Available online at www.sciencedirect.

com

ScienceDirect
Procedia Manufacturing 10 (2017) 578 – 589

45th SME North American Manufacturing Research Conference, NAMRC 45, LA, USA

Geometric Complexity Based Process Selection for Hybrid

Manufacturing
Anay Joshi and Sam Anand*
Department of Mechanical & Materials Engineering
Center for Global Design and Manufacturing
University of Cincinnati, Cincinnati, OH 45221, USA

Abstract

Hybrid manufacturing (HM), which combines additive manufacturing (AM) and subtractive manufacturing (SM)
operations in a single machine is gaining popularity among engineers because of its capability of leveraging the
advantages of both processes. As a result, a part with complex geometries and organic designs can be manufactured
with the desired surface finish with minimal setup. However, there is a need for choosing the correct manufacturing
method for a given product and developing the optimum process plan. This paper presents a novel approach for
decision making between the three manufacturing processes, along with an optimal part division methodology for
HM. The proposed process plan minimizes geometric complexities and optimizes the manufacturing resources. A
novel metric, called complexity score, has also been developed for quantifying the part’s geometric complexities
which includes various parameters based on Design for Manufacturing (DFM) rules. The proposed methodology
will be useful for detecting potential manufacturing difficulties before the actual fabrication of a product and
identifying the best manufacturing process for fabrication. Three case studies representing different manufacturing
methods are presented for validating the practical application of the research.
© 2017 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
© 2017 The Authors. Published by Elsevier B.V.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-reviewunder
Peer-review underresponsibility
responsibility
of of
thethe Scientific
organizing Committee
committee of 45th
of the NAMRI/SME.
SME North American Manufacturing Research Conference

Keywords: Hybrid additive manufacturing; Geometric complexity; Decision making; Design for manufacturing

* Corresponding author. Tel.: +1-(513)-556-5596; fax: +1-(513)-556-3390.

E-mail address: sam.anand@uc.edu

2351-9789 © 2017 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of the 45th SME North American Manufacturing Research Conference
doi:10.1016/j.promfg.2017.07.056
 Anay Joshi and Sam Anand / Procedia Manufacturing 10 (2017) 578 – 589 579

1. Introduction

The prime focus of today’s industry is to launch a new product in the market in the shortest possible time. DFM is
an approach by which the total development time is significantly reduced and it plays a key role in any product
development process. It is used to predict manufacturing constraints, time, and costs during early design phase and
also helps find solutions to minimize or eliminate some of these difficulties [1]. The salient concepts of DFM for
additive and subtractive processes are used in this paper to form the basis of decision making.
Additive manufacturing processes fabricate a part using a layer-by-layer material addition approach. With the
availability of AM, designers can now produce highly complex parts with applications ranging from the aerospace
industry to the medical industry. However, there are some inherent limitations to additive manufacturing, for
example, cusp height error due to layering results in low surface finish [2]. Subtractive manufacturing refers to
conventional manufacturing processes such as milling and turning operations, which realize a part by removing
material from a blank of the required material. Subtractive machines produce parts with good surface finish,
however the throughput of these processes is low compared to additive manufacturing process when it comes to
complex geometries. Moreover, there are manufacturability constraints for different SM processes. A common
example for milling process is the recommendation to use clearance between part features so that large cutters can be
used to increase material removal rate [3]. Hybrid manufacturing machines, which have material removing tools as
well as additive manufacturing tools, provide a good compromise between AM and SM. This opens an entirely new
domain for engineers who can use hybrid machines to compensate the manufacturing limitations of both additive
and subtractive processes to produce a net shaped part as quickly as possible.
However, this presents the necessity of optimal process planning and decision making of part decomposition for
AM and SM to optimize the manufacturing resources and time and minimize geometric complexities. This paper
provides a process planning approach to decide on the correct manufacturing method based on geometric
complexities of the part. This method can be incorporated in a cost model to check whether investments for new
manufacturing facilities are justifiable based on the product portfolio of an industry.
The rest of the paper is structured as follows. The next section presents previous research carried out in AM, SM,
HM, and decision making procedures between these processes. In subsequent sections, methodologies to identify
important DFM parameters and combining them to calculate complexity score are explained. Various case studies
are also presented to demonstrate practical applications of the process plan developed in this paper. The paper
concludes with the summary and directions for future work.

2. Literature Review

The motivation to introduce additive deposition heads in CNC machines to create HM machines is to improve
part quality and reduce processing time. With commercially available machines and CAM software [4-6], hybrid
manufacturing can play an important role towards integrating additive processes in the manufacturing industry. This
section presents an overview of the literature available on hybrid manufacturing, design for manufacturing rules and
decision making between AM and SM.

2.1. Hybrid Manufacturing

Hybrid manufacturing in the form of laser cladding and milling operations was used by Jeng and Lin [7] to
maintain accurate dimensions and smooth surface finish for mold fabrication. Karunakaran et al. [8] demonstrated
hybrid manufacturing on a single machine by retrofitting a welding process representing additive manufacturing on
a CNC machine which represented subtractive manufacturing. Manogharan et al. [9] introduced a new method
called AIMS (Additive system Integrated with subtractive MethodS). It uses sacrificial support structure which not
only supports overhangs in additive manufacturing but also acts as a fixture during subtractive manufacturing.
Economic models of hybrid manufacturing were presented by Gunther et al. [10] and Manogharan et al. [11].
580 Anay Joshi and Sam Anand / Procedia Manufacturing 10 (2017) 578 – 589

2.2. Design for manufacturing

DFM rules for different SM processes are typically used in industry during the design stage of product
development. Various papers have established DFM rules to design parts for manufacturability and assembly [1].
Design for manufacturability handbook by Bralla [3] is a comprehensive book which presents manufacturing
recommendations for forming, machining, casting, and assembly processes. Chiu and Kremer [12] reviewed Design
for X (DfX) methods in detail and also recommended its sequential application during design stage.
Design for additive manufacturing (DFAM) rules are relatively new design guidelines for reducing difficulties
while fabricating a part by various additive manufacturing processes which take into consideration the limitations of
AM processes. Ranjan et al. [13] developed design guidelines for additive manufacturing which includes geometric,
thermal and support structure parameters. Ranjan et al. [14] integrated DFAM guidelines with topological
optimization. Guido and Zimmer [15] developed process independent design guidelines for additive manufacturing.
Nelaturi [16] et al. demonstrated the manufacturability maps of a part contour based on resolution of the machine.

2.3. Decision making between manufacturing processes

Kerbrat et al. [17] presented hybrid and modular approaches for part divisions to increase manufacturability, and
developed manufacturability indices based on octree decomposition of a CAD model to compare the original and
modified models. Zhu et al. [5] developed a process plan for HM which facilitates remanufacture and inspection of
parts. This method identifies the manufacturable section of a part and decomposes the features accordingly. The use
of weighted producibility factors and design difficulties have been demonstrated in [18]. Ranjan et al. [13] also
developed producibility index for comparing different part designs to be fabricated by additive manufacturing.
The parameters used for decision making for hybrid manufacturing in the current literature are inadequate
because critical geometric difficulties are not considered and thus, need to be more practical. This paper takes into
account the detailed manufacturing constraints for both the processes and also provides a process plan decision
during the product development phase, prior to manufacturing phase.

3. Methodology

This section presents a step by step approach to identify and quantify the proposed complexity score using
DFM/DFAM parameters for a given part. The complexity score is calculated for manufacturing by AM only, SM
only and a combination of part divisions for HM. To accommodate this requirement, every DFM/DFAM parameter
is calculated at each division of CAD model. The steps to calculate the complexity score and the associated decision
of AM, SM, or HM are presented below.

3.1. Additive manufacturing parameters

For additive manufacturing, the parameters to calculate complexity score are based on DFAM guidelines. The
following parameters are considered in this study: sharp corners, thin regions and thin gaps in each layer, support
structure volume, and volume and height of a part.

3.1.1. Sharp corners

Deposition heads or laser beams used for additive manufacturing are circular in shape and have a finite
deposition size or sintering size making fabrication of sharp corners in each layer very difficult. Thus the sharp
corners must be identified for each build layer and must be minimized or eliminated during design phase.
 Anay Joshi and Sam Anand / Procedia Manufacturing 10 (2017) 578 – 589 581

15° 15°

255° 255°

Figure 1. Sharp corners for AM.

The approach for identifying and counting the number of sharp corners is adapted from [13] within our group, in
which a CAD model is converted to an STL file which is a standard file format used by AM machines. Each facet of
the STL file is sliced [19] along the planes orthogonal to the build direction and are separated by a distance equal to
user defined slice thickness. The points resulting from intersection of these planes and planar STL facets are
arranged in a cyclic order and joined by line segments. Next, angles between consecutive line segments are
calculated. The angles which are smaller than the user specified threshold value are counted as sharp corners. For
example, with the threshold angle of 20°, sharp corners of 15° are identified in the CAD model shown in Figure 1.
To calculate the number of sharp corners in an entire model, the sharp corners identified in all slices are added
together. To calculate the number of sharp corners in a divided part, the sharp corners identified in slices
corresponding to respective AM portion of the division are added together.

3.1.2. Thin regions and thin gaps

AM deposition heads have a finite deposition size making fabrication of thin regions very difficult. Thin regions
are also prone to and increase thermal deformations [13]. On the other hand, thin gaps may be fused together
combining different features separated by it. Thus, thin regions and thin gaps in each layer must be avoided.
A ray casting based algorithm is used to identify and count the number of thin regions and thin gaps in a CAD
model. Prior work on using ray casting for Boolean operations has been reported in [20]. The first step is to convert
the CAD model of a part to an STL file. Intersection points of a facet with horizontal planes and facet normals are
stored layer by layer. To find thin regions, rays originating from the contour points resulting from slicing process are
traversed along the direction opposite to the facet normal. The distance traveled by the ray before intersecting the
same contour is calculated which represents the overall size of a feature in a part. If this distance is less than a
certain threshold value, the existence of a thin region is identified. This is illustrated in Figure 2b. To find thin gaps,
rays are initiated in a direction along the facet normal and points of intersection are calculated with neighboring
contours. This distance represents the gap between features. This is illustrated in Figure 2c.
The total number of thin gaps and thin regions in a part is the summation of the thin regions and thin gaps
identified in each layer. To calculate the number of thin features of a part, the summation from appropriate AM z-
coordinates is performed.

Thin region Thin

gaps
a) b) c)
Figure 2. (a) CAD model; (b) Identified thin regions; (c) Identified thin gaps.
582 Anay Joshi and Sam Anand / Procedia Manufacturing 10 (2017) 578 – 589

3.1.3. Support structure volume

Additive manufacturing processes conventionally fabricate a part along z-axis. Thus the overhanging features
require a support below them in order to maintain their spatial position and functionality. Support structure
contributes to material wastage and increase manufacturing time and therefore must be minimized. When an STL
facet whose normal angle with the build direction is more than a threshold value, support structures are necessary.
The method used to calculate support structure volume is adapted and modified from [21] within our group,
which is based on a voxel based algorithm. A voxel is a small cube with user defined size and voxelization
represents a three dimensional (3D) decomposition of a CAD model. An angle criterion is incorporated in this
method to accommodate the threshold angle for generating support structures. In this method, the CAD model is
converted to an STL file, and the STL facets which require support based on the angle criterion are identified. A
point mesh is generated on these facets with its size depending on the voxel dimension. The CAD model is then
voxelized which provides a 3D array of filled and empty voxels. The filled voxels which occupy support structure
point mesh are marked as ‘support structure required voxels’. Each column (column represents the column of voxel
array along build direction of the part, which is z-axis) is then scanned for support structure required voxels and all
empty voxels below it are counted until filled voxel or the substrate is encountered. The substrate in this case is the
last voxel along a direction opposite to build direction. Finally, all support structure voxels are added and multiplied
by the volume of a single voxel to calculate total support structure volume required for a part.
Figure 3 shows an example of a bracket with an angle criterion for support to be 45° with XY plane. Thus, any
facet with the normal angle more than or equal to 135° with the positive z- axis requires support below it. The left
leg is at an angle of 28° with positive z-axis and thus does not require support, whereas the right leg makes an angle
of 62° with positive z-axis and require a support structure which is represented by voxels centroids.
To calculate the support structure for each
z z part division of HM, the total support structure
required for the part to be fabricated by AM is
calculated. As shown in Figure 3b, the blue
horizontal lines show the division above which
x
z the part is to be additively manufactured. Red
voxel centroids represent the support structure
required for additively manufactured part and
a)) bb) green voxel centroids represent the support
y)
a) x required to maintain the spatial position of
b x
additively manufactured part and its support.
Figure 3. Representation of support structure by voxel centroids: (a) Support
structure for an entire part; (b) Support structure required for split subsections

3.1.4. Volume and height of a part

The volume and height of a part represent the time required for fabrication of an additively manufactured part
[22]. The volume and height of a part can be minimized by different approaches such as redesign, topological
optimization and use of lattice structures.
The height of a part in the build direction is directly calculated from the CAD model. For calculating the volume
of a part, it is voxelized, and all filled voxels are summed up. The volume of the part divisions to be manufactured
by AM is calculated by counting voxels corresponding to appropriate z coordinates of the AM portion of the part
model.

3.2. Subtractive manufacturing parameters

Different subtractive manufacturing operations, such as turning, milling and drilling have specific DFM rules. In
this section, DFM rules for a 3-axis CNC milling machine are considered and detailed methodologies to identify and
calculate the manufacturing constraints are explained. Following constraints are considered for SM: internal corners,
 Anay Joshi and Sam Anand / Procedia Manufacturing 10 (2017) 578 – 589 583

free form surfaces, tool inaccessibility, and machining volume.

3.2.1. Internal corners

Sharp corners occurring inside a slot or pocket or any non-convex corners are difficult to manufacture by CNC
end mill because of its circular shape. Sharp corners and points may also lead to tool breakage [3].
To calculate the number of internal sharp corners, the part is converted to an STL file and it is sliced at a distance
of ‘depth of cut’ defined by the user instead of slice thickness as in AM. First, the non-convex features such as slots
or pockets in a part are identified, because only non-convex sharp angles are difficult to manufacture as shown in
figure 4. Non-convex features in a model are identified by extension of the approach presented by [23]. To find the
non-convex features, sliced points are arranged consecutively to form a contour and the cross products of
consecutive line segments starting from the bottom left point are calculated. If the cross product vector is in a
direction opposite to tool approach direction, then non-convex features are identified. This method is shown in
Figure 5.
Once the non-convex features have been identified, the next step is to check whether the corner angle is sharp. To
do so, the angle between consecutive line segments forming a corner in a slice is calculated. If the angle is within
the threshold, sharp internal corners are identified. Figure 6 shows the contour of an STL slice of a circular and
rectangular cross-section. As seen, the triangular faceted approximation of a circular cross section results in larger
angles between consecutive STL facets than a prismatic shape.
The total number of internal corners in an entire model is calculated by adding internal corners identified in all
slices and the number of internal corners for part divisions are summation of internal corners in slices corresponding
to appropriate SM division.
Cross
product
vectors 160°
Tool path 90°
Tool approach
direction

Figure 6. STL points of circular and rectangular cross

section
Figure 4. 2D contour of one Figure 5. Identification of non-convex
machining pass features

3.2.2. Number of free form surfaces

Machining free form surfaces on a 3-axis CNC machine is time consuming and complex. It may require tool
changes, reorientation and changes in fixtures thus increasing processing time. Therefore, the number of free form
surfaces in a part must be minimized.
In this paper, free form surfaces are detected in a CAD environment (NX is
used as a CAD tool) using NX Open API (Application Program Interface)
CC [24].
Detection of free form surfaces is a three step process. First, equidistant
point grid is formed on each face of a CAD model. Second, face normals at
BB each point are identified and angles between consecutive normals are
calculated. Third, the difference in the angles obtained in second step are
calculated. If this difference in angles are zero, the face is either planar or has
A a constant radius of curvature such as a cylinder or a sphere. Otherwise the
A face is identified as a free form surface. As shown in Figure 7, Faces A and B
are identified as planar and cylindrical surfaces respectively. These are easier
Figure 7. Identification of free form
surfaces
584 Anay Joshi and Sam Anand / Procedia Manufacturing 10 (2017) 578 – 589

to manufacture compared to face C which is identified as a free form surface.

For calculating number of free form surfaces in all SM divisions, faces within the z coordinate corresponding to
particular division are taken into consideration. Therefore, only those surfaces which correspond to SM portion are
detected.

3.2.3. Tool inaccessibility

The tool axis of a 3-axis CNC machine is along z-axis. The faces with surface normal between 0° to 180° with
positive x-axis are accessible for a tool, however, rest of the faces are inaccessible. Undercuts are difficult to
manufacture as it requires re-orientation of the part, possibly increasing the number of fixtures as well as setup and
processing time. Undercuts may also require special tools to manufacture [3]. Internal features are difficult to
manufacture with CNC machines.
To quantify inaccessibility, a CAD model is voxelized and the voxels inaccessible from positive z direction are
identified below which the part is inaccessible. Every column of voxel is scanned along the z direction for a
condition of empty voxel below a filled voxel. If this is identified, the empty voxel is inaccessible to the tool and the
column is identified as an inaccessible column. This is shown in Figure 8 in which red voxels and all voxels below
these are inaccessible from the top (i.e. along tool approach direction). Likewise, every column is scanned for
inaccessibility and the total number of inaccessible columns are counted. The percentage of inaccessibility is
calculated according to equation 1 as follows:

Number _ of _ inaccessible _ columns

% _ inaccessibility × 100 (1)
Total _ number _ of _ columns

To calculate inaccessibility per part division, only the voxels corresponding to z coordinates of the SM portion
are considered.

Figure 8. Methodology to calculate inaccessibility

3.2.4. Machining volume of stock

Machining volume of raw material represents the time required for machining. The entire volume to be machined
can be calculated by summing up all empty voxels. To calculate machining volume per division, the number of
empty voxels corresponding to z coordinates of the SM portion are summed up.

The next section summarizes the methodology and explains the procedure to combine parameter values to
calculate complexity score, which is used to interpret the geometric complexity of a part.

3.3. Complexity score

To begin with, the CAD model is analyzed using appropriate methods such as STL slicing, voxel discretization or
 Anay Joshi and Sam Anand / Procedia Manufacturing 10 (2017) 578 – 589 585

CAD API to calculate complexity score parameters. Table 1 shows methods used to calculate parameters and their
units.
Complexity score parameters are based on DFM rules of additive and subtractive manufacturing processes and
represent manufacturing constraints and time. Three main steps for calculating the complexity score are
normalization of parameters, decision of parameter weights and weighted summation.
Table 1. Summary of methodology.
Start Convert/ Method Parameter Unit
Convert to STL and slice Sharp corners for AM No unit- Integer
Thin regions and gaps for AM No unit- Integer
Internal corners for SM No unit- Integer
CAD Model Convert to STL and voxelize Volume to AM mm3
Support structure for AM mm3
In-accessibility for SM %
Volume to machine (SM) mm3
NXOpen API Free form surfaces for SM No unit- Integer

Complexity score parameters have different units and their values vary from zero (for example, absence of sharp
corners) to a large value (for example, volume of a part to be additively manufactured). This necessitates
normalization of each parameter score by mapping its values between 0 and 1. Each parameter value is divided by
the maximum value of the same parameter which normalizes the parameter values with maximum possible value of
1.
There are different manufacturing constraints for different processes and therefore a common approach for
comparing different parameters is needed. Thus, parameters are grouped together in categories and their weights
from 0 to 1 are decided such that summation of weights is always unity.
In this paper, three categories are considered, namely ‘impossible to manufacture’, ‘difficult to manufacture’ and
‘time-based parameters’ with weights as 0.55, 0.35 and 0.1 respectively. After this step, individual parameter
weights are calculated according to equation 2.

Weight _ of _ the _ group

Weight _ of _ individual _ parameter (2)
Number _ of _ parameters _ in _ a _ group

Table 2 shows the individual weights of parameters used in this paper. The complexity score is calculated by
taking the weighted summation of each parameter.

Groups Weights Parameters Individual weights

Internal corners (SM) 0.183
Impossible to manufacture 0.55 Sharp corners (AM) 0.183
Thin regions and gaps (AM) 0.183
Free form surfaces (SM) 0.117
Difficult to manufacture 0.35 Inaccessibility (SM) 0.117
Support structure (AM) 0.117
Volume to machine (SM) 0.033
Time dependent parameters 0.1 Volume of part (AM) 0.033
Height of part (AM) 0.033

Table 2. Individual weights of complexity score parameters.

If there are ‘n’ parameters, ‘m’ divisions, and ‘P’ represents the absolute value of the parameter, then the
complexity score for ith division can be calculated according to equation 3 as:
586 Anay Joshi and Sam Anand / Procedia Manufacturing 10 (2017) 578 – 589

n
Pij
Complexity _ Score _ i ¦Wj × max (P1 j, P2 j, P3 j,..., Pmj )
j 1
(3)

where,
i = ith division Wj = weight of the jth parameter
j = jth parameter Pij = parameter value

4. Results

This section explains the systematic application of the decision-making process plan described earlier on three
different case studies. The steps of the algorithm are as follows:

x Generate CAD model of a part to be manufactured.

x Decide the number of divisions along planes perpendicular to build direction / tool axis.
x Identify and quantify DFM/DFAM parameters for each division of a part.
x Calculate complexity scores for AM, SM and every combination of HM.
x Plot complexity scores along divisions and identify the optimal manufacturing process.

4.1. Test case 1

Generation of a CAD model: In this case study, a CAD model of an impeller was analyzed. Traditional
manufacturing processes for an impeller are casting or milling. Analyzing its geometric complexity for additive
manufacturing and 3-axis milling process leads to the final process plan. This model, with a height of 67 mm and a
base diameter of 140 mm, was adapted from Grabcad [25].
Decision on the number of divisions of a CAD model: The model was divided as shown in Figure 9. The analysis
was performed assuming the following processes: Figure 9a to be manufactured by AM only. Figures 9b-e to be
manufactured by HM. Figure 9f to be manufactured by SM only. For HM, the raw material is first machined to
realize the lower portion of the split followed by additive deposition of the top portion.
Identification and quantification of complexity score parameters: The following constraints were identified in the
model for subtractive manufacturing based on algorithms described earlier. Fins are free form surfaces and the lower
portion of fins are inaccessible for a 3 axis CNC machine. Several internal corners were identified for CNC
machining. The only constraint for AM identified was the requirement of support structures for fins. There were no
sharp corners, thin regions and thin gaps for AM. Figure 10 shows support structures required for AM (angle
criterion of 125°), internal corners detected for SM, and tool inaccessibility for SM. Table 3 shows absolute values
of parameters for the impeller.

a) b) c)
a) b) c) d) e) f) Figure 10. (a) Support structure for AM; (b) Internal corners
for SM; (c) Inaccessibility for SM
Figure 9. Divisions of a CAD model of an impeller

Table 3. Absolute complexity score parameter values for a CAD model of an impeller.
Subtractive Manufacturing Additive Manufacturing
 Anay Joshi and Sam Anand / Procedia Manufacturing 10 (2017) 578 – 589 587

Number Support Volume Height

Number of of Volume to Number Number of structure of part of AM
Inacces- free form internal machine of sharp thin volume mm3 mm3 part
Division Ibility- % surfaces corners mm3 (*103) corners features (*103) (*103) mm
0= entire AM 0 0 78.4 189.3 67.8
1 5.6 30 117 88.9 0 0 83.4 113.2 54.3
2 16.6 30 296 281.3 0 0 73.5 63.6 40.7
3 27.2 30 480 492.3 0 0 16.7 33.7 27.1
4 33.9 30 675 737.1 0 0 0.0 13.2 13.6
5=entire SM 35.9 32 860 995.2

Plotting the graph of complexity scores: Complexity score was

calculated and plotted (Figure 11) for each division. Complexity
score of 0.176 was the lowest for AM, resulting in a decision to
manufacture this component entirely by AM rather than SM or
HM.

Figure 11. Complexity scores for test case 1

4.2. Test case 2

Generation of a CAD model: This case study used the CAD model shown in Figure 12 for analysis.
Decision on the number of divisions of a CAD model: The model was divided into 5 split sections and each
section was individually analysed.
Identification and quantification of complexity score parameters: Several thin regions and sharp corners were
identified for AM. Practically, there was no geometric complexity for a 3-axis CNC process due to the absence of
internal corners, inaccessible regions and free form surfaces.
Plotting the graph of complexity scores: As expected, the complexity score of 0.033 was the lowest for SM and
led to the decision to manufacture this component entirely by SM. The graph of complexity scores is shown in
Figure 13.

4.3. Test case 3

Generation of a CAD model: This test case analysed a CAD model of a faucet shown in Figure 14.
Decision on the number of divisions of a CAD model: The model was divided into 10 split sections and these
sections were analysed individually.
Identification and quantification of complexity score parameters: This model exhibited geometric complexity for
both AM and SM. The constraint for AM was the requirement of support structures and SM constraints included
tool inaccessibility, large machining volume and presence of free form surfaces.
Plotting the graph of complexity scores: As seen from Figure 15, the complexity score of 0.166 was the lowest
for HM with the first division. That means the fabrication should be started with rectangular shaped stock followed
by CNC machining to form a part till the first division. The rest of the part is to be manufactured by additive
manufacturing. The process flow of the manufacturing sequence for this part is shown in Figure 16.

Figure 12. CAD model for test Figure 13. Complexity scores for test case 2
case 2
588 Anay Joshi and Sam Anand / Procedia Manufacturing 10 (2017) 578 – 589

Figure 14. CAD model of Figure 15. Complexity scores of test case 3
test case 3

Figure 16. Process plan for manufacturing a faucet with hybrid manufacturing

5. Conclusions and future work

This paper provides an initial approach to automate the decision making process for additive, subtractive, and
hybrid manufacturing processes. A systematic approach was developed by assessing complexity scores which
consisted of the parameters representing geometric complexity and manufacturing time. To calculate these
parameters, different approaches were adopted, such as STL slicing, voxelization, and the use of NX Open API.
Case studies for AM, SM, and HM were demonstrated for validating the research in identifying optimal process
selection and sequence. The method provides the flexibility of changing the complexity parameter weights
according to user preference and available manufacturing facilities.
The methodology used in this paper can be extended to include more DFM/DFAM rules and manufacturing
constraints such as geometric tolerances for decision making. Further research on orientation and redesign steps to
make the process plan dynamic is under way. Investigation regarding optimal and automatic number of divisions
based on complexity scores can be undertaken. Multi axis CNC and additive machines can also be considered for
which some complexity parameters will have to be modified, such as multi-axis slicing for additive manufacturing
followed by support structure generation for each axis and collision detection of print head. Feature-wise
decomposition, feature-wise complexity score calculation, and inclusion of process parameters can help increase
current research’s practical applicability.

References

[1] G. Boothroyd, Product design for manufacture and assembly. Computer-Aided Design, 26 (1994) 505-520.
[2] P. Kulkarni, D. Dutta, An accurate slicing procedure for layered manufacturing. Computer-Aided Design, 28 (1996) 683-697.
[3] J. G. Bralla, Design for manufacturability handbook, second ed., McGraw-Hill, New York, 1999.
 Anay Joshi and Sam Anand / Procedia Manufacturing 10 (2017) 578 – 589 589

[4] J. M. Flynn, A. Shokrani, S. T. Newman, V. Dhokia, Hybrid additive and subtractive machine tools–Research and industrial developments.
International Journal of Machine Tools and Manufacture, 101 (2016) 79-101.
[5] Z. Zhu, V. Dhokia, S. Newman, A novel process planning approach for hybrid manufacturing consisting of additive, subtractive and
inspection processes. In IEEE International Conference on Industrial Engineering and Engineering Management (2012) 1617-1621.
[6] NX Hybrid Additive Manufacturing [online], n.d. NX CAM. Available at: https://www.plm.automation.siemens.com/en_us/products/nx/for-
manufacturing/cam/hybrid-additive-manufacturing.shtml.
[7] J. Jeng, M. Lin, Mold fabrication and modification using hybrid processes of selective laser cladding and milling. Journal of Materials
Processing Technology, 110 (2001) 98-103.
[8] K. P. Karunakaran, S. Suryakumar, V. Pushpa, S. Akula, Low cost integration of additive and subtractive processes for hybrid layered
manufacturing. Robotics and Computer-Integrated Manufacturing, 26 (2010) 490-499.
[9] G. Manogharan, R. Wysk, O. Harrysson, R. Aman, AIMS – A metal additive-hybrid manufacturing system: System architecture and
attributes. In Procedia Manufacturing, 43rd North American Manufacturing Research Conference, 1 (2015) 273-286.
[10] M. Lutter-Günther, S. Wagner, C. Seidel, G. Reinhart, Economic and ecological evaluation of hybrid additive manufacturing technologies
based on the combination of laser metal deposition and CNC machining. Applied Mechanics and Materials, 805 (2015) 213-222.
[11] G. Manogharan, R. A. Wysk, O. L. Harrysson, Additive manufacturing–integrated hybrid manufacturing and subtractive processes:
Economic model and analysis. International Journal of Computer Integrated Manufacturing, 29 (2016) 473-488.
[12] M. Chiu, G. Okudan, An investigation of the applicability of DfX tools during design concept evolution. Journal of Product Development, to
be published, 13 (2011) 132-167.
[13] R. Ranjan, R. Samant, S. Anand, Design for manufacturability in additive manufacturing using a graph based approach. In ASME
International Manufacturing Science and Engineering Conference (2015) V001T02A069-V001T02A069.
[14] R. Ranjan, R. Samant, S. Anand, Integration of design for manufacturing methods with topology optimization in additive manufacturing.
Journal of Manufacturing Science and Engineering, (2016).
[15] G. A. Adam, D. Zimmer, Design for additive Manufacturing—Element transitions and aggregated structures. CIRP Journal of
Manufacturing Science and Technology, 7 (2014) 20-28.
[16] S. Nelaturi, V. Shapiro, Representation and analysis of additively manufactured parts. Computer-Aided Design, 67 (2015) 13-23.
[17] O. Kerbrat, P. Mognol, J. Hascoët, A new DFM approach to combine machining and additive manufacturing. Computers in Industry, 62
(2011) 684-692.
[18] B. Ramon, Tool and manufacturing engineers handbook, Volume VI, Design for manufacturability, Society of Manufacturing Engineers,
1992.
[19] M. Eragubi, Slicing 3d cad model in stl format and laser path generation. International Journal of Innovation, Management and Technology,
4 (2013) 410.
[20] C. C. Wang, Y. Leung, Y. Chen, Solid modeling of polyhedral objects by layered depth-normal images on the GPU. Computer-Aided
Design, 42 (2010) 535-544.
[21] R. Paul, S. Anand, Optimization of layered manufacturing process for reducing form errors with minimal support structures. Journal of
Manufacturing Systems, 36 (2015) 231-243.
[22] P. Kulkarni, A. Marsan, D. Dutta, A review of process planning techniques in layered manufacturing. Rapid Prototyping Journal, 6 (2000)
18-35.
[23] S. Joshi, N. N. Vissa, T. Chang , Expert process planning system with solid model interface. The International Journal of Production
Research, 26 (1988) 863-885.
[24] NX Open [online], n.d. NX Open. Available at: https://docs.plm.automation.siemens.com/tdoc/nx/10/nx_api/#uid:index.
[25] GrabCad [online], Grabcad. Available at: https://grabcad.com/library/mechanical-part-34.