Evaluation of a Short-Run Injection Molding Process with Somos

NanoForm 15120 Tooling

Alexander K. Do
University of California, Berkeley
Department of Mechanical Engineering

processes have limited their viability as methods of direct

ABSTRACT production.
A low-cost, short-run injection molding process was In recent years, RP has been used in direct and
investigated using a tool directly fabricated from indirect methods to fabricate injection mold tools, with
stereolithography with Somos NanoForm 15120 in high- varying results. Examples of such work include research by
resolution build mode. The tool produced 150 small, flat Dickens and Smith [1], and Harris et al. [2-4]. The ability to
parts (2” wide by 3.5” long), cost less than $2,500 to fabricate injection molding tools directly from RP would
implement, and had a lead time of three weeks from order to dramatically simplify production of injection-molded parts.
final injection-molded parts. Results from the investigation However, the processes for doing so are still in their infancy
indicate that the process is viable for short-run injection and there is a lack of extensive knowledge and literature
molding but challenges and deficiencies exist. Primarily, the about them. In this research, a low-cost, short-run injection
SLA industry must address anisotropy in the dimensional molding process is investigated using a tool directly
accuracy of the high-resolution stereolithography process of fabricated from stereolithography with Somos NanoForm
NanoForm. Thermal and mechanical properties of 15120.
NanoForm tooling are also found to be unique in comparison
to conventional injection molding practices, requiring a 2. REVIEW
learning curve on part of the processor.
2.1. Direct and Indirect Tooling Methods
1. INTRODUCTION An indirect RP tooling method refers to a process in which a
rapid prototyped master part is used as a casting pattern for a
1.1. Economics of Injection Molding final molding tool. One example of indirect tooling is the
Injection molding remains a very prominent manufacturing spray metal tooling method [5], in which an RP model is
technology globally for the production of goods in industries sprayed with fine droplets of molten metal, creating a metal
varying from consumer products to medical devices. Tooling shell about 0.08” thick. The shell is backed with an epoxy
for the injection molding process represents a large fixed cost filling, allowing for the integration of cooling channels and
and long lead time within the total product development an ejection system. Another casting method which uses a
process. Interviews conducted with local and global industry sintered metal-powder mixture is the 3D-Keltool process [6].
operators revealed that while advances in technology and The 3D-Keltool process can be used to produce a cavity and
growth in foreign markets in recent years have lowered tool core set or to build an electrical discharge machining (EDM)
costs from hundreds of thousands of dollars to tens of electrode for creating hard metal tooling. Indirect methods
thousands, lead times can still span 8 weeks or longer. Due have been proven with accurate and durable tools, but the
to the costs and lead times, conventional injection molding availability of processors is limited – for example the 3D-
processes are generally not economical for production runs of Keltool process requires an operating license.
less than 500 parts. For such batch sizes, alternative tooling Direct tooling methods, by contrast, use RP
strategies must be investigated to fabricate parts with the technologies to produce the final mold core and cavity. In
quality and properties achieved by high-pressure injection addition to Dickens’ work in stereolithography, other
molding. processes such as Z-Corporation 3D printing have been
studied in injection molding [7]. While direct tooling
1.2. Rapid Prototyping in Mold Tooling methods offer the ability to easily produce a tool in-house or
through a service bureau, the limiting factors in material
The set of fabrication processes classified as “rapid properties have prevented them from become industry
prototyping” (RP) technologies have shown to be effective standards for short-run injection molding.
low-cost tools for prototyping parts, primarily in areas of
design validation. Generally, RP refers to any process which
2.2. Stereolithography
uses a layer-based solid freeform fabrication method to create
physical parts directly from digital CAD data. Prototypes and The stereolithography (SL) process by 3D systems became in
models built from RP have been used for interaction between 1987 the first commercially available layered manufacturing
disciplines in product development: design, engineering, process for producing parts directly from CAD data, and 19
marketing, buyers, and manufacturing. However, material years later it still remains the standard for precision and
properties, quality factors, costs, and lead times of RP accuracy in rapid prototyping. Since its introduction, the
Alexander K. Do Page 2 of 18

process capabilities and available materials have dramatically 3. EXPERIMENTAL WORK

improved. Attainable feature sizes are now within 0.005” on
an SLA Viper [8] machine and layer thicknesses of 0.001” 3.1. SLA Direct Tooling Molding Process
are possible on an SLA 7000 [9]. Elastomeric materials The total design-and-manufacturing process used for the
allow for parts with a Shore A hardness of 70 and 75% short-run injection molding study is similar to the process
elongation at break (Somos ULM17220 [10]), while particle- which would be taken for a standard production-quantity
filled composite materials from Somos and 3D Systems can mold:
attain a tensile modulus of 1,700,000 psi (Somos ProtoTool 1. Design and validate part:
[10], 3D Systems Bluestone [11]). Per-part costs have gone -Design for manufacturability (DFM) analysis
down at least an order of magnitude since commercialization -Moldflow analysis
and it has recently become viable to use SL models in many -Rapid prototyping
more applications. 2. Design mold and ejection system:
The chemical process for stereolithography is based -Cavity scaling factor
upon the photopolymerization of a monomer resin with -Parting line
ultraviolet (UV) light providing the activation energy. -Runners, gating, venting
Typically, a table or stage submerged into the resin forms a 3. Partition cavity/core features to interchangeable
platform for the part, which is polymerized as a UV laser inserts:
traces out each layer using a hatching pattern such as STAR- -Mechanism to secure inserts
Weave or ACES [6]. The stage lowers itself into the resin as 4. Fabricate SL inserts and mold base:
a new layer of uncured resin is spread over the part for -Outside service bureau or in house
hatching. The polymers in stereolithography apparatus 5. Determine injection parameters by testing:
(SLA) systems consist of acrylates, epoxies, and vinylethers -Injection pressure and speed
which are cured by free radical or cationic polymerization. -Clamping force
During polymerization, the polymer chains are cross-linked -Holding and cooling time
to create a thermosetting part.
The primary novelty in this process occurs during Step
2.3. Somos NanoForm 15120 3, the partitioning of cavity and core features to inserts.
Stereolithography resins have long been studied for direct Generally, cavity and core features are be formed by a single
injection mold tooling [1], but limitations in accuracy, or multiple metal parts in a production tool which are pressed
durability, and heat resistance hindered practical applications or bolted into the mold assembly. Similarly, designing a
[12]. Composite materials recently introduced by materials mold assembly to use SL cavity and core inserts which can be
manufacturers DSM Somos and 3D Systems offer secured into a metal base allow the complex geometric
dramatically increased material properties in the areas features to be rapidly manufactured by SLA. Securing these
previously listed. The materials employ a matrix of inserts into a base rather than fabricating the entire tool from
thermosetting polymer around a reinforcing type of ceramic stereolithography allows for a number of benefits:
which is used to provide stiffness and heat resistance while • Reduced amount of volume to be built on SLA,
maintaining the manufacturability of stereolithography. The reducing lead time and cost
composite materials available at the time of this writing • Use of a generic mold base for a family of
consist of ProtoTool and NanoForm by DSM Somos and interchangeable SL inserts, reducing tool costs
Bluestone by 3D Systems. • Installation and mating of the assembly to standard
Of the three composite materials, NanoForm is platen / nozzle interfaces, which create better
particularly interesting not only because it is a lower cost than distribution of clamping force through the hard
Bluestone and ProtoTool, but it has been reported by SLA metal mold base
operators to be a more “user friendly” material. Specifically, • Simpler installation of alignment, ejection, and
the silica “nanoparticles” suspended in NanoForm do not cooling systems
settle into the uncured resin – unlike ProtoTool which • In the case that the tool fails or needs considerable
requires a mixing unit to maintain particle suspension and adjustment in geometry, a new insert can be
Bluestone which requires vat stirring. One of the most fabricated quickly at a low cost
compelling reasons to investigate NanoForm specifically over
the other two is that while it offers lower strength and thermal Fabrication of the SL inserts can be done in-house
resistance properties, it is the only composite resin which can with SLA equipment, but with the widely available number
be run at high-resolution. The high-resolution build mode of service bureaus it can be very simple to use a third-party
enables a 0.002” layer thickness and 0.003” beam diameter – vendor for this process. Some service bureaus such as
contrasted to 0.004” layer thickness and 0.010” beam Quickparts.com offer electronic online quoting systems [13],
diameter in standard build modes [8]. The material properties and many other vendors will accept files by email and return
for NanoForm are listed in Appendix 9.2.1. quotes within hours. This enables a particularly good lead
time for fabricating the inserts, typically as fast as one week
including shipping.
Alexander K. Do Page 3 of 18

The process, assuming access to a molding press and a project became shelved for one year but was revived when
stock of molding materials, can cost anywhere from $1,000 to other research work demonstrating the viability of direct SL
$6,000 or more depending on size, complexity, and materials. tooling was discovered.
A summary of the approximate costs and turn time:
• Aluminum mold base: $500 - $3,000 depending on 3.2.2. Interchangeable Cavity and Core Inserts
material, machining time, in-house capabilities The mold assembly consisted of aluminum blocks for
• SL inserts: $500 - $3,000 depending on material, alignment and distribution of clamping force, SL cavity and
volume, z-height (build time), layer thickness core inserts secured by flange screws, two hardened steel
• Alignment and ejection hardware: $100 locating pins, and two hardened steel ejector pins. No active
• Turn time: 1-2 weeks cooling was designed into the mold assembly – instead the
mold would need to be air-cooled between shots. The cavity
3.2. Design of Mold Cavity Set (Case Study) and core inserts for the mold assembly were designed to
minimize volume and thickness so that the cost of the SL
3.2.1. Project Background inserts would be minimal. The x and y-dimensions of the
The motivation for this case study originated from an cavity and core inserts were 6” x 4”, and the z-dimensions
academic design project for an injection mold which was to were 0.4” and 0.25” respectively. The aluminum base parts
produce a pilot series of a particular tile from ABS plastic. measured 8” x 6” x ~1.25”. The pockets in the aluminum
The part was designed with unique geometry so that six base were also designed with 0.015” of relief above the
pieces could be assembled to form a rhomboid building block length and width dimensions of the inserts, so that any
for the visualization of various zonohedra – in lay terms, thermal expansion of the inserts would be unbounded. The
mathematical Legos. In particular, these tiles had unique locating pins were pressed into the aluminum blocks to align
features which would be difficult to machine into a mold: the cavity and core. The ejection method is discussed in
sharp angles, well-defined teeth, and two parting planes. further detail in Section 3.2.5. The aluminum used was 6061-
Figure 1 shows a computer-aided design (CAD) rendering of T6 for its ease of machining. The inserts were fabricated by
tile design, showing the teeth on each edge, with a secondary Protogenic Inc. with Somos NanoForm 15120 on an SLA
parting plane for the teeth on the left side of the tile. For a Viper at a high-resolution build with 0.002” layer thickness.
sense of scale, each of the tile edges are 2 inches long, and STL files were sent to the service bureau to make the parts,
the teeth are 1/8” wide. exported from SolidWorks CAD software with 0.00058”
deviation tolerance and a 2.1˚ angle tolerance to ensure that
limitations in accuracy were at the machine-level and not in
the 3D data. Except for basic support removal and
UV/thermal post-curing, no added post-processing, such as
polishing or finishing inside diameters of holes, was used on
the inserts so that issues resulting from the SLA process
could be analyzed. An analysis of the stresses in the SL
inserts predicted a factor of safety (FOS) of 3 before failure
from compressive yielding (Appendix 9.1.1) during injection,
and an FOS greater than 10 for failure during clamping.
Figures 2a and 2b show the mold assembly for the cavity and
core, respectively.

Figure 1 - CAD rendering of zonohedra tile

The original design was adapted for
manufacturability, considering both injection molding and
tool machining requirements. There were some geometric
features which could not be sacrificed during the redesign,
and it became clear that the machining would not only be
very difficult, but it would be quite expensive as well.
Conventional CNC machining could not produce the required
geometry, and the most viable option for machining the
cavity and core was with wire-EDM, a slow and costly
process. Since neither the designers nor machinists could be
sure of how the final part would behave after molding, the
money for an expensive tool could not be invested. The
Alexander K. Do Page 4 of 18

(a) shrinkage between 0.001 – 0.002 in/in. Compensating for

thermal expansion by calculation was shown to be an
effective way to determine a compensated shrinkage value
which could predict the total part shrinkage. In addition, the
photopolymerization process in stereolithography imparts
shrinkage from the actual area activated by the laser path,
introducing an additional design compensation. The mold
cavity in this experiment was designed to target the design
parameters of the final part by accounting for thermal
expansion of the molding cavity, a prediction of an adjusted
shrinkage value, and compensation for SL shrinkage.
Appendix 9.1.2 shows the derivation of a dimensional
scaling equation which accommodates for each of the known
factors,
Cd = Pf
(1 + C.T .E.Steel (Tm '−T0 ')) ,
(1 − S SL )(1 + C.T .E.SL (Tm − T0 ))(1 − S ASTM )
(b) where Cd is the designed cavity dimension, Pf is the final part
dimension, C.T.E.Steel is the coefficient of thermal expansion
of the molding steel used in the ASTM D955 shrinkage test
for a given material, Tm’ is the molding temperature used in
the test, T0’ is the initial temperature of the steel mold (room
temperature), SSL is the expected cure shrinkage of the SL
insert, C.T.E.SL is the coefficient of thermal expansion of the
SL insert, Tm is the expected molding temperature used for
molding the parts, T0 is the initial (room) temperature of the
SL insert, and SASTM is the published ASTM D955 shrinkage
value of the molded material. The known and estimated
values (explained in 9.1.2) used to estimate the required
cavity scaling are presented:
• Tm’ = Tm = 130˚ F
• T0’ = T0 = 70˚ F
• SSL = 0, compensated for by vendor
• C.T.E.Steel = 7x10-6 in/in-˚F
• C.T.E.SL(Tm – T0) = 28x10-6 in/in-˚F (122˚ F – 70˚ F)
Figure 2 – mold top with cavity insert (a), mold bottom + 47x10-6 in/in-˚F (130˚ F – 122˚ F) = 0.0018 in/in
with core insert and locating/ejector pins (b) • ABS Shrinkage SASTM: 0.005 in/in

3.2.3. Size and Shrinkage The final cavity scaling factor derived from the scaling
During injection molding, the thermoplastic material is equation was 1.0036, which was rounded up (due to lack of
heated to molecular flow, which occurs well above the glass significant figures and large estimations) to a uniform cavity
transition temperature, Tg, and a crystalline melting point Tm scaling percentage used in the SolidWorks CAD model of
(if the polymer is semi-crystalline). Since the rate of thermal 0.4%.
expansion of molten thermoplastics above these points is
discontinuous, mold cavities are commonly designed with a 3.2.4. Releasability: Draft and Layer Thickness
scaled increase in volume corresponding to a predicted Several researchers believe that the primary cause of SL tool
amount of linear part shrinkage, usually published in failures result from tensile stresses on the core side during
reference texts or supplier datasheets based on the results of ejection [2,3,14]. The force required to eject a part is largely
ASTM D955 tests. due to static friction, which is a result of the coefficient of
The final size of the part, however, is not actually static friction between the plastic and the core and the normal
determined by a shrink from the size of the molding cavity, force imparted by part shrinkage. Through design of the tool,
but from the size of the heated molding cavity at molding layer thickness and draft angle can be adjusted to optimize
temperature. Since most high-production molds are made these parameters; thinner layers reduce the surface roughness,
from hardened steels with minimal thermal expansion, any and larger draft angles reduce the projection of normal force
expansion which might occur in the cavity during molding is onto the core during ejection. However, thinner layers can
factored into the end-results of ASTM D955 tests. Harris [2] also increase the build time (and cost) and there is a finite
found that differences in thermal expansion between limit to what many commercial SLA machines are capable of.
aluminum and SL molds could impart a difference in Accordingly, draft angles which are too great may
 Alexander K. Do Page 5 of 18

compromise the design of the part. Harris’ study of ejection 3.2.6. Gating and Venting
forces in SL5190 tooling [3] shown in Figure 3 demonstrate The part was gated from the sprue (eliminating the need for a
that there is a linear relationship between ejection force and runner), with a flat gate at the parting line which was 0.280”
draft angle and a non-linear relationship between ejection wide and 0.032” thick, shown in Figure 4b. This gate
force and layer thickness, with layer thickness having a much thickness was smaller than the recommended 40 to 90% of
bigger impact than draft angle. wall thickness [15], which was 0.120” around the perimeter
of the part and 0.080” in the center. This small gate thickness
was a result of a CAD design error and was originally
intended to be 0.050” thick. Between molding runs the gate
thickness was eventually increased by approximately 0.010”
to 0.020” which aided in molding.
The high-resolution 0.002” build thickness allowed
the addition of parting line vents to allow removal of air from
the cavity to aid in part fill. Primary vents must be thin
enough to prevent plastic from leaving the cavity, and Bryce
[15] recommends vents for ABS which are 0.002” deep.
Four vents were located at each of the part corners which
were 0.25” wide by 0.15” long. The short length was
designed to reduce clogging of the vents by moisture or other
condensation from the plastic and mold-release spray. These
vents opened into larger channels, referred to as secondary
Figure 3 - Ejection force as a result of draft angle
vents, which were 0.020” deep, running to outside
and layer thickness [3]
atmosphere. Figure 5 shows the vents designed into the
From our particular vendor, the marginal cost from cavity insert. Additional venting was located at the teeth
0.004” layer thickness to 0.002” was 33% more so 0.002” features, with a different venting strategy used for each tooth
was chosen as the layer thickness. A draft angle of 1˚ on all block to determine the minimal venting required at the teeth.
faces would not compromise the geometric features in the The four combinations used were: no venting, 0.002” deep
design so that was used as a minimum value for draft. pressure relief gaps (lower right of cavity in Figure 5) which
did not go to outside atmosphere, primary vents which went
3.2.5. Ejection directly to outside atmosphere, and primary vents which fed
Examination of the mold geometry suggested that most forces into a secondary vent leading to outside atmosphere (upper
resulting from part shrinkage would be concentrated around right of cavity in Figure 5).
the pin on the core side of the mold. To aid ejection, the
mold was designed with one 3/16” ejector pin on each sides
of the core pin for manual ejection, shown in Figure 4a.
There was early speculation that shrinkage of the teeth
0.020” deep
features on the cavity side would cause sticking during mold
opening, but a cooled and hardened sprue could be used
during mold separation as an ejector to keep the part on the
core side.

(a) (b)
core pin sprue
s 0.280” wide

0.002” deep

Figure 5 - Mold cavity design showing venting system

ejector pins 0.032” thick
3.2.7. Total Cost and Time to Develop
Figure 4 - ejector pins from above (a), Once the design was finalized, it took a couple of weeks to
gate location and dimensions (b) locate and consult with vendors for the machining and
stereolithography process. The effort it took to modify the
original mold design to use cavity and core inserts was
minimal. Here is a breakdown of the costs and lead times for
each part of the mold assembly:
Alexander K. Do Page 6 of 18

• Local shop-machined aluminum: preheated with a heating plate to 130˚F. Polytek Pol-Ease
o raw material – $100 2300 mold release spray was used sparingly to aid in part
o setup and machining – $1100 ejection. Based on molding results, the gate was expanded
o machine time – 4 hours twice by filing material away from the core insert. It is
o lead time – one week estimated that the gate thickness was increased by 0.005” to
• SL from Protogenic Inc.: $1200 0.010” each time.
o 0.002” layer thickness During the initial molding trials, the mold was tested
o machine time – 24 hours total, under a wide range of injection parameters:
plus 5-6 hours curing
o lead time – one week Barrel Temperature 450 – 470˚F
• Alignment and ejection hardware: $20 Nozzle Temperature 450 – 477˚F
• TOTAL COST: Less than $2,500 Clamping Force 12 – 16 tons
• TOTAL LEAD TIME: One week Injection Pressure 6,000 - 9,000 psi
Ram Speed 7.3 – 9.0
3.3. Mold Testing Methodology Shot-and-Hold Time 2 – 25 s
Preheat 0 –140˚F
The mold cavity and core were tested on a Morgan Press
GT100, a small 20-ton manual injection molding press, to Table 1- Operating range of mold in experiment
evaluate tool quality, part quality, and process repeatability.
The mold setup is shown in Figure 6. The blend of ABS used After a large number of shots, the ideal operating conditions
was a standard unfilled material, Lustran 648. for the mold were determined:

Barrel Temperature 465 – 470˚F

Nozzle Temperature Barrel temp. + 5˚F
Clamping Force 16 tons (maximum
attainable)
Injection Pressure 7,000 psi
Ram Speed 8.5 – 8.8
Shot Time 2.5 – 3 s
Preheat None during steady
state, 140˚F to bring
up cold mold to
temperature
Table 2 - Injection Molding Parameters Used
The temperature and pressure parameters are at the lower end
of the plastic manufacturer’s recommendation (Appendix
9.2.2), while the mold temperature and injection speed are
comparable to the recommended levels. Once the optimal
parameters were determined, the tool was tested until it was
used for 150 total shots. Each part was batched by parameter
grouping and evaluated to determine quality.

3.3.1. Tool Quality

Figure 6 - Mold loaded into Morgan Press GT100 The primary areas of investigation of the tool were in
The injection molding parameters for the tool were dimensional accuracy and tool life. For dimensional
determined by beginning at minimal energy parameters (low accuracy, 24 dimensional parameters were measured on the
temperatures, clamping force, injection speed, and injection cavity insert and 17 parameters on the core with a Mitutoyo
pressure) in order to protect the tool, and adjusting these 500-133 digital caliper to be compared with measurements in
parameters based on part quality indications. The first set of the SolidWorks CAD environment. It should be noted that
parameters used was: 450˚F barrel and nozzle temperature, while the caliper has an instrument resolution to half-
12 tons of clamp force, and 6,000 psi injection pressure. thousandths of an inch, it was often difficult to align the
Since the Morgan Press uses a pneumatic ram to force caliper edges with certain parameters, creating a random error
injection, it does not have control over the actual flow rate of as much as +/– 0.005” in some instances. The measures to
(measured in cubic feet per minute). Instead, the ram evaluate tool life were the total number of shots until the tool
adjustment is made with a unitless setting on an orifice which was rendered unusable, visible wear, and any characteristic
controls the pneumatic airflow to the ram. The initial setting failures which occurred such as fracture. Other factors which
for the ram was 7.3. The tool was loaded into the press and were investigated were whether the ejection and venting
Alexander K. Do Page 7 of 18

systems had been properly designed. During the tooth block breaking off of the core. The fracture and tooth
experimentation, it was determined that mold temperature block are shown in Figure 7.
played a significant role in determining the part quality, so a
Mannix IRT4 infrared thermometer was used to measure the (a)
mold temperature at various locations between injection
cycles.

3.3.2. Part Quality

The intended purpose of the parts was to study their usability
in the assembled rhomboid form, and learn how to adjust
certain design parameters to produce a revised tool. The
parts were primarily examined for aesthetic qualities such as
warp, sink, fill, and surface quality, and dimensional accuracy
to verify whether they would be within the predicted size.
The same Mitutoyo caliper was used to measure several
dimensional parameters on the final molded parts. For
shrinkage calculations, cavity features corresponding to the
measured part dimensions were measured between molding
cycles to attain the heated cavity dimensions. The teeth on
the final parts were also studied to find out if any particular
strategy of tooth block venting worked better: no vent,
pressure relief only, primary to outside atmosphere, primary
plus secondary.

3.3.3. Process Repeatability

For process repeatability, the primary criterion used was to
see whether part quality could be reasonably controlled from
shot to shot. As part of this test, the cycle times (injection, (b)
part cooling, open close, mold cooling) were adjusted to
determine if a steady-state molding cycle could be
implemented on this tool. Unfortunately the injection timer
on Morgan Press consisted of analog electronics which were
in disrepair and could not be used to control the shot time.
All of the timing was controlled by hand, and measured with
a digital watch or timer. This limited the timing resolution to
being within one second only.

4. RESULTS

4.1. Tool Quality

4.1.1. Before Molding

After the SL inserts arrived from the vendor, the most
noticeable defect was a number of dimples on the top Figure 7 - Core insert showing brittle fracture (a),
surfaces of both the cavity and core, but much more tooth block after separation from core (b)
prominent on the cavity. These dimples were on large flat
areas of the parts, and were caused by surface dewetting of It appears as though the cavity was pulling up on the
uncured resin – since they were difficult to photograph are tooth block of the core during uneven mold opening (not held
not shown in this paper. As a result, the shutoff land of the level) and gouged the crack into the core before delamination
mold was not completely flat. of the entire block. Figure 7a shows that outside of the
The inserts were fit into the aluminum bases and gouge, the crack propagated along three or four cured layers
appeared to fit well enough (they were very snug on the of the core. Also visible in Figure 7a is a difference in color
locating pins) until the mold assembly was closed together to from the exposed material (from fracture) to the surface of
see how well it fit. The closure between the cavity and core the inserts. Following the fracture, it was suspected that the
was very snug at the two sliding shutoff areas where the core and cavity inserts were not within the original design
parting line jumps from one plane to another. Upon opening tolerance, so they were measured.
the mold, the core insert fractured instantly, with an entire The width of the gaps which interfaced with the tooth
blocks from the core were 0.002” smaller than the original
Alexander K. Do Page 8 of 18

dimensions while the block which fractured was 0.003” wider

at the base than designed. In total, 24 cavity parameters and
17 core parameters were measured, resulting in 11 parameters
in the cavity and 4 in the core which were more than 0.005”
from the designed dimensions. Two of the larger offsets,
coming from approximately 0.010” overshoot of the plate
thicknesses of both inserts, were later excluded from analysis
since the vendor stated they would not be within specification
without post-processing (as a result of the necessary support
base during the SLA process). The offsets from the design
specifications of the remaining 39 measurements ranged from
+0.005” to –0.027”, with a mean value of –0.004” and a
standard deviation of 0.006”.
Following the observations in dimensional accuracy,
the inserts needed some manual rework with sandpaper and
other polishing agents to fit with each other, the locating pins,
and ejector pins. After reducing its size, the fractured block
was bonded back to the insert with a cyanoacrylate adhesive. Figure 9 - Mold core after shearing of core pin
4.1.2. During and After Molding After the core pin broke off, the tool did not undergo
any failures which prevented molding and performed
reasonably well – after injecting 150 parts over several
different molding sessions, the tool was still considered
usable. From visual examination of the tool and the parts,
there were no indications of surface wear anywhere in the
cavity and the core except for a very small section of the gate
at the cold slug which chipped away. Surprisingly, the
bonded tooth block did not ever fail during the course of the
molding.
The ejection system did work as designed, though
only one pin needed operation to eject the part: the one
between the two teeth blocks on the core. The sprue was
needed as an ejection aid as well, since the part was also
sticking to the cavity side at the teeth. When the barrel
temperature was increased, the part started to stick more to
the teeth on the cavity side, inhibiting use of the sprue as an
Figure 8 - Shot 1 in mold (sprue removed) ejection aid. However, with additional cooling time this
The first molding experiments began after the adhesive had problem was mitigated.
fully set. The first injection parameters, coupled with a 25 While it was found that applying the mold release did
second shot-and-hold time produced an encouraging result, have a minor beneficial effect on releasability, the most
shown in Figure 8. However, not enough care was taken drastic effect on ease of ejection was cooling time. The parts
during mold opening to ensure that the sprue remained on the consistently ejected better when the part’s temperature was
core side, causing the part to bend upward with the cavity cooler than the inserts’ glass transition temperature (Tg),
(anchored to the sprue by the gate), which in turn 176˚F (Appendix 9.2.1). During one trial in which the mold
immediately sheared the core pin from the core insert. While was opened rather quickly, the temperature of the part at the
a small fillet was designed in to reduce stress concentration, center of a flat triangular half-section (0.080” wall thickness)
the pin broke just above the fillet. The pin broke nearly was approximately 180˚F, while the temperature at one of the
straight across, but managed to gouge one or two layers teeth sections (0.120” wall thickness) was 200˚F. The
below into a flat bowl-like shape on the core, shown in Figure temperature of the SL mold at that particular tooth block was
9, and a corresponding flat dome-like shape on the fractured measured at 190˚F, more than 10 degrees above Tg. Under
pin. Molding was continued with the absence of this feature these conditions, ejection was quite difficult as the part would
in order to continue testing the tool. It was not believed that bend up with the ejection pin while the teeth on the part stuck
the pin would have remained set if bonded back to the core, to the blocks on the core.
due to the perpendicular orientation to rapidly flowing Upon further study of the temperature distribution of
viscous plastic during injection. the mold, it was found that the temperatures were quite
variable spatially and temporally throughout an injection and
cooling cycle. From measuring the mold temperatures during
different part cooling and mold cooling times, it was found
Alexander K. Do Page 9 of 18

that the amplitude of a thermal cycle for a given section was With persistent use, the SL inserts began to change
generally at least 60˚F in areas which contacted the ABS color in certain areas, shown in Figure 11.
plastic. Specifically, the temperature of the repaired tooth
block typically went from 120˚F to at least 190˚F during the (a)
injection cycle. Without an instrument to continuously
monitor the temperature through the molding cycle, the peak
temperatures could not be determined, so instead the mold
was opened with minimal cooling time in some instances to
record the temperatures of the hotter areas. Figure 10 shows
a thermal map of the mold temperatures after part ejection,
with each measurement taken immediately after part ejection
under steady-state molding. The map shows a high spatial
variation of temperatures in the mold, with as much as 60-
degree difference between molding areas and shutoff on the
inserts. It should be noted that during peak temperatures this
difference was closer to the order of 100˚F.

(a)

95˚

98˚
110˚
125˚

106˚ (b)
160˚
145˚
160˚ 147˚

(b)

90˚ Figure 11 - Visible discoloration in cavity (a) and core (b)

120˚ After 90 shots, a small hairline crack was discovered

in the core of the mold, seen in Figure 12a. The crack
measured about 0.21” and extended from the corner of the
105˚
90˚ 140˚ base of the repaired tooth block along the approximate
outline of the cavity parting line. The crack was very
85˚ difficult to see in most lighting conditions and there was no
noticeable gap. Another similar crack was discovered after
103˚ 100 shots in the cavity, pictured in Figure 12b. This crack
was considerably larger, measuring nearly 0.84” long. The
crack ran from a corner of the parting line vent down to the
part top (base-layer of the cavity), along the edge of the
cavity wall, and into the shutoff area where the cavity meets a
Figure 10 - Mold temperature distribution after part tooth block from the core. It is not known when the cracks
ejection in cavity (a), core (b) (˚F) first developed, as they were not obvious. Once they were
discovered, neither crack seemed to propagate with continued
cycling for the rest of the molding experiment.
Alexander K. Do Page 10 of 18

(a)
(a)

(b)
(b)

Figure 13 – Typical part molded from optimal

Figure 12 - Hairline cracks in core (a), and cavity (b) parameters, top (a), and bottom (b)

4.2. Part Quality The largest problem in quality was flash. Even the
best parts had minimal amounts of flash at the parting line
4.2.1. Aesthetic Quality (~1/32”), but flash at the end of the tile opposite from the gate
was a very large problem. Often times the flash would crack
In several aspects the part quality was generally impressive the mold open, causing sometimes inches of material to flow
once the final injection parameters had been determined; they into the shutoff area. In addition, the sliding shutoff at the
had good fill, the teeth features were very well-defined, the two tooth blocks on the core had flash which was likely
parts were flat (any warp was minimal and less than 1/32”), caused by tool rework – since the tooth blocks on the core
and sink on most surfaces were minimal. Figure 13 shows an were sanded by hand to fit with the cavity, too much material
average part molded under optimal conditions. The primary was taken off and as a result there was flash occurring
exception was the cylindrical stump in the middle, which had through the resultant gap between the core and cavity. There
severe problems with sink and fill. However, with the was also more flash that occurred on top of those tooth
removal of the core pin, the wall thickness in this area was blocks, at the upper parting plane in the mold. Higher
greatly increased from the original design and would account clamping force was attempted to control the flash, but the
for those problems. The surface quality of the cavity side machine limitation was reached at 16 tons of force. The
(which was left unprocessed) was smooth and glossy, while clamping force applied to the aluminum mold base was also
the finish on the core side (which had been textured with visibly uneven, as the heating plate used was centered to the
fine-grit paper) was smooth and matte. lower platen, causing a 2” overhang of the mold bottom off
the side of the heating plate. After the 90th shot, the heating
plate was disconnected and moved directly underneath the
Alexander K. Do Page 11 of 18

mold bottom to eliminate the overhang. This improved the verified through repeated measurement was that the width of
flash at the opposite end of the gate but did not eliminate it. the individual teeth experienced a “negative shrinkage” from
Defects which were prominent in earlier runs (before the original cavity dimensions. The average tooth width on
identifying the optimal parameters) were generally the molded part was 0.1131”, compared to an average cavity
characterized by flow lines. In many cases the flow lines width of 0.1086”. When these features were measured on the
were fairly deep and marked with ripples, appearing to be cavity after ejection, they were found to be only 0.001” to
formed by cold-flowing plastic. To mitigate the flow lines 0.002” larger with the tooth block at 120˚F.
the gate was opened up twice and the barrel temperature was
increased. However as the temperature was increased, the 4.3. Process Repeatability
tendency for sink increased as well. Repeatability in part quality was fairly difficult to attain, and
The quality of the teeth features did not appear to was not achieved to satisfaction. The variability in flash was
favor any one venting strategy over another. Generally, fill fairly high, and the nature of the sink tended to vary a bit
was not a problem and as stated earlier there was a consistent from part to part. Dimensional stability of the length and
amount of flash (though minimal) in these areas. width parameters was very repeatable, with parts lining up
with each other when placed back-to-back.
4.2.2. Shrinkage In the earlier cycles, the heating plate was left on at
The dimensions of the cavity during molding were 130˚F throughout the process, but it was causing the mold to
demonstrably different than when at room temperature, so cool very slowly. It was found that preheating a cold mold to
measurements of the cavity during molding were used to 140˚F and then turning off the heating plate worked better,
determine the dimensional accuracy of the molded parts. bringing the mold to steady state more quickly. A somewhat
However, it was only possible to measure the heated cavity steady state cycle was achieved in spite of variations in flash
dimensions after the mold had cooled enough to open, so the and sink – hindered slightly by both the timing of manual
measurements may not correspond to the actual sizes operations (such as opening the mold) and thermal control of
imparted onto the molten plastic. Because the cavity and the mold assembly. The ideal shot time was somewhere
core inserts were cooling as the measurements were taken, it between 2 and 3 seconds, with too short of a time causing
was necessary to take the measurements over multiple cycles short-shots and too long of a time causing potentially large
and very quickly. amounts of flash. The approximate cycle times attained
Of the measurements, the three largest parameters under the optimal molding conditions (Table 2) are presented
were used for calculating part shrinkages from the cavity at here:
room temperature and when heated. The two parameters
which expanded the most were the corner-to-corner length Injection and packing 2.5 s, +/– .5 s
and width of the cavity, which expanded 0.011” and 0.014” Holding 10 s
respectively. The normal distance between the tile edges was (no packing pressure)
also measured and used for shrinkage calculations, presented Part cooling in mold 45 s
below in Table 3. The measurements are also compared in Mold opening and ejection 20 – 30 s
the table to the original design dimensions to gauge error in Mold cooling and hopper refill 20 – 30 s
scaling the cavity size. Alignment and clamping 10 – 15 s
Many of the other parameters measured were too TOTAL 2 min
small in relation to the instrument error to use for meaningful
shrinkage results. However, an interesting result which was Table 3 – Cycle time under steady state

Heated % Shrinkage, Unheated

Part % Shrinkage Design % Error
Cavity Heated Cavity
Length 3.3550” 3.3800” 0.70 3.3685” 0.40 3.4027” 1.4
Width 2.0690” 2.0910” 1.0 2.0775” 0.41 2.1028” 1.61
Normal 1.7810” 1.8045” 0.86 1.7980” 0.50 1.7888” 0.011
Table 4 - Part dimensions, compared to heated cavity, cavity at room temperature, and original design dimensions
Alexander K. Do Page 12 of 18

cavity was press fit into the slightly larger block. It is not
5. DISCUSSION believed that this error could be accounted for beforehand in
the design, as the sliding shutoff required a very narrow
5.1. Tool Design and Quality tolerance. The offset errors of the female features were not
consistent with the original design parameters to neither an
5.1.1. Dimensional Accuracy absolute amount nor linear ratio. The offset magnitudes
The primary issue with the tool was that the dimensional ranged from 0.002” to 0.013” and the percent error from
measurements revealed serious anisotropy in accuracy. A 0.2% to 5%. As a result, it is highly unlikely that a designer
histogram of offset errors is presented in Table 5. The would be able to compensate for this in a CAD model.
histogram shows the majority of the offsets fall within the bin It is not known whether the problems in accuracy
for [0, –0.005”). The table also shows a strong weighting were a result of improper configuration of the SLA machine
toward negative offset errors, corroborating the mean value used to build the part, but that is not believed to be the case
of –0.004”. From these two parts, the resultant natural since the accuracy of male features was within expectation
tolerance (following the +/– 3σ model [16]) of NanoForm in and considerably better than that of female features. At the
high-resolution is determined as (+0.014”, –0.022”). time of this writing, interviews with three SLA processors
revealed that there has not been much industry experience
Bin Frequency with the NanoForm material built in high-resolution. More
–0.030” 0 investigation into the interaction between material properties
–0.025” 1 and process control are necessary to determine the cause of
the anisotropy.
–0.020” 1
The errors in plate thickness which were excluded
–0.015” 0 from analysis can have an effect on the mold construction as
–0.010” 2 well, as they affect hybridization with other components such
–0.005” 8 as the ejection system and any metal inserts secured to the
0” 21 base. For example, the ejector pins are mounted onto an
0.005” 6 ejector plate from below, which must be aligned such that the
More 0 pins are flush with the surface of the core; any variations in
the core thickness will require adjustment of the ejection
Table 5 - Histogram data for cavity and core offset errors system. It is therefore recommended to request finishing of
the insert bases from the service bureau or to finish them
Since there was a limit to measurement accuracy with
independently with a milling machine.
calipers for many of the dimensions, as well as a process
capability limitation outlined by the vendor (+/– 0.005”), all
5.1.2. Thermal Management
offsets which were greater in magnitude than 0.005” were
compared. What was found was that of these 13 larger The second-most important issue arising from the study was
offsets, 11 of the errors were from “female” type features, thermal management; the variability in temperature affected
consisting of inside diameters of holes or widths of cycle times, part quality, repeatability, and possibly tool life.
cavity/gap features. The other types of features which were Without an active cooling system, the cycle times in this
characterized were “male” features, which consisted of study were much longer than what could be expected from a
diameters and widths of protruding material, and “depth” metal tool, even one without active cooling as well. The
features, which were dimensioned along the layer axis (z- thermal conductivity of NanoForm is 0.09 lbs./s-˚F (0.7
axis). This characterization of the 41 parameters showed the W/m-K) (Appendix 9.2.1), compared to that of aluminum: 25
following trends: lbs./s-˚F (200 W/m-K) [17] – the ratio between the two is
• female features smaller than designed (all); greater than 275! An active cooling system, such as liquid
mean offset –0.008”, standard deviation of 0.006” coolant pumped through copper piping between the insert and
• male features very accurate (most); the base, may reduce the cycle time but could also introduce
mean offset –0.001”, standard deviation of 0.004” more stresses into the tool that would affect its life.
• depth features within 0.002” layer resolution (all The results shown in Figure 10 and knowledge of the
except plate thickness) thermal cycling amplitude suggest that large internal stresses
could have been occurring during the injection and cooling
• hole centers on target (x-y location)
cycles, since the thermal gradients were great and the cooling
rate was uneven within the mold. While fully constrained
Included as a subset of the depth features were also the
thermal expansion would cause compressive stresses which
primary vents, which were successfully built to one layer.
would not lead to crack propagation, differences in cooling
Conversation with an industry operator confirmed that they
rate would create tensile stresses which would lead to crack
had seen a similar issue in dimensional accuracy, though no
growth. It is very likely that such stresses would have caused
publications on the matter are known to exist at the time of
the two fractures pictured in Figure 12. Both of the cracks
this writing.
initiated from stress concentration points and traced along the
The resulting anisotropy in accuracy can be identified
outline between cavity and shutoff; the cooling rate at an
as the cause of the initial tooth block fracture; the smaller
Alexander K. Do Page 13 of 18

interface with the molded part would be expected to be strength (Appendix 9.2.1) – and the inherent anisotropic
slower than at the shutoff. nature of parts made from layer-based fabrication processes
Research by Ahrens, Riberio Jr. and Beal [18] to such as stereolithography.
improve cooling in SL molds has shown increased efficiency The core pin is believed to have fractured during
in cooling, but has yet to demonstrate improvements in ejection, and not during injection. While it is possible that
quality, repeatability, and tool life. Active pumping of heated the viscous flowing plastic could have initiated the crack on
liquid may help to stabilize the mold temperature as well. contact with the pin – the initial strategy of using low-energy
Ultimately, finite element analysis of the tool is necessary to parameters may have been detrimental, as the initial molding
understand whether active cooling and/or heating would temperature was relatively low – the even wall thickness and
improve or degrade the tool life. vertical nature of the through-hole in part #1 do not
Another alternative method to aid thermal corroborate this. The crack initiated at a stress concentration,
management of the mold is electroplating of the core and as it started the layer directly above the “fillet.” The word
cavity inserts. A process by RePliForm Inc. deposits 0.002” fillet is used colloquially due to the discrete reduction of
to 0.003” of copper and nickel on the top surface of an SL geometry to 0.002” layer thicknesses, especially apparent in
insert. Metal-coating of the core and cavity would not only very small fillets. The fracture angle of the core pin (straight
increase their impact and abrasion resistance, but it would across) in Figure 9 suggests that the crack propagated by
also transfer heat much better, likely reducing cooling time delamination, as only three layers were spanned from the
and thermal gradients. This would likely increase the tool point of initiation to the apex of the gouge. This is to be
life, but uniformity of the coating over sharp or narrow contrasted to more typical brittle failures in non-layered
geometric features is questionable. In addition, differences in materials which propagate at an angle of maximum shear
thermal expansion could lead to delamination due to fatigue. stress.
For these specific inserts, the cost is estimated to be less than An interesting observation is that the discoloration
$140 per insert, with an expected turn-around time of two shown in Figure 11 did not match up with any visible wear on
days [19]. the surface or to the distribution of temperature after ejection
shown in Figure 10. It does appear as though the
5.1.3. Cavity Size and Shrinkage discoloration corresponds to the high-stress locations in the
While the problems with dimensional accuracy prevented the cavity and core, as evidenced by the prominent discoloration
final part dimensions from being reasonably within about the gate and other low-thickness areas.
predictions, the 0.4% scaling factor used to design the cavity Because of the visible surface dewetting in the shutoff
agreed fairly well with the 0.40%, 0.41%, and 0.50% part areas, the flatness of the shutoff was questionable and may
shrinkage from the cavity size at room temperature. The have led to insufficient clamping, which could account for the
0.7% – 1.0% shrinkage values resulting from the heated flash problems. Three SLA operators confirmed that surface
cavity measurements are larger than the term produced by the dewetting is a known problem when highly viscous SL resins
ratio: are spread across large flat areas, and the results are even
(1 + C.T .E.Steel (Tm '−T0 ')) , more apparent in the ProtoTool material. One strategy to
mitigate this would be to reduce the shutoff area either by
(1 − S ASTM ) trimming the perimeter of the inserts or by creating a stepped
which should produce a shrinkage factor compensated for platform for the shutoff. This change would lead to a greater
mold material. Given the values determined in Appendix apparent clamping pressure from a reduction of the shutoff
9.1.2, the compensated shrinkage should be on the order of area. Another strategy would be to try an upright build
0.54%, which is comparable in magnitude but lower the 0.7% orientation, building the inserts on their sides such that the z-
- 1.0% attained values. Some deviation from the predicted height would be 4”, eliminating the surface dewetting on the
compensated shrinkage value is expected since the thermally flat surfaces and likely aiding ejection. However, the
insulating properties of NanoForm creates high variability in appreciable increase in z-height, and consequently build time,
mold temperature, resulting in local variations in cavity would increase the cost of the inserts to $1,600, 33% more
expansion and prolonged cooling times which could increase than when built flat. This orientation would also make the
shrinkage and/or sink. In addition, continued thermal vents difficult to build, as the beam tracing (0.003” beam
expansion of the cavity after the injected parts have set their thickness) is more difficult to consistently control than layer
shape is expected through continued heat transfer between the thickness.
hot plastic and the mold.
5.2. Process Repeatability
5.1.4. Tool Wear and Other Quality Issues
Many of the problems with repeatability are believed to have
The most prominent types of tool wear which affected part been caused by limitations in the equipment. Primarily, the
quality were the fractures of the tooth block and core pin. In Morgan Press does not support closed-loop control of
the tooth block, the fracture line shown in Figure 7 and cycles parameters other than the barrel and nozzle temperature. The
to failure (one) qualify a very brittle type of failure. This is other parameters are controlled by analog potentiometers or
consistent with both the brittle nature of the material – a high mechanical valves.
tensile modulus, low tensile strength and high compressive
Alexander K. Do Page 14 of 18

The problems with flash are believed to be a result of There are a number of particular strengths in the
the following factors: injection molding process investigated:
1. Insufficient clamping force. • The primary DFM restriction in high-resolution
The Morgan Press GT100 claims a 20-ton clamping mode is that feature sizes are limited by 0.002” layer
capability, but no more than 16 tons of force could thickness and limited by a 0.003” laser spot
be achieved according to the dial indicator. • Low cost and quick lead time of tool inserts
2. Uneven clamping force. • For a family of similar parts, a single mold base can
The off-center location of the nozzle seat in the be combined with different cavity and core inserts to
mold likely caused an uneven clamping force reduce the total cost of tooling
imparted from the platen assembly.
3. Poor control of shot weight. Of course, a number of weaknesses are specific to this
Since injection in a Morgan Press is driven by a process:
pneumatic ram and does not use a screw, it is • Dimensional accuracy, as discussed
impossible to independently control the volume of • Tool failure, while sometimes reparable, can be
the shot, the injection rate, and the injection instantaneous; areas of stress concentration can be
pressure. In particular, the ram sits above the barrel particularly vulnerable
of semi-molten plastic. Depending upon the exact • The maximum tool size at high-resolution is
amount and phase of plastic in the barrel, a different recommended 5” (L) x 5” (W) by 10” (H); this is
shot amount is delivered for the same injection time. meant to reduce parallax error, as a Viper can
4. Poor control of shot timing. actually build up to 10” x 10” x 10”
Using a digital timer or stopwatch to coordinate
• Crystallinity (and effectively shrinkage) in semi-
hand-operations on the machine was incredibly crystalline polymers cannot be easily controlled
difficult, as the difference between 2 seconds and 3
seconds was often the difference between short-
The final point can be a serious limiting factor in the types of
shots and excessive flash. materials which can be accurately and repeatedly molded.
Since a metal such as aluminum is 275 times more thermally
In addition to the flash, the second major theme in
conductive than NanoForm, the two materials will impart a
repeatability was mold temperature. As mentioned earlier, different thermal cycle upon the molded plastic and as a
unpredictable behavior of core and cavity temperatures
result a different percent of cyrstallinity, χc, in semi-
affected the part quality and cycle time, and therefore overall
crystalline polymers. As the percent cyrstallinity increases,
repeatability. Minor variations in sink in the final parts may not only will the mechanical properties of the molded part
be acceptable at a prototyping stage, but are not acceptable
change, but the final amount of shrinkage will increase as
for a high-yield, tight-tolerance project. For low part
well. Harris [2] showed that the resultant difference in
volumes however, longer cycle times do not pose a critical
shrinkage from SL tools is appreciable in a semi-crystalline
disadvantage to the process.
material such as polyamide 66 (Nylon PA66). In addition,
Finally, the gate thickness is believed to have limited
variability in temperature throughout the mold cavity and
the range of injection parameters under which the mold
core are likely to induce uneven cooling, and consequently
produced good parts. The gate thickness, which was less than
shrinkage.
30% of the wall thickness, was simply too small. This caused These unique process characteristics bring about a
the part to cool very fast. As the barrel temperature was
learning curve. As stated by DSM ProtoComposites
increased to reduce viscosity, the resulting increased
Marketing Manager Eva Montgomery, “tool design and
temperature variations in the mold caused repeatability issues processing optimization remain critical differentiators relative
such as increased sink and problematic ejection for
to meeting expectations” [20]. While it is the intention of this
insufficient cooling times. In this case, a larger gate would
research to provide information relative to these issues, much
have allowed for more a flexible window of processing of the process knowledge may not be transferable to or from
parameters.
other short-run or production-volume tooling strategies.
By addressing all of these issues with a larger gate, a
Even if success is achieved in a NanoForm process, it is not
more advanced molding press, and an active heating and believed that injection parameters can be carried over to a
cooling system, it is believed that part repeatability can be
production tool. However, since the NanoForm material
drastically improved.
properties are not as high-performing as those in metals, tool
design parameters such as gate and vent geometry may work
5.3. Process Viability and Quality equally well in a high-production tool.
If the repeatability can be addressed with more precise
molding equipment, nothing in this research indicates that 5.4. Comparison to the Industry Standard
injection molding with NanoForm will not work. 150 parts
The most popular and likely lowest-cost alternative to this
were produced on the tool, demonstrating a viable tool life
process in short-run injection molding is known as “rapid
and low cost-per-part ($16).
injection molding.” The term generally describes the
fabrication simple aluminum molds made by CNC-
Alexander K. Do Page 15 of 18

machining. The most prominent company in this area is first, measure the absolute offset, and compensate
Protomold [21], which offers a web-based quoting system the design of the core features accordingly. The
complete with DFM checks for undercuts, draft angles, wall expectation in this strategy is that the core features
thickness, and tool radius. The company has developed will be accurate enough to mesh with the known
software that automates the process planning portion of cavity size.
machining the molds, which is the most labor-intensive (and 5. Core Pin
expensive) part if done manually. In terms of cost and lead Core features with high-stress concentrations should
time, the short-run process in this experiment was be made out of aluminum and pressed into the base.
comparable with Protomold, which quoted $2,365 for 25 of These features could also be attempted from RP
the zonohedra tiles (at an equivalent surface finish) in 15 processes such as metalized parts.
days. Protomold typically charges $1,800 to $5,000 for a 6. Finishing
batch of anywhere from 25 to 250 parts, depending largely on The plate thicknesses should be finished to
part size, complexity, and finish requirements. Protomold specification to ensure compatibility with ejection
will also run larger batch sizes from 2,000 to 100,000+ parts. pins and other core features.
The largest caveat of the Protomold model is that the
customer does not ever receive or get access to the molding 6. FUTURE WORK
tool, which may be a critical part of the design evaluation The following are recommendations for further research work
process. Part redesign and continued molding favors the SL in the area of NanoForm injection molding tools:
process, since a new cavity or core insert could be built • FEA analysis of mold tool regarding stresses
independently for a relatively low cost. The natural tolerance • Fabrication of another set of NanoForm inserts using
of CNC machining is likely more accurate than that which lessons learned from the case study
was achieved in this study. However, the DFM requirements
• Use of a more precise molding press
for CNC-machined molds such as minimum radius (0.016”),
• Experimentation with cooling channels relating to
minimum wall thickness (0.035”), and minimum draft angles
repeatability and tool stress
(varies with depth) can impose on the designed geometry.
This process can be done in-house, given the right
expertise and equipment. However, it can be quite costly, as 7. CONCLUSIONS
most machine shops will not have software as advanced as
that which Protomold has developed. The local machine 7.1. Process Viability
shops quoted at least $5,000 to CNC-machine the zonohedra In this study 150 parts were injection molded from ABS with
mold with aluminum. Another consideration as well is that a tool built from NanoForm 15120 in high-resolution. The
during the course of the project, it was much more difficult to cost of the tool was less than $2,500 and the total lead time
locate, get information, and secure pricing from machine was less than three weeks, including molding time and costs.
shops for the aluminum parts than it was to perform the same Although only one of the parts was completely “usable” as
tasks for SLA services. This is likely because of the more molded (since all others lacked a through-hole), a large
“digital” nature of RP technologies over even CNC- number were post-processed and used to study design
machining. The SLA services also worked on a faster lead changes for a second-round. Despite the learning curve and
time than the machine shops. challenges, it is concluded that injection mold tooling made
from NanoForm 15120 is viable for use in short-run injection
5.5. Lessons Learned for the Next Iteration molding. It is believed that with fine control of process
The following are a list of simple design considerations parameters, the process repeatability will be acceptable for a
which were learned as a result of this project: majority of small to medium-sized parts (1 – 5 inch cross
1. Gating section). The process also compares favorably to accepted
The gate was too small. A gate thickness of 80% of alternatives in cost, lead time, and feature size.
the wall thickness will be used in the future.
2. Ejection 7.2. Additional Findings
Pushing down on the sprue during ejection was more One of the most important findings in this investigation is
difficult than expected. A negative draft in the cold that the SLA industry must account for the problem of
slug will keep the sprue and part on the core side anisotropy in dimensional accuracy. Until such time it is
during mold opening. An ejector pin will be placed inadvisable to use sliding shutoffs or other features which
at the cold slug, and the part will only be ejected require a natural process tolerance finer than (+0.014”, –
once the mold temperature is below Tg. 0.022”). Potential users should be aware that NanoForm
3. Shutoff Area tooling can undergo instantaneous failure from brittle crack
The shutoff area should be decreased to reduce propagation and/or delamination, especially at stress
surface dewetting and improve clamping pressure. concentration points. Finally, it is advised to use injection
4. Sliding Shutoff molding equipment with fine control of process parameters
At present the best strategy by which to fabricate the when working with NanoForm tooling.
sliding shutoff feature is to build the cavity insert
Alexander K. Do Page 16 of 18

9. SLA 7000 Data Sheet, 3D Systems.

8. ACKNOWLEDGEMENTS http://www.3dsystems.com/products/datafiles/sla700
For their contributions to this research, the authors wish to 0/datasheets/SLA_7000_system_rev_0204.pdf.
acknowledge the following individuals and corporations: 10. Materials Data Sheets page, DSM Somos.
o Dan Odell, UC Berkeley, http://www.dsm.com/en_US/html/dsms/pd_product_
o Xiaorui Chen, UC Berkeley, data_sheets.htm.
o Lisa Pruitt,UC Berkeley, 11. Bluestone Data Sheet, 3D Systems.
o Tim Nakari and Jason Boh, Protogenic, http://www.3dsystems.com/products/datafiles/accura
o Paul and Bret Bordner, Laser Reproductions, /datasheets/DS-Accura_Bluestone_nano-
o Brian Bauman, DSM Somos, composite.pdf.
o Mike Daniels, Production Robotics, 12. Williams, Jim and Kaufmann, Charles, “Injection
o Erik Appelbom, Jatco, and Molding with Composite Stereolithography Resins,”
o Tom Hagen, Poly-Tek SME Rapid Prototyping 2005, Dearborn, Michigan,
This project was funded by the Ford Motor Company, 2005.
CITRIS, and in by the California Energy Commission under 13. Quickparts. http://www.quickparts.com.
Grant Award DR-03-01. 14. Jacobs, P.F., “Recent advances in rapid tooling from
stereolithography,” Proceedins of 2nd National
REFERENCES Conference on Rapid Prototyping and Tooling
Research, Buckinghamshire College, 1996.
1. P.M. Dickens and P. Smith, “Stereolithography 15. D. M. Bryce, Plastic Injection Molding … mold
Tooling,” Proceedings of the 1st European design and construction fundamentals. Dearborn,
Conference on Rapid Prototyping, 1992, 309-317. Michigan: Society of Manufacturing Engineers,
2. R.A. Harris, H.A. Newlyn, R.J.M. Hague, P.M. 1998.
Dickens, “Part shrinkage anomalies from 16. Wright, Paul Kenneth, 21st Century Manufacturing.
stereolithography injection mould tooling,” Upper Saddle River, New Jersey: Prentice-Hall Inc.,
International Journal of Machine Tools & 2001.
Manufacture, 2003, Vol. 43, 879-887. 17. MatWeb, Material Property Data,
3. R A Harris, H A Newlyn and P M Dickens, http://www.matweb.com.
“Selection of mould design variables in direct 18. C. H Ahrens, A. S. Riberio Jr, V. E. Beal, “Heat flux
stereolithography injection mould tooling,” canals (HFC) technique: an alternative to cool down
Proceedings of the Institution of Mechanical stereolithography moulds,” Journal of the Brazilian
Engineers, Part B, Journal of Engineering Society of Mechanical Sciences and Engineering,
Manufacture, 2002, Vol216 No. B4, 499-505. 2003, Vol. 25, No. 3
4. R. Harris, N. Hopkinson, H. Newlyn, R. Hague and 19. From a presentation at the DSM Somos Rapid
P. Dickens, “Layer thickness and draft angle Tooling Roundtable session on June 8, 2005.
selection for stereolithography injection mould Information available directly from RePliForm Inc.
tooling,” International Journal of Production www.repliforminc.com.
Research, 2002, Vol. 40, No. 3, 719-729. 20. E. Montgomery, “Composite SL Materials Paving
5. Martyniak, S. A., “Prototype and Limited the Way Fore Advancements in Rapid Tooling,”
Production, Spray Metal Tools for Injection and Time Compression Technologies, May 2005.
Blow Molding,” Proceedings of the SME Rapid 21. Protomold. http://www.protomold.com.
Prototyping and Manufacturing ’93 Conference, 22. Morgan Instrustries, Cutting Costs in Short-run
Dearborn, Michigan, 1993. Plastics Injection Molding. Long Beach, California:
6. Gebhardt, Andreas, Rapid Prototyping, Concinatti, Twelfth Printing, 1998.
Ohio: Hanser Gardner Publications, Inc., 2003.
7. P M Hackney and K P Pancholi, “Analysis of the
application of the Z-Corps three-dimensional
printing system for rapid tooling for plastic injection
moulded components,” Fifth National Conference
on Rapid Design, Prototyping, and Manufacturing,
Centre for Rapid Design and Manufacture,
Buckingham Chilterns University College, and
Lancaster Product Development Unit, Lancaster
University, UK, 2004.
8. Viper Data Sheet, 3D Systems.
http://3dsystems.com/products/datafiles/viper/datash
eets/Viper_final_rev_0303.pdf.
Alexander K. Do Page 17 of 18

C0=Cd(1-SSL). (9.1)
9. APPENDIX However, the SLA vendor assured before the part would be
made that the design would not be needed to compensate for
9.1. Molding equations shrinkage, such that SSL could be set to zero in the prediction
of cavity scaling.
9.1.1. Clamp Force and Injection Pressure
The clamping force, Fc, required to keep a mold closed must Cavity Expansion
be greater than the distribution of the pressure of molten Differences in mold tooling materials necessitate that thermal
plastic in the mold over the projected area on the parting expansion be accounted for in order to more accurately
plane, A. The maximum pressure in the cavity is a product of estimate the cavity dimensions. The relationship between a
the injection pressure, PI, against a viscosity factor, η [22], cavity size at room temperature, C0, and the heated cavity
which is a function of not only time, but also geometry of the size, Ch, can be expressed as:
sprue, runner, and gate system. An equation for clamping Ch=C0(1+C.T.E.SL(Tm-T0)), (9.2)
force Fc is shown below: where C.T.E.SL is the coefficient of thermal expansion of the
stereolithography insert, Tm is the molding temperature, and
Fc ≥ ηPI • A [22] T0 is room temperature, assumed to be 70˚ F. While it is
Since it is difficult to characterize the viscosity factor, it is difficult to predict the actual temperature of the cavity during
more practical in industrial settings to set the clamping force molding, an approximation is sufficient since the final effect
to exceed the injection pressure against the projected area. A will be on order of 0.05% per 10 degree F. The Lustran 648
measurement in SolidWorks of the area of the molded part datasheet recommends molding between 110 - 150˚F,
projected onto the parting plane was approximately 3.5 in2. yielding a thermal expansion over that 40-degree range of
Due to the 20-ton clamp force limitation in the Morgan Press 0.2%. With an approximation of Tm = 130˚ F and T0 = 70˚ F,
GT100, the maximum injection pressure for this part before the calculation of thermal expansion based upon the
exceeding clamping could be 11,400 psi. properties listed in Appendix 9.2.1 is shown:
To ensure that the SL inserts would not fail during Expansion = 28x10-6 in/in-˚F (122˚ F – 70˚ F)
clamping and injection, a rudimentary comparison between + 47x10-6 in/in-˚F (130˚ F – 122˚ F) = 0.0018 in/in
these forces and the maximum compressive strength of the Based upon the likely molding range of 110 - 150˚ F, the
material is sufficient since tensile and shear stresses can be predicted thermal expansion of the tool is 0.18 ± 0.1%.
neglected in the loading configuration of the inserts. The
maximum compressive strength of NanoForm 15120 as Part Shrinkage
reported by Somos (Appendix 9.2.1) is 34,000 psi. A Part shrinkage, measured by ASTM D955 standards, is
measurement of the total shutoff area (area loading the inclusive of the thermal expansion of the molding material
clamping force) for the cavity insert in SolidWorks was (usually hardened steel). The part shrinkage, Spart, is related
reported as 16.6 in2. At 20 tons of clamp force, this creates to the original cavity dimensions and final part dimensions
an average normal stress of 2.4 kpsi, allowing a rather large through the relationship:
FOS (factor of safety) > 10. Simply comparing the maximum Pf=C0(1-Spart), (9.3)
injection pressure of 16,300 psi, corresponding to a plastic where Pf is the final part dimension.
pressure of 11,400 psi in the cavity area to the maximum Since Harris [2] showed that there can be some
compressive strength allows for an FOS of approximately 3. discrepancy between measured shrinkage values due to
A more rigid analysis of stresses in the cavity during injection thermal expansion, an adjusted shrinkage, Sadj, which
would require finite element analysis due to the complexity in accounts for thermal expansion of the steel must be used to
geometry. The actual injection pressure used was in the predict the thermal contraction of plastic from barrel
region of 6,000 - 9,000 psi, and a maximum clamping force temperature to room temperature:
of 16 tons was applied to the mold. Pf=Ch(1-Sadj), (9.4)
where Ch is the heated cavity dimension at molding
9.1.2. Calculation of Cavity Size temperature.
There are three independent volumetric changes of the cavity Substituting 8.2 into 8.4, we get:
which affect the final part size: Pf=C0(1+C.T.E.SL(Tm-T0))(1-Sadj). (9.5)
1. Tool shrinkage after curing Combining the two RHS equations from 9.3 and 9.5 yields
2. Thermal expansion of the cavity during molding the following relationship between Spart and Sadj:
3. Part shrinkage from above Tg to room temperature 1-Spart=(1+C.T.E.SL(Tm-T0))(1-Sadj). (9.6)
The adjusted shrinkage can now be expressed in material
Tool Shrinkage After Curing properties which can be measured or are already known, Spart
Stereolithography parts are known to shrink after final curing. and C.T.E.SL:
DSM Somos reports that NanoForm has final cure shrinkage 1 − S part
S adj = 1 − . (9.7)
(SSL) between 0.001-0.0015 in/in [20]. The relationship
1 + C.T .E.SL (Tm − T0 )
between the designed cavity size, Cd, and final cavity size at
room temperature, C0, can be expressed as:
Alexander K. Do Page 18 of 18

However, we cannot determine the part shrinkage before the 9.2.2. Lustran ABS 648
tool is made, so we must use another set of known properties. Information obtained from Lanxess Lustran datasheet
Without loss of generality, this relationship can be held true
for any molding material. In particular, if the molding
material is a hardened tool steel, the Spart becomes the ASTM Typical Injeciton Molding Conditions
D955 published shrinkage SASTM: Barrel Temperature at Nozzle 475 – 500 °F
1 − S ASTM Melt Temperature 475 – 510 °F
S adj =1− , (9.8)
1 + C.T .E.Steel (Tm '−T0 ') Mold Temperature 112 – 150 °F
Injection Pressure 10,000 – 16,000 psi
with Tm’ and T0’ as the range of temperatures used in the steel
rather than SL mold. Injection Speed High

Final Equation Typical Properties for Natural Resin

Combing equations 9.1, 9.5, and 9.8 algebraically, we obtain Mold Shrinkage (ASTM D955) 0.004 – 0.006 in/in
the following relation: Tensile Modulus (ASTM D638) 340 ksi
(1 + C.T .E.Steel (Tm '−T0 ')) Coefficient of Linear Thermal 51 x 10-6 in/in/°F
Cd = Pf . (9.9) Expansion (ASTM D696)
(1 − S SL )(1 + C.T .E.SL (Tm − T0 ))(1 − S ASTM )
While we do not know the coefficient of thermal expansion
of the actual tool used in the ASTM shrinkage tests, we can
estimate that for a standard molding steel such as P20 tool
steel which is commonly used for cavities and cores [15], the
CTE will be approximately 7x10-6 in/in-˚F [17], and molding
temperatures Tm’ and T0’ will be on par with Tm and T0. This
yields a total thermal expansion of 0.042%. At this point, all
the parameters which are required to predict the required
cavity scaling above the designed part dimensions are known.

9.2. Material Properties

9.2.1. Somos NanoForm 15120

Information obtained from DSM Somos

Property ASTM NanoForm

Method 15120 UV +
Thermal
Postcure
Tensile Strength D638M 7.7 ksi
Modulus of Elasticity 856 ksi
Flexural Strength D790M 18.7 ksi
Flexural Modulus 645 ksi
Compressive Strength D695-02a 34 ksi
Compressive Modulus 678 ksi
CTE 32°F – 122°F E831-00 28 x 10-6
in/in-°F
CTE 122°F – 212°F 47 x 10-6
in/in-°F
Tg E1545-00 176 °F
HDT @ 66 psi D648-98c 517 °F
HDT @ 264 psi 240 °F
k (Thermal Conductivity) 0.09 lbs./s-˚F
(0.7 W/m-K)