Developing Topology Optimization with Support Structure and Overhang

Constraints of Additive Manufacturing

Project report submitted

In fulfillment of the requirement for the degree of

Master of Technology

IIT KHARAGPUR

By

Manoj Nehra (16ME31004)

DEPARTMENT OF MECHANICAL ENGINEERING

INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR

2020-21
 CERTIFICATE

This is to certify that this report entitled ‘Developing Topology Optimization

with Support Structure and Overhang Constraints of Additive Manufacturing’
submitted by Manoj Nehra, to the Indian Institute of Technology, Kharagpur, is
a record of bonafide research work carried out under my supervision and is
worthy of consideration for M.Tech Project, and that this work has not been
submitted elsewhere for a degree.

Supervisor Name: Prof. P. Saha

Signature :

Date:
HIGHLIGHTS
• Conveys AM support volume/mass constraint in TO
• Combines support structure sensitivity with performance sensitivity
• Put forward topological sensitivity for support structures in optimization
algorithm

Abstract
There has been significant growth in the domain of Additive Manufacturing and
Topology Optimization and interest in integrating both is on rise. But, Topological
Optimization certainly results in complex designs that contradicts the sole purpose of
optimization from a cost perspective if Additively manufactured. For example, end
results of Topology Optimization may need a large %v of material in the form of
support structures before we can manufacture it using AM, which in turn increases
the fabrication and post-manufacturing clean-up costs.
In this study, we put forward a Topology Optimization framework with end results
requiring significantly reduced volume of support structures material. We will
introduce the concept of ‘support structure topological sensitivity’ integrated with
performance sensitivity to result in a Topology Optimization methodology that
maximizes performance (compliance measurement is taken as the performance
parameter), subjected to Additive Manufacturing support structure constraints. To
demonstrate the efficiency and robustness of this approach some examples are
shown with ANSYS simulations.

Introduction
The human – machine relationship is an intricate balance stretching eons. Our need
and our hunger for a better life has fueled the development and growth of our
civilization right from the first wheel to our interplanetary missions. And we have
ensured that our machines evolve along with us. We began with horse carts and now
have metal boxes moving at hundreds of miles an hour. We began with mud houses
and now have created iron towers stretching into the sky for us to live in. All this
advancement needed hordes of advanced technologies to support it. These
technologies in turn need a battery of complex machinery to run the system. And as
our machines began getting complex making them became all the more difficult. This
led to the research into complex manufacturing methods to satisfy our technological
need. As the methods and their end products became difficult to build their cost
sky-rocketed and everyone joined the race to reduce these high costs. In today’s
competitive world reducing costs is the most important aspect of any business as
reduced costs are synonymous to increase profits.

As our material usage went up we began searching for means to reduce this without
impacting our performance and efficiency. If we could build a bridge traditionally
costing 100 Crores in 60 Crores by optimizing the design so that it meets all the
safety and structural conditions and in turn reduced the amount of raw materials we
use, then it’s a huge margin for cost reduction. This problem and possible solution
gave birth to the idea of topology optimization and forms its basis.

Topology optimization is an advanced mechanism that has been used by engineers

for many years to make their designs efficient thus saving and reducing material
usage and finally reducing costs for manufacturing. For example we can minimize
the amount of used material and the strain energy of structures while maintaining
their mechanical strength by studying the stress distribution in the structure and
removing excess material from the sections where the body is not under stress. This
not just reduces the material quantity but ultimately brings down the cost. It can be
formally defined as a calculation method which optimizes the distribution of mass
within a defined body, by fulfilling given constraints (subjected loads, supports,
reaction forces, torsion etc.) experimentally or theoretically established and in turn
minimizing the cost function. For such an optimization procedure, the three main
elements are design variables, the cost function and the constraints.

Topology optimization, as a structural or operational design method, has seen

tremendous growth in the past few decades. Topology optimization is being preferred
over size and shape optimization, as a freeform material distribution scheme as it
enables the creation, coalescence and distribution of the interior solids and voids
during the structural evolution and thus providing us with a much larger design space
to be explored, and enhanced structural performance can be expected compared
with size and shape optimization.
2.1 Impact of using Topology Optimization
The major reason for us doing the whole exercise of topology optimization is to
reduce costs. Let us explore to what extent this can be done. We can write costs as
–

The 3 factors which we can reduce to ultimately reduce costs are -

a. Amount of material used b. Cost of Material c. Production Cost

Of these 3 we shall explore ‘a’ first. Production costs depend on the machining
process that we use. Topologically optimized parts are a little complex than their
normal counterpart and have less machinability. Therefore it is seen that additive
manufacturing is a better alternative to manufacture optimized parts in comparison to
traditional manufacturing processes. We will explore this in the later section and also
cover production cost while we study the various manufacturing method types used
in the additive manufacturing process.

2.2 Impact of Topology Optimization on Amount of Material Used

Case Studies
The impact on material usage is immense and can be gauged from the following
case studies references –

1) Altair Engineering, in its official report, presented that they redesigned an

engine-mount bracket for Volkswagen AG, Wolfsburg, Germany, that reduced
part mass by 23% and produced annual material-cost savings of $35,200.
Volkswagen used Altair's OptiStruct software to build the new design which
was found to have the same strength and stiffness as the old design.
However, the weight had been reduced to 730 g from 950 g. The cost savings
were based on production of 200,000 units per annum at a total decrease in
weight of 44,000 kg and savings in the material cost of $0.80/kg.
2) AMAERO Engineering used their proprietary software to topologically optimize
and redesigned an actuator component which showed 58% reduction in mass
along with 51% reduction in max stress, 52% increase in stiffness and a 200%
increase in buckling load.
These 2 case studies show the huge impact that topology optimization has on
material usage and part properties.

2.3 Topology Optimization Process

Topology optimization is carried out on CAD (computer-aided design) software.
There are licensed products available by reputed companies like Altair, Siemens,
Ansys etc. The mechanical specific software modules of the above-mentioned
software are instrumental in getting this done. The process that is generally followed
is –

1) Model construction –

The primary 3-D model is first built with all the boundary conditions and loads
clearly incorporated. This model will be our basis for the whole topology
optimization process and we’ll be reducing this to the optimized form

2) Internal stress mapping –

The whole part is mapped with a locational grid and the internal stress distribution
is studied. This step shows us the sections of the part where the stress is
negligible and hence the removal of material can be carried out accordingly

3) Trimming of material –

This step involves repetitive iterations in which the part is carved out layer by
layer and the desired optimized part is achieved. We can define various process
parameters like number of iterations, what feature to optimize (mass, volume
etc.), percentage of feature to retain and many more

4) Post-processing –

The optimization produces surfaces that are not smooth and are hard to machine.
Post optimization involves providing finishing touches to the generated model
without spoiling the optimization output.

These are the four basic buckets which constitute the topology optimization process
for a part. Though high-end procedures have come up to optimize complex
structures like car chassis, engine parts, rotor mechanism etc. the basic framework
remains the same.
Proposed Approach
Out of all the advancements in Topology Optimization and Additive Manufacturing
there is still a lack of a robust framework to integrate both and in this research we
have tried to suggest one possible framework that can be used specifically for
minimizing the material required in support structures.
We will first start with a general minimizing compliance problem as stated by the
Equation (1)
Minimize J
Ω⊂Ω0
given, (1)
Ω <= Vf|Ω0|
Kd = f
In Equation (1), Ω0 is the initial volume of the object, Ω is the volume to be computed
as in Topology, and Vf is the required volume fraction; f is the external force vector, K
is the stiffness matrix, d is the displacement vector and J = fTd is the compliance that
must be minimized.

3.1 Topology Optimization Example in ANSYS without AM constraints

To understand more about Topology Optimization process flow we performed a
numerical simulation in ANSYS.
 Fig1: CAD design of the object
Consider this 3-hole bracket with two holes constrained with cylindrical support and a
load is applied on the 3rd as 50N in x-direction and 100N in Z-direction. Material of
the object is steel with density of 7850kg/m^3, Tensile yield strength of 2.5E+08 and
Young’s modulus equal to 2.E+011.
Following are the results for 50%V retained and 75%V retained

Fig 2: Topology Optimized with 50%V retained

 Fig 3: Topology Optimized with 75%V retained

3.2 PareTo Level-set Method

There are a lot of Topology Optimization methodologies available each with its own
pros and cons. The objective is to find Pareto-optimal topologies. In the present
context, a topology Ω is ‘Pareto-optimal’ if no other topology Ω′ exists with smaller
compliance and identical volume. In this study, we will be using the level-set based
Pareto Topology Optimization (PareTO) method because of the following benefits:
1. In level-set methods, the boundary of the design is clearly defined using level
set function in equation(2) at all times, making it easier to impose support
structure constraints,
2. It works with the topological sensitivity analysis applied to various constraints
and performance measures aka Compliance and hence can be used to
accommodate support volume sensitivity.
Equation Formulation:
The pseudo-time t is introduced into the Lipschitz continuous level set function
Ф(x) , which is defined over a reference domain D⊂Rd (d 2 or 3) including all
admissible shapes Ω ( Ω ⊂ D) . For d=2, the scheme can be defined as
follows:
Ф(x,t) < 0 ⇔ ∀ x ∈ D \ Ω (void region)
 Ф(x,t) = 0 ⇔ ∀ x ∈ ∂Ω ∩ D (boundary region)
Ф(x,t) > 0 ⇔ ∀ x ∈ Ω \ ∂Ω (solid region)
(2)

3.3 Limitations of Overhang Constraints

In this section we will first discuss how support structures are generated
algorithmically so as to understand how support constraints can be imposed in the
optimization algorithm. In Additive Manufacturing supports are generated based on
the concept of overhang

As we can see from the definition that it is theoretically possible to eliminate all
support structures by removing all overhang surfaces. But this is not considered as
an optimal strategy in Topology Optimization because of following reasons:
1. Elimination of all Overhang Surfaces:
a. In case of 2-D problems it is very much possible to remove all the
overhang surfaces but in case of general 3-D scenarios it is nearly
impossible to eliminate all the overhang surfaces as it might lead into
design distortion.As was also suggested in a study by V. H.
Coverstone-Carroll, “… there will probably be cases where it is not required
for all supports to be removed and so the user should be able to have partial
control on the strength of the penalty function.”.

2. Support volume penalty and Overhang Constraint

a. One problem with such penalization is that if two overhang surfaces
have equal angle then those will be penalized equally which is not
expected because based on the surface area they might have large
differences in the support volume required so to avoid such scenarios a
direct constraint is required on support volume.
3. Penalty on Overhanging Surfaces
a. If we consider an object with two surfaces as shown in fig 4, there is
one overhanging surface and one opposing surface. To reduce support
structure volume, both of these surfaces should be penalized, for
example, by moving them closer to each other as illustrated. By
penalizing the overhanging surface, only half the problem is addressed.
 b. These limitations suggest that we must seek an alternate, and
fundamentally different method to impose constraints on support
structures during Topology Optimization.

Fig 5: Moving either of the surfaces to change the support volume

In this study, we propose an approach that depends on: 1) Continuously evaluating

the support volume as the iterations of TO increases 2) impose volume constraints
on support structures through sensitivity analysis

Now, for the first step of continuously evaluating the volume we have to assume that
the supports are vertical in nature so the volume will be simply the integral of the
length of structures over the boundary with a given parameter to account for density
difference(fill ratio).

S = ⋎ * ∫β≥β^lm dΓ (3)
S : Net Volume of the supports
β : Subtended angle
lm : Length of support structure at point m
⋎ : Fill ratio (relative material density) of supports
Note: In our case we will take fill ratio value as 0.5 as it is not important for optimality

In some scenarios even though the angle is greater than the threshold but still we
don’t need supports because of the short overhangs.In case of Fused Deposition
Modelling the approximated formula for overhang is given by:

h(mm) = 5 + 40(1 -β/pi) for 3*pi/4< β ≤ pi

= ∞ for 0 < β ≤ 3*pi/4 (4)

Now, on the boundary at any point with any given angle alpha, if the overhang
distance is less than h then there will be no requirement for support structure. When
we implement this equation in simulations during TO it looks for a self-supporting
boundary within the range from eq (4).

Fig 6: Optimizing for self-supporting boundary

3.4 Incorporating Support structure constraint

Initial and most common thought is to implement an absolute and easy constraint as
given by eq 5:
S ≤ Smax (5)

But we really don’t have any knowledge about Smax , so we will place relative
constraints on Smax . As we know in the PareTo method we store multiple solutions
for different volume ratios, that is we will first solve our TO problem without the
support volume constraint and calculate a reference support volume i.e Suno(v) for
different volume ratios v. For example refer to fig 7 for unconstrained volume for
3-hole brackets.
 Fig 7: Support Volume vs Volume ratio for unconstrained case

Now we will try and incorporate a soft constraint i.e., this will be used to prioritize the
solutions within the feasible space, rather than limiting this space. Let us consider an
parameter k where 0<k<1 and formulate the constraint as given in eq 6
S(v) ≤ k*Suno(v) (6)
Summarizing the approach,

Minimize J
Ω⊂Ω0
given, (7)
Ω <= Vf|Ω0|
S(v) ≤ k*Suno(v) (soft)
Kd = f
The variable k will be used to trade-off between performance and Additive
Manufacturing costs as shown in the numerical experiments done in ANSYS ahead.
In the next section we will use ANSYS simulations to solve the above equations. We
will be using topological sensitivity for support volume and topological sensitivity for
performance to design our framework.
3.5 Performance Topological Sensitivity
In this study we take Compliance as the performance parameter and compute its
sensitivity w.r.t Topological change. To do that, we will start by making a hypothetical
small hole in the object and now the performance topological sensitivity is given by
the rate of change of compliance w.r.t volume of the hole. Now for the case of
2-D, we have the formulation given by equation 8:

Tφ(p) ≡ limε⟶0(φ(p;ε) - φ)/(πε2) (8)

When we have the compliance as performance parameter Eq(8) becomes as

follows:

TJ(p) = 4*𝞂/(1+v) : ε - (1-3v)/(1-v2)tr(𝞂)tr(ε) (9)

where,
𝞂 = stress matrix
ε = strain matrix

The process to compute using any simulation software is as follows:

1. Create a Finite Element Analysis Mesh over the design space
2. Compute all the stress and strains in the body
3. Now use equation 9 to calculate the topological sensitivity field
An example is shown in the fig(8) for Topological sensitivity of performance using
MATLAB:

Figure 8: (a) Initial Structure, (b) topological change(small hole) (c) sensitivity curve.
Now, we will be using the PareTo analysis with TO sensitivity as level-set and draw
the curve with changing volume ratio. At every iteration all the above 3 steps are
repeated. For our example resulting field is shown below:

Figure 9: (a) Initial topology, (b) performance TO sensitivity (2D) (c) 3D view.

3.6 Support Volume Sensitivity

For this we will be using sensitivity based on surface angle i.e the rate of change in S
w.r.t V
where, S = support volume and V = hole volume
We will be using already available methods i.e. the topological-shape sensitivity
method [5],[1],[4],[3] to formulate support volume sensitivity. We have two possible
scenarios i.e 1) Interior Hole 2) Boundary Hole
1. Interior Hole
The topology derivative as proposed in [5] is given by eq 10

Ts(p∈Ω) ≡ limr➝0,delta➝0 (S(Ωr+delta)-S(Ωr))/(V(Br+delta)-V(Br))

(10)

where, S(Ωr) is support volume and V(Br) is hole volume for a hole of radius r
Now from eq 10 we have support volume sensitivity as:

Ts(p∈Ω) = 3(pi - βˆ - sin(βˆ)cos(βˆ))(sin(βˆ)-sin3(βˆ)/3)/pi

(11)
 2. Boundary Hole
Ts(p ∈ ∂Ω) = 0.5(1-cos(αp))
(12)

Now with eq 11 and 12 we can calculate sensitivity at each possible point in the
design space.

3.7 Combination
To combine both the above calculated and normalized sensitivities into our Topology
Optimization formulation to impose support constraints we will use augmented
Lagrangian Methods. This method has similarities to the penalty methods in which
we change the constrained optimization problem with a number of unconstrained
problems and a penalty term in the objective function. In this method we combine the
constraint and our objective into one single equation.

L = J + Lg
Lg = n*g + 0.5*γ*(g2) n+γg>0 (13)
= 0.5*n2/γ n+γg≤0
where,
g = S(v)/(k*Suno(v)) - 1 ≤ 0 (14)
n = Lagrangian multiplier
γ = penalty parameter(updated during iterations)
Now, If we take the topological derivative of the eq(13), we will get the effective
sensitivity,

T = TJ + wSTS (15)
where,
wS = n + γg for n + γg>0 (16)
= 0 for n + γg≤0

Now, If we look at eq(16), it gives the sensitivity value as 0 if g < -n/γ. As the
iterations increases, mass/volume is calculated continuously using eq(15), while n
and γ are changed with each repetition as explained in the literature[8][9].
Now as we can see this is unresponsive to the normalization of the TO sensitivity
values, but it will reduce the computational cost and hence make it fast and robust.

3.8 Proposed Framework

Combining all these steps together, final framework is as follows:
1. Calculate the support volume for unconstrained case i.e. calculate Sunc(v)
2. Use simulation software to carry out mesh analysis(FEA) on the design space
i.e. Ω
3. Compute TJ and TS and then use augmented lagrangian methods to evaluate
the combined T field and then smoothen and normalize T.
4. Consider topology(T) as a level-set function, and iterate to a new topology Ω
with the fixed-point iteration method, and the topology is updated. Now we
have to repeat it till we reach the convergence point.
5. Perform steps 2,3,4 till we reach the required volume or acceptable volume
for support structures by decreasing the volume ratio component

Numerical Experiment Example

To illustrate the impact of incorporating the support volume constraint based on the
sensitivity approach into our Topology Optimization process, we have performed
some simulations in ANSYS with an object as a 2D MBB. Now as we know we don't
need any support structure for unoptimized MBB beam but once we run Topology
Optimization on it some holes come into picture and hence need for support
structure. Results of the analysis are shown in the figures below.
 Fig 10: Compliance Curve for the beam

Fig 11: Unconstrained Case Support structure volume w.r.t volume fraction
Now using fig 9 as reference for Sunc , and impose support volume constraint.
Particularly we are interested in the impact of k(relative constraint eq 6) on the
optimum design or final topology for v.f of 0.65. Fig(12) shows the results before and
after implementing the support volume constraint. We can see that if we keep on
increasing the constraint parameter this will try to reduce the number of internal
holes and this is what we expect.

Fig 12: Effect of constraint on optimization of 2D MBB

Conclusion
In this study we tried to incorporate Additive manufacturing specific constraints into
the Topology Optimization phase to take advantage of the natural synergy between
the two technologies. Additive manufacturing promises unlimited possibilities in
terms of shape, size and design. Structures that are forged and optimized using
topology optimization can be executed through AM processes without having to
force-fit into the considerations of ‘traditional’ manufacturing methodologies.
Topologically optimized structures are often complex and hard to machine. If they are
to be machined through traditional manufacturing methods then they have to be
amended and often this results in loss of optimization. As we saw before additive
manufacturing provides us with the flexibility to create complex structures faster and
in an accurate manner and therefore the topologically optimized parts can be created
with better ease.
We suggested an approach using Topological Sensitivity of Performance and
Support Volume sensitivity and combined them with a penalty parameter. We used
the unconstrained volume to be a reference and performed the optimization by
initially setting a volume fraction parameter. And the final effectiveness of the
framework was shown with a numerical example of a 2D MBB and we have also
compared the results with normal Topology Optimized part if manufactured using
Additive manufacturing as it will require support structures to create holes. We have
assumed the support structures to be vertical for simplifying the calculations. But due
to the lack of literature in the domain it can’t be clearly compared with other methods
based on any efficacy parameter. Finally, this research is just the beginning in
incorporating other constraints like material induced anisotropy, surface roughness,
inter-layer fusion, microstructure distortions etc that are induced due to different
Additive manufacturing processes. And this method can be combined with other
optimization methods to find an optimum build direction to further decrease support
requirements.

References
1. A. A. Novotny, “Topological-Shape Sensitivity Method: Theory and
Applications,” Solid Mechanics and its Applications, vol. 137, pp. 469–478,
2006.
2. J. Sokolowski and A. Żochowski, “On Topological Derivative in Shape
Optimization,” SIAM journal on control and optimization, vol. 37, no. 4, pp.
1251–1272, 1999.
3. A. A. Novotny, R. A. Feijoo, and E. Taroco, “Topological Sensitivity Analysis
for Three dimensional Linear Elasticity Problem,” Computer Methods in
Applied Mechanics and Engineering, vol. 196, no. 41–44, pp. 4354–4364,
2007.
4. I. Turevsky and K. Suresh, “Generalization of Topological Sensitivity and its
Application to Defeaturing,” in ASME IDETC Conference, Las Vegas, 2007.
5. R. A. Feijoo, A. A. Novotny, E. Taroco, and C. Padra, “The topological-shape
sensitivity method in two-dimensional linear elasticity topology design,” in
Applications of Computational Mechanics in Structures and Fluids, CIMNE,
2005
6. Bendsøe M (1989) Optimal shape design as a material distribution problem. Struct
Optim 1(4):193–202
7. Bourdin B (2001) Filters in topology optimization. Int J Numer Methods Eng
50(9):2143–2158 Brackett D, Ashcroft I, Hague R (2011) Topology optimization for
additive manufacturing. Proceedings of the Solid Freeform Fabrication Symposium,
Austin, TX
8. Bruns T, Tortorelli D (2001) Topology optimization of non-linear elastic structures and
compliant mechanisms. Comput Methods Appl Mech Eng 190(26-27):3443–3459
9. Cloots M, Spierings A, Wegener K (2013) Assessing new support minimizing
strategies for the additive manufacturing technology SLM. Proceedings of the Solid
Freeform Fabrication Symposium, Austin, TX
10. Gaynor AT, Guest JK (2014) Topology Optimization for Additive Manufacturing:
Considering Maximum Overhang Constraint. 15th AIAA/ISSMO Multidisciplinary
Analysis and Optimization Conference, AIAA Aviation
11. Guest J, Prevost J, Belytschko T (2004) Achieving minimum length ´ scale in topology
optimization using nodal design variables and projection functions. Int J Numer
Methods Eng 61(2):238–254
12. Guest JK (2009a) Imposing maximum length scale in topology optimization. Struct
Multidiscip Optim 37(5):463–473
13. Guest JK (2009b) Topology optimization with multiple phase projection. Comput
Methods Appl Mech Eng 199(1-4):123–135
14. Guest JK (2014) Projection-based topology optimization using discrete object sets.
ASME 2014 International Design Engineering Technical Conferences and Computers
and Information in Engineering Conference, American Society of Mechanical
Engineers, Buffalo, NY, pp 1–8
15. Guest JK (2015) Optimizing the layout of discrete objects in structures and materials:
A projection-based topology optimization approach. Comput Methods Appl Mech
Eng 283:330–351
16. Guest JK, Smith Genut LC (2010) Reducing dimensionality in topology optimization
using adaptive design variable fields. Int J Numer Methods Eng 81(8):1019–1045
17. Mercelis P, Kruth JP (2006) Residual stresses in selective laser sintering and selective
laser melting. Rapid Prototype J 12(5):254– 265
18. Poulsen T (2003) A new scheme for imposing a minimum length scale in topology
optimization. Int J Numer Methods Eng 57(6):741– 760
19. Rozvany G, Birker T (1995) Generalized Michell structures - exact least-weight truss
layouts for combined stress and displacement constraints: Part I - general theory for
plane trusses. Struct Optim 9(3-4):178–188
20. Sigmund O (1997) On the design of compliant mechanisms using topology
optimization. Mech Struct Mach 25(4):493–524
21. Sigmund O (2007) Morphology-based black and white filters for topology
optimization. Struct Multidiscip Optim 33(4-5):401–424
22. Svanberg K (1987) Method of moving asymptotes - a new method for structural
optimization. Int J Numer Methods Eng 24(2):359– 373
23. Zhou M, Lazarov BS, Wang F, Sigmund O (2015) Minimum length scale in topology
optimization by geometric constraints. Comput Methods Appl Mech Eng
293:266–282
24. Ahn SH, Montero M, Odell D, Roundy S, Wright PK. Anisotropic material properties of
fused deposition modeling ABS. Rapid Prototyp J 2002;8:248–57.
doi:10.1108/13552540210441166.
25. Amir O, Sigmund O. Reinforcement layout design for concrete structures based on
continuum damage and truss topology optimization. Struct Multidiscip Optim
2013;47:157–74. doi:10.1007/s00158-012-0817-1.