256412
256412
256412
Cover: Topology Optimization and Cast Simulation of Rear Lower & Upper Control
Arms.
iv
Cast simulation and manufacturing constraints for detailed part optimization
Master’s thesis in Automotive Engineering
ALIREZA NILIPOUR TABATABAEI
CHETAN KRISHNASWAMYREDDY
Department of Applied Mechanics - Chalmers University of Technology
Abstract
Structural optimization and cast simulation tools has gained significant importance
in automotive industry. The industry is slowly transforming into CAE driven de-
sign process, as the acceptance of CAE simulations in providing new and inspiring
lighter designs with shortened development cycles is demonstrated in recent times.
Optimization has become an integral part of design, there has been a demand for
optimization of aluminum cast parts where engineers are challenged to guarantee
both functionality and cast-ability of the component. On the other hand, casting
simulation finds its applications in later part of design process particularly used by
foundries for optimization of casting process by mold flow and solid simulation.
The purpose of this master thesis is to propose a methodology with emphasis on ap-
plication of manufacturing constraints in the topology optimization and investigate
the possibility of integrating cast simulation tools in optimization driven product
development. In addition, to decrease the lead development time and number of
iterations in design modifications to verify cast-ability of the components. Use of
casting simulation among engineers is rather new and it helps CAE and design en-
gineers to understand the process and identify the defects and problems in early
concept phase, so that there is enough time and flexibility to make changes and
evaluate different concepts and aid them in choosing the best possible design. Op-
tistrut is a finite element based structural analysis software, used to study structural
topology optimization and NovaFlow&Solid is a finite volume based CFD simula-
tion tool, used to study solidification and filling process in casting for evaluating the
generated concepts from different manufacturing constraints.
v
Preface
This master thesis in automotive engineering at Chalmers University of Technology
is carried in cooperation with Volvo Car Corporation under weight and optimiza-
tion department for fulfillment of 30 credits in Gothenburg during spring 2017. The
examiner and academic supervisor was Professor Ragnar Larsson, Head of Division
Material and Computational Mechanics, Department of Industrial and Materials
Science, Chalmers University of Technology. Andreas Carlsson was supervisor in
industry, Optimization Engineer Volvo Cars, Gothenburg.
This thesis was one part of thesis cluster proposed by Optimization Culture Arena,
Where the objective is to develop optimization culture as a natural part of compo-
nent development process and demonstrate capabilities of CAE driven optimization.
vii
Acknowledgements
We would like to thank all the people who contributed in some way to the work
described in this thesis. First and foremost, we thank Andreas Carlsson, our super-
visor at Volvo Car Corporation for accepting us for this thesis work and giving us
intellectual freedom in work, supporting our attendance at RE-OPTIC conferences,
encouragements and guidance throughout the thesis.
We would like to thank Ragnar Larsson, our academic supervisor and examiner at
Chalmers University of Technology for his insightful comments and guidance during
thesis.
We would like to express our deepest gratitude to Harald Hasselbald, Volvo Car
Corporation for providing valuable knowledge and feedback through discussions and
meetings.
We received generous support from Robin Larsson and Chrisat Schuette at VCC,
Pontus Andresson at NovaCast Systems AB, Jakob Olofsson at Jönköping Univer-
sity.
Further, we would like to thank Weight Management & Optimization team for won-
derful fika time and Hrnic Adnan for informative guided tour to durability test
centre at VCC.
viii
Abbreviations
RLCA Rear Lower Control Arm
RUCA Rear Upper Control Arm
NFS NovaFlow&Solid
TO Topology Optimization
DFM Design For Manufacturing
FE Finite Element
DOC Drive Over Curb
BIP Brake In Pothole
ROC Rearwards drive Over Curb
x
Contents
List of Tables xv
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Theory 4
2.1 Topology optimization . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1.1 Responses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.2 Objective function . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.3 Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.3.1 Draw direction . . . . . . . . . . . . . . . . . . . . . 9
2.1.3.2 Member size constraint . . . . . . . . . . . . . . . . . 10
2.1.3.3 Symmetry constraint . . . . . . . . . . . . . . . . . . 10
2.2 Aluminum casting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.1 Die casting . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.2 Permanent mold casting . . . . . . . . . . . . . . . . . . . . . 12
2.2.3 Sand casting . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.3.1 Different sand casting types . . . . . . . . . . . . . . 13
2.2.3.2 Sand mold, core and pattern formation . . . . . . . . 14
2.2.3.3 Gating system . . . . . . . . . . . . . . . . . . . . . 14
2.2.4 Mold filling . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.5 Solidification . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.6 Casting defects formation . . . . . . . . . . . . . . . . . . . . 16
2.2.6.1 Mold filling related defects . . . . . . . . . . . . . . . 16
2.2.6.2 Solidification related defects . . . . . . . . . . . . . . 17
2.3 Modeling of casting process . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.1 Early phase casting simulation . . . . . . . . . . . . . . . . . . 19
2.3.2 Solidification simulation post-processing . . . . . . . . . . . . 20
2.4 Design for casting - Guidelines for design realization . . . . . . . . . . 21
2.4.1 Design of junctions . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4.2 Draft angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
xi
Contents
3 Method 25
3.1 FEM pre-processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2 Parameter Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.1 Problem formulation Study . . . . . . . . . . . . . . . . . . . 26
3.3 Methodology for selecting optimal ISO threshold value . . . . . . . . 27
3.4 Cast simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.4.1 Solidification simulation setting in NovaFlow&Solid . . . . . . 30
6 Discussion 48
7 Future work 50
Bibliography 51
xii
List of Figures
xiii
List of Figures
xiv
List of Tables
xv
List of Tables
xvi
1
Introduction
This master thesis is a part of Optimization Culture Arena under weight manage-
ment and optimization department at Volvo Car Corporation in Gothenburg. The
thesis mainly focuses on developing a generic understanding and methodology which
integrate manufacturing constraints and casting simulations results in component
optimization and development process using topology optimization and casting sim-
ulation tools. This chapter provides the necessary background information, followed
by purpose and limitations of this thesis project. Subsequently, method workflow
and framework of a thesis is explained.
1.1 Background
Structural optimization and cast simulation tools has gained significant importance
in automotive industry. The industry is slowly transforming into CAE driven de-
sign process, as the acceptance of CAE simulations in providing new and inspiring
lighter design with shortened development cycles is demonstrated in recent times.
Optimization has become an integral part of design, there has been a demand for
optimization of aluminum cast parts where engineers are challenged to guarantee
both functionality and cast-ability of the component. On the other hand casting
simulation finds its application in later part of design process particularly used by
foundries to optimization of casting process by mold filling and solidification simu-
lation.
Topology optimization tends to create design proposal which are hollow with rib
structure. In general the material is added on outer area of design space which
make them complicated to realize the design into castings. In the realization phase,
design engineer tries to answer question on how feasible the new design is terms of
manufacturing. It is followed by consideration of casting process design guidelines
to modify the new component accordingly to pass the cast-ability assessment. In
many cases, material is added during design realization phase which has no struc-
tural significance and out weights the benefits of topology optimization.
The Thesis focus on RLCA and RUCA components which are manufactured by
aluminum castings. However, directional structural stiffness requirements, strength
events, load cases and design space are well known for these components. A Linear
isotropic material behaviour is assumed for optimization and solidification simu-
lation is carried to investigate the casting feasibility and how close the optimized
1
1. Introduction
structure is to the design realization and also a need to reduce number of iteration
between design and foundry in later stages of product development where their is
less flexibility for optimal changes, it more natural to include topology and casting
simulation in early phase of component development process.
1.2 Purpose
The main purpose of this master thesis is to improve casting development process by
identifying relations between topology optimization and casting philosophy and pro-
pose a methodology which encourages the use of topology optimization and casting
simulations tools in early phase of component development process for chassis com-
ponents. The casting constraints origin from studying the process of casting method.
The casting simulation is then used to investigate impacts of added manufacturing
constraints in final designs.
1.3 Limitations
Three dimensional FE-meshed design space with no temperature dependencies, vi-
bration free, time independent, multiple loading and linear elastic isotropic material
within Hook’s range is considered. Local approximation methods are used in sensi-
tivity analysis for reducing the computation time. In order to formulate directional
stiffness structural constraint, finite element problem with linear static behavior
is preferred. Since it seems very difficult to formulate every studied casting re-
quirements into topology optimization problem, only those manufacturing related
constraints facilitated by Optistruct tool are used in this study. NovaFlow&Solid
software is utilized as a casting process simulation tool. Although the software is
able to model both filling and solidification process of the casting, only the solidi-
fication process is simulated. This is mainly due to the fact that filling simulation
parameters are process dependent and differs from one foundry to other. The flow
constrains are difficult to formulate. Influence of gravity is not considered in solidi-
fication and material model for temperature dependent properties like density, heat
transfer coefficients are limited to standard data from Novaflow material library.The
limitations of concept selection methods also applies.
1.4 Method
The Thesis work started with literature study on theory of structural optimization,
aluminum casting process and casting simulation. Later on software training on both
tools NovaFlow&Solid 6.0, and Optistruct were done at Volvo Cars by studying
tutorials and attending workshops. For making an optimized casting component
where the design is close to manufacturing, topology optimization is performed
by implementing different available manufacturing constraints. With the help of
parameter study considering different formulation with various combinations best
setting is finalized. A method for finding the optimum threshold filter value for
2
1. Introduction
efficient extraction of results from density plots file is also proposed. Optimal setting
to import filtered high density elements using function to NovaFlow&Solid where
its re-meshed and boundary condition are setup for solidification simulation. This
is continued by investigation on the solidification defects formed during this phase.
The reasons of defects formation are then interpreted. A criteria for comparing the
cast-ability is defined for evaluating the best design proposal which is tried upon
RUCA. The solidification simulation of current RLCA and proposal from previous
thesis is compared in terms of casting feasibility.
3
2
Theory
This chapter provides the necessary theoretical materials as well as concepts used
in this thesis work. Section 2.1 briefs on basis of structural and topology optimiza-
tion, we focus more on responses and constraints used in optimization setup. In
addition, we classify the manufacturing constraints of interest to include in problem
formulation. Sections 2.2 & 2.3 give a short description of the aluminum casting
process, defects and simulation theory in detail. Eventually, section 2.4 addresses
on Design For Manufacturing (DFM) for casting which captures the best practices
and guidelines used by design engineer in design realization phase.
where ρ is the design variable with assumed material density and penalization power
p. In Optistruct, parameter value is defined with Opti-control card which is always
equal to p − 1. In order to see the effect of penalization it must take values p > 1,
intermediate values of ρ contribute less to stiffness and more to mass of structure.
Hence intermediate values will be eluded in setting up the optimization problem.
4
2. Theory
5
2. Theory
Optistruct uses move limit bounds to reach convergence with minimum number
of FE Analysis which enables large design variables changes in the first few iter-
ations. In sensitivity analysis, second derivative of structural responses from the
FEM analysis with respect to design variable is calculated. Next update of design
variable is obtained by explicitly solving the optimization problem with sensitivity.
Topology uses dual method which solves the optimization problem with Lagrange
multipliers related to active constraints. This method is highly efficient with more
design variable and less number of constraint which is true in the case of topology
optimization.
2.1.1 Responses
Response are predefined variables types used to define objective and constraints in an
optimization setup. There are numerous structural responses from which Optistruct
allows to choose, for our study we will be mainly looking in detail of 4 responses (a)
mass and volume, (b) weighted compliance, (c) compliance, (d) mass and volume
fraction.
Mass and volume are global responses which can be defined for parts of interest
of the structure or whole structure. Mathematically given by equation below:
N
ρe Me0
X
M = M (ρ) = (2.4)
e=1
6
2. Theory
N
ρe Ve0
X
V = V (ρ) = (2.5)
e=1
Where, M e, V e and ρe are mass, volume and normalized design parameter of eth
element respectively and N is total number of elements.
Weighted compliance is also a global responses defined for whole structure used
to consider multiple load steps in a topology optimization. It is the sum of the
compliance of each individual sub-load step. It is a single scalar value for the whole
structure which approximates the structural performance or strain energy stored in
the structure, lower the weighted compliance implies stiffer the structure.
L L
X 1X
CW = W i Ci = Wi (U (ρ) × F (ρ)) (2.6)
i=1 2 i=1
1 1 F (ρ)T F (ρ) 1 1 1
C = U (ρ)T F (ρ) = T
= f = Constant (2.8)
2 2 K(ρ) 2 K(ρ) K(ρ)
7
2. Theory
min
PN
C(ρ) = F (ρ) × U (ρ) = U (ρ)T × K(ρ) × U (ρ) = ρe p ue T Ke ue
e=1
subjected to
(2.10)
V = fv V0 = N
P
e=1 ρe ve
F = KU
0.01 < ρ < 1
e
min
V = N p
P
e=1 ρe ve
subjected to
PN
C(ρ) = e=1 ρe p ue T Ke ue < Cbound (2.11)
F = KU
N odal displacement < U B
0.01 < ρ < 1
e
2.1.3 Constraints
Constraint is a restriction placed on the optimization problem, they are generally
a limit value associated with a response which is expressed as inequality function
of design variable and this condition must always be satisfied for the design to be
valid. All the created responses can be used as a constraint expect the response
associate with objective. There can be multiple constraints for an optimization
problem. When all constraints are satisfied the design is feasible. Optimum design
is one which gives the minimum of objective function satisfying all constraint equa-
tions, violation of any of constraints results in an in-feasible design.
8
2. Theory
In most common cases it is not suggested to have through holes in castings as a flow
9
2. Theory
10
2. Theory
mization problem by using Pattern grouping options which link topology variables
together in such a way that facilitates the formation of desired reinforcement pat-
terns. One-plane, two-plane, three-plane and cyclical symmetry pattern grouping
can be achieved.
11
2. Theory
Permanent mold casting consists of molds and cores made of steel or similar metals.
The advantage of this method is that the mold can be reused. The procedure of
permanent mold casting is fairly straight forward so that the molten aluminum is
poured into the prepared cavity and after the metal has solidified the two mold halves
are pulled away from the formed part [9]. It should be mentioned that depending on
the shape of the cavity and used material, a liquid molten metal is subjected to gas
pressure or vacuum. High-volume production of castings with uniform wall thickness
and limited undercuts is the most tendency of using this casting method. For those
aluminum alloy castings where maximum mechanical properties are required, heat
treatment is used to improve mechanical properties [9]. Transmission case and intake
manifold are good examples of permanent mold aluminum castings in automotive
applications.
Sand casting is the most economical shape casting manufacturing process since there
is no need of using high-cost metallic tooling [10]. Typically it is performed in the air
atmosphere with sand mold at room temperature. Sand casting normally begins with
the creation of a pattern as a replica of the desired shape. It is followed by pressing
the pattern into the fine sand mixture to form a mold cavity. After preparation of
the sand mold, liquid metal is poured through channel(s) into the cavity which relies
on gravity until it gets completely filled with molten metal. The solidification is then
started to transfer the poured metal from liquid to solid phase. Once the casting
part is ready the sand mold should be broken to take the formed cast component
out. This process compared to two other mentioned casting methods is slow and
therefore is limited to have a high rate of production. However, casting quality of
the sand cast parts is highly determined by foundry techniques [9]. Since the sand
mold is used, the rough surface texture cast part with low dimensional accuracy has
been produced which usually needs machining and finishing operations to ensure
accuracy and surface quality of the final product. Moreover, components with thin
ribs and walls have a limited capability to be manufactured in sand casting. In
automotive industry, rear lower and upper control arm, RLCA and RUCA, in rear
wheel suspension and steering knuckle arm in vehicle suspension system are examples
of aluminum sand castings. The flowchart below shows the step-wise typical sand
casting process:
12
2. Theory
Sand in sand casting method can be either wet or dry. Green sand casting refers
casting with wet sand that contains water or oil and organic bonding compounds
such as clay. ’Green sand’ statement origins from the fact that the sand mold is
uncured even when it is filled by molten metal. On the other hand, in dry sand
casting the sand mold is baked at a specific temperature to make the mold stronger
as ensures accurate size for casting products [11].
13
2. Theory
Sand mold creation consists of packing sand around the pattern and holding the
sand around the pattern by using box called flask which is removed after the metal
has solidified.
In case of having hollow components or producing internal cavities and re-entrants
as a result of casting, sand cores are used to create negative spaces in the final piece.
Cores are made from sand with special binders and different making types exist [12].
They are placed into the mold after building the mold cavity and in addition to the
desired shape of cores, there are extensions called core prints for correct positioning
of the cores into the sand mold. In some sand castings due to the disability of
making the exact desired cavity forming by pattern cores are used, an example can
be a tight corner which cannot be perfectly shaped by sand. Normally sand cores
are disposable units which will be destroyed to be able to get it out of the casting
product.
As mentioned before, pattern is the replica of the part to be cast. It is used to form a
cavity in the molding material where the liquid metal is poured later [13]. The sand
cast patterns can be made out of different types of materials. Wooden, different
types of plastics or even metals patterns are used in different casting conditions. In
order to have a good pattern material selection, various parameters such as size and
complexity of the shape, casting method, number of required casting products to
be manufactured and characteristics of casting should be taken into considerations
[14].
It should be noted that, patterns are commonly made slightly larger than the antic-
ipated casting part to compensate shrinkage in solidification phase which occurs for
nearly every metal alloy and also to compensate thermal contraction of the metal
during cooling to room temperature. ’Shrink rule’ is the phrase used for estimating
amount of shrinkage in each metal according to molding process. For instance, the
shrink rule for aluminum sand casting with hand packed sand is 1/8 inches [10].
The liquid metal delivery system in casting is called gating system that must be
considered in the casting process design. Typically a gating system includes a pour-
ing basin, down sprue, runner and ingates. Based on the part to be cast, gating
system is designed and built. The aim of using gating system is to fully feed the
cavity from different regions before solidification takes place which has significant
influences on the casting quality. With pouring molten metal into pouring basin of
gating system the filling process is begun.
14
2. Theory
2.2.5 Solidification
Solidification is changing the material phase from liquid to solid in a casting process.
When the temperature drops, solidification starts. Pure poured liquid metal solid-
ifies at a freezing point which is a constant temperature whereas, solidification for
alloys does not occur in an exact temperature point. Depending on the composition
of the alloy solidification takes place over a temperature range. This cooling down
in the temperature highly affects mechanical properties and the geometric relation
between volume and surface area of the casting product. Besides, most of the casting
15
2. Theory
defects are solidification related defects which lead to investigations on the thermal
effectiveness of the mold design. In thermal effectiveness of the sand casting mold
two important issues are investigated. First, solidification progress must take place
from the mold walls upward and towards the liquid feeders to have a good quality
in casting products. This progression in the solidification phase is called directional
solidification [15]. Secondly, the potential of providing desirable micro-structural
constituents, shape and grain size by the mold design [10, 16]. Solidification plays
an important role on the quality of the casting product, also time taken for produc-
tion of the casting resultant shows the importance of solidification stage in casting
industry [17].
16
2. Theory
filling which cause turbulence and thereby decrease the risk of having gas
entrapped in the casting [10].
• Gas porosity
This defect is not specifically solidification related defect and can be counted
as both, filling and solidification related, defect. During pouring or solidifying
the metal absorbs gasses which can not be rejected completely from the mold.
Consequently, another void type is created which is called gas porosity [10].
The absorbed gasses may be either originated from chemical reactions between
17
2. Theory
molten metal and sand mold materials or presented inside sand and mold
cavity. Some of the absorbed gasses are rejected when liquid metal solidifies
but the rest remain in the bubble shape. Gas porosity defects tend to form
in long freezing range alloys casting such as aluminum based alloys and since
aluminum has a high gas solubility, gas porosity in aluminum is always a
concern [10].
• Hot tears and cracks, residual stresses and distortion
During solidification process when thin liquid film contracts, the hot tears
and hot cracks can be formed. This is as a result of stresses in temperature
reduction. On the other side, residual stresses shows the state of the casting
when it has reached room temperature. All these phenomena are related to
the strains happening due to thermal expansion, volume variations and solid
phase transformations. Moreover, as long as there is no plastic deformation
in casting, no distortion will form. But when the thermal stress is higher
than yield strength of the cast metal, plastic deformation and nonuniform
contraction takes place, therefore distortion in the casting can be found [10].
• Macro-segregation
Macro-segregation defect can be controlled by taking care of local solidification
time and the rate of cooling of the casting. This natural phenomenon affects
final mechanical properties of the casting [10].
18
2. Theory
the effectiveness of the selected orientation of the casting since different orientation
selection bring different solidification defects in the cast product. These are only
examples of casting modeling results usages. In below the most important benefits
of using cast simulation software in the early phase of the production are itemized
[18]:
• Standardization of the pouring method. Testing and optimizing the variant of
feeding.
• Cast-ability measurement of various designs, particularly for structural and
topology optimized designs.
• Investigation and prediction of casting possible defects, for instance, solids
shrinkage percentage.
• Analysis of casting temperature changes to reduce energy consumption in the
casting processes, especially for casting parts in mass productions.
• Optimization of the complete casting process to provide good casting quality
at the lowest cost.
19
2. Theory
20
2. Theory
• Shrinkage It has been explained in section 2.2.6.2 that during solidification al-
most every metal alloy shrinks and the percentage of shrinkage for aluminum
alloys is approximately between 3% to 7% of its volume. For prediction of
the shrinkage defects the cast simulation software needs to identify the liq-
uid pools surrounded with solidified regions through solidification process and
those regions where there is high risk of having shrinkage voids due to material
contraction. The studies revealed this fact that long freezing range alloys like
aluminum alloys tend to have too many small shrinkage voids over the casting
part [10]. The improvement in casting shrinkage results can be done by modi-
fying gating systems, risers, chills and all other filling parameters. If the filling
simulation is not modeled, modification of the casting design based on temper-
ature contours will be a good way of improving casting volume shrinkage after
solidification simulation. It should be mentioned that all of the design-wise
modifications are based on creating directional solidification in the casting
part. These adjustments are introduced in details in section 2.4.
• Hot spot The liquid trapped pools inside already solidified regions are hot
spots’ results in the solidification simulation. The software predicts the hot
spots result when the liquid phase reaches zero percent. After that, the number
and size of the hot spots do not change by temperature reduction of the casting.
The hot spots result represents directions of solidification but it should be
noted that a number of hot spots are not as the same as a number of shrinkage
and it proofs the fact that not all of the hot spots form shrinkage defects and
gravity also has an impact on shrinkage formation. The location of formed
hot spots are also predicted by the software.
• Solid phase The solid phase results represent the progression of the solidifica-
tion visually. The filled volume value in percentage is calculated based on the
solid phase. The areas without material after solidification shows the effect
of gravity shrinkage on the casting; However the gravity influence calculation
can be turned off when only solidification is simulated to prevent calculating
massive gravity shrinkage on top of the casting. In fact, by expelling gravity
influence from solidification calculation the result will be orientation indepen-
dent. Moreover, areas which become solid at last can be indicated from the
solid phase results. These last solidified areas are those where thermal modu-
lus can not be predicted and the software shows them transparent in a thermal
modulus optical result. In design point of view, thinner parts of the casting
component become solidify sooner than thicker parts and this differentiation
in solidification progression forms solidification defects. A general guidance
for design realization of casting components which avoids defects formation
will be discussed in section 2.4.
21
2. Theory
could bring weak areas in the casting products. Accordingly, modification on the
casting’s design is performed for eliminating defects causing because of the shape
of the casting. Although the aim of this master thesis is to provide a design out
of the topology optimization that does not encounter with manufacture disability,
sometimes due to limited capability of the optimizer or after understanding of the
solidification process with cast simulation results, changes are required. These mod-
ification will be applied to design in design realization phase to either omit minor
design related defects or prepare the design for manufacturing. The general guide-
line below can be applied to all the casting designs, some of them are fulfilled after
implementing manufacturing constraints into optimization, though.
22
2. Theory
23
2. Theory
possibility of making pattern and cores easier, the orientation of the mold cavity in
casting process is also specified.
24
3
Method
25
3. Method
26
3. Method
at the hard points and critical regions. (b) Improve the design of existing compo-
nents to their optimal where goal is trying to reduce weight of component satisfying
the requirement or to increase stiffness of component with the same weight where
optimization is difficult due to less flexibility and in some cases to find a completely
new solution, where maximum mass or volume fraction is an input to your formula-
tion of optimization problem. Formulation of an optimization can be done in many
ways and it is very important to choose the best formulation in order to understand
objective, constraints and direction of getting the optimal solution. The flow chart
below shows all possible combination of problem.
27
3. Method
density which is based on the strain energy that means the higher strain energy in
an element the higher normalized density and closer to 1.
The important output files generated from topology optimization solver, Optistruct,
are .hist, .mvw, .sh files and .OSSmooth file. Shape file (.sh) contains the material
density values, void size parameters and void orientation angle for each element and
its IDs in the analysis for the final iteration.
The density plots can be opened in HyperView for better visualization of the colour
density counter plots as shown in Figure 4.3 and 5.5. With default colour settings
red colour indicates load bearing elements with a density of 1. Blue refers to less
important elements with a density of 0.01. The density contour plot shows the struc-
turally important regions in design volume. However, there is a challenge involved
on understanding and interpreting intermediate density elements such as elements
with a density of 0.3 or 0.7.
Interpreting the topology optimization result has always been a major challenge for
integrating optimization with cast simulation tools and design realization phase.
ISO value is a parameter in Optistruct which aids in interpreting designs and un-
derstanding load paths. The default settings of ISO value is 0.505.
ISO threshold value is manually varied between 0 and 1 in order to decide a solution.
This threshold value is chosen by the designer where he feels confident. By choos-
ing this value the designer is excluding all elements with a density value below the
threshold, which implies that their respective contribution to the overall compliance
of the structure will also be neglected. This is compensated by assigning all elements
above the selected ISO threshold with the new standardized density value of 1, which
also implies that the compliance of the structure is increased. This demand to have
a method to select an exact ISO value can stabilize these effects. This value plays
an important role as it will be used in OSSmooth function to extract and import
the final design geometry from topology results into casting simulation software as
.STL file exchange or .IGES format is used to generate and export surfaces into
CAD tools for design realization.
In this thesis an automatic process that uses an excel template which calculates the
ISO value for every topology result is proposed. The following steps and formulation
are used to get the optimum ISO threshold value.
• Open .sh file in Microsoft Excel program using text import wizard choose ‘fixed
with’ radio button to import the element id and its corresponding normalized
element density into columns. Sort the data with increasing value of density.
• Paste the sorted data in Sheet1 of template with element ids and densities in
A, B columns, respectively. The template itself calculates the threshold value
which can be used for further extraction.
The Sheet2 of the template contains the element id in first column and its corre-
sponding element volume in the second column. This can be extracted from .fem
file which gives coordinates of each nodes and element ids with corresponding node
ids for tetra4 solid element. With this data volume of tetrahedron can be computed
using distance formula.
Sheet3 of the template file contains all the calculation details. Vlookup excel func-
tion is used to arrange the data in sheet1 and sheet2 such that all elements which
form the design and Non-design space are arranged with element id and their cor-
28
3. Method
responding volume sorted with normalized densities from largest to smallest. The
sum of product of volume and its corresponding density is then computed. This
value is searched with the cumulative sum of volume of elements in order by row.
The ISO value of the matching row is the threshold. The method is validated by
calculating mass with the sum of volume of all elements above the threshold and it
matches closely with result of optimized mass.
Figure 3.3: Template to find the threshold ISO value for OSSmooth extraction.
The whole process is automated and can be easily implemented for another compo-
nent, which is very crucial for integrating topology optimization.
29
3. Method
30
3. Method
31
4
Topology optimization for Rear
Upper Control Arm
Linear Tetra4 elements are used with average element size of 1.5mm considering
mesh quality criteria as per mesh guideline for solid chassis components. Hard
point 1 assembles with the sub-frame and hard point 7 to the knuckle. Two nodes
are created at pt1 and pt7 hard point coordinate as shown in Figure 4.1. These
two nodes constitute reference nodes in connecting couplings. The couplings at pt1
and pt7 are for each reference node grouped is created by node based surface set
solely containing the nodes on the surfaces with a normal perpendicular to the bush-
ing axis. A local rectangular coordinate system is defined with its Z axis passing
through between pt7 and pt1, oriented in the direction towards pt1. Furthermore,
three defined hard point nodes are transformed into this local system.
32
4. Topology optimization for Rear Upper Control Arm
The FE model of actual RUCA component is not available as a given input. Hence,
the information about the compliance of the actual structure is not clear. It is best to
consider minimize volume fraction with the given displacement constraint. Output
of this run gives us the optimum value of compliance and mass as it is a well-known
fact that optimal solution is a trade-off between compliance and mass. It is also
observed that formulation with minimized compliance with displacement constraint
does not work as the optimization solver will stop soon as it fulfills the displacement
requirement, “optimal design is probably not reached if the requirement is too easy
to fulfill, since the optimization stops too early”. This gives a hint of adding a
volume fraction constraint for the next problem formulation.
The following conclusions can be derived out of this study:
• Minimize volume fraction as objective with displacement constraint is the best
formulation as it gives the optimal compliance and minimum mass.
• Compliance is a direct measure of displacement. It is observed that for unit
load case the compliance is inverse of 12 the displacement that is C = ( 12 0.06)−1 ∼
=
30 (scaled values) which follows the definition of compliance in theory.
• In order to use minimize compliance as an objective, displacement requirement
is not enough. It is recommended to use volume fraction constraint which in
turn satisfies the displacement requirement. It can also be considered that if
compliance requirement is fulfilled, nodal displacement requirement is taken
care automatically.
33
4. Topology optimization for Rear Upper Control Arm
34
4. Topology optimization for Rear Upper Control Arm
All the concepts generated with different manufacturing constraints have the same
structural performance with global compliance response 30 (scaled values). It can
also be seen that the problem with no manufacturing constraint is heaver than
the problem formulation with split draw and split draw with no-hole constraints
with same compliance value and nodal displacement. This is interesting observation
which confirms that by applying manufacturing constraints, the Optistrcut solver
takes different formulation which can lead to better optimized solution. One more
observation is when we combine a split draw with no-hole with symmetric con-
straints, instead of forming one single shear plane on auto mid plane as in case
of split draw with no-hole constraint, the optimization converges with two parallel
shear planes. However, when no manufacturing constraint is applied it gives a hollow
structure with ribs on outer surface of design space as shown in Figure 4.3. The for-
mulation with split draw no-hole and stamped does not converge to optimal. It can
be seen that all formulations in RUCA resulted in less weight than the previously
optimized and redesigned component with single manufacturing constraint. Now, it
is required to continue with 4 best concepts to next step for casting simulation in
order to understand the cast-ability of these designs.
The cast simulation results for solidification of different topology optimized RUCA
designs are presented in Figure 4.5 to 4.7. It needs to be mentioned that all the
simulations are done without performing any CAD realization phase.
35
4. Topology optimization for Rear Upper Control Arm
Figure 4.5: Volume shrinkage indication of RUCA showing defective regions for
different casting design. Optimization with (a)No manufacturing constraint
(b)Split draw (c)Split draw No-hole (d)Split draw No-hole Symmetry.
Figure 4.6: Hot spots of RUCA showing directional solidification for different
casting design. Optimization with (a)No manufacturing constraint (b)Split draw
(c)Split draw No-hole (d)Split draw No-hole Symmetry.
36
4. Topology optimization for Rear Upper Control Arm
After optimizing RUCA with different manufacturing constraints, it has been de-
cided to evaluate the different topology optimization designs by considering different
requirements. This concept evaluation is mainly proposed when more than one opti-
mization result exists and a decision regarding the final selected concept with respect
to cast-ability needs to be made. In this evaluation, categorized requirements are
weighted based on the results of both topology optimization and cast simulation
and that concept wins which has collected higher weight summation. The evalu-
ation matrix with considered requirements for four final RUCA concepts is shown
below:
37
4. Topology optimization for Rear Upper Control Arm
Figure 4.8: Concept evaluation matrix for four topology optimized RUCA
designs with different manufacturing constraints.
Note that an objective function for all four concepts coming from topology opti-
mization is minimizing mass. The Figure 4.8 shows how different topology opti-
mized concepts have been weighted based on structural and cast-ability criteria.
The ranking method gives higher weight to the best concept in each aspect.
Out of four different concepts for RUCA, concept 3 and concept 4 have received
higher total weight. It means that these two concepts are closer to the casting
process and they are the best-optimized concepts by considering cast-ability re-
quirements. When it comes to comparing the two final concepts, the cast-ability
evaluation can be taken into account. In other words, in our evaluation system the
concept which fulfills the cast-ability requirements more than the other would be
a better choice to continue with. This can be a good technique to differ the final
concepts.
After evaluating concepts with the suggested matrix and having the result of cast
simulation for the selected concept(s), design modifications in development phase
can be accomplished to make the design ready for manufacturing phase.
38
5
Topology optimization for Rear
Lower Control Arm
Rear Lower Control Arm is studied which is an aluminum cast component bolted
to sub-frame, knuckle, damper and leaf spring attachments. RLCA is subjected to
many forces and moments through adjacent components.
Hard point 3 and 4 assemble to the sub-frame, whereas pt6 and pt18 assemble to the
knuckle. The pt56 engages the damper and the pt68 attaches to the spring element
(Leaf or Air spring) as shown in Figure 5.1. Second order tetra4 solid elements are
used with average FE grid element size of 6mm and mesh quality criteria to be used
for the RLCA are as per the standard mesh guideline for solid chassis components.
At each hard point coordinate, a node is created for couplings connection where the
loads and boundary condition are applied. The coupling uses RBE2 element for pt3,
pt4, pt6 and pt18 for each hard point grouped into a node based surface set solely
containing the nodes on the surfaces with a normal perpendicular to the coupling
axis. The damper interface couplings are a part of the clevis bracket which is mod-
elled rigid coupling for simplicity and also the pre-tension of bolts are neglected.
The c-beam element with steel material properties is used which connects the series
of coupling and forms the spring connection at hard point pt56.
39
5. Topology optimization for Rear Lower Control Arm
Load application and stiffness requirement The six leading load cases channels
are selected from load extraction from chassis rig cycle test and different dynamic
loading events.
• Torsional load is unit load of 1000N applied at pt18 along positive x direction.
• Damping load is unit load of 1000N applied at pt56 along negative z direction.
• Spring load is unit load of 1000N applied at pt68 along negative z direction.
• Drive over curb – DOC is set of load and moments applied at all hard points.
• Rearwards drive over curb – ROC is set of load and moments applied at all
hard points.
• Brake in pothole – BIP is set of load and moments applied at all hard points.
The structure stiffness when loading in negative Z direction in three different points,
are of interest named as Torsional, Leaf spring attachment and Damper attachment
stiffness. Displacements are measured which must be less than 1.0989, 0.33, 0.033
(scaled values) for unit loads at respective loading points are pt18, pt68 and pt56.
Table 5.1: Static linear analysis of current design of RLCA, compliance in FE units
and displacements in millimeter (scaled values).
Following conclusions can be derived out from the FE analysis of current RLCA
component:
• The displacement requirement at hard points of interests is satisfied for tor-
sional, damping and spring loads cases.
• The load case DOC is dominating load case and load case BIP is the second
dominating load case.
• The requirement displacement constraints will not be enough to define an
effective optimization problem.
• The original component is 9% of current design volume.
40
5. Topology optimization for Rear Lower Control Arm
In the case of RLCA, all the load cases and its corresponding compliance and dis-
placements are available, since the design volume of this component is given from
packaging space analysis. Therefore, the topology optimization problem can be
formulated in many different ways as listed below in Table 5.2 with the results of
optimization.
The dotted blue line shows the reference compliance performance of the current
component with their respective load cases and all the curves which are on this line
41
5. Topology optimization for Rear Lower Control Arm
or below this line are good problem formulations. The following conclusions can be
derived out from this study:
• In problems with minimize volume fraction as objective all formulations except
constraint with 3 displacements result in stiffer solutions than the reference
component. Hence, the best formulation can be chosen considering the criteria
of minimum volume fraction. The formulation with 6 reference compliance
constraints which give 7.285 percent of volume fraction is chosen as the best
formulation.
• It is also observed that problem with minimize compliance with volume frac-
tion constraint obtained from previous runs reference component results quali-
fies individual compliance requirements. Which indicates that in problem with
multiple load cases and displacement requirements this method can be used to
convert the complex formulation into single constraint and objective problem.
• Minimize volume with 36 displacement constraints captures the classical trade
off optimization problem of compliance and weight.
• Compliance decreases with increase in mass fraction constraint.
Figure 5.3: Trade off between compliance and volume fraction for different
formulation.
The trade-off between compliance and volume fraction is shown in the figure above.
The case with minimum volume fraction and weighted compliance gives more op-
timal solution. The case with problem formulation of minimizing volume fraction
with 6 reference compliance as constraints gives the best solution.
42
5. Topology optimization for Rear Lower Control Arm
between each member generated, (e) single draw direction, (f) split draw direction,
(g) split draw with no-hole with uniform thickness.
43
5. Topology optimization for Rear Lower Control Arm
The results of the optimization have been plotted in Figure 5.5. It is seen that
formulation with minimum member size of 73 millimeter looks lighter compared to
the original component and has more defined ribs. The result is also closer to casting
in terms of production compared to formulation with no manufacturing constraints
and split draw that results in many thin ribs with hollow design. Hence, it is not
considered for further casting simulation.
In section 4.3 and 5.3 solidification simulation results are presented. The general
simulation setup has been already mentioned in 3.4.1. It should be explained that as
long as the filling simulation is not conducted in our simulation there is no need of
setting gating systems and melt pouring type. Plus, due to lack of filling simulation
it is recommended to cancel out the gravity influences on the shrinkage calculation
to get a result without interfering of filling parameters. By having this setup the
solidification results are orient-independent and investigations on the orientation of
casting are therefore eliminated.
For interpreting the casting feasibility of the RLCA which its design is way more
complicated than RUCA, two different designs are cast simulated. The first design is
result proposed by previous master thesis where topology optimization of same com-
ponent, RLCA, has been studied without considering any manufacturing constraints
[27]. Thus, the final proposed design (I) has passed manual design realization phase
to ensure the mold-ability of the component. Due to interests in checking the cast
simulation results on this component and investigating the effectiveness of the de-
sign realization phase, solidification of casting setup has been simulated. Finally, the
second design (II) is a proposed topology optimized RLCA component with imple-
menting manufacturing constraints during the optimization phase by current study.
Note that the geometry of design (II) has been imported directly from optimizer to
cast simulation software without passing any design realization phase. The casting
design proposals and their solidification simulation results are shown below.
44
5. Topology optimization for Rear Lower Control Arm
Figure 5.7: Volume shrinkage indication of RLCA showing defective regions for
different casting design.
45
5. Topology optimization for Rear Lower Control Arm
46
5. Topology optimization for Rear Lower Control Arm
in 14 different directions. Each direction creates an isolated liquid pool and gives
raise to porosity which reduces the quality of casting. The hot spots with minor
differences are predicted at the same locations which reconfirms the directional so-
lidification inspection. Thickness variation, sharp edges and the junction of walls
are the main causation of having many solidification directions, these have been
figured out after checking geometric dimensions and thermal modulus. It should be
again noted that filling simulation is not considered. The shrinkage prediction of
design proposal (I) reveals the fact that even after considering the standard design
guidelines for interpreting the optimization results by the design engineer, redesign-
ing to remove shrinkage defects in the rib intersections is needed at regions Sb 5, Sb 6,
Sb 7 and Sb 8 predicted from simulation as feeders at these regions are not practical.
It can also be suggested to have ingates for filling simulation in regions labeled as
Sb 2 and Sb 3 since these are thickest regions as seen from thermal modulus and also
shrinkage volume is high which can be compensated during filling. It is also ob-
served from hot spots prediction result that porosity is high and distributed on load
carrying ribs which results in lower quality casting.
From casting simulation of design proposal (II) it can be seen that shrinkage is
concentrated on the diagonal rib and lower right region. Thermal modulus suggests
the need of having a gating system at these regions to compensate the large volume
shrinkage happening. Hot spots on thin ribs can be neglected which can be removed
during later design realization phase. It can be concluded that solidification results
are valuable inputs for design engineer to perform CAD modification in early re-
alization phase. It also gives the design engineer basic clues of gating and feeding
points.
47
6
Discussion
In this work we have presented an approach to handle the casting component devel-
opment process using topology optimization and casting simulation. The method
is based on first identifying the best optimization formulation for the given loads
and design space and then implementing casting constraints such as draw direction,
minimum member size and penalization factor to find the optimized structure.
Firstly it was intuitively believed that the topology optimization without manu-
facturing constraint would give lighter structure compared to optimization with
manufacturing constraint which was true in case of RLCA. However, It was discov-
ered that topology optimization of RUCA resulted in lighter model with split draw
and no-hole constraints which gives a single shear plane at mid surface, when com-
pared to hollow ribbed structure with no manufacturing constraint while achieving
same structural performance. It was realized that all the optimization for RUCA
with manufacturing constraints was under the targeted weight. Whereas, in case of
RLCA optimization with min member size below 73 and discreet parameter below 1
with no manufacturing constraints resulted in lighter weighted structure compared
with the existing component.
It was also noted that there was large number of intermediate density elements in
these solution and when the results were extracted to casting simulation tools these
designs were heavier than the optimized weight from topology.
The work also provides a generic template to find out the threshold ISO value for
accurate extraction of density plots to CAD realization.
Ideal casting simulation setup for evaluating the topology optimized design is pro-
posed and its benefits are shown on RUCA. The solidification simulation was done
on the current RLCA from where the number of gates, feeders and their position was
identified based on shrinkage and hotspots regions which was followed by filling and
solidification simulation. It was observed that shrinkage was reduced by 98% and
hotspots decreased from 15 to 11. The same method was implemented to optimized
RLCA which showed 97% reduction in shrinkage and number of hotspots regions
decreased from 23 to 17.
Moreover, the casting simulation of optimized RLCA (II), directly extracted from
topology optimization without considering CAD realization, was also done. From
the result of casting simulations it is observed that some of the ribs do not fulfill the
minimum thickness requirement and they are hard or even sometimes impossible to
be manufactured by sand casting method. It is also seen that a few of hot spots
are formed only due to lack of gradual thickness variation, which are removed after
design realization phase. When mold-ability of RLCA component is concerned it
can be generally said that creation of cavity mold for this design is not an easy
48
6. Discussion
work. However, comments from cast experts and production group could avoid such
difficulties.
49
7
Future work
50
Bibliography
51
Bibliography
52