Design of Machines and Structures, Vol. 10, No. 2 (2020), pp. 123–132.

https://doi.org/10.32972/dms.2020.023

BRIEF OVERVIEW OF GENERATIVE DESIGN SUPPORT SOFTWARE

KRISTÓF SZABÓ – GYÖRGY HEGEDŰS

University of Miskolc, Department of Machine and Product Design
3515 Miskolc–Egyetemváros
szabo.kristof@uni–miskolc.hu

Abstract: The new generative design process that has emerged in recent years has become
available in more and more software. With the classic rapid prototyping procedures, it was
not possible to test the designed products during long-term operation. However, the advent
of metal powder printing and additive technology already allows for the long-term testing of
designed prototypes, and even, if the deviation from the required properties of the product is
negligible, the production of the final products. As a result, with the generative design pro-
cess, design engineers also can create products that seemed unthinkable so far.
Keywords: shape optimization, topology optimization, generative design

1. INTRODUCTION
The shape of a product is primarily determined by the functions to be implemented
by that, which may be slightly influenced by the manufacturing process properties
of the product (e.g. machined or cast parts). During the development of a given pro-
duct, it is advisable to perform various engineering tests and then improve the pro-
duct by its analysis. This can happen by the reduction of the weight of the product,
the of the load capacity of the product, or the change of the final geometry. This
iterative design-development process can also be considered as the search for an op-
timal solution; however, this process is time consuming and requires a more experi-
ence. If the traditional manufacturing technology operations are considered (e.g.
prismatic milling) when developing a product, solutions may have which are limited
by manufacturing technology. Most CAD/CAM modules in computer-aided design
systems, or most custom CAD and CAM software, support modelling procedures or
manufacturing processes that conform to the traditional design approach. The built-
in algorithms that provide shape optimization in the software have been designed to
meet the limitations of the traditional design approach. However, using additive
manufacturing technology, products that were previously thought to be unmanufac-
turable can also be produced (e.g., inaccessible surfaces, “native” elements). It
should be noted that additive manufacturing is a collective term, including the long-
known rapid prototyping processes, which mostly work with polymeric raw materi-
als, however, due to the development of technology, the processes working from
metallic raw materials e.g. by developing and disseminating the properties of
124 Kristóf Szabó – György Hegedűs

Selective Laser Melting (SLM), Laser Metal Deposition (LMD) and the base metal
powders, the product thus produced is not a prototype but an end product [1]. If
additive manufacturing technology is considered during the design process, produc-
ing the geometry of the part to be optimized with traditional modelling methods is
an extremely time-consuming operation. Generative design software and modules
were created in order to solve these kinds of issues.

2. SHAPE OPTIMIZATION AND TOPOLOGY OPTIMIZATION

Optimization during mechanical design means the best solution selected under the
given operating conditions, which is selected from the set of solutions produced.
Mathematically, optimization means the determination of the maximum or minimum
of an objective function. In general, the optimization searches for the best value of
an objective function in each acceptable range, where both the acceptable range and
the objective function can be of different types. Often the objective function defined
at the design phase is multi-objective and multivariate, so optimization can be de-
fined as the calculation of several target values.
During shape optimization, the shape of the margin surface of the generated body
model is changed so that the best possible value of the objective function is obtained
by observing the optimization conditions. The CAD model data also includes design
variables, design parameters, material properties, and boundary conditions for finite
element tests. During shape optimization, the design variables are selected from the
parameters describing the geometry, this parametric data set (e.g. dimensions, shape
features, part history) is automatically generated in the applied parametric integrated
design systems. Since a model of a part can be produced in several ways (e.g. choice
of shape features, sequence of operations, dimensions of sketches), it is important to
note that this can affect the optimization task, so the geometry must be carefully
prepared already during modelling.
In case of topology optimization, the goal is to determine the optimal design of
the part within a given volume under predefined boundary conditions and loads.
During topology optimization, the elements of the initial volume are deleted from
the design space, considering the design variables and the objective function. A de-
tailed elaborate initial geometry is available for shape optimization, while it is not
necessary for topological optimization. There may be names for some software that
refer to shape optimization, but the result produced is the result of a topology opti-
mization. The difference between shape and topological optimization is made clear
by the applied procedure. Topological optimization procedures require computation-
ally intensive numerical and finite element algorithms, and their proliferation has
become possible in recent decades. In practice, several methods have been deve-
loped, which can be divided into two major groups:
(1) gradient type methods:
− material distribution method: SIMP (Solid Isotropic Microstructure/Material
with Penalties) [2],
− homogenizing methods OMP (Optimal Microstructure with Penalization) [3],
 Brief overview of generative design support software 125

− discrete, global methods [4],

− topological derivative and level set method (LSM – Level Set Method) [5],
− method using Sudden Death Method, e.g. ESO (Evolutionary Structural Op-
timization) [6];
(2) non-gradient-type, heuristic methods [7].
One of the most efficient and computationally simple methods is the SIMP
method, most software developers use this method in their systems for topological
optimization.

3. THE MECHANICAL GENERATIVE DESIGN PROCESS

Generative design can be applied not only in mechanical but also in other fields, like
in architectural, furniture, artistic works. The design process can be generalized;
however, the approach focuses on mechanical design processes. Due to the spread
of additive and hybrid manufacturing technology, there has been an increasing em-
phasis in recent years on the emergence of generative design methods in mechanical
design systems. Generative design is a new type of design process, the main features
of which are artificial intelligence-based software and machine learning, which al-
lows the shape and composition of a part to be determined by physics-based simula-
tion and other analysis methods, taking into account expected requirements and op-
timizing objectives (e.g. minimum cost and/or weight). This new design process dif-
fers from traditional methods in that the generative algorithm evaluates and changes
the product model for the next analysis iteration without user intervention and results
in far more solution variants for a given function. The traditional design process re-
quires additional – user-driven – iterations until we get to the manufacturing process,
which increases the product implementation time.
In the generative design process, the software uses nature-based evolutionary ap-
proaches using machine learning. The user only needs to define the design variables
(e.g. materials, design space, weight, manufacturing technology, production cost).
Knowing this, the software produces all possible combinations by knowing the vari-
ables and parameters. Compared to traditional design processes, the number of ver-
sions produced can be orders of magnitude larger. Since each of the generated solu-
tion variants meets the prescribed manufacturability conditions, the individual that
represents the final solution can be selected from the set.
Another feature of generative design is that it can be applied at an early stage of
the design process without an existing conceptual design being available. As a result,
the generative algorithm creates completely new solutions by considering the man-
ufacturability aspects, thus significantly reducing the time of the part testing process.
In the traditional design process, shape or topology optimization targets an already
manufactured version, which removes material that is unnecessarily operational, in
a way that ignores manufacturability and manufacturing costs. As a result, further
modelling, conventional simulation, and testing may be required. In generative de-
sign, simulation is integrated into the design process. In the generative design pro-
cess, the characteristics of the manufacturing technology can be specified as a design
126 Kristóf Szabó – György Hegedűs

variable (e.g. additive manufacturing, 2.5–5D milling, and casting), so the software
only produces solutions that meet the specified variables. Considering additive and
hybrid elaboration, generative design can be used to combine multi-component pro-
ducts (function combination) into a product that was not previously possible due to
traditional manufacturing technology, which can lead to further cost savings for later
use, e.g. for the supply chain or for maintenance and installation.

4. EXAMPLES OF SOFTWARE SUPPORTING GENERATIVE DESIGN

Software supported generative design appeared in the first half of the 2010s. During
these years, AutoDesk has developed optimized structures for Airbus, where Jesse
Coors-Blankenship was the chief engineer in the development department. Using the
experience gained here, he founded a company called Frustum, where he developed
a kernel to support generative design called TrueSOLIDTM, a product that is available
in several design software. The other big developer is AutoDesk, where a product to
support generative design was created from a project called Dreamcatcher running
in its research department, which is also available. Recognizing the need for gene-
rative design, products from newer software developers have become available.
These are characterized by extensive optimization capabilities (e.g., size, weight,
strength, material quality, cost, schedule, manufacturability) as well as cloud-based
service. Computational performance in iterative processes of the generative design
process is much more costly than traditional tasks, so cloud-based computing pro-
vides cost-effective access to many generated variants. Table 1 summarizes and
briefly presents the best-known products that support generative design.

Table 1
Generative design software [8]
Developer Product Properties
A cloud-based application that combines a voxel-based
design algorithm with finite element analysis. It was a
Frustum Generate
stand-alone product until November 2018, after which it
was acquired by PTC.
A generative, function-based application that provides
nTopol-
Element instant feedback during design as it optimizes the shape
ogy
of the object as well as the manufacturing process.
ParaMat- A cloud-based design platform that focuses primarily on
CogniCAD
ters additive manufacturing processes.

Procedures supporting topological optimization and generative design are also avail-
able in simulation software supporting engineering development (e.g. FEM, CFD)
(Table 2).
 Brief overview of generative design support software 127

Table 2
CAE software supporting generative design [8]
Developer Product Properties
It allows to run in parallel to perform large-scale
Altair OptiStruct
optimization tasks quickly.
ANSYS topology optimization algorithm can be
ANSYS
ANSYS started from the desktop, so it can be integrated
Mechanical
into the simulation workflow.
The Tosca structure optimization package integrates
Dassault Tosca Structure, with the CAE environment in cooperation with the
Systèmes Tosca Fluid ABAQUS, ANSYS and MSC Nastran finite element
solvers.
PAM–STAMP, Offers built-in generative design techniques as well
ESI Group ProCAST, as special shape optimization in the SYSTUS simula-
SYSTUS tion package
It offers a variety of processes, from shape and
MSC MSC Nastran
topology optimization to process management
Software Optimization
solutions.

Modules supporting generative design are also available in parametric design sys-
tems (Table 3). These are basically the products of the developers described above,
which are available in an integrated way in the systems.
Table 3
Integrated system including generative design module [8]
Developer Products Properties
Provides access to optimization settings and calcula-
Fusion 360,
Autodesk tions, as well as cloud computing resources for pre-
Inventor
mium subscriptions.
Dassault Provides access to the TOSCA optimization package
TOSCA suite
Systèmes for CATIA and SOLIDWORKS CAD software.
Grasshopper relies on a visual programming
Robert
language and environment to automate design.
McNeel & Rhino
Users logically connect parts by pulling operations
Associates
as planned.
Vanderplaats uses R&D GENESIS for optimization.
It converts the results to free-form (B-rep) objects,
PTC Creo Simulate
thus avoiding polygon models used elsewhere. From
Creo 7.0, it uses Frustum technology.
Integrates the Frustum Generate kernel for genera-
Siemens NX, Solid Edge tive design. Users can modify the generated results
with convergent modelling.
solidThinking Topological optimization is integrated into the CAD
Altair
Inspire modelling process.
128 Kristóf Szabó – György Hegedűs

The advantage of modules available in integrated systems is that models generated by

generative design can be used directly in the given system, so there is no need to ex-
change product data between different software, which can cause conversion errors.

5. CASE STUDY OF GENERATIVE DESIGN

On the 12th of June in 2013, General Electric published a call for tenders on the
grabcad community interface, which required the optimization of an aircraft engine
mount, assuming additive manufacturing technology [9]. Considering the require-
ments in the tender and using the generative design method based on the available
model, we produce a couple of possible solutions with AutoDesk Fusion 360 and
Solid Edge 2020 software, the results of which are described in this chapter.
The material of the component to be tested is Ti–6Al–4V, the assumed yield
strength Re = 131 ksi, the operating temperature To = 75 F, the minimum wall thick-
ness t = 0.05 in (the values given in the tender were used when specifying the units).
The component analysis is examined for four load cases:
(1) a load of 8000 lbs on the pin surface in the opposite direction to the y axis,
(2) a load of 8500 lbs on the pin surface in a direction parallel to the z axis,
(3) a load of 9500 lbs in the yz plane, enclosing an angle of 42° in case of the load
F1,
(4) a torque of 5000 lb-in through a midpoint of the axis of symmetry of the pin
surface and arousing about an axis parallel to the z axis (Figure 1).

In case of topological optimization, the objective function (minimum weight) was

performed with a given safety factor (ns = 2), the original weight of the part was
m = 2.038 kg.

Figure 1
Load cases to be tested for the clamping element.

In addition to additive manufacturing technology, Autodesk Fusion 360 software

also supports the ability to set criteria for manufacturing products made with milling
(2.5D–5D) machining, 2-axis machining, and casting technology. This means addi-
tional solutions in the number of versions generated. In the case study, 5D milling
 Brief overview of generative design support software 129

and additive manufacturing were considered for machining technologies, and the
tabular results generated in the software are illustrated in Figure 2.

Figure 2
Results of generative design using AutoDesk Fusion 360 software

The generated solution variants can be compared to each other, users can analyse the
selected results, which basically allows the load distribution on the part (Figure 3).

Figure 3
Results of load distribution of generated versions
in AutoDesk Fusion 360 software

The load distributions observed in the figure above are in the ideal range, the mag-
nitude of arising stresses in some of the fixing holes is low, in practice the display of
130 Kristóf Szabó – György Hegedűs

the load distribution in this case provides only useful information, numerical values
can be determined by finite element analysis. The Premium version of Solid Edge
2020 provides the ability to perform generative design processes, however, services
are more limited, and operations require more user intervention, which results in
more time. For manufacturing processes, additive as well as conventional (e.g. cast-
ing, turning) machining can be selected; however, it is not possible to specify more
than one machining mode in one test.

Extraction x Extraction y Extraction z Extraction No

Manufacturing Additive Manufacturing Additive Manufacturing Additive Manufacturing Additive
Volume (mm3) 241,533 Volume (mm3) 304,937 Volume (mm3) 198,612 Volume (mm3) 126,785
Weight (kg) 1,088 Weight (kg) 1,374 Weight (kg) 0,894 Weight (kg) 0,571

Figure 4
Results of generative design using Solid Edge 2020 Premium software

If test of multiple manufacturing methods is required, the user must create additional
generative tests (Figure 4). If the generated versions are available, it will also be
possible to compare them. The results of the load distribution in case of the different
variants are illustrated in Figure 5.

Figure 5
Results of load distribution of generated versions using
Solid Edge 2020 Premium software
 Brief overview of generative design support software 131

It is not possible to examine the numerical values of the load distribution, like Auto-
Desk Fusion 360 software. The results obtained are only for information, which al-
lows the identification of load-critical regions; however, the load distribution is not
ideal, as previously illustrated in Figure 5. The disadvantage is that the software does
not allow the tabular display of the results and the simultaneous visual comparison
of the generated versions, and the user does not even have information on the nu-
merical values of the characteristic geometric data (e.g. volume), maximum of von
Mises stress and global displacement is not available for the user. Comparing the
results of the two softwares, it is seen that we get different solutions even with similar
configuration parameters.

6. SUMMARY
In this paper, we have briefly reviewed the various software applications that support
the generative design process. The solutions of two large companies (AutoDesk,
Frustum) developing a generative module, which is widespread in mechanical de-
sign systems, were examined through a case study. Comparing the two generative
design modules, the cloud-based technology used in AutoDesk Fusion 360 software
shortens the time of the generative design process and the number of generable so-
lutions can be significantly increased by adjustable machining parameters within a
given test (one test – many solutions). The main feature of the Frustum Generate
generative module integrated into Solid Edge 2020 Premium software is that it can
be run on a given machine, so the computing capacity required for the generated
models is limited by the user's computer. Another characteristic disadvantage is that
only one manufacturing mode can be specified at a time in the tests, which reduces
the number of solutions that can be generated within a test. If you want to create
multiple solution variants, you need to run multiple studies (many solutions – many
studies), which increases the time of the design process.

ACKNOWLEDGEMENT
The described article/presentation/study was carried out as part of the EFOP–3.6.1–
16–00011 Younger and Renewing University – Innovative Knowledge City – institu-
tional development of the University of Miskolc aiming at intelligent specialisation
project implemented in the framework of the Szechenyi 2020 program. The realiza-
tion of this project is supported by the European Union, co-financed by the European
Social Fund.

REFERENCES
[1] Gibson, I., Rosen, D., Stucker, B. (2010). Additive manufacturing technolo-
gies. New York, Springer, ISBN 978-1-4419-1119-3,
https://doi.org/10.1007/978-1-4419-1120-9
132 Kristóf Szabó – György Hegedűs

[2] Zuo, K., Chen, L., Zhang, Y., Yang, J. (2007). Study of key algorithms in
topology optimization. Int. J. Adv. Manuf. Technol., 32, pp. 787–796,
https://doi.org/10.1007/s00170-005-0387-0.
[3] Rozvany, G. (2001). Aims, scope, methods, history and unified terminology
of computer-aided topology optimization in structural mechanics. Struct. Mul-
tidisc. Optim., 21, pp. 90–108., https://doi.org/10.1007/s001580050174.
[4] Bendsøe, M. (1995). Optimization of Structural Topology, Shape, And Mate-
rial. Berlin, Springer, ISBN 978-3-662-03117-9,
https://doi.org/10.1007/978-3-662-03115-5.
[5] Novotny, A., Sokolowski, J. (2013). Topological Derivative in Shape Optimi-
zation. Berlin, Springer, ISBN 978-3-642-35244-7, https://doi.org/10.1007/
978-3-642-35245-4.
[6] Huang, X., Xie, Y. (2010). Evolutionary Topology Optimization of Continuum
Structures. Wiley, Chichester, ISBN 978-0-470-74653-0.
[7] Tajs-Zielińska, K., Bochenek, B. (2018). A heuristic approach to optimization
of structural topology including self-weight. AIP Conference Proceedings
1922, 020001, https://doi.org/10.1063/1.5019028.
[8] An Introduction to Generative Design – A Digital Guide from the Editors of
Cadalyst. Cadalyst, Longitude Media, (2018) https://cadalyst.tradepub.com/
free/w_cada04/prgm.cgi.
[9] https://grabcad.com/challenges/ge–jet–engine–bracket–challenge