Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/272092215

Optimal topology for additive manufacture: A


method for enabling additive manufacture of
support-free optimal...

Article in Materials and Design July 2014


DOI: 10.1016/j.matdes.2014.06.015

CITATIONS READS

30 709

5 authors, including:

M. Leary Maciej Mazur


RMIT University RMIT University
55 PUBLICATIONS 495 CITATIONS 21 PUBLICATIONS 76 CITATIONS

SEE PROFILE SEE PROFILE

Milan Brandt
RMIT University
188 PUBLICATIONS 1,576 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Laser-assisted Surface Modification and Organo-silane Coatings for Corrosion Resistance of


Magnesium Alloy AZ91D View project

Laser weldability of SLM Ti6Al4V View project

All content following this page was uploaded by M. Leary on 10 February 2015.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
Materials and Design 63 (2014) 678690

Contents lists available at ScienceDirect

Materials and Design


journal homepage: www.elsevier.com/locate/matdes

Technical Report

Optimal topology for additive manufacture: A method for enabling


additive manufacture of support-free optimal structures
Martin Leary a,, Luigi Merli b, Federico Torti b, Maciej Mazur a, Milan Brandt a
a
RMIT Centre for Additive Manufacturing, School of Aerospace, Mechanical and Manufacturing Engineering, RMIT University, Melbourne, Australia
b
Department of Mechanical Engineering, Politecnico di Milano, Milan, Italy

a r t i c l e i n f o a b s t r a c t

Article history: Topology optimisation enables profound insight into the optimal material distribution for a given
Received 16 January 2014 structural objective, applied loading and boundary conditions. The topologically optimal geometry is
Accepted 6 June 2014 often geometrically complex and incompatible with traditional manufacturing methods. Additive
Available online 20 June 2014
manufacture can accommodate signicantly more complex geometries than traditional manufacture; how-
ever, it is necessary that specic design rules be satised to ensure manufacturability. Based on identied
design for additive manufacture rules, a novel method is proposed that modies the theoretically optimal
topology as required to ensure manufacturability without requiring additional support material. By
assessing the manufacturing time and component mass associated with feasible orientations of the
proposed geometry, an optimal orientation can be identied. A case study is presented to demonstrate
the usefulness of the proposed method.
2014 Elsevier Ltd. All rights reserved.

1. Introduction Despite the enhanced geometric freedom associated with


additive manufacture, it is necessary that specic design rules be
Additive manufacture involves the progressive addition of satised to ensure manufacturability [79]. Design for additive
material to generate component geometry; this differs fundamen- manufacture includes requirements associated with minimum
tally from traditional, subtractive manufacture whereby material is feature size, manufacturable inclination angle, allowable bridging
progressively removed from an initial geometry as required [1]. distance, and the robust accommodation of heat transfer. Typically,
Additive manufacture has been labelled as a disruptive technology design for additive manufacture constraints are either accommodated
[2] due to the associated capacity to economically manufacture at by intuitive modication of the intended geometry [9], or by the
very low batch sizes; and, the capability to manufacture highly use of support material to enable acute inclination angles and to
complex geometries. This latter capability provides an opportunity transfer heat as required. The use of support material extends
to physically implement topologically optimal geometries, which the envelope of feasible geometries, but incurs cost and time
are often highly complex, and therefore incompatible with penalties [7].
traditional manufacturing methodologies. A novel method is proposed in this work that modies the
Topology optimisation enables identication of the optimal theoretically optimal topology as required to enable additive
structural connectivity for a specic design scenario, boundary manufacture according to the identied design for additive manu-
conditions and available spatial envelope. Topology optimisation facture constraints. The proposed method requires no manual
is not based on a priori assumptions of material distribution, intervention and results in a geometry that is manufacturable
resulting in complex truss networks with high structural efciency without necessitating the use of support material. By assessing
[35]. The geometric complexity of the topologically optimal the manufacturing time, component volume and platen support
design is typically incompatible with traditional manufacturing base associated with orientations of the proposed geometry, an
methods [6]. Additive manufacture therefore provides an opportunity optimal orientation can be automatically identied.
to manufacture components that more closely approximate the
optimal geometry than traditional methods.
1.1. Topology optimisation

Topology optimisation refers to the search for geometry that


Corresponding author. Tel.: +61 428955563. optimises an objective function, such as minimal mass or cost,
E-mail address: martin.leary@rmit.edu.au (M. Leary). subject to associated boundary conditions and constraints, such

http://dx.doi.org/10.1016/j.matdes.2014.06.015
0261-3069/ 2014 Elsevier Ltd. All rights reserved.
M. Leary et al. / Materials and Design 63 (2014) 678690 679

Nomenclature

N number of optimisation iterations () DV increase in volume due to additional support structure


Ak total area associated with DL and DR at iteration k (%)
nelx number of horizontal elements () DR, DL infeasible domain, subscript indicates right and left
nely number of vertical elements () respectively ()
f volume fraction (%) O feasible void domain ()
p penalisation power () q voxel density value (%)
rmin lter radius (m) qdis discrete voxel density value (%)
tmin offset thickness (m) qt density threshold (%)
/ local gradient (deg) Cint internal sub-boundary of infeasible points ()
/feasible feasible build angle (deg) Cext external sub-boundary of infeasible points ()
S rolling average span () (xi, yi) coordinate of point i in the boundary C ()
Sn number of iterations of rolling average () b base width (m)
a inclination angle of modied geometry (deg) bmin minimum required base width (m)
nk infeasible point of interest at iteration k () 
b base width of the manufactured part (m)
ni, nf initial and nal point of an infeasible domain () e base threshold (m)
X design domain () Hmax maximum height from the platen (m)
Xint internal domain () Hmax maximum height from the platen of the manufactured
dbridge maximum bridging distance (m) part (m)
H orientation angle with respect to platen (deg) hp vertical distance of the point p to platen (m)

as applied loads, allowable spatial envelope or maximum for non-trivial scenarios the layout is complex and therefore
allowable stress [10]. Topology optimisation is distinct to size or impractical with traditional manufacturing methods [6].
shape optimisation, in which the connectivity within a design
domain is specied a priori [11]. Topological optimisation results
in an optimal material distribution that is independent of a priori 1.3. Additive manufacture
assumptions of domain connectivity, and therefore provides a
signicant opportunity for innovative structural design [12]. Additive manufacture is dened by ASTM: F2792 12a as the
In the early 1900s, Michell [13] showed that the limits of process of joining materials to make objects from 3D model data,
economy of material possible are attained if all elements of a usually layer upon layer, as opposed to subtractive manufacturing
loaded frame display equal strain magnitude. Subsequently, numer- methodologies, such as traditional machining. Additive manufac-
ous approaches have been presented in the literature in the eld of ture can utilise numerous source materials, including liquids,
optimal continuum structures [14]. For example, the homogenisa- sheet, wire and powder. Independently of the source material,
tion method [15,16], Solid Isotropic Microstructure with Penalisa- additive manufacture begins with a digital representation of the
tion (SIMP) method [1719], genetic algorithms [11,20], and intended geometry, which is processed as required to generate
evolutionary methods [2124] have been presented. In this work, the tool path of the additive manufacturing system. From this tool
the SIMP method [25,26] is used to solve a 2D topology optimisation path, additive manufacture generates near-net component
problem with the objective of compliance minimisation. geometry without requiring molds, xtures, or tooling.
The fundamental differences between subtractive and additive
manufacture enables a range of novel advantages; consequently,
1.2. Solid Isotropic Microstructure with Penalisation (SIMP) additive manufacture has been labelled as a disruptive technology
[1,28]. Additive manufacture can accommodate highly complex
The 2D topology optimisation problem will be solved with the geometries; for example high-value jet engine [29], aerospace
Matlab implementation of the SIMP method, as initially developed [30], and automotive [31] components that are incompatible
by Sigmund and Petersson [27]; and, computationally enhanced by with traditional manufacturing methods. The unique capability of
applying a method that permits ltering of either the sensitivities additive manufacture to generate highly complex geometries
or densities [26]. The design domain is assumed to be rectangular provides an opportunity to manufacture components and systems
and is discretised by square voxels; where nelx and nely are the that are close to their theoretically optimal geometry. In particular,
number of voxels in the horizontal and vertical directions, respec-
tively. The prescribed volume fraction, f, denes the ratio between
target volume and initial design domain volume; the lter radius,
rmin, denes the radius of the sensitivity lter area; the penalisation
power, p, is a factor introduced to force the element density toward
limiting values of either q = 0 (void) or q = 1 (solid) in order to
generate more distinctly binary designs.
A classic application of the SIMP method is a cantilever beam
with point load, for example [25,26]. This scenario (Fig. 1) will be
used as a case study to demonstrate the novel method presented
in this work. Fig. 2 identies the results of the SIMP method for a
cantilever beam with point load at various iterations, N. It is
apparent that for small values of volume fraction, f, the optimal
geometry is a highly efcient, yet complex truss network [35].
Although such as truss network may be geometrically dened; Fig. 1. Topology optimisation boundary conditions cantilever beam.
680 M. Leary et al. / Materials and Design 63 (2014) 678690

Fig. 2. Topological optimisation results based on SIMP method for the following parameters: nelx = 100, nely = 40, f = 0.5, p = 3, rmin = 1.5; algorithm iteration, N, identied
above.

topologically optimal geometries often include highly complex components with relatively high structural integrity. Although this
truss networks and internal void structures [35]. work refers to the FDM method, the associated results are general-
Polymeric additive systems, are signicantly less complex that izable to other methods of additive manufacture.
metallic systems; for example, Selective Laser Melting (SLM) is sub- A critical technical limitation associated with FDM is the
ject to over 130 processing parameters of relevance [3335]. SLM achievable inclination angle, whereby if the angle between platen
involves signicant thermal gradients, in the order of 104 K/mm and surface, /, is excessively acute, the manufacture will be
[36]. Grain growth occurs in the direction of the thermal gradient compromised, for example [45]. Leary et al. [7] report three
[36], and microstructure and associated mechanical properties are distinct zones of interest to this work1 (Fig. 3):
determined by the temperature prole [35]. Consequently, the
control of heat transfer during manufacture is critical to the robust (1) Robust zone, 40 6 /
additive manufacture of metals. Polymeric materials will be consid-
ered in this work (Section 2), due to their lower complexity; however, Build surface has no identiable structural defects. Volume is
this work provides insight that can be extended to metallic systems, self-supporting.
and an initial assessment of the inuence of the proposed method (2) Compromised zone, 30 6 / < 40
on the temperature prole has been completed (Section 5.3).
Build surface self-supporting and attached to the specimen, but
Despite the dimensional freedom associated with additive man-
includes identiable defects.
ufacture, there are technical constraints that must be satised in
order to generate robust geometry [9,37,38]. In particular, the min- (3) Failed zone, / < 30
imum feature size [6], bridging distance, manufacturable inclina- Complete delamination of the build material to the bulk speci-
tion angle [7] and, for metallic systems, accommodation of heat men. Volume is not self-supporting.
transfer [35] are critical design constraints. Minimum feature size Equivalent manufacturability data is available for metallic
refers to the minimum section size, tmin, that can be feasibly additive systems, for example [45]. When surfaces are detected
manufactured. Bridging distance, dbridge, refers to the maximum to lie outside the robust zone, a dedicated support material is
physical gap that an additive manufacture method can accommo- used to ensure manufacturability. The use of support material sig-
date without compromising manufacturability. Manufacturable nicantly increases the number of geometries that are feasible
inclination angle refers to the minimum inclination to the with the FDM process; however, the use of support material
platen at which an unsupported structure can be successfully incurs signicant disadvantages, including: increased build time
manufactured [7,39]. and cost; post-processing costs and potential contamination of
Despite the opportunities enabled of additive manufacture, specialised polymers, including medical and re-retardant aero-
there exists a lack of methods [9] appropriate for accommodat- space grades.
ing the associated manufacturing constraints. Design for tradi- A novel method has been developed in this work to identify
tional manufacture is reasonably well established with methods geometries that are not within the robust zone, and to modify
of topology optimisation [6]. However, Brackett et al. report that this infeasible geometry as required to ensure support-free
no methods are presented for incorporating specic additive man- manufacturability. The novel method consists of the following
ufacturing constraints within the topology optimisation process sequential stages (Fig. 3):
[40]. This work contributes to the available literature with a novel
method that modies the optimal topology as required to enable 1. Generate optimal topology
additive manufacture without the use of support material.
The SIMP topology optimisation method is applied based on a
widely utilised Matlab implementation proposed in the literature
2. Implementation of proposed method [25,26].

This work will focus on the Fused Deposition Modelling (FDM)


method, whereby a component is manufactured with an extruded
polymer lament [4143]. FDM is commercially signicant as it 1
Based on a FORTUS 900mc system with ABS material, 0.254 mm layer thickness
enables the cost-effective manufacture of large-scale functional and T16 extrusion nozzle.
M. Leary et al. / Materials and Design 63 (2014) 678690 681

2. Dene topology boundary

The SIMP method generates a continuous density voxel eld


associated with each element of the design domain, X. Based on
a threshold value, qt, this continuous density eld is converted to
a binary voxel array. From this binary voxel array an array of
internal and external boundaries are generated (Section 2.1).

3. Boundary smoothing

The internal and external boundaries consist of horizontal and


vertical segments that correspond to the square voxels used in
the SIMP method. These boundaries are converted to a curvilinear
geometry based on a rolling average method (Section 2.2).

4. Measure local gradients

The local gradient / is calculated in relation to the platen


(horizontal) plane for every point associated with the identied
boundary domains (Section 2.3).

5. Identify infeasible domains

The local gradient is used to identify domains where the FDM


build is: robust, compromised or failed, according to the zones
identied in Section 3. Infeasible domains with length below the
bridging distance are agged as being feasible for manufacture
(Section 2.4).

6. Apply novel methods to enable manufacture of infeasible domains

The novel methods developed in this work (Section 3) are


applied to remove infeasible domains and ensure manufacturabil-
ity of the proposed geometry. Pertinent manufacturing attributes
associated with the manufacturable geometry are quantied,
including: component volume, maximum build height and the
base width of the component in contact with the platen.

7. Identify optimal orientation for manufacturability

By assessing the manufacturing attributes associated with


various orientations of the geometry to the build platen, the
optimal component orientation is identied (Section 5.4).

2.1. Dene topology boundary

The SIMP method generates a continuous voxel eld with a


density q 2 [0, 1] associated with each voxel of the design domain,
X. From this continuous voxel density, a binary geometry must be
constructed. The Matlab function generateBoundary.m developed
by the authors achieves this outcome with the following logic2:

 A threshold value, qt, is applied to convert the voxel density


eld from a continuous, q, to a discrete binary variable, qdis,
as follows:

0; if q < qt
qdis 1
1; if q P qt
 The Matlab function bwboundaries.m, is then used to extract the
boundary elements from the binary matrix and arrange this
data in a matrix cell array of external and internal boundaries.
The bwboundaries.m function implements the Moore-Neighbour Fig. 3. Application of the proposed method to the cantilever beam problem (Fig. 1).

2
The Matlab code referred to in this work is available by contacting the
corresponding author.
682 M. Leary et al. / Materials and Design 63 (2014) 678690

zones identied in Section 2. Domains with / > /feasible and length


not exceeding the bridging distance, dbridge, are agged as being
feasible for manufacture.

3. Novel method to enable manufacture of infeasible domains

A novel method has been developed to identify infeasible


domains and modify this infeasible geometry to enable additive
Fig. 4. Connectivity assessment logic.
manufacture. The modied geometry is iteratively dened accord-
ing to the support inclination angle a and the offset thickness tmin,
associated with the selected additive manufacturing process.
tracing algorithm2 and is capable of tracing parent and child
Custom algorithms have been developed to accommodate infeasi-
boundaries by assessing the connections of every element with
ble domains with internal and external boundary.
either 4 or 8 neighbourhood elements (Fig. 4). Better results
were observed with 8-element connectivity.
3.1. Internal boundary method
 A custom algorithm boundary2Nodes.m checks every boundary
element to identify and store the nodal coordinates associated
The method to ensure support-free manufacture of an infeasible
with the boundary sorted in a clockwise order.
domains with an internal boundary is an iterative process that leads
to the denition of three new domains: two domains, DL and DR,
2.2. Boundary smoothing
include the infeasible sub-boundary Cint and became new
subdomains to be iteratively assessed; the third domain, O, is
The internal and external boundaries generated by generatePer-
compatible with support-free additive manufacture. The steps for
im.m consist of horizontal and vertical segments corresponding to
processing the internal domain Xint are formalised below (Fig. 5):
the square voxels of the SIMP method. The function smoothEdge.m
was developed by the authors to smooth the boundaries as follows:
1. The rst point, ni, of the infeasible sub-boundary Cint, is selected
(Fig. 5a). From ni, lines l and r are dened with inclination of a
 The Matlab function smooth.m is applied individually to x- and
and a, respectively (Fig. 5b). The intersection of l and r with
y-data points of the boundary with a moving average of S
the original boundary of Xint is identied (Fig. 5b). The function
points. The location of boundary points associated with bound-
buildSupport.m implemented in MATLAB requires as input a
ary conditions (constraints or loads) are omitted from this
matrix containing the Cartesian coordinates and feasibility of
smoothing operation.
each point of the internal domain Xint. Infeasible points of the
 To avoid generating sharp corners at the beginning and end of
domain with length lower than dbridge are not considered (Sec-
the smoothed boundary (due to insufcient points within the
tion 1.3). For every infeasible point of coordinate (xp, yp) the
span S) a second smoothing operation was applied in the oppo-
two line equations at the support inclination angle, a, are
site direction and with an offset starting point.
dened as follows:

2.3. Measure local gradients y mx cl y mx cr 2

To ensure manufacturability, boundary domains with unfeasible where


inclination angle must be identied and removed. The Matlab func- c yp mxp ; cr yp  mxp ; m tana 3
tion boundaryGradient.m developed by the authors calculates the
local gradient, /, in relation to the platen (horizontal) plane for 2. Three new domains (DL, DR, O) are dened from the original
the identied boundary (Section 3). The gradient associated with boundary and lines l and r (Fig. 5). Domain DL and DR contain
each point, i, of the generic boundary is calculated as the slope of the infeasible points and domain O is feasible according to the
the segment connecting i + 1 and i  1 points, evaluated in clock- associated manufacturability constraints.
wise order. This method may incorrectly evaluate the gradient of To create the left side domain, DL, the intersection between the
sharp corners, and such points must be specically accommodated line l and the boundary of the original domain Xint is detected.
for the identication of infeasible domains (Section 2.4). The rst point (xi, yi) of the boundary after the intersection point
is identied by counting anticlockwise (index i) until the fol-
2.4. Identify infeasible domains lowing condition is satised.
(
x < yc
m
l
The local gradient is used to identify domains where additive 4
manufacture is: robust, compromised, or failed, according to the x 6 xl

Fig. 5. Method to enable additive manufacture of internal boundaries.


M. Leary et al. / Materials and Design 63 (2014) 678690 683

Fig. 6. Method to enable additive manufacture of domains DL and DR.

The intersection point can then be evaluated as the intersection


between the line and the segment connecting (xi, yi) and (xi1, yi1):
   1  
yint 1 m cl
 5
xint 1 mp cp

where mp and cp are the gradient and the y-intercept of the bound-
ary segment, respectively:

yi  yi1
mp ; cp yi  mp  xi 6
xi  xi1
The left hand domain, DL, is dened as:
2 3
xint;L yint;L
6x yp 7
6 p 7
6 7
6 7
DL 6 ... ..
. 7 7
6 7
6 7
4 xi yi 5
xint;L yint;L
The right hand domain, DR, is dened with the same logic as for
DL, but with index i increasing in a clockwise direction:
2 3
xint;R yint;R
6x yp 7
6 p 7
6 7
6 7
DR 6 ... ..
. 7 8 Fig. 7. Method to enable additive manufacture of external boundaries.
6 7
6 7
4 xi yi 5
3.2. External boundary method
xint;R yint;R
The domain O = Xint  DL  DR is dened as: The previous logic is only suitable for closed boundaries (holes)
since the intersection between lines with inclination a and a and
2 3
xint;L yint;L the boundary of the internal domain is always assured. For exter-
6 xp yp 7 nal boundaries a logic that considers the limitation of the design
6 7
6 7 domain X is necessary. The proposed method accommodates
6 xint;R yint;R 7
6 7 external boundaries as follows:
6x yi;R 7
O6
6
i;R 7
7 9
6 .. .. 7 1. Identify the leftmost point and the rightmost point (L, R in
6. . 7
6 7
6 7 Fig. 7a) of every infeasible external sub-boundary Cext. Dene
4 xi;L yi;L 5
a line with inclination a from the leftmost point and a from
xint;L yint;L the rightmost point.
3. The total area Ak associated with DL and DR at iteration k is 2. Evaluate the intersection points IL, IR between the lines and the
dened as (Fig. 5): external boundary or the platen, i.e. x-axis (Fig. 7a).
Z Z 3. If IL, IR in Step 2 are found, calculate the areas AL, AR of the new
Ak dA dA 10 closed domain build from the line and the external boundary if
DL DR
the infeasible domains belongs to that area, otherwise a penalty
4. Steps 13 are repeated for all infeasible points nk within the is applied to that area (AL, AR = 1).
boundary Cint. The identied DL and DR that minimise area Ak 4. If the intersection point in Step 2 is not found, no new domain is
are selected. dened.
5. Generate lines l2 and r2 with offset tmin from l and r respectively 5. Select the minimum area between AL and AR (Fig. 7b).
and identify the intersection points with the DL and DR bound- 6. Evaluate the intersection point between the offset line and the
aries (Fig. 6). external boundary or the platen; build the new closed domain D
6. New domains DL2 and DR2 are dened by the intersection of the from the offset line and the external boundary or from the line,
new offset lines l2 and r2 and the boundary of DL and DR (Fig. 6). the external boundary and the platen, as shown in Fig. 7c.
7. Steps 1 to 6 are repeated for DL2 and DR2 until they are empty or 7. If the new domain is not empty iterate point 16 as for internal
no longer include infeasible points. boundaries.
684 M. Leary et al. / Materials and Design 63 (2014) 678690

Fig. 8. Optimal orientation investigation: Orientation angle h and maximum height


Hmax.

4. Identify optimal orientation for manufacturability

By assessing relevant manufacturing parameters versus orien-


tation angle; the optimal component orientation for support-free
manufacture may be identied. The MATLAB function orientOpt.m
was developed to assess the manufacturing parameters for with
support-free geometry rotated in XY plane by increments of Dh
degrees.
The coordinate transformation for the generic point (xp, yp) from
xy coordinates system to the oriented coordinate system (x0 p, y0 p)
is dened as:
!   !
x0p cos h sin h xp
 11
y0p  sin h cos h yp

The following manufacturing parameters are considered


relevant to the selection of optimal orientation for support-free
additive manufacturing (Section 3):

 Volume increase, DV, associated with the modied geometry to


enable support-free additive manufacture, dened as:

AoptStruct  AorigStruct
DV 12
AorigStruct
where AoptStruct is the total area of the support-free structure, and
AorigStruct is the total area of the topologically optimal structure.
 Maximum height from the platen, Hmax (Fig. 8).
 The base width at the platen, b, evaluated as:

b jxmax  xmin j; for ymin < y < ymin e 13


where e is a threshold depending on the machine settings.

5. Case study: additive manufacture of support-free


topologically optimal cantilever beam Fig. 9. Method to enable additive manufacture a = 45, tmin = 1 mm.

The novel method developed to enable additive manufacture of and has been manufactured to conrm that the modied geometry
support-free topologically optimal geometry has been is manufacturable by additive methods, and to compare the out-
implemented for a cantilever beam with point load. The steps of comes of the support-free method with additive manufacture
the proposed method and associated results for a range of using support material.
manufacturability parameters are identied. Numerical stress
analysis has been conducted to verify the inuence of the 5.1. Support-free additive manufacture
additional structure generated by the proposed method on the
associated stress distribution. The optimal topology for additive The methods developed in Section 5 to enable support-free
manufacture has been identied for various scenarios of interest additive manufacture have been applied to a cantilever beam with
M. Leary et al. / Materials and Design 63 (2014) 678690 685

Fig. 10. Increase in volume, DV, associated with varying support inclination angle, a, and offset thickness, tmin.

Table 1
Summary of additional volume DV associated with angles a and offset thickness tmin
of Fig. 10.

a () tmin (mm) DV (%)


30 1 4.67
30 3 7.08
60 1 19.58
60 3 30.95

Fig. 12. Von Mises stress associated with 2D cantilever beam with point load.
point load (Fig. 1). Fig. 9 details the method, including: optimal Support-free geometry for H = 0 with manufacturability parameters: tmin = 2 mm,
topology, identication of infeasible domains, and iterative modi- a = 45.
cation according to a = 45 and tmin = 1 mm, until no infeasible
domains exist. Fig. 10 identies the support-free geometry devel-
oped for variations of parameters a and tmin, with associated
different load transmission and stress distribution in the manufac-
increase in volume. It is apparent that increasing the inclination
tured component.
angle of modied geometry, a, results in an increase in the associ-
A numerical stress analysis has been conducted for the
ated additional volume required for support-free manufacture, DV,
geometry associated with this case study in order to evaluate the
as does the increase in offset thickness, tmin (Table 1).
variation in observed stress due to the geometric modications
for support-free manufacture. A unit load has been applied, and
5.2. Inuence of modied geometry on stress distribution iso-contours of Von Mises (rVM,) stress are plotted (Figs. 1113).
For this case study, it is apparent that the geometric modications
The additional material introduced to ensure support-free addi- to enable support-free manufacture do modify the stress distribu-
tive manufacture results in increased section sizes and additional tion; however the maximum observed stress is not increased. As
load transmission paths; these modications will result in a

Fig. 13. Von Mises stress associated with 2D cantilever beam with point load.
Fig. 11. Von Mises stress associated with 2D cantilever beam with point load. Optimised geometry for H = 276 with manufacturability parameters: tmin = 2 mm,
Topological optimal geometry. a = 45.
686 M. Leary et al. / Materials and Design 63 (2014) 678690

Inial geometry provides a high


resistance to heat transfer

The modied geometry lowers heat transfer


resistance, lowering peak and average tempreratures

Fig. 14. Qualitative comparison of temperature distribution in initial (above) and modied (below) geometry subject to a unit heat ux at approximately 90% completion of
the manufacturing process.

Fig. 15. Optimal orientation investigation indicating: Maximum height Hmax, Increase in volume DV and, base width b (with e = 0) related to orientation angle.

required, the effect of additional material on stiffness and associ- 5.3. Inuence of modied geometry on temperature distribution
ated tolerance considerations can be assessed [46].
In Section 6 it is identied that the geometric modications for Polymeric additive systems have been considered in this work,
support-free manufacture enable a signicant reduction in due to their lower manufacturing complexity in comparison with
manufacturing time in comparison to additive manufacture with metallic systems. However, many of the outcomes of this work
a dedicated support material. It may provide a competitive can be transferred to metallic additive systems by specifying
advantage to manufacture the component with support-free the associated feasible inclination angle and minimum offset
geometry, and then mechanically remove the additional material thickness, for example [44]. In addition to constraints associated
with post-manufacture operation. In such cases the nal geometry, with self-supportability, metallic additive systems are subject to
and associated stress distribution will be as for the topologically complex thermal effects that directly determine microstructure
optimal geometry. and associated mechanical properties [35]. To provide insight into
M. Leary et al. / Materials and Design 63 (2014) 678690 687

Table 2  Minimum build height (H = 0).


Summary of Fused Deposition Modelling process parameters associated with this  Minimal increase in build volume (H = 276).
work.
 Near-minimal increase in build volume, while maintaining base
Parameter tmin width (H = 90).
Machine type Fortus 900mc 1 mm
Build material Polycarbonate (PC) 6. Manufacturability experiment
Support material Stratasys support material
Filling logic: Sparse interior
Fine surface nish
The support-free orientations identied in Section 7.3 were
Slice height 0.254 mm experimentally assessed and compared to traditional additive
Part thickness 10 mm manufacture with and without support material. Build time,
support material consumption and build material consumption
were been quantied. Additive manufacture was completed with
Fused Deposition Modelling (FDM) with process parameters
the effect of the proposed method on the heat transfer during metal- dened in Table 2.
lic additive manufacture, a thermal analysis has been completed for
the initial, and modied geometry with unit heat ux to the upper 6.1. Orientation 1: Minimum build height (H = 0)
surface at approximately 90% completion of the manufacturing
process (Fig. 14). From this qualitative comparison, it is apparent Orientation 1 (H = 0) is identied to achieve the minimal build
that the additional material added by the method proposed in this height, Hmax. The associated increase in volume, DV, is approxi-
work acts to reduce the heat transfer resistance through the part mately 12%, and the base width, b, is the global maximum of 90
geometry, resulting in lower peak and average temperatures. voxels (Fig. 15).
For Orientation 1 (Fig. 16), additive manufacture with support
material provided robust outcomes; however manufacture
5.4. Optimal orientation without support failed catastrophically due insufcient lament
adhesion. Support-free additive manufacture by the proposed
The point loaded cantilever beam was assessed for optimal ori- method enabled robust manufacture with: no use of support
entation according to the parameters of: build height, Hmax, material; an increase in the consumption of build material of
required increase in build volume, DV, and, base width, b, versus approximately 14% in comparison with the use of support
orientation angle. Three orientations were identied for t = 2 mm, material; and, a reduction in build time of approximately 70% in
and a = 45, as being of interest to assess experimentally (Fig. 15): comparison with the use of support material.

Fig. 16. Orientation 1 (H = 0). Details: (a) Machine support. (b) Optimal support. (c) No support Failed.
688 M. Leary et al. / Materials and Design 63 (2014) 678690

Fig. 17. Orientation 2. Minimal increase in build volume (H = 276).

6.2. Orientation 2: Minimal increase in build volume (H = 276) approximately 5% in comparison with the use of support material;
and, a reduction in build time of approximately 76% in comparison
Orientation 2 (H = 276) is identied to achieve the minimal with the use of support material.
increase in volume, DV. The associated height is approximately
97 voxels, which is near the associated global maxima of 100 vox-
7. Summary
els, and the base width, b, is approximately 6 voxels, which is near
the associated global minima of 0 voxels (Fig. 15).
Topology optimisation enables profound insight into the
For Orientation 2, additive manufacture with support material
optimal material distribution by enabling identication of struc-
provided robust outcomes; however manufacture without support
tural connectivity that is independent of a priori assumptions
and support-free manufacture failed catastrophically due to insuf-
[36]. Since the seminal work of Michell [13], topology optimisa-
cient adhesion to the build platen (Fig. 17). The theoretical
tion has achieved a high level of sophistication in the eld of
reduction is signicant for support-free manufacture, i.e. approxi-
optimal continuum structures [24]. The identied optimal
mately 77% reduction in build-time in comparison with the use
topology consists of complex truss networks that are challeng-
of support material, however it is apparent that sufcient base
ing for traditional manufacture [25]. Additive manufacture
width must be provided to ensure robust connection between
provides an opportunity to manufacture components with high
platen and component.
complexity, however there are few design rules to aid the
designer in ensuring that additive manufacturability constraints
6.3. Orientation 3: Low increase in volume, while maintaining base are satised [9,40]. In particular, no specic design rules have
width (H = 90) been presented that modify the theoretically optimal topology
to ensure manufacturability without requiring additional
In response to the failure of the support-free manufacture in support material.
Orientation 2, Orientation 3 (H = 90) was selected to provide This work presents an automated method to modify topologi-
near-minimal increase in build volume, while maintaining a cally optimal geometries as required to enable support-free manu-
reasonable base width. The associated increase in volume, DV, facture. The method has been applied to polymeric additive
is approximately 4%, and the base width, b, is 30 voxels systems as they are signicantly less complex that metallic
(Fig. 15). systems. In addition to being subject to feasible inclination
For Orientation 3 (Fig. 18), additive manufacture with support constraints [44], metallic systems are subject to signicant thermal
material provided robust outcomes; however manufacture with- loading that must be accommodated to ensure robust outcomes.
out support failed due insufcient lament adhesion, although Initial analysis indicates that the additional material added by
the failure was less catastrophic than for Orientation 1, and, unlike the method proposed in this work acts to reduce heat transfer
Orientation 2, the component was robustly connected to the platen resistance, resulting in lower peak and average temperatures.
throughout the build. Support-free additive manufacture by the Further work is required to develop the proposed method for
proposed method enabled robust manufacture with: no use of sup- metallic additive manufacture; however it provides a logical
port material; an increase in build material consumption of extension to the current work.
M. Leary et al. / Materials and Design 63 (2014) 678690 689

 Any increase in inclination angle of modied geometry, a, or


offset thickness, tmin, results in an increase in the additional
volume required for support-free manufacture, DV. It is
therefore desirable to use materials with large feasible build
angle, /feasible, and minimum allowable offset thickness.
 The proposed method to enable additive manufacture results in
additional material being added to the topologically optimal
geometry. This additional material results in increased section
sizes and additional load transmission paths. These modica-
tions change the stress distribution within the component;
however, for the scenarios investigated, there was no increase
in the observed maximum stress.
 Although the method was applied to polymeric additive manu-
facture, initial analysis indicates that the method is benecial to
metallic systems as it acts to reduce heat transfer resistance,
thereby reducing peak and average temperatures.
 Signicant reduction in manufacturing time can be achieved by
the use of the proposed method, when compared to additive
manufacture with support material. As the proposed method
does not require support material it is also advantageous
for applications with internal voids (which cannot be accessed
to remove material post-manufacture), and for medical grade
materials (where support materials may pose a risk of
contamination).
 The method enables the optimal component orientation to be
identied considering the relative importance of manufacturing
time, material consumption and requirements for stability of
the component on the platen during manufacture.

References

[1] Zhai Y, Lados DA, LaGoy JL. Additive manufacturing: making imagination the
major limitation. JOM 2014;66(5):80816.
[2] Hopkinson N, Hague R, Dickens P, editors. Rapid manufacturing: an industrial
revolution for the digital age. John Wiley & Sons; 2006.
[3] Bendse M, Sigmund O. Topology optimization: theory, methods and
applications. Springer; 2003. 370 pages.
[4] Bendse M. Optimization of structural topology, shape and material. Springer;
1995.
[5] Seepersad CC. A robust topological preliminary design exploration method
with materials design applications. Doctoral dissertation. Georgia Institute of
Technology; 2004.
[6] Chang KH, Tang PS. Integration of design and manufacturing for structural
shape optimization. Adv Eng Softw 2001;32(7):55567.
[7] Leary M, Babaee M, Brandt M, Subic A. Feasible build orientations for
self-supporting fused deposition manufacture: a novel approach to space-
lling tessellated geometries. Adv Mater Res 2013;633:14868. Trans Tech
Publications, Switzerland.
[8] Chahine G, Smith P, Kovacevic R. Application of topology optimization in
modern additive manufacturing. In: Solid freeform fabrication symposium,
Austin, Texas; 2010.
[9] Doubrovski Z, Verlinden JC, Geraedts JM. Optimal design for additive
manufacturing: opportunities and challenges. In: ASME 2011 international
Fig. 18. Orientation 3. Near-minimal increase in build volume, while maintaining
design engineering technical conferences and computers and information in
base width (H = 90).
engineering conference. American Society of Mechanical Engineers; 2011. p.
63546.
[10] Sigmund O. Topology optimization: a tool for the tailoring of structures and
materials. Philos Trans R Soc A Math Phys Eng Sci 2000;358(1765):21127.
8. Conclusions [11] Jakiela MJ, Chapman C, Duda J, Adewuya A, Saitou K. Continuum structural
topology design with genetic algorithms. Comput Methods Appl Mech Eng
2000;186(2):33956.
The method proposed in this work has been demonstrated to [12] Pedersen P, Pedersen NL. Reections on truss and continuum topology optimal
enable support-free additive manufacture of topologically optimal designs. In: IUTAM symposium on topological design optimization of
components. This outcome enables signicant opportunities for structures, machines and materials. Netherlands: Springer; 2006, January. p.
6776.
enhanced design and manufacture of structurally optimal compo-
[13] Michell AGM. LVIII. The limits of economy of material in frame-structures.
nents. The following conclusions can be drawn: London Edinburgh Dublin Philos Mag J Sci 1904;8(47):58997.
[14] Eschenauer HA, Olhoff N. Topology optimization of continuum structures: a
review. Appl Mech Rev 2001;54(4):33190.
 The proposed method can successfully modify a topologically
[15] Bendse M, Kikuchi N. Generating optimal topologies in structural design
optimal geometry as required to enable support-free additive using a homogenization method. Comput Methods Appl Mech Eng 1988;
manufacture of FDM structures. To accommodate the geometric 71(2):197224.
complexity of topologically optimal structures requires distinct [16] Suzuki K, Kikuchi N. A homogenization method for shape and topology
optimization. Comput Methods Appl Mech Eng 1991;93(3):291318.
logical processes for internal and external boundaries of two- [17] Bendse MP. Optimal shape design as a material distribution problem. Struct
dimensional geometries. Optim 1989;1(4):193202.
690 M. Leary et al. / Materials and Design 63 (2014) 678690

[18] Zhou M, Rozvany GIN. The COC algorithm, Part II: topological, geometrical and [35] Krauss H, Eschey C, Zaeh MF. Thermography for monitoring the selective laser
generalized shape optimization. Comput Methods Appl Mech Eng 1991;89(1): melting process. In: Bourell D, Crawford R, Seepersad C, Beaman J, Marcus H,
30936. editors. Proceedings of the 23rd annual international solid freeform fabrication
[19] Rozvany GIN. Aims, scope, methods, history and unied terminology of symposium. Austin, TX: The University of Texas at Austin; 2012.
computer-aided topology optimization in structural mechanics. Struct [36] Rehme O, Emmelmann C. Rapid manufacturing of lattice structures with
Multidiscipl Optim 2001;21(2):90108. selective laser melting. In: Lasers and applications in science and
[20] Chapman CD, Jakiela MJ. Genetic algorithm-based structural topology design engineering. International Society for Optics and Photonics; 2006. p. 61070K.
with compliance and topology simplication considerations. J Mech Des [37] Hague R, Campbell I, Dickens P. Implications on design of rapid manufacturing.
1996;118(1):8998. Proc Inst Mech Eng Part C J Mech Eng Sci 2003;217(1):2530.
[21] Xie YM, Steven GP. Evolutionary structural optimization. London, [38] Vayre B, Vignat F, Villeneuve F. Designing for additive manufacturing. Proc
England: Springer-Verlag; 1997. ISBN 3-540-76153-5. 200 pp. CIRP 2012;3:6327.
[22] Xie YM, Steven GP. A simple evolutionary procedure for structural [39] Bablani M, Bagchi A. Quantication of errors in rapid prototyping processes,
optimization. Comput Struct 1993;49(5):88596. and determination of preferred orientation of parts. Trans NAMRI SME
[23] Xie YM, Steven GP. Optimal design of multiple load case structures using an 1995;XXIII:31924.
evolutionary procedure. Eng Comput 1994;11(4):295302. [40] Brackett D, Ashcroft I, Hague R. Topology optimization for additive
[24] Rong JH, Xie YM, Yang XY. An improved method for evolutionary structural manufacturing. In: 22nd Annual international solid freeform fabrication
optimisation against buckling. Comput Struct 2001;79(3):25363. symposium; 2011, August. p. 34862.
[25] Sigmund O. A 99 line topology optimization code written in Matlab. Struct [41] Crump S. Apparatus and method for creating three-dimensional objects. US
Multidiscipl Optim 2001;21(2):1207. Patent 5,121,329. Minneapolis, Minn.: Stratasys, Inc.; 1992
[26] Andreassen E, Clausen A, Schevenels M, Lazarov BS, Sigmund O. Efcient [42] Bellini A, Geri S. Mechanical characterization of parts fabricated using fused
topology optimization in MATLAB using 88 lines of code. Struct Multidiscipl deposition modeling. Rapid Prot J 2003;9(4):25264.
Optim 2011;43(1):116. [43] Ahn SH, Montero M, Odell D, Roundy S, Wright PK. Anisotropic material
[27] Sigmund O, Petersson J. Numerical instabilities in topology optimization: a properties of fused deposition modelling. ABS Rapid Prot J 2002;8(4):24857.
survey on procedures dealing with checkerboards, mesh-dependencies and [44] Wang D, Yang Y, Yi Z, Su X. Research on the fabricating quality optimization of
local minima. Struct Optim 1998;16(1):6875. the overhanging surface in SLM process. Int J Adv Manuf Technol 2013;65(9
[28] Anderson C. Makers: the new industrial revolution. Crown Business; 2012. 272 12):147184.
pages. [45] Levya GN, Schindela R, Kruth JP. Rapid manufacturing and Rapidtooling with
[29] Halchak J, Wooten J, McEnerney B. Layer build of titanium alloy complex- Layer Manufacturing (LM) Technologies, state of the art and future
geometry components for rocket engines. In: Ti 2011 Proceedings of the 12th perspectives. CIRP Ann Manuf Technol 2003;52(2):589609.
world conference on titanium, vol. 3, Beijing, China; 2012. p. 17158. [46] Mazur M, Leary M, Subic A. Computer aided tolerancing (CAT) platform for the
[30] Brandt M, Sun S, Leary M, Feih S, Elambasseril J, Liu Q. High-value SLM design of assemblies under external and internal forces. Comput Aided Des
aerospace components: from design to manufacture. Adv Mater Res 2011;43(6):70719.
2013;633:13547. Trans Tech Publications, Switzerland.
[31] Cooper DE, Stanford M, Kibble KA, Gibbons GJ. Additive manufacturing for
product improvement at Red Bull Technology. Mater Des 2012;41:22630.
Further reading
[33] Baufeld B, Biest OVD, Gault R. Additive manufacturing of Ti6Al4V
components by shaped metal deposition: microstructure and mechanical [32] Kruth JP, Levy G, Klocke F, Childs THC. Consolidation phenomena in laser and
properties. Mater Des 2010;31:S10611. powder-bed based layered manufacturing. CIRP Ann Manuf Technol
[34] Krauss H, Zaeh MF. Investigations on manufacturability and process reliability 2007;56(2):73059.
of selective laser melting. Phys Proc 2013;41:80815.

View publication stats

You might also like