize the iterations needed between the design engineer and the FE-experts
by providing a better initial design.
Topology optimization is very useful in the concept generation phase, it will show
the user where material is needed and where it is useless. This might seem obvious
for experienced engineers however this is not the case. The topology optimization
will generate geometry suggestions using FEM calculations which even an experi-
enced engineer has difficulties to generate.
The developed method during this master thesis has integrated topology op-
timization as a step in the design process which the user can follow. Where a
thorough description will provide engineers with no or brief experience in topol-
ogy optimization the useful steps to carry out a topology study of a component.
The main advantage of the topology optimization is that it provides the user de-
sign suggestions which a user probably not generate, based on FEM calculations
where the material is needed. One disadvantage with the topology optimization is
that the users need to be certain of the load cases acting on the component, since
it computes where material is needed based on the load case. Which means the
suggested design is not relevant if the input assumptions are incorrect. Another
disadvantage is if no fixed geometries are defined such as mounting points and con-
nection pipes, then several concepts and therefore several topology optimizations
needs to be performed. Lastly the CAD modelling of the geometrical boundaries
the topology optimization requires can be time consuming and difficult to define.
The main difference with the design process used at Scania today is in their
concept development projects, the concepts are evaluated based on ranking func-
tionalities and what the project members think is the best. And in their prod-
uct development projects, the product is modelled up in CATIA directly without
any proper concept evaluation. Whereas the topology optimized shape can be
evaluated and then modelled in CATIA where adjustments are made to make it
7.3 Research question 3 53
The main difference between the developed working process and a traditional
working process is the usage of design optimization and simulation driven design.
The developed method is applicable for similar products but probably not general
enough to apply on all kind of products and systems. The main reasons for this
non generality was the need to be specific in order for a design engineer to be
able to use it and for the authors to validate the method on a case study. The
method was developed with Ulrich and Eppingers design process and Scanias de-
sign process as foundations, enabling the possibility for combining parts from the
three methods into a customized method for product development if needed. The
developed working process contains the same steps of concept development, detail
design and testing and refinement which both Ulrich and Eppingers and Scanias
method contains. However, the contents within the steps are different due to the
focus on design optimization. The developed method lacks the step of system level
design which Ulrich and Eppingers method contains however the surrounding com-
ponents should be evaluated in Phase 2 when defining the geometrical boundaries
for the topology optimization step. Neither does the developed method contain
the step of production ramp-up. Mainly because it was not needed to validate the
impact when using the design optimization method, it does not need to change
when using design optimization and the step used by Scania was considered to be
complete. Since the sequence of activities are the same for the developed method,
an integration of it into the traditional working method can easily be done.
54 Conclusion
[1] Inge Vierth, Samuel Lindgren, and Hanna Lindgren. Vehicle weight, modal
split, and emissions-an ex-post analysis for Sweden. Sustainability, MDPI,
2018.
[3] Scania CV AB. Product development process, 3:rd edition, STD4303, 2013.
[4] Timothy W. Simpson and Joaquim R. R. A. Martins. Multidisciplinary De-
sign Optimization for Complex Engineered Systems: Report From a National
Science Foundation Workshop. Journal of Mechanical Design, 2011.
[5] Karl T. Ulrich and Steven D. Eppinger. Product Design and Development:
Fifth Edition. McGraw-Hill, 2012.
[6] David G. Ullman. The Mechanical Design Process. McGraw-Hill, 2013.
[7] M Lindahl and J Tigerström. En liten lärobok om miljöeffektanalys. Institu-
tionen för maskinteknik, Högskolan i Kalmar, Kalmar, 3rd edition, 2006.
[8] Dejan Lukic, Mijodrag Milosevic, Stevo Borojevic, Mica Durdec, Jovan Vuk-
man, and Aco Antic. Manufacturing cost estimation in the conceputal process
planning. Machine Design, 8(3), 2016.
[9] Jorge D. Camba, Manuel Contero, and Pedro Company. Parametric CAD
modeling: An analysis of strategies for design reusability. Computer-Aided
Design, 2016.
[10] Jianxin Jiao and Mitchell M. Tseng. Fundamentals of product family archi-
tecture. Integrated Manufacturing Systems, 2000.
[11] Kristian Amadori, Mehdi Tarkian, Johan Ölvander, and Petter Krus. Flex-
ible and robust CAD models for design automation. Advanced Engineering
Informatics, 2012.
55
56 Bibliography
[27] Singiresu S. Rao. The finite Element Method in Engineering. Elsevier Science
& Technology, fourth edition, 2005.
[28] G. R. Liu and S. S. Quek. Finite Element Method: A Practical Course.
Elsevier Science Ltd., first edition, 2003.
[29] Brian J Schwarz and Mark H Richardson. Experimental Modal Analysis.
Vibrant Technology Inc., 1999.
[30] Santosh Reddy Sama and Guha P. Manogharan. Sand Casting Design Rules.
Pennsylvania State Univesrity, 2017.
[31] ASM International. Casting design and performance. Materials Park, Ohio,
1:st edition, 2009.
[32] John Campbell. Complete casting Handbook. Elsevier Ltd., UK, first edition,
2011.
[33] C.W. Ammen. The complete handbook of sand casting. Tab Books, USA,
first edition, 1979.
[34] ViChi Luu. Methodology development for parametric CAD modeling in Ca-
tia V5 to aid simulation driven design using turbine volute as a case study.
Technical report, Department of Managemen and Engineering, Linköping,
2015.
[35] W.L. Robertson. Lubrication in practice. Taylor and Francis, 2:nd edition,
1984.
58 Bibliography