Segmented Object Manufacturing

brontidegrrrΜηχανική

14 Νοε 2013 (πριν από 3 χρόνια και 6 μήνες)

60 εμφανίσεις

Segmented Object Manufacturing



K.P. Karunakaran,
Saurabh Agrawal,
Pankaj D
.

Vengurlekar
,

Onkar S. Sahasrabudhe
,
Vishal Pushpa

and Ronald H. Ely


Computer Graphics Laboratory, Department of Mechanical Engineering

Indian Institute of technology Bombay, Po
wai, Mumbai 400076, INDIA


Rapid Prototyping (RP)

is a technology based on “divide
-
and
-
conquer” strategy that
enables
automatic physical realization of a design without any special tooling. However, e
xisting RP processes
suffer from staircase defect as the
y
are all based on 2.5 axis kinematics
.

T
o minimize the
error
due to

staircase defect,
they build the parts from
v
ery thin layers

of 0.010mm to 0.300mm thickness
.
Therefore, hundreds of layers are required to produce
typical

object
s

making RP

a slow and co
stly
process.
To overcome these limitations, a new RP process called
Segment
ed
Object
Manufacturing

(
SO
M)

is

proposed
in this paper.
SO
M
make
s

use of 3 axis kinematics in conjunction with a novel
slicing method
.
Slicing in
SO
M is based on
certain visibilit
y considerations

and is independent of the
part accuracy.

As there will be only a few
thick
layers

in
SO
M
, the part

can be produced
faster and
cheaper with
accuracy

comparable to that of CNC machining
.


Keywords
:

Rapid Prototyping (RP),
Computer Numerical
Control

(CNC)
, Slicing,
Visibility
, Large
Prototyping
, Evaporative Pattern Casting

(EPC)



1.

INTRODUCTION


Rapid Prototyping (RP)
is the process of manufacturing physical objects in a layer by layer
manner directly from their CAD models without any human

intervention and without the use
of any tools, dies or fixtures specific to the geometry of the objects being produced.
RP,
which is based on “divide and conquer” strategy, is known more appropriately by the term
Layered Manufacturing (LM)
. The
principle
of RP is illustrated in Figure 1.

The CAD model
of the object shown in Figure 1a is sliced by parallel planes. The edges of these slices (Figure
1b) are squared (Figure 1c). A complex 3D object is thus decomposed into several 2D objects
or slices which are

simple to manufacture. These slices are physically realized in one of
several ways, stacked and joined together as shown in Figure 1d to get the physical object. It
is obvious from Figure 1e that the staircase effect limits the accuracy of the prototype.
For
better surface finish, the object can be polished (Figure 1f).
Stereo
-
lithography (SLA)
,


2

Laminated Object Manufacturing (LOM)
,
Fused Deposition Modeling (FDM)
,
Selective
Laser Sintering (SLS)

and
3D Printing

(3DP)

are some of the most popular RP proces
ses.


Since t
he 3D object is approximated into several 2D slices in RP
, the tool movement is
either along Z axis or in XY plane

during th
e course of building the object; in other words, all
these

popular RP processes are based on 2.5 axis kinematics. Such

a system will invariably
have staircase defect as s
een

in Figure 1e. In order to bring the size of the stair
-
steps within
acceptable accuracy, very thin slices have to be used. These slice thicknesses are as thin as
0.010mm (for patterns of jewelry) and g
enerally not more than 0.300mm. Therefore,
hundreds of layers are required to produce any object making RP a very costly and slow
process. Because of this high cost and time, RP is still not viable for many applications in
spite of being completely automat
ic “from art to part”.





Figure
1
Principle of Layered Manufacturing




The
slices of
all commercially available RP machines are of
uniform thickness
,

have
their edge surfaces vertical

and their bottom and top contours are

the same
. This type of

3

slicing is called
uniform slicing of 0
th

order edge

surface

(
Yan and Gu, 1996; Karunakaran et
al.
,

2000
a
)
. As the number of slices is very high in these conventional RP machines owing to
very thin slices, researchers have been explo
ring various ways to reduce it. This led to the
proposals for
adaptive slicing

by several researchers. Adaptive slicing will result in less
number of slices than uniform slicing for the same accuracy. In

adaptive slicing,
the slice
thickness at any locatio
n depends on the local geometry, particularly, normal and curvature.
Furthermore, in addition to 0
th

order edge surfaces, researchers have considered the use of 1
st

order, 2
nd

order or even higher order edge surfaces as illustrated in Figure 2; the 1
st

ord
er edge
will be a ruled surface, the 2
nd

order edge will be a quadratic surface and so on. For a given
accuracy required, higher the order of edge surface
,

less will be the number of slices
(Karunakaran et al. 2000a)
.



Order of Approximation

for the Edge

Surface

Zeroth

First

Higher

Uniform

Slicing




Adaptive

Slicing




Figure
2


Various
S
licing
M
ethods



Kulkarni and Dutta have presented an adaptive slicing method for 0
th

order edge surface
(
Li Hui

et al., 1998)
. Although adaptive slicing al
gorithms for 0
th

order edge surface usually
consider vertical or squared edge surfaces, in some processes such as

FDM

they are actually

parabolic. More r
ecently Pandey et al. have used this fact in their algorithm for uniform
slicing for FDM; they report l
ess number of layers for the same accuracy
(Pandey et al.,
2003)
. Tyberg and Bohn have proposed
local adaptive slicing

in which they arrive at
different slice thicknesses for different features at the same height
(Tyberg and Bohn, 1999)
.

4

They have demonstr
ated its implementation for FDM process.

Karunakaran et al. have used
adaptive slicing with 1
st

order edge in a
H
ybrid
Layered Manufacturing

(H
LM
)

process for
metallic dies and molds
(Karunakaran et al. 2000a)
.
Similarly Taylor et al. also have used
adapti
ve slicing with 1
st

order edge for making plastic and wooden parts in a process called
Solvent Welding Freeform Fabrication Technique (SWIFT)
.
T
hey
apparently
realize the ruled
edge surfaces using
a flat end mill in 5
-
axis machining

(Taylor et al., 2001)
.



Horváth et al. have reported use of
an advanced
adaptive slicing
method
for
manufacturing
expanded
polystyrene
(thermocole)
prototypes
(Horváth et al., 2002;

Broek et
al., 2002)
.
They have used a
thin
hot
blade

elastically
buckled
into a spline by compr
essing it

at the ends
. The spline shape can be varied by varying the pressure at the ends and it can be
oriented such that the vertical plane containing the spline is always normal to the edge
surface.

The
edge
surface
produced by this blade on the thermoc
ole sheet
is the sweep of this
variable spline which can be upto
cubic

degree

(4
th

order)
.

Due to this,
t
hey could
achieve
fairly thick layers

and hence they call their process as
Thick
-
Layered Manufacturing

(TLM)
.
However,
the edge surface is
actually an
approximation of the
desired
surface as it is a
collection of the closest matching splines; TLM cannot produce internal features, requires a 6
axis kinematics and is not yet fully automatic. Furthermore,
the layer thicknesses are likely to
be higher and mo
re accurate than TLM

in
Shape Deposition Modeling (SDM)
,
Hybrid RP

(
HRP
)

and
Segmented Object Manufacturing (SOM)

proposed in this paper

(
Krishnan
Ramaswamy, 1997;
Sangkyun Kang

et al.
,
2003;

Junghoon Hur

et al.
,
2002
)
.




The
objects built using the
above

slicing methods in which the order of the edge surfaces
have
a
finite limit, be it uniform or adaptive slicing,
inherently produce only approximation
s

of the object
s
.
If and only if the order of the edge is same as that of the original model,
it
is
possib
le

to obtain

the exact shape physically
. In such a situation, splitting of the
3
D object
will no

longer

be dictated by accuracy but by manufacturability considerations
. This situation
also
leads to
thick slices.
Some processes having this characteristic ar
e reviewed in the next
section and the proposed SOM process is introduced.
This is followed by the description of
visibility
-
based slicing which is fundamental to SOM process.
Finally
a hypothetical
machine
to realize
SOM
along with

its operating steps

is
presented
. The paper ends with a fairly
complex

illustration and conclusions.




5

2.

SEGMENTED OBJECT MANUFACTURING


Manufacturing
objects in thick layers
has been the dream of researchers for quite so
metime.
Probably the
first attempt towards this goal

was
in
SDM

process
(Krishnan Ramaswamy,
1997)
.

SDM makes use
of two deposition heads, one each for depositing
model
material
and
a suitable
support

material
. The
slices of the
object
are

obtained by
split
ting it

wherever
its

normal just becomes horizontal
, i.e
., wher
ever its Z component changes
its
sign
.

To that
extent, SDM also uses visibility considerations for slicing.

In any slice, the normal
s

of the
object may be upward or downward. In all regions of the slice where the normal
s

are

downward, support will b
e required
there
and hence the support head
will
deposit material
filling those regions

extending to the bounding box of the object
. Since
any such

deposition is
only near
-
net
, machining is used to finish it.
This
will be

followed by the deposition of

mode
l
material

and
again finish
ing it

using machining. Thus each slice is built by deposition and
machining of support and model materials

alternately
until

the entire slab of the slice is
complete
.

E
ssentially,
the previous region(s) deposited and machined ac
t as mold cavities to
hold the subsequent depositions.

Mold

SDM

is a variant of SDM to create fugitive mold
cavities out of wax; it uses a sacrificial photopolymer as support material. This mold is used
for making tiny casting such as miniature turbine fro
m ceramic gel or polymer
(
Sangkyun
Kang

et al.,
2003)
.
In SDM, s
licing and the subsequent process planning to determine
(i) the
various regions for any layer, (ii)
the order in which these regions are
to be
deposited and
(iii)
the tool path for deposition
and machining of each of these regions
,

are
all
too involved
.

SDM requires a 5 axis kinematics.
The focus of research of the SDM team seems to be on the
manufacture of miniature parts and
multi
-
material and embedded structures

and it is not clear
whether f
ull automation has been achieved
.

In
SOM
proposed
by
the authors, use of support
material is avoided by performing machining from both sides of the
stock
. Furthermore, they
are able to manage with 3 axis machining instead of the costly and complex 5 axis
m
achining.




The research group of K.
Lee has proposed a
H
ybrid
RP

(HRP)

process which
also
aims
at building objects with minimum number of slices

(
Junghoon Hur

et al.
,
2002)
.
They first
identify
and separate machinable
features and suppress them. The resu
lting geometry is only
sliced

for HRP
.
Each slice which is quite thick is
buil
t

through

the
near
-
net material
deposition and net
-
shape machining.
Although this process claims to produce objects with
minimum number of slices, it requires
fair amount of
user

input to determine

the machinable

6

features and
the levels at which slicing is to be done
. Therefore, it still has a long way before
evolving into an automatic system.



While the interlayer bonding in SDM and HRP is through fusion at the interface, it is

to
be achieved using adhesive spray in SOM.
The slices in all the RP processes discussed so far
including SDM and HRP have horizontal top and bottom surfaces; however, this may not be
the case in SOM enabling it to use even higher layer thicknesses.
It is

interesting to note that
the slicing methods used in
SDM
, HRP and SOM belong to the category of adaptive slicing
wherein the order of edge is same as that of the object. All three
processes
make use of the
concept of visibility directly or indirectly. Fur
thermore, due to the use of thick slices, these
process
es

make use of machining extensively for finishing.

In other words, these emerging
processes deviate from the traditional additive or generative approach to a hybrid approach to
make the best use of bo
th.
Several researchers have addressed the machining of these layers
or machining a surface in layers
(
Chen and Song, 2001;
Li Hui

et al., 1998; Taylor et al.,
2001)
.



Any RP process
in general
is based on “Divide and conquer” strategy. In the beginning

of
RP
, the segmentation was for treating a 3D object
as

a set of 2D slices
of uniform thickness
which
a
re

easy to manufacture. Since this require
d

too many slices, higher order slicing
methods were proposed with suitable methods of manufacturing the increa
singly complex
slices.
In all such strategies, the slices have horizontal top and bottom surfaces; however,
SOM proposed in this paper make
s

use of slices which may not have this charact
eristic
. In
fact,
it is more appropriate to call
the resulting
pieces
as
segment
s

rather than
slice
s
.

Similar
segmentation
approaches
can be observed in a few other applications.
“100 day engine
project” carried out by
Ford
is one such example
(
Ion Gibson
,

1997
)
. In order to reduce the
engine development time, they split the

engine casting into slices of appropriate thicknesses
manually; these slices were machined and then joined by brazing.
Another example is
Space
Puzzle Molding

process from Protoform of Germany which can automatically design the
injection molding dies of v
ery complex objects in pieces that constitute the die halves and
inserts

(Protoform, 2004;

Enimco,
2004
)
.

These pieces fit together in a special frame like a
3D jigsaw puzzle.
Molds are manually assembled and disassembled during each shot.
CNC is
typically

used to make the puzzle elements; however, direct m
etal RP processes such as SLS
and

indirect processes such as
Direct AIM
, an aluminum
-
filled epoxy casting process, have
also been used

to make these pieces
.
Chen and Rosen also have proposed a method of

7

a
utomatically obtaining the injection mold in pieces

(Chen and Rosen, 2001 & 2003)

from
the CAD model of the plastic object
.

They make use of the concept of visibility extensively

(
Woo,
1994)
.
Karunakaran et. al. ha
ve

developed a software program called
Opt
iLOM

which
eliminates grid cutting and decubing operation
s

in LOM
-
RP
(Karunakaran et al., 2000b &
2002)
.
In order to extract the LOM prototype from inside a box,
OptiLOM splits
the material
inside and
surrounding the object into the minimum number of extra
ctable pieces
;

w
hen
the
combined STL file of all
the
se

pieces
and the object
are
made in LOM machine
,

there will be
no grid cutting and decubing.

The stock halves and the plugs calculated by OptiLOM
essentially are the mold halves and inserts.
Therefore, O
ptiLOM also can be used to design
the injection molds.
OptiLOM has been available as an optional module of
Magics RP

since
2002.

While the above three works aim at obtaining the molds of an object, albeit in pieces,
SOM aims at splitting the object
itself
into segments each of which satisfy certain
manufacturability
criteria
, viz
.
, cutter access to the entire
surface

of the segment either from
top or bottom
.

Interestingly, Dongwoo and K. Lee too have addressed the problem of
splitting an object such as a st
amping die into pieces machinable from two opposite directions
(Dongwoo Ki and Kunwoo Lee, 2002)
. However, they aim at splitting the object into a
minimum number of such machinable pieces so that they can be machined individually and
then glued together
; t
he pair of machining directions corresponding to each piece could be
different in their method.
While the goal in the present work of the authors also is to split the
object into pieces machinable from two opposite directions, they restrict this machining
direction along Z axis so as to enable the development of a completely automated process

involving slicing, machining and gluing
.



Huge sheet metal stamping dies used in automotive industry have very complex geometry
with a lot of hollowness and ribs as
they have to be light for ease of handling. These are
machined from their near
-
net castings obtained from
Lost Foam Casting

process, also known
as
Evaporative Pattern Casting

or
Full
-
Mold Casting
. These lost foam patterns are made of
expanded polystyrene
(
commercially known as
thermocole
)

with adequate allowances on the
forming surfaces. As these patterns are complex, these are presently segmented
manually
into
machinable shapes arbitrarily. The number of pieces that constitute any such pattern and the
amou
nt of machining will depend on the skill of the person who segments it. The genesis of
our present research was a desire to develop an automatic process to manufacture these
patterns.
After considering
different adaptive slicing methods of various levels o
f
approximations of the slice edge surfaces (starting from ruled surface approximation)
, the

8

authors have f
inally come up with what is called
Visibility Slicing (
V
-
slicing
)

in which

no
surface approximation takes place. In fact, V
-
slicing is essent
ially a
process
of segmenting an
object geometry into pieces having certain machinability characteristics.



The
commercially available
RP processes discussed so far
are based on 2.5 axis
kinematics; SDM and TLM require 5 axis and 6 axis kinematics respectively.

A
s against
th
e
s
e

RP

processes
, the proposed
Segmented Object Manufacturing (SOM)

will
require only
a
3 axis kinematics in conjunction with
this

novel slicing method that is not dictated by the part
accuracy but is based on certain visibility considerations.

Note that the top and bottom
surfaces of any slice of SOM need not be horizontal unlike in the conventional RP systems.
The slice thickness in the proposed SOM will be very high and it will have no staircase
defects. As there will be only a few slices or
segments, the parts can be produced at low cost
quickly with its accuracy comparable to that of CNC machining.






(b) V
-
slices

(c) V
-
slices



(a) Object

(
d
) V
-
slices

(e
) V
-
slices

and hori. Levels

Figure
3


Illustration of
V
-
slicing



9

3
.
VISIB
ILITY
-
BASED SLICING


In the
conventional
slicing strategies
,

the slice thickness and the part
accuracy
are closely
related
. As against this, visibility is used as the criteria for determining the
slice

thickness,
(
segment geometry
, to be more precise),
in
the proposed
Visibility
-
based Slicing (V
-
slicing)
.
The object is split into
visib
le

slices (v
-
slices)
, also known as
segments
.

The intersection of
any

vertical ray
with the v
-
slice
will
be always a pair of points. When the faces encountered
by the ray happ
en to be vertical, one gets a line segment as intersection in which case the end
points of this line segment can be treated as the pair of intersection points.
This characteristic
of v
-
slice ensures its machinability by a
vertical
cutter from two opposite
directions.

Figure 3
illustrates the concept of v
-
slicing for an object shown in Figure 3a.


An object need not have a unique set of v
-
slices and hence some more variants are
possible
as

shown in Figures 3b
-
e
.
Figures 3b & 3c are the two possible sets of v
-
slices. T
he
raw material used for realizing the
se

v
-
slices will be equal but
Figure 3
d

will
require the least
amount of raw material.

Therefore, after obtaining the v
-
slices, a post
-
processing is done to
transfer materials among these v
-
slices so as to mi
nimize the total raw material requirement.




(a) Bounding box of the object in 1
st

setting

(b) Blank at the end of 1
st

setting



(c) Blank at the end of 2
nd

setting

(d) Blank at the end of 3
rd

setting

Figure
4


Setting
s

R
equired in CNC
M
achining
for the
S
ame
O
bject



10



(a) Examples of visible and invisible faces

(b)
Prism obtained by extruding the face upward



(
c
)
Invisible patch
p
I
obtained by recursively
collecting the invisible faces

(
d
)
Solid
1
S
obtained by extruding the invisible
patch
p
I
till the bottom of the bounding box of
the manifold solid
org
S



(
e
)
Segment resulting from
(
org
S

-

1
S
)

(
f
)
Segment resulting from
(
org
S

1
S
)

Figure
5


Algorithm for Visible Slicing


The proposed v
-
slices can be correlated with the number of setups required in CNC
machining to produce the object by scooping out material from its bounding box. Figure 4a
shows the
blank of this object in 1
st

setting. Figure 4b shows the blank at the end of 1
st

setting. After reversing the object, the remaining surfaces are machined except the eye
-
end
hole (Figure 4c). Machining this hole requires a separate setting as shown in Figur
e 4d.
Therefore, CNC machining, which is purely a subtractive process, requires three settings to
make this piece from a blank. The same object can be made through SOM in just two V
-
slices (Figure 4d), each requiring machining from top as well as bottom; h
ence this object
Invisible face

Visible face


11

will have essentially four settings in SOM. SOM has one setting more since the eye end hole
is realized in two settings of different layers.


If the slicing is accurate enough, the horizontal surfaces of the object can be obtained
during
the slicing operation itself whereas the non
-
horizontal surfaces will require machining
in scan milling. Therefore, after obtaining the set of v
-
slices that have the least heights, the
authors prefer to split them further if any of the slices have large ho
rizontal surfaces.
Accordingly, the preferred set of slices for this object will be the one shown in Figure 3e.
This is obtained from Figure 3d by splitting the bottom slice at its horizontal surface.



Algorithm

for Visibility Slicing


A face of the solid

will be called
invisible face

if
(i)
its normal is upward and
(ii)
it is
shadowed by
its
other faces; otherwise, it will be called a
visible face
.
These are illustrated in
Figure 5
a
.

A contiguous set of invisible faces is called
invisible patch
.
While the

segmenting
can proceed in top
-
down or bottom
-
up manner, the building will happen in a bottom
-
up
manner only. We have chosen segmentation in the top
-
down manner in this algorithm.
Let
S
be the set of v
-
slices or segments.

The follo
wing algorithm
s

will convert the object
O
into
the set of v
-
slices
S
:


Algorithm 1

Algorithm for Determining the V
-
Slices

Initialize
S

with
O
.


For each

member of
S
, say

i
S
,

{

status =
Segment(
i
S
,
segments
S
);


If status = true, then continue

as
i
S

is already a v
-
slice;

Remove
i
S
from
S
and add its segments
segments
S

at the end of
S
;

}



The above algorithm produces v
-
slices but they could be more in number with the
possibility of combining some of th
e segments into one segment without affecting the
visibility.
This post
-
processing is done by the following algorithm:




12

Algorithm
2

Algorithm for Post
-
Processing to Combine V
-
Slices Wherever Possible

For each member of
S
, say

i
S
,

{

For each member of
S
, say

j
S
,

{

Continue if i = j;

Continue if
i
S

and
j
S

do not overlap along Z direction;

new
S

=
i
S

U

j
S
;

status =
Segment(
new
S
,
segments
S
);


If status = true, // This means that
new
S

is a v
-
slice

{

Replace
i
S

by
new
S
;

Remove
j
S

from
S
;

}

}


The following function
Segment

takes a manifold solid
org
S

as input. If
org
S
is already a
v
-
slice, i
t returns “
status = true

;

otherwise, it returns

status = false


and also

calculates the segments
segments
S
of the

original

solid
org
S
. Note that
segments
S
will be an array of
manifold solids but th
ese
may or may

not be v
-
slices.


Function 1

Function to split the given solid
org
S

into its segments
segments
S


status
Segment (
org
S
,
segments
S
)

{

Status = false; // Initially

assume

that
org
S
is not

a v
-
slice.


Step

1: Identifying the first invisible face:

----
-----
------------------------------------

For every face
i
F

of the
input manifold solid

org
S
,

{

Let

'
i
F

be the projection of
i
F

on the top of the bounding box of
org
S
. Make an extruded solid
P

between
i
F

and
'
i
F

(see Figure 5b)
.



For every face
j
F

of
org
S
,

{


If
(i=j), Continue;


If (
j
F

is below
i
F
), Continue;

If
j
F

intersects
P
,
break

this loop since
j
F

is the first
invisi
ble face;

}


if (j > number of faces of
org
S
),


13

{

Status = true; // Declare that the input solid as a v
-
slice.

Return from this function
since the
object is
already
a v
-
slice;

}

}


Step

2: Recursively growing the first
invisi
ble face

j
F

into an
invisi
ble patch

p
I
:

----------
---------------
---------------
--------
-------
----
-----

Initialize the invisibl
e patch
p
I
with
j
F
;

while (true)

{


For each of the three neighboring faces of
j
F
, say
i
F
,


{

For every face
k
F

of
org
S
,

{

If
k
F

is not same as
i
F
, continue;

If
k
F

lies outside the X and Y extents of
i
F
, continue;


If
k
F

is below
i
F
, continue;

If the projections of
i
F

and
k
F

in XY plane intersect

{

add face
i
F

to
the
invisi
ble patch,
p
I
;

Set
j
F

=
i
F
;

}

}

}

If no
ne of the three

i
F

is added to
p
I
, break
the while loop as

constructi
on

of
th
e

invisi
ble patch
p
I

is complete

(see Figure
5
c
)
;

}


Step

3: Ob
taining the segments
segments
S

from the
invisi
ble patch:

-
----
----
------------------
------------------------
------------

Make a solid
1
S

by extruding
p
I

till the bottom of the bounding box

of
org
S

(see Figure 5
d
)
.


Calculate (
org
S

-

1
S
) and (
org
S


1
S
).
These are shown in Figures 5e
and 5f.
These two solids are two segments

of
org
S
.


If these are non
-
manifold solids, split them into manifold solids.
A
ll these manifold
solids
will
be

returned as
segments
S
.

Note that all
the elements of
S
need not be v
-
slices.

No
te also that
1
S
and
(
org
S


1
S
)

are same in the illustration of Figure 5

(d and f)
; however, this
may not be the case always.




14

4
.

A HYPOTHETICAL MACHINE FOR
SEGMENTED OBJECT

M
ANUFACTURING


A
hypothetical
S
OM

machine is
presented
in this section using w
hich

these
v
-
slices can be
physically realized, stacked and joined to get the object.
This machine will be suitable for
making objects out of soft materials such as expanded polystyrene, polyurethane or synthetic
wood

automatically
.
Its schematic diagrams
are
shown in Figures 6 and 7. Table 1 is at the
bottom and is fixed. Table 2 is at the top and it can be dropped smoothly on Table 1
under the
influence of a suitable

counterweight and retracted

as and when required
.
The lowering and
retraction
of Table 2
are
switching functions.
So this action can be implemented using the
miscellaneous code (M code)

of a CNC system.
Between these two tables, there is a frame
which can slide up and down.
This slide

will
be referred

as
vertical slide

or
Z axis
. A XY
stage an
d a slicing unit are fitted onto this frame. The slicing cutter could be a horizontal hot
wire or a band saw.
If the raw material is expanded polystyrene, then hot wire will be the
ideal tool for slicing it. If objects are to be made of relatively harder m
aterials such as
polyurethane or synthetic wood, then band saw will be the suitable tool for slicing.
The two
extreme positions of the slicing cutter are shown in Figure 7.
The shuttling of this horizontal
slicing cutter is a toggle function which also can

be implemented using the

M code
of a CNC
system.
These two units can be moved up and down together by the Z axis

control
. The
working head of the
XY
stage

carries a double
-
sided end
-
milling cutter
(shown in Figure 8)
and a glue spraying nozzle
. This cutte
r is operated by a pneumatic turbine.
The tool head
could as well be a turret with multiple tools.




Table 2 (can drop by gravity and
can be retracted)




Hot wire C
h

for slicing





Vertical rotary end milling cutter
C
v

with motions

possible in X, Y
and Z directions


Table 1 (fixed)


Figure
6

Proposed RP
M
achine for
P
olystyrene
O
bjects



15

Each
v
-
slice is realized using the following steps
. These are also illustrated in Figure 9
.
Note that the Z coordinates of the

free surfaces of the
material stocks

on both tables at their
ex
treme positions at any time are known and their

changes are kept track of either through
software control or with the help of a touch sensor similar
to

a popular RP process called
Laminated Ob
ject Manufacturing (LOM)
.


Step
0

Table 2 is at the top most position.
Table 1 and Table 2 have vacuum clamps to hold
the stock. The stock is mounted on Table 2 using vacuum. Vacuum in Table 2 is off in
the beginning.


Step 1

The vertical cutter has two cu
tting tips as shown in
F
igure

8
. The top

tip

of
the
vertical cutter scan
-
mills all the surfaces
of the first v
-
slice
visible to it on Table 2.








Raw material mounted on Table 2.






Pneumatically driven end mill capable of moving in all 3
dire
ctions


Slicing wire moving together with the XY stage up and down. It
can in addition shuttle horizontally to effect slicing. Dotted line
shows its position at the other extreme.


Table 1. This is fixed.


Figure
7


Schematic
D
iagram of the
P
roposed RP
M
a
chine for
P
olystyrene
O
bjects


Step 2

The
vertical
cutter homes in XY and the Z
-
slide goes to the required Z value so that
the hot wire
/ band saw

is at the correct level for slicing

the first v
-
slice.
The top table
falls gently due t
o gravity onto the bot
tom table;

t
he rate of fall is
dependant on

the
counterweight
.
Vacuum is applied at the bottom table.



Step 3

The hot wire
/ band saw

shuttles from one end to the other slicing the block.



16

Step 4

The top table retracts upward.

Now the top surface of the
fi
r
st
v
-
slice

is exposed and
is ready for machining
. The
hot wire
/ band saw

shuttles
back to its home

end
.


Step
5

The bottom
tip of the vertical
cutter machines the
top
surfaces
of the first v
-
slice on
the
bottom

table
.


Step 6

The bottom of
second v
-
slice

on the top table
is machined by the top cutter
.
Glue is
selectively applied on this surface at the required places
.

Note that the same XY head
carrie
s

the glue nozzle in addition to the vertical cutter.


Step 7

The
vertical
cutter homes in XY and the Z
-
s
lide goes to the required Z value so that
the hot wire
/ band saw

is at the correct level for slicing
the second v
-
slice
.
The top
table falls gently due to gravity onto the bottom table. A brief delay is given to allow
glue setting
.






(a) Pneumatic
tw
in
-
c
utter

(
b
)
Hexagonal turret with p
neumatic
c
utter


Figure
8

Two Possible Arrangements of the
Pneumatic Cutter


Step8

The hot wire shuttles from one end to the other slicing the block
.


Step
9

The top table retracts upward
.
Now the top surface of the
se
cond v
-
slice

is exposed
and is ready for machining
.
The
hot wire
/ band saw

shuttles
back to its home

end
.


17


Step

10

The bottom cutter machines the surfaces of top of
second v
-
slice

on the top table
.


Step

11

Ste
ps 6 to 10 are repeated when the number layers

are more than two
until the whole
prototype is built.

Finally, t
he
vertical
cutter home
s in XY
, vacuum on the bottom
table is cut off and the part is taken out
.






(a) Step 0

(
b
) Step
1

(
c
) Step
2

(d
) Step
3





(e
) Step
4

(
f
) Step
5

(
g
) Step
6

(
h
) Step
7





(i
) Step
8

(j
) Step
9

(k
) Step
10

(l
) Step
11

Figure
9

Steps in Making an Object


18

5
.

ILLUSTRATIVE EXAMPLE




(a)
Gear lever housing to be built

(b)
Exploded
view of the v
-
slices or segments



(c)
1
st

v
-
slice made on FDM 1650 RP machine

(d)
2
nd

v
-
slice made on FDM 1650 RP machine



(
e
)
3
rd

v
-
slice made on FDM 1650 RP machine

(f
)
4
th

v
-
slice made on FDM 1650 RP machine


(
g
)
V
-
slice assembled into gear

lever housing

Figure
10


Illustration of the Manufacture of a Gear Lever Housing Using SOM Principle


Gear lever housing
, a fairly complex object shown in Figure 10a, was taken for illustrating
the principle of SOM. The V
-
slicing program of the authors
was able to split this object into 4
v
-
slices or segments. These segments are shown in exploded view in Figure 10b. Each of
these v
-
slices was
built using FDM 1650 RP machine; each could have been
made
in 2
settings using a 3 axis CNC machine

as well
. Thes
e four physical v
-
slices are shown in

19

Figures 10c
-
f. The final
physical
object shown in Figure 10g was obtained by gluing these
four v
-
slices.
The same part which has a height of 136.084mm would be built in 536 slices
each of 0.254mm thick in a commercial
RP process like FDM as against just 4 v
-
slices in
SOM.



The machine being built by the authors will be able to produce this object automatically
as explained in the previous section. The authors have developed the software for
automatically generating t
he cutter path for machining the v
-
slices using a single ball nose
end mill. However, it is desirable to develop software that would make use of cutters of
different diameters and shapes intelligently. Furthermore, more fine
-
tuning of the post
-
processing p
art of v
-
slicing
algorithm
is desirable to transfer material among layers to
minimize height.



6
.

CONCLUSIONS


Existing RP machines produce 3D objects by assembling their 2D approximations called
slices. Hundreds of thin slices constitute the object so
as to make it to a reasonable accuracy.
As against this costly and slow process, a new RP process called Segmented Object
Manufacturing (SOM) was presented in this paper. It makes use of visibility considerations to
split the 3D object into a few chunks of

material called v
-
slices or segments which are
automatically machinable from two opposite directions. As no geometric approximations are
introduced during slicing, the accuracy and finish of the objects produced using SOM
machine will be comparable to tha
t of CNC machining. The feasibility of this method was
illustrated using a hypothetical SOM machine for soft materials. When this SOM machine is
built, it will be useful for making large non
-
metallic prototypes automatically, accurately,
quickly and econom
ically. Particularly it will be useful for manufacturing EPC patterns of the
stamping dies.



The principle of SOM can be used for the manufacture even hard materials using CNC
milling machines semi
-
automatically; the blocks of the required thickness for
each v
-
slice can
be made using face milling and machining of both sides of each v
-
slice can be done in two

20

settings. These v
-
slices can be joined using fastening, or adhesive bonding or brazing
depending on the application requirements.



It is interesting

to note that SOM and a few other RP processes (like SDM, HLM and
TLM) that aim at manufacturing objects in thick layers heavily depend on machining. In
other words, the conventional wisdom of RP being an additive or generative process may no
longer hold g
ood if these processes become fully automatic.



R
EFERENCES


1.

Broek, J.J., Horváth, I., Smit, B., Lennings, A.F., Rusák, Z. and Vergeest, J.S.M. (2002)
Freeform Thick Layer Object Manufacturing Technology for Large
-
Sized Physical
Models,
Automation in C
onstruction
, Vol. 11, pp. 335
-
347.


2.

Chen, Y.H. and Song, Y. (2001) The Development of a Layer Based Machining System,
C
omputer
A
ided
D
esign
, Vol. 33, pp. 331
-
342.


3.

Chen, Y. and Rosen, D.W. (2001)
A
Region Based Approach to Automate Design of
Multi
-
P
iece Molds with Applications to Rapid Tooling,
Proceedings of

ASME Design
Engineering Technical Conference
, September 9
-
12
, Pittsburgh, Pennsylvania
.


4.

Chen, Y. and Rosen, D.W. (2003) A Reverse Glue Approach to Automated Construction
of Multi
-
Piece Molds
,

Journal of Computing and Information Science in Engineering
,
Vol. 3, No. 3, pp. 219
-
230.


5.

Dongwoo Ki and Kunwoo Lee (2002) Part Decomposition for Die Pattern Making,
Journal Material processing Technology
, Vol.
130
-
131
, pp. 599
-
607.


6
.

Enimco (2004)

A brief Description of Space Puzzle Molding,
http://www.enimco.com/puzzle.html
.


7.

Horváth, I.
,

Vergeest, J.S.M.
,

Broek,

J.J.,
Zolt
á
n

Rus
á
k

and

Smit
, B. (2002) Tool Profile

21

and Tool Path Calculation for Freeform Thick Layered Fabrication,
Computer Aided
Design
, Vol. 30, pp. 1097
-
1110.


8.

Ion Gibson (1997) Email to RPML Group on Ford’s 100 Day Engine Project,
http://rapid.lpt.fi/archives/rp
-
ml
-
1997/0366.html
.


9.

Junghoon Hur, Kunwoo

Lee,
Zhu
-
hu and Jongwon Kim (2002) H
ybrid
R
apid
P
rototyping
S
ystem

Using Machining and Deposition,

Computer Aided Design
,

Vol 34, pp. 741
-
754.


10.

Karunakaran, K.P., Shanmuganathan, P.V., Jadhav, S.J., Bhadauria, P. and Pandey, A.
(2000a) Rapid Prototyping of
Metallic Parts and Moulds,
Journal of Materials Processing
Technology
, Vol. 105, pp. 371
-
381.


11.

Karunakaran, K.P., Shivmurthy Dibbi, P. Vivekananda Shanmuganathan, D.
Satyanarayana Raju and Srinivasarao Kakaraparti (2000b) Optimal Stock Removal in
LOM
-
R
P,
I

MECH

E Journal of Engineering Manufacture
, Vol. 214, Part B, pp. 947
-
951.


12.

Karunakaran, K.P., Shivmurthy Dibbi, P. Vivekananda Shanmuganathan, Srinivasarao
Kakaraparti and D.Sathyanarayana Raju (2002) Efficient Stock Cutting in Laminated
Manufactu
ring,
Computer Aided Design
, Vol 34, pp. 281
-
298.


13.

Krishnan Ramaswamy (1997) Process Planning for Shape Deposition Manufacturing,
Ph.D. Dissertation
, Department of Mechanical Engineering, Stanford University.


14.

Kulkarni P. and Dutta, D. (1996) An Ac
curate Slicing Procedure for Layered
Manufacturing,
Computer Aided Design
, Vol. 28, pp. 683
-
697.


15.

Li Hui
, Lee Yong Tsui and Thomas Gong

(1998)
Layered 5
-
Axis Machining,
Conference
on Sculptured Surface Machining
, University of Michigan, Ann Arbor
, Nov
ember 9
-
11
.


16.

Pandey, P.M., Reddy, N.V. and Dhande, S.G. (2003) Real Time Adaptive Slicing for
Fused Deposition Modeling,
International Journal of Machine Tools & Manufacture
,
Vol. 43, pp. 61
-
71.


22


17
.

Protoform (2004) Description of Space Puzzle Moldin
g,
http://www.protoform.com
: Web
site of Protoform, Germany
.


18
.

Sangkyun Kang, James P. Johnston, Toshiyuki Arima, Minoru Matsunaga, Hideaki
Tsuru, Fritz B. Prinz

(2003)
Micro
-
Scale Radial
-
Flow Compressor Impeller Made of
Silicon
Nitride: Manufacturing and Performance,
Proceedings of ASME Turbo Expo
2003
, Power for Land, Sea, and Air, June 16
-
19, Atlanta, Georgia, USA.


19.

Taylor, J.B., Cormier, D.R., Joshi, S. and Venkat
araman, V. (2001) Contoured Edge Slice
Generation in Rapid Prototyping via 5
-
Axis Machining,
Robotics & CIM
, Vol. 17, pp. 13
-
18.


20.

Tyberg, J. and Bohn, J.H. (1999) FDM Systems and Local Adaptive Slicing,
Materials &
Design
, Vol. 20, pp. 77
-
82.




21.

Taylor, J.B., Cormier, D.R., Joshi, S. and Venkataraman, V. (2001) Contoured Edge Slice
Generation in Rapid Prototyping via 5
-
Axis Machining,
Robotics & CIM
, Vol. 17, pp. 13
-
18.


22.

Woo, T
.

C
. (1994)

Visibility
M
aps and
S
pherical
A
lgorithms
,

Computer

Ai
ded Des
ign
,
Vol
.

26
,

pp
.

6
-
16.


23.

Yan X. and Gu P. (1996) A Review of Rapid Prototyping Technologies and Systems,
Computer Aided Design
, Vol. 28, pp. 307
-
318.