Composites Extruder Head Development - Final Report (7)

Composites Extruder
Head Development
Final Report
Colin Biery (720)216-7625 handsvod@rams.colostate.edu
Ryan Dunn (303)229-8358 rysdunn@gmail.com
Michael Hansen (720)427-1687 mikelangelo.mh77@gmail.com
Logan Rutt (303)495-8382 lrutt@rams.colostate.edu
Tristan Vesely (925)876-2343 tvesely@rams.colostate.edu
Colorado State University, Mechanical Engineering,
Senior Practicum Projects Program
April 28, 2016
Advisor: Dr. Don Radford

2
Table of Contents
Tableof Contents............................................................................................................................................................................................................................................................................2
ExecutiveSummary......................................................................................................................................................................................................................................................................2
Acknowledgements.......................................................................................................................................................................................................................................................................5
Introduction.......................................................................................................................................................................................................................................................................................6
Background........................................................................................................................................................................................................................................................................................6
Composites Properties........................................................................................................................................................................................................................................................6
Composites Manufacturing...............................................................................................................................................................................................................................................6
AdditiveManufacturing......................................................................................................................................................................................................................................................7
Current Solutions:..................................................................................................................................................................................................................................................................7
Design Problem Analysis...........................................................................................................................................................................................................................................................8
Problem Statement................................................................................................................................................................................................................................................................8
Objectives....................................................................................................................................................................................................................................................................................8
Design Constraints.................................................................................................................................................................................................................................................................8
Design Summary.............................................................................................................................................................................................................................................................................9
Final Design and Analysis.......................................................................................................................................................................................................................................................10
Fiber Placement System..................................................................................................................................................................................................................................................10
Fiber-Filament Integration............................................................................................................................................................................................................................................11
Pellet Stock Integration...................................................................................................................................................................................................................................................13
NozzleMaterial Selection...............................................................................................................................................................................................................................................13
Integrator Block Material Selection.........................................................................................................................................................................................................................14
Thermal Insulation Block Material Selection.....................................................................................................................................................................................................14
Matrix Selection forIntegration System Material Selection.....................................................................................................................................................................15
Safety Considerations.......................................................................................................................................................................................................................................................15
Design for Manufacturingand Assembly..............................................................................................................................................................................................................15
Evaluation........................................................................................................................................................................................................................................................................................16
Testingand Refinement..................................................................................................................................................................................................................................................16
Results........................................................................................................................................................................................................................................................................................19
Manufacturingand Assembly Evaluation.............................................................................................................................................................................................................25
Unit Cost Analysis................................................................................................................................................................................................................................................................26
Project Development Cost..............................................................................................................................................................................................................................................27
Deviations from Original Plan.....................................................................................................................................................................................................................................27
Conclusions and Recommendations................................................................................................................................................................................................................................29
Appendices Appendix A: ASTM D2344 – Short Beam Shear TestingBrief......................................................................................................................................31
Appendix B: ASTM D7264– Flexural Stiffness TestingBrief...................................................................................................................................................................32
Appendix C: Mechanical Testing Results..............................................................................................................................................................................................................33
Appendix D: Hypothesis TestingResults from Minitab Statistical Software..................................................................................................................................36
Appendix E: Composites MoldingSOP...................................................................................................................................................................................................................38
Appendix E: Maximum bendingstress of glass fibers calculation........................................................................................................................................................40
Appendix F: CambridgeEngineeringSelector Material Selection Stages of Nozzle...................................................................................................................41
References........................................................................................................................................................................................................................................................................................42

3
Executive Summary
Composite materials are useful because of their high stiffness and low weight when
compared to more common metals or polymers. Composites are a combination of a more
ductile polymer and a stiffer fiber reinforcement. The fiber strengthens and stiffens the
material while the polymer provides toughness and protects the fiber from damage and
transfers stresses to the structurally capable fibers. Currently, composites are difficult and
expensive to produce, particularly for one off, or low volume parts. They require special
molds manufactured to a high tolerance. Furthermore, complex composite parts are often
manufactured by hand in a time consuming process. The expense of composite parts has
limited their availability and application in engineering.
3D printing produces parts by building up layer upon layer of material using a computer
controlled machine. Fused deposition modelling (FDM) is the most common, low cost, form
of 3D printing. It uses a heated extruder to deposit thermoplastic in layers in order to
create a part. FDM allows parts to be produced with almost no overhead cost and with no
tooling needed to define the geometry. This allows the quick, inexpensive creation of
prototype and low volume parts. However, because the parts are thermoplastic, they are
limited by the mechanical, and thermal, properties of the material. This greatly reduces the
possible applications of FDM manufacturing.
The goal of this project was to use a modified FDM process to create composite parts by
designing new extruder heads capable of inserting continuous reinforcing fibers into the
thermoplastic layers. This new method of composites manufacturing would be more
flexible than traditional molding techniques and would increase the amount a part’s fiber
orientation can be tailored to a specific application. Finally, the process would be designed
to fit onto a commercial FDM printer in place of a standard extruder head and to use
multiple types of fiber and thermoplastic feedstock in order of decreasing financial cost
from commingled tow to 3D printer filament with dry fiber, and finally thermoplastic
pellets and dry fiber.
Creating a new method of composites manufacturing was a daunting challenge which was
broken down into three smaller segments to make manageable. The first challenge was to
determine a method of depositing layers of fiber onto the flat build plate used by an FDM
printer using a material stock in the form of commingled tow, which comes as intermingled
strands of fiber and thermoplastic. The first section of the new extruder head design
included special nozzles for a standard hot end extruder being machined that pushed the
pre-mixed fiber and thermoplastic together using a tapered channel. The outlet of the
nozzle was rounded which allowed a printer to operate with the extruder at 90 degrees to
the build plate without the fiber being broken on a sharp corner. Lastly, the nozzles had a
large flat area on the bottom which applied heat and pressure to the layers and forced them
together.
Using these extruder head nozzles it was determined that a composite of comparable
quality to molded composites could be printed with commingled tow (a mixture of glass

4
fiber and thermoplastic strands). However, applying enough pressure to successfully
consolidate the layers caused a loss of control over the layer’s geometry. This led to a
tradeoff between the strength and stiffness of the final part and its shape.
Once a method to deposit layers of thermoplastic with reinforcing fiber was determined
the next challenge was to develop a means of combining dry fiber and the thermoplastic
filament normally used in FDM printers. An integrator block was designed and
manufactured to combine the materials. The dry fiber is fed into the top of the block and
travels through before exiting out the nozzle at the bottom. About midway through, the
fiber channel is intersected by another channel through which viscous thermoplastic is
forced by a commercial hot end extruder. The theory was that if the thermoplastic was of a
low enough viscosity it would be pushed in between the fibers. However, this proved to be
a challenge as the fibers tended to stick to just one side of deposition. This led to prints
which had clear, dry, bundles of fiber surrounded by plastic rather than the being evenly
distributed. The results were parts with poor properties due the inexistence of the matrix
material between fibers to transfer stresses throughout the material.
The final segment of the problem was designing a compact means of combining
thermoplastic pellets with the dry fiber to create a composite. To achieve this a drive
system was designed which attached to the thermoplastic channel of the integrator. The
extruder uses an auger to force pellets through a heated barrel. The barrel applies heat
until the thermoplastic achieves a low enough viscosity enough to be pushed through the
channel into the integrator. The operational function of the pellet drive system was
successfully tested as it heats up and drives thermoplastic down into the integrator.
In order to determine the quality of a printed composite mechanical testing was used. A
short beam shear test (ASTM D2344) was used to determine how well the layers were
consolidated together while a flexural stiffness test (ASTM D7264) provided values for
stiffness and gave an indication of the effectiveness of a load transfer between fibers and
matrix. The printed beams of a standardized size were compared to compression molded
beams of the same size. The results showed that it was possible to get comparable
properties to the molded beams using commingled tow and the tapered nozzles. Photo-
microscopy was also carried out in order to examine the dispersion of fibers within the
composite samples. Samples printed using commingled tow showed good wetout
illustrated by fiber dispersion within layers but clear bands of thermoplastic between the
layers. The samples printed using the integrator had clear bundles of still dry fiber rather
than a dispersion of fiber through the print.
Ultimately the attempt to combine composites with FDM printing yield successes and
difficulties. Quality composites were produced using commingled tow but at the expense of
geometric control. The fiber-filament integrator did print out the two material in one layer
but they were poorly wetout and the resulting part had poor properties. Lastly a compact
pellet extruder was designed and manufactured but a tight schedule and 3D printer issues
kept it from being tested.

5
Acknowledgements
The Composites Extruder Head Development Team would like to thank the following
organizations and individuals for their contributions to the success of these findings:
Dr. Don W. Radford
Dr. Mitch Stansloski
Kent Warlick
Patrick Rodriguez
Patrick Jackson
Kevin Hedin
Paul Colasuonno
Coroba Plastics
Colorado Waterjet

6
Introduction
Fiber reinforced thermoplastic composites are incredibly useful materials due to their
impressive specific modulus as well as their specific strength. Specific modulus is measured
by Modulus of Elasticity divided by density, and specific strength is measured by tensile
strength divided by density. Unfortunately, composite manufacturing is a difficult and
costly process that makes composite parts expensive. In contrast additive manufacturing is
a relatively simple manufacturing process, but creates weaker parts. Combining the ease of
additive manufacturing techniques with the performance of composites would enable
designers to rapidly create components that meet structural requirements through a
procedure which allows complex shapes. This will eliminate lag time for prototypes and
reduce market-level manufacturing times. The proposed method is an extruder head
capable of manufacturing consolidated thermoplastic composites, with continuous
reinforcing fiber, through 3D printing.
Background
Composites Properties
Composite materials offer mechanical properties that traditional materials cannot compete
with. They have remarkable durability and resistance to fatigue [2]. Continuous fiber
reinforced thermoplastic matrix composites function by transmitting external energy
through the thermoplastic matrix material to the hard, brittle fiber reinforcements. The
fibers take the applied load while the matrix protects them from damage. Properties of
composites depend on the properties of the matrix, reinforcement, and the ratio of matrix to
reinforcement, which is stated as the fiber volume fraction [2,8].
Fiber orientation is one factor that influences the properties of a composite. Fibers are
categorized by their aspect ratio (length divided by diameter), where continuous fibers
have high aspect ratios [1]. Composites are most effective when fibers are continuous and
aligned, increasing their ultimate tensile strength and stiffness. Continuous fiber
composites have anisotropic material characteristics, and fail at lower stress values when
transversely loaded [2,8].
Consolidation is an important issue when dealing with composite materials. Consolidation
describes how effective the thermoplastic is at wetting all of the fibers. Proper
consolidation uniformly arranges the fiber reinforcement throughout the material.
Transfer of energy between the matrix and fiber is accomplished through proper wetting of
the composite. Proper wetting is also critical to adequate bonding between the matrix and
fibers [2]. Inadequate wetting out of the composite results in insufficient mechanical
properties. Without the fibers the thermoplastic has a much lower strength and a lower
modulus of elasticity. If the fibers are not distributed evenly through the thermoplastic
matrix consistent material properties throughout the composite are not achieved.
Composites Manufacturing
The manufacturing process for composites can be costly and time consuming.
Manufacturability is a limiting factor for commercialization of these materials, as the

7
process requires multiple steps and bulky molds
[7]. The tooling required to create composite
components is expensive to design and
manufacture and is not adaptable. The
manufacturing process also often requires time
consuming manual operations [6].
Additive Manufacturing
Additive manufacturing (AM) refers to the process
of building 3-D objects by adding layer upon
layer of material [10]. The most common and
commercially available type of AM is fused
deposition modeling (FDM). FDM generally uses thermoplastic filament as the stock
material. The filament is fed into a heated extruder head where it is melted and then
extruded onto a base plate. Currently, most of these printers move in the x-y plane to create
a layer and then move in the z-direction to begin printing the next layer [10].
FDM manufacturing requires no tooling or user interaction to create finished parts. 3D
printing software reads stereolithography (STL) files and generates G-code directly from
them. This form of AM is extremely useful for developing geometries; however, it is at a
disadvantage when developing structural parts due to both the inherent limitations of the
thermoplastic polymer and the incomplete fusion between subsequent layers.
Current Solutions:
There are several ways that composites are being
implemented into AM. These include using hot
end extruder heads to pull and consolidate fibers
[7], using plastic filament pre-impregnated with
chopped fibers [4], laying printed plastics and
fibers in series using multiple extruder heads [3],
and using robotic controls in combination with
additive technologies to produce continuous
fiber reinforced thermoplastics [12].
A laboratory scale extruder head, developed in
Switzerland, is capable of processing continuous composite lattice structures [7]. The
method of manufacturing uses a two-stage extrusion head to manufacture the composite
seen in Figure 1.
There are multiple companies that are selling thermoplastic filament pre-impregnated with
short chopped fibers [4]. This filament can be used in many printers but does not benefit
properties to the same level as traditional composite manufacturing with continuous fiber.
There are few commercially available options for printing continuous fiber composites [3].
The most well know being the Mark One, by MarkForged [3]. It uses a dual head extruder
system to print nylon with one head and commingled glass and nylon with the other. This
method consolidates when the fiber head passes over previously printed layers of nylon.
Figure 1 – Commingled tow extruder head developed by ETHZ
Structures [7]
Figure 2 - Mark Forged MarkOne Printer [3]

8
Design Problem Analysis
Problem Statement
Composites production is a time-intensive and expensive process when creating
complicated parts. Molds are created for a single part and cannot be used to manufacture
anything else [6, 8] and most composites are laid up into the mold by hand. Manufacturing
the molds and tooling is difficult and requires a high level of precision in order to create
quality composites.
Additive manufacturing is a rapidly growing field that is continually incorporating new
methods of production and new material choices. Fused deposition modeling (FDM) is
incredibly easy to use and can create unique, and complex, shapes for virtually no overhead
cost [10]. It is versatile and capable, but the parts created are weaker than thermoplastic
parts created with traditional methods [14] and therefore not useable in many
applications.
Being able to produce composite materials via additive manufacturing, particularly FDM,
creates opportunities to save money by avoiding costly production techniques. Companies
utilizing composites stand to benefit from the application of additive manufacturing
because they are paying for the expense of current production methods. One of these
companies, Boeing®, uses carbon fiber thrust reversing cascade baskets for their jet
engines. There is one company in the world which produces the majority of baskets and
they use an expensive hand-laying process. The proposed solution is to quickly build
effectively consolidated composites, which are mechanically comparable to standard forms
of composite manufacturing, using a modified fused deposition modeling process.
Objectives
Table 1 - Design Objectives
Objective Name Priority* Method of
Measurement
Objective
Direction
Target
Consolidation 5 ASTM D2344 –
Short Beam Shear
Test
Compare material
properties to
molded
equivalent values
Properties are
comparable to that of
molded equivalent [9]
Adjustable Fiber
Volume Fraction
4 ASTM D2584-
Resin Burnout or
Photo Microscopy
Controllable %
Fiber per Volume
or Area
Up to 50% fiber by
volume
Composites Stiffness 4 ASTM D7264
Flexural
Properties Test
Maximize Properties are
comparable to that of
molded equivalent [9]
Operating
Temperature
3 Optical Inspection
of Printed Beams
Optimize Balance between
viscosity and
degradation
Deposition Rate 2 Deposited weight
per time
Optimize Optimal rate for
consolidation
* Priority is weighed on 1-5 scale with 5 most important

9
Design Constraints
Table 2 - Design Constraints
Constraint Method of Measurement Limits
Material Stock Form Thermoplastics and
Reinforcing Fibers Stock
Commingledtow, TP filament, dry fiber,
TP pellets
Hot End Nozzle
Size
Dimensions (mm x mm x
mm)
54 x 65 x 65
Hot End Temperature
Capability
HeadTemperature (degrees
C)
Up to 500 C
Commercial Software Compatible slicing and
controls software
Marlin, Cura, Slic3r
Modification Limits Compatibilitywith standard
formsof FDMstock material
Headcan still effectively print parts with
standard thermoplastic filament and
standard filament with chopped fiber
Budget Dollars Spent $2000
Safety Possibility of Serious Injury 0
Design Summary
The ultimate purpose of this project is to develop a new extrusion system that produces a
high quality composite by utilizing fused deposition modelling technology. In order to
tackle a project of this scope and complexity it was necessary to break it into smaller
problems which could be solved one at a time. The different sections of the project were
developed in order from the nozzle back to the material infeed.
Based on current research, some commonly used engineering analysis was deemed
unnecessary to predict the performance of the extruder heads. Current Finite Element
Analysis is not where it needs to be to accurately predict mechanical performance of 3D
printed components, including 3D printed composite components [19]. After researching
current attempts of 3D printing composites, it was found that the best approach to design
was determined through experimental procedures [13, 14, 16, 18, 19, 20]. In addition, it
was seen that modeling the behavior of molten matrix surrounding reinforcing fiber to
predict wet-out, consolidation, and void creation through Computational Fluid Dynamics
(CFD) was beyond the scope of this project. There were no resources found that predicted
this behavior through CFD to create a better design [7, 13, 14, 16, 18, 19, 20]. Therefore, in
place of a conventional finalized design concept and analysis, the Composites Extruder
Head Development Team took the path of experimental analysis to determine how each
component of an extrusion system behaves in regards to composite materials.
The first problem was how to effectively layout and consolidate composite layers onto the
flat build plates used by FDM printers. The solution to this problem was to use specially
machined brass nozzles fitted onto commercially purchased hot end extruders.
Commingled tow was used in order to test the fiber placement. The nozzles were EDM
machined with a tapered channel which forces fiber and plastic together. Each nozzle also
has a rounded exit so that the head can print at a 90 degree angle to the build plate. The

10
large flat area on the bottom of the nozzle applies pressure and heat to the layer to create
consolidation.
The second problem was how to combine dry fiber and thermoplastic in a controlled fiber
volume fraction. An integrator was designed in which fiber was pulled through a central
channel. Another channel meets the fiber channel at a 45 degree angle. A commercial hot
end forces viscous thermoplastic through the channel so that it meets with the fiber and is
pulled out through one of the specialized nozzles on the bottom of the integrator.
The final problem that needed to be solved was how to use lower cost forms of
thermoplastic feedstock in the process. The thermoplastic filament used in FDM must be
manufactured and thus there is an added cost and limitations on availability of various
thermoplastics. An extruder was designed which utilized an auger as a lead screw which
draws pellets from a 3D printed hopper. The auger forces pellets through a heated barrel
which melts them. The melted thermoplastic is then further forced into the integrator to
combine with the dry fiber.
Final Design and Analysis
Fiber Placement System
The goal of the fiber placement system is to provide a method for consolidating the
composite as it exits the head in a controlled and quality manner. It seems counter-intuitive
that one could lay out fiber perpendicular to a build plate due to the likelihood of the fibers
fracturing after heating the matrix material up around them. However, after analysis it was
determine that creating a radius at the outlet that doesn’t allow the max bending stress to
act on the fibers would be ideal. Refer to appendix D for the detailing of the mathematical
analysis used to determine that a radius of 0.18 mm is the smallest corner that the fibers
can safely travel around.
Another challenge that arose
when testing the fiber placement
system was that in traditional 3D
printing nozzles, the channel is
stepped down to final outlet
diameter with chamfered edges.
Although this is fine for extruding
thermoplastic, the fibers catch on
the edges and abrade the brass
and also bend the fibers leading
to clogging of the nozzle. The
first mitigation plan was to not
step down the nozzle and leave it as a 2 mm channel throughout. This was successful as
fibers were laid out continuously and reaming the nozzle channel allowed for a high
surface finish and a reliable operation life of the brass. Having a nozzle outlet diameter this
large however, meant that there was less geometric control over the deposition. To
overcome both the challenges of a large exit diameter and stepped channels, the nozzle was
Figure 3 - Nozzle manufacturing drawing showing radiused outlet and tapered
channel

11
machined using a wire electrical discharge machining method to achieve a tapered channel.
This was the most successful design because based on optical inspection, the material being
extruded was coming out of the nozzle smoothly and therefore less force was needed to
tack down the deposition as it turned corners.
As results obtained from optical
microscopy discussed in later sections
show, the material being extruded using
a tapered and radiused nozzle allowed
for adequate consolidation based on the
flat of the nozzle applying pressure and
also the brass reheating previous layers
before depositing on top of them. A
layering effect was still observed as is
became clear layer to layer consolidation
would not be as impressive as in
composites molded and pressed together
although this was an expected result. Based on the successes of the fiber placement system
in extruding commingled tow onto a flat build plate, it was determined that the system was
one that did not have to be modified as the team moved into the second iteration of the
design and began combining dry fiber with thermoplastic filament.
Fiber-Filament Integration
To continue to lower the cost of composites production it is necessary to move away from
using the more expensive and less available commingled tow and begin integrating dry
fiber and thermoplastic filament. The integration point was designed as an intersection of
two channels that the fiber and filament travel through respectively. Two hot ends are
utilized in the design, one that comes as a part of the commercially produced Pico hot end
that heats up the matrix material to a lower viscosity for wetting out and a second located
near the nozzle outlet to control the temperature of the deposition to allow for quality
printing. Through testing and thermal analysis, best results were obtained when the hot
end operated at 290o C and the outlet temperature is kept at 290o C.
Modified barrels that are often utilized in commercial 3D printing applications were
utilized as screws that allow for the hot end and fiber inlet to be attached to the block. This
cut down on complexity of machining precision mounting screws and allowed for easy
sourcing of parts. A concern was having the surroundings of the extruder head heat up to a
point of degradation when running at high temperatures. To control the amount of heat
flowing from the extruder head, a ceramic block was designed to thermally isolate the
system. The material was a glass-mica ceramic exhibiting high machinability but care was
taken in design to ensure there were not any features that would have created stress
concentrations or features too difficult for traditional machining processes.
A small hole was drilled into both the integration block and the thermally isolating ceramic
that were connected with a small dowel pin. The purpose of the pin was to eliminate
rotation of the integrator due to torsion on the head. When press fit in, the two parts were
Figure 4 - Fiber placement system laying out continuously
reinforced commingled tow onto flat build plate

12
allowed almost no movement creating a singular head that didn’t exhibit any unwanted
rotation. The mounting system for the integrator head was two-fold as it needed to be
attached to both the Zmorph printer and the gantry printer manufactured by Team
Cascade. Mounting to the Cascade printer was simplistic in nature as the space available for
placement was large but mounting to the Zmorph required 1/8” aluminum plates to be
manufactured. The plates were cut with a water jet and allowed for a simple replacement
of the carriage already on the Zmorph for traditional print heads.
Dry fiber comes commercially available with a chemical coating put on it called sizing for
better handling, inter-constituent bond strength, and wetout. The fiber used in testing was
sized for epoxy thermosets and therefore bonding the fiber to a thermoplastic matrix
becomes a challenge. Unfortunately, fiber sized for thermoplastics has only recently
become available and wasn’t available to the team for testing. A three dowel pin system
shown in figure 8 allows the fibers to break from their sizing and spread out. This system
was designed to sit atop the ceramic thermal isolation block in line with the fiber infeed. In
addition, this three dowel pin system acted as a small tensioning system. Apply small
tension applied to the line of fiber minimized broken fibers catching along the channel wall,
and reduced clogging of the extrusion process as observed during testing.
The temperature of the extruder head is controlled by a PID control system that is
implemented in the configuration file of the printer. The values for the PID control were
obtained by running a calibration test for each separate heating source. The calibration
values for the PID were then tested with a K-type thermocouple and proved to be accurate
for the extrusion process. Refer to Appendix I for detailing views and parts list of
integration system.
Figure 5 - CAD cross-sectional model displaying
integration point
Figure 6 - Final extruder head product with Zmorph mounting plates

13
Pellet Stock Integration
Thermoplastic filament used in 3D printing is one of the more expensive forms of plastics
commercially available. In order to continue to reduce manufacturing costs, the Pico hot
end was replaced with an extrusion system that utilizes plastic pellet feedstock instead of
plastic filament. This system uses a method of extrusion similar to current injection
molding. A quarter inch auger screw was modified to act as the lead screw in the extruder.
Pellets fall into a heated aluminum barrel from a 3D printed hopper where the auger drives
them through a heated barrel. The auger screw is driven by a planetary NEMA-17 stepper
motor with a 5:1 gear reduction ratio to increase the torque supplied. The barrel interfaces
with the fiber-filament integration block previously described.
Nozzle Material Selection
The nozzle of the extrusion process has a great importance in applying consolidation heat
and pressure. Selecting a material for the nozzle was crucial in analyzing all factors that
would positively affect its consolidation ability. The nozzle must have good thermodynamic
properties, must resist wear, and must be readily machinable. The computer application,
Figure 7 - Fiber filament integrating head laying out dry fiber and
ABS on Zmorph printer Figure 8 - Fiber in spread state going
through sizing break
Figure 9 - CAD model of pellet extrusion system
with mounting
Figure 10 - Final product pellet
extrusion system
Figure 11 - Hopper and auger system onpellet
extruder

14
Cambridge Engineering Selector, was created to select the correct material for an
application. The following outlines the material selection process resulting in the team’s
choice of brass as the most suitable nozzle material.
Stage one of the selection process aimed to set limits to values involving thermal properties
and cost. Minimum service temperature, thermal conductivity, and thermal expansion were
the main factors the team pinpointed as the most critical so that the material still
transferred heat at high operating temperatures without deforming too much. Stage two of
the selection process aimed to ensure the material chosen has a good level of machinability
by analyzing its surface hardness and ductility. Stage three of the selection process
analyzed how well the material resisted wear in the desired nozzle shape. The graph put
materials into categories of wear resistance ranging from ‘very poor’ to ‘very good’ and
only the materials in the very good were chosen. After running all three stages, the
applicable materials list was narrowed down to two: aluminum-bronze and brass. Based on
procurement availability, it was decided that brass was the best material to proceed with.
The beneficial results of the CES analysis are two-fold. Most importantly it proved that
brass was the ideal material for the team to use for the nozzle based on its ability to be
machined, its high thermal properties, and resistance to wear. It also is important because
brass is already one of the most widely used nozzle materials and has stock parts easily
accessible. This allows the team to order stock brass nozzles and machine them to desired
geometries instead of having to order custom nozzles made out of a less accessible
material. This helps cut down on material cost, manufacturing cost, and lead time.
Integrator Block Material Selection
The integrator block is the key component that combines the matrix material with the
reinforcing fiber. Key requirements for this component are abrasion resistance, high
machinability, and adequate heat transfer coefficient. For this iteration in the design
process, it is important to test concepts for the in-head consolidation method. As a result, a
low cost material selection is necessary due to the high probability of future redesign.
Aluminum 6061-T6 was the material selected create the integrator head. 6061-T6 is an
artificially aged aluminum alloy that is highly machineable, low in cost, and processes heat
transfer characteristics that are adequate to test the integration concept. This material
would not be adequate for manufacturing carbon fiber PEEK composites. The
recommended temperature for 3D printing PEEK is 500 centigrade, which is over %50 of
the melting temperature (in Kelvin) for aluminum, and could result in creep failure of the
extruder head. Once a proof of concept is finalized, 416 free machining stainless steel
would be an adequate material selection for carbon fiber PEEK.
Thermal Insulation Block Material Selection
The mounting system for the Zmorph printer requires thermal isolation of the gantry
system. The insulation block requires a material that has low thermal conductivity and a
material that is also machineable. Ceramic glass mica was selected for this component for
its thermal capabilities and machinability. This ceramic is capable of being machined using

15
traditional manufacturing techniques, and provides thermal resistance to delay heat
transfer to the gantry system.
Matrix Selection for Integration System Material Selection
Due to the small size and geometry of the fiber-matrix integration system, the matrix
material must be easily removed from the components. Acrylonitrile butadiene styrene
(ABS) chemically reacts with acetone and dissolves the thermoplastic buildup in the
integration head. This matrix material was selected for this reason. Using ABS and acetone
allows for adequate cleaning of the matrix and fiber channels. The modulus of elasticity of
the thermoplastic matrix are extremely low when compared to the reinforcing fiber, so the
substitution of ABS for polypropylene as the matrix material should not show difference in
stiffness of the composite.
Safety Considerations
There were 3 main safety risks included in this project. The first was the danger of working
in a machine shop; many of the parts had to be designed and manufactured in the EMEC,
with large machinery that if not treated responsibly can pose huge dangers to the
individual working on it. Extreme care was taken in machining all parts. The next safety
risk was the individual moving parts that the two printers have. If care was not taken
around the stepper motors and belts, they could pose the risk of catching hair, fingers, or
other body parts in them. This was mitigated with full enclosures that surrounded each
printer, isolating the moving parts and protecting the user. The final safety consideration in
our project was the high temperatures reached by the 3d print heads. The isolation
provided by the enclosures minimized this risk, but care was still taken around the print
heads when these enclosures were moved out of the way for various troubleshooting and
changes made as prints were in progress.
Design for Manufacturing and Assembly
Due to budget and time constraints, it was found to be advantageous to not outsource the
manufacturing of all of our components to a third party. As a result a design for simplicity,
assembly and functionality was extremely important.
The integration block requires a vertical channel that has a high surface finish, to minimize
abrasion on the fibers. A 2mm ream through the entire vertical channel is required to
achieve this goal. The matrix input channel requires a 45 degree (measured from the
vertical fiber channel) 2mm channel to intersect the vertical fiber channel. The 45 degree
angled channel can be manufactured using a 45-45-90 triangle block to offset the stock
material, and then drilled to the precise depth for intersection. The Holes used to connect
the mounting pin for the insulation block, the Pico hot end, and nozzle are all M6X1 blind
hole taps. A hole depth of 17mm was used for the mounting screw of the external hot end
to provide easier access to the integration point of the two channels. This is needed to
polish the intersection point of the two channels before operation to minimize possible
abrasion of the fibers.

16
The insulation block is ceramic glass mica which is a machineable ceramic, however the
material still has a low fracture toughness. This must be taken into consideration for design
and machining operations to minimize the possibility for fracture. All features within the
component are holes, one 6mm and five 3mm through holes, to minimize stress
concentrations. The overall geometry is a simple rectangular cube.
The assembly of the fiber matrix integrator head is modular. This was done so adequate
cleaning of separate components may be conducted without full disassembly of the
extruder head. The nozzle, external Pico hot end, and input for the fiber channel are all
capable of being separately removed for proper cleaning and maintenance. The modular
assembly also allowed for quick interchanging of different nozzles without having to
completely clean the entire extruder head. It is necessary for the vertical channel to be
clean at the beginning of testing to minimize breaking and clogging fibers during the
extrusion process. The 2mm channel provides a large enough diameter to plunge most of
the excess material through the channel after a test has been conducted.
Evaluation
Testing and Refinement
The design objectives of this project revolve mostly around creating a composite material
that could realistically be used in engineering applications. The nature of composites, much
like other engineering materials, doesn’t allow for critical quality analysis without scientific
testing and experimentation. Therefore, laying out and performing mechanical tests was
necessary to judge the effectiveness of a composites extrusion systems.
Without reference, data acquired using extruded composites
may not prove any usefulness of the material so it is important
to set a baseline for structural values based on traditional
composites manufacturing. Researching material values of
composites presents a unique challenge due to the large
amount of variables that greatly affect each material produced.
Fiber volume fraction, matrix composition, fiber composition,
fiber sizing, and production methods all come into play. In
order to eliminate any biases in tested samples of the team’s
own production and those determined in academic research,
baseline values were chosen to be determined via compression
molding by the team’s own accord. This means that the
thermoplastic resin, glass fiber, and testing fixtures used to
measure the quality of extruded composites would be identical
to those of the reference values.
Figure 12 - Sample Beams
Produced Using ABS Matrix
Material

17
Figure 14 - Failure modes of short beams loaded in shear
Composites production can be an intricate process due to the
factors stated above along with processing parameters such as
molding temperature, pressure, and time. A standard process to
follow had to be established to ensure repeatability of any data
collected. Through academic research as well as consultation with
composites professors and graduate assistants at Colorado State
University, the process was finalized after documented
experimentation and trials. A brief of the processing parameters
includes:
 Compression hot press platen temperature: 230 C
 Pressure applied in compression gradient: 0 – 280 psi
 Total time under heat and/or pressure: 89 min.
 Total time under maximum pressure: 21 min.
 Total cooling time: 45 min.
Refer to Appendix C for the complete Standard Operating Procedure written by the team
for compression molding 50% Vf GF-ABS composites.
Following the successful production of reference beam samples, they needed to be tested
structurally to determine any stiffness and strength values that could be analyzed. In
accordance with advisory consultation, it was determined the best values to experimentally
determine included short beam shear strength and stiffness of the sample beams. Short
beam shear strength is a process utilizing a traditional 3-point bend test but with a small
length to thickness ratio of the samples at 32:1. This high aspect ratio actually discourages
samples from failing due to bending stress and increases the likelihood of samples failing
due to interlaminar shear. One of the challenges of producing composites using an
extrusion based process is ensuring that layering isn’t a weakness of the material in
comparison to molded materials. Refer to Appendices A and B for ASTM testing standards
regarding short beam shear strength and flexural stiffness testing respectively. Beams can
be expected to fail in one of three ways as a result of running a short beam shear strength
test. Interlamimar shear is when layers split, bending would occur when the fibers actually
fracture allowing cracks to propagate along the surfaces of the samples, and inelastic
deformation occurs when neither of the other two do and the sample is deformed. Failure
is defined in the test as observing a load drop-off of 40% which would typically represent
bending failure. If the load doesn’t drop, failure is defined when the loading point travels
the thickness of the beam.
Figure 13 - Honacomp
Compression Molder used in
Reference Composites
Production

18
Figure 15 - ABS-GF Sample 1-1 Exhibiting
Interlaminar Shear and Inelastic Deformation
The chart of short beam shear strength vs. time of sample 2-2 demonstrates several key
findings. There exists a small dip in load cell force at around 1.2 ksi but after running
several tests and seeing it each time and consulting with research assistants working with
the equipment it was determined that is a simple artifact of the machine and doesn’t have
an effect on the data. However, another dip exists around 1.7 ksi and this phenomena didn’t
occur every test and if it did, it wasn’t always in the same spot. After analyzing it was
determined that it is likely a drop of load due to the resin in the sample inelastically
deforming under the reaction pins and load pin but it quickly begins escalating again. The
test needed to be watched closely for a rapid escalation in load force after the load levelled
out because it would represent a switch from measuring a reaction force in the sample and
into a reaction force of the steel pins due to being pinched between the sample and the
loading pin. Allowing the test to continue could result in damage to the testing fixture so
best judgement was used to stop a test after a significant period of no load escalation even
if the head had not yet travelled the beam thickness or the load hadn’t dropped 40%.
The testing machine measures only the load being applied to the sample, Appendix A
details the connection between load and short beam shear strength and includes a
conversion function:
𝑆𝐵𝑆𝑆 = 0.75 ∗
𝑃𝑚𝑎𝑥
𝑡 ∗ 𝑤
Where:
SBSS = Short Beam Shear Strength Pmax = Maximum Applied Load t = thickness w =
width
Stiffness is the one of the most important mechanical behaviors to analyze composites
because that is the main benefit to using them in structural design. Measuring stiffness is a
similar process to short beam shear strength in that they both use 3 point bend tests,
although in the case of stiffness the samples to be tested have a considerably longer
geometry. The cross sections remain the same but the sample is tested with a reaction pin
span of 60 mm so samples were consistently cut to 75 mm to allow for 7.5 mm of overhang
on each side. Figure 17 shows a polypropylene – glass fiber composite in the flexural
stiffness testing rig. Appendix B outlines ASTM D7264 the standard test procedure for
polymer matrix composites in bending. For the three point portion of the test, failure
occurs when the sample fails in flexure on either the top surface in tension or the bottom in

19
compression. Failure can be observed by the testers as a steep drop in applied load and can
often be observed via audible cracking of the reinforcing fibers. Similarly to short beam
shear strength, it is inappropriate to report raw load data and to account for any slight
geometric variations, flexural stress was calculated from the load data to obtain useable
results via the following equation:
𝜎 =
3𝑃𝐿
2𝑏ℎ2
Where:
σ=Flexural Stress, P=Applied Load, L=Sample length, b=sample width, h=sample thickness
Results
Following are the results of the mechanical testing run on the ATS testing machine of
composites produced by compression molding. Refer to Appendix C for detailed results of
mechanical testing.
Table 3 - Statistical results of molded composites in shear
The standard deviation of the short beam shear strength of the molded composites was
higher than expected, this represents that although the process and materials used
remained constant throughout each test not every sample was created equally. It does
create a target value for the results of the short beam shear strength of commingled tow
samples.
Table 4 - Statistical results of commingled tow samples in shear
Commingled Tow Testing – Feed Rate Study
Peak Load Short Beam Shear Strength (ksi)
Average: 103.55 2.68453418
SD: 8.48746976 0.220037689
Sample 1 Avg: 2.673318625 F1000 mm/min
ABS-GF Molded Composites Testing
Peak Load (lbf) Short Beam Shear Strength (ksi)
AVG: 117.9 3.111365236
SD: 26.341644 0.69515247
Figure 17 - Long beam tested
under flexure
Figure 18 - Graph of flexural stiffness vs.time

20
When printing with a consistent layer height of 0.30 mm and varying feed rates an average
SBSS was found to be 2.68 ksi but 2.77 when the feed rate is 600 mm/min. This allows
testing to move forward with a specified feed rate so that multiple variable aren’t being
testing in similar tests. The next variable to attempt and determine experimentally is layer
height.
Table 5 - Statistical results of commingled tow samples in shear
Commingled Tow Testing – Layer Height Study
Sample Peak Load Short Beam Shear Strength (ksi)
Average: 62.29 1.591343
SD: 9.67 0.34
Printing with a layer height varying from 0.42 mm to 0.75 mm shows that the average SBSS
drops significantly from when using a layer height of 0.30 mm. These assumptions are
based on only short beam shear strength and not on flexural stiffness so long beams from
the same sample batches were also tested to fail under bending stress.
Table 6 - Statistical results of molded samples in flexure
Long Beam Flexural Stress – Molding Technique
Sample Peak Load Max Flexural Stress (ksi)
AVG: 28.742 35.9125364
SD: 8.47760 10.59259514
Running the long beam samples of the ABS molded composites yields an average flexural
strength of 35.91 ksi. The commingled tow samples from the feed rate study and layer
height study to compare to the molded composites.
Table 7 - Statistical results of commingled tow samples in flexure
Long Beam Flexural Stress – Layer Height of 0.3 mm with varying Feed Rate
Sample Peak Load Flexural Stress
CT1-1 – Feed=200 mm/min 30.56 26.64542
CT2-1 – Feed=600 mm/min 37.37 23.6136
CT3-1 – Feed=1000 mm/min 17.44 26.49978
AVG: 28.45667 25.58627
SD: 10.13012 1.709932
When the layer height of the print remains at a constant of 0.30 mm the data suggests an
average flexural stress of 28.45 ksi. Following these tests were the flexure results of the
samples with varying layer heights.

21
Table 8 - Statistical results of commingled tow samples in flexure
Long Beam Flexural Stress – Feed rate of 600 mm/min with varying layer height
AVG: 12.72 14.96186
SD: 1.508347 3.000702
When the layer height of the test was increased the data suggests a trend in lowering
flexural stress which is consistent with a drop in short beam shear strength as well when
the layer height is dropped.
To distinguish whether or not the results from the commingled tow samples provide a
realistic alternative to composite parts created using traditional molding techniques,
hypothesis testing of the data is necessary. Two sample t-tests can show if the data from
two different groups is statistically different from one another. T-testing is only an
applicable process if the affiliate data is normally distributed, so Minitab was used to
determine normality.
Refer to Appendix D for the detailed Minitab output on the normality test. The results of
running normality tests of the short beam shear strength and flexural stiffness of the
commingled tow samples yields p-values of 0.494 and 0.503 respectively. Because the p-
value is greater than 0.05, one can fail to reject the null of no distribution meaning that the
data is in fact normally distributed. Therefore, t-testing the data is an appropriate action.
Observing the data suggests that there may be a correlation between layer height and
mechanical behavior so hypothesis tests were run with two batches of commingled tow
samples tested to failure in a flexural stiffness test. The first batch was printed with a layer
height of 0.3 mm and the second ranging from 0.42 to 0.75 m. The following is the result of
the hypothesis test. Refer to Appendix D for the detailed hypothesis testing outputs of the
flexural stiffness t-tests and the short beam shear tests.
The data suggests based on a resulting p-value of 0.000 that there is in fact a difference in
the flexural stiffness’ of the two different batches. Though it is of note that there were only
3 data points available for analysis therefore it would be inappropriate to make any
absolutely conclusive statement but the trend is suggested. Figure 20 shows that here is
also a discernable difference between the short beam shear strengths between the sample
printed with layer heights of 0.30 mm and those printed at higher layer heights. This result
makes sense as the key component to quality consolidation of a composite is pressure and
a lower layer height there is a larger pressure on the extruded layer.
The results of the two sample t-test comparing the short beam shear strengths of molded
composite samples to commingled tow extruded samples yields a P-value of 0.094.
Working on a confidence interval of 95.0, the data fails to reject the null hypothesis that
there is no difference strengths. Therefore, the short beam shear strength of the extruded
samples is not significantly weaker in shear than those of traditional molded techniques.

22
The results of the two sample t-test comparing flexural stiffness of molded composite
samples to commingled tow extruded samples yields a P-value of 0.100. Working on a
confidence interval of 95.0, the data fails to reject the null hypothesis that there is no
difference strengths. Therefore, the flexural stiffness of the extruded samples is not
significantly weaker than those of traditional molded techniques.
While strength testing yields good results for comparing structural integrities, micro-level
analysis is required to determine why the similarities and differences in the materials exist.
Therefore, samples created using commingled tow were mounted in an acrylic compound
and polished to allow for microscopic viewing.
The images generated on a microscope of varying magnifications of the commingled tow
samples offer important findings. Figure 22 shows a view of the sample magnified 5x and
shows a layering phenomenon exhibited by somewhat clear fiber boundaries with a section
of thermoplastic matrix lying outside of it. Ultimately fiber dispersion is desired throughout
the whole part but the fact of the matter is when creating parts with fused deposition
modeling a complete elimination of layering effects appears unrealistic. In that regard the
area of thermoplastic only material is smaller in sample 1 than in sample 3 (figure 21). This
is an expected result as the feed rate of sample 1 was 200 mm/min compared to 1000
mm/min in sample 3. This means the time in thermal contact of the material is longer at
the slower feed rate. One intent of designing an extrusion nozzle with a large flat region not
Figure 19 - 100x magnification of commingled tow sample 3 Figure 20 - 400x magnification image of commingled tow sample 2
Figure 22 - 50x magnification of commingled tow sample 6
Figure 21 - 100x magnification of sample 1

23
just behind the exit but in front as well was to allow previously extruded layers to reheat
with the hopes of enhancing layer to layer consolidation when a new layer passes over.
That means at slower feed rates the nozzle is in contact with the previous layer for longer
allowing for more heat transfer and more consolidation with the warmer matrix material.
Figure 19 shows a highly magnified view of sample 3 which was printed at a feed rate of
600 mm/min. Noting that the image comes from a section that is already fiber dense, the
consolidation and dispersion of the fibers looks to be of high quality. Small dark sections
can represent a third constituent, most often air voids, but the relative area of them is small
showing that little air stayed trapped in the deposition. Equally spaced fibers surrounded
by matrix material allows the matrix to transfer applied stresses to the fibers which leads
to a stiff part. This falls in line with the findings that structurally parts printed at 600
mm/min showed the most desirable properties though the difference in each was not
tremendous. Overall, the results obtained from photomicroscopy of the commingled tow
samples offer the reason for the small drop in mechanical properties over the molded parts
being related to interlayer consolidation but not in a catastrophic fashion.
Figure 22 shows a sample printed with a layer height of 0.70 mm and it exhibits the
poorest consolidation of all the samples examined. This result correlates with theories
provided regarding successful consolidation needing pressure placed on the part as well as
results obtained from mechanical testing. With a higher layer height, there is clearly less
downward pressure on the deposition which caused less dispersion in layer and interlayer
as well as the presence of a relatively large air void represented as a large dark section in
the image.
Following encouraging results from testing with commingled tow, the team felt
comfortable moving forward with testing the fiber filament integration system. It was
deemed best to test the integrator on the Zmorph printer so alterations to the firmware of
the printer were required to allow for different temperature settings as compared to
traditional thermoplastic printing parameters. Because the dry fiber and filament aren’t
pre-mixed as such in commingled tow, analysis was needed to determine the speed of both
the print head’s travel and the extrusion speed. Because the fiber is tacked to the build
plate, the mass flow rate of the fiber is directly proportional to the print speed.
The matrix material chosen for testing the integrator head was acrylonitrile butadiene
styrene (ABS), a common thermoplastic used in 3D printing. ABS was chosen for several
reasons but the two main reasons being that it adheres to itself and to glass build plates
such as the Zmorph’s fairly well and it also dissolves in acetone making cleaning the print
head straightforward. It was seen in testing commingled tow material that matrix material
often sticks to the walls of the infeed channels when printing and although it may not be
enough to alter the steady state process of material flow it does create additional obstacles
for fibers to travel across leading to fiber degradation.
Initial testing of the integrator head showed difficulties in keeping the tow of fiber intact
while passing over previously extruded material. The cause of this problem was an excess
amount of tension being put on the fiber on subsequent passes to the initial pass. The g-

24
code that controlled the print head was adjusted to allow for a larger turn in the y direction
after each x direction pass to lower the tension on the fiber. This allowed for successful
samples of the desired geometry to be printed and examined.
Examining the samples created with the integrator system yielded mixed results. The fibers
looked to be laid out continuously with no fractures and on turns there seemed to be little
folding of the fibers, all of which are desirable outcomes. Unfortunately, there seemed to be
a lack of adequate wet out amongst the fibers laid on the new pass. When pulling fiber and
matrix straight through the extruder and not pressing it to the build plate, matrix did a
decent job reaching the far side of the tow band and the wet out was much better than
when the fibers are immediately laid flat upon extrusion. It was beneficial though that the
thermoplastic tended to group together on the underside of the extrusion and the fibers on
top because that allowed for subsequent passes to apply additional heat, pressure, and
matrix material to the undistributed fibers and make the strength of the composite higher
on the inside layers than the top. An additional pass over the top of a completed part with
only the matrix material extruding could create a situation where each layer of the part is
given the opportunity to consolidate by an additional pass going over the top of it and
creating a quality composite.
Figure 23 shows the dry fiber and filament sample with low magnification, it is clear that
fibers were not dispersed throughout the part. There are 8 clear passes, 4 side by side,
which is a worse layering effect than what was observed with commingled tow. Figure 24
shows more troubling conclusions that the fibers are not only clumped together but almost
no matrix was able to wet the fiber bundles as seen with the large dark regions
representing air voids. It is unclear whether there was not adequate pressure to
consolidate but it is unlikely that is the case because the same amount of pressure was
utilized to consolidate the commingled tow samples fairly well. Much more likely is that as
predicted, the fibers did not promote enough bonding to the matrix material.
Figure 23 - 40x magnification of integrated sample Figure 24 - 100x magnified view of integrated sample

25
Figure 25 shows an image magnified 20x and shows some interesting results. The darkest
shades on the right likely show air voids but there are several other shades darker than the
matrix material which is shown as the bright groupings while the bright singular circles
show the fibers. It is possible that the medium dark shades represent the chemical sizing
that coats the fibers as it is mostly present only directly surrounding the fibers.
The fibers used in testing of the integrator block came from a roving spool of e-glass fiber
sized for use in epoxy composites. This means that the film former portion of the sizing is
chemically similar to an epoxy matrix which is a thermoset instead of a thermoplastic.
Film former is what allows the fibers to bid together in handling but also promotes their
separation when introduced to the matrix material [5]. Thermosets are heavily crosslinked
in comparison to the more amorphous thermoplastics and cannot be reformed. Although
the time in which the fiber is heated before introduction to the matrix is much lower than
what is required to completely cure epoxies, it is a possible scenario that the fiber sizing is
curing a very small amount and that is just enough to help bind the fibers together. This is
also a concern because the print head operates at 290o C and most commercial epoxies cure
at a temperature far below that, that is not to say however, that the fibers reach that
temperature because they move through the system relatively quickly and the ceramic
block thermally isolates the area above the fiber infeed channel.
It was important for the integration iteration of the project to determine how to control the
fiber volume fraction of the deposition. Refer to Appendix H for the detailed formulation for
volume fraction based on printing speed and feed. Measuring the dimensions and mass of
Figure 26 - Printing continuous fiber reinforced ABS through integration
system on Zmorph printer
Figure 27 - Printed continuous fiber
reinforced ABS composite
Figure 25 - 200x magnified view of integrated sample

26
the extruded sample come out with approximately 62.61 mm3 of fiber and 260 mm3 of ABS.
This means the printed sample is 24% fiber and 76% ABS. Based on the print speed of 1.00
mm/s and extrusion rate of 0.7 mm/s the formula yields a theoretical volume fraction of
19% fiber. Our actual volume fraction makes sense to be higher because the density
method of volume fraction deduction assumes no air voids in the extrusion. According to
the microscopy of the printed samples, there were clearly air voids which would
theoretically be replaced by matrix material and therefore the volume fraction would drop.
Manufacturing and Assembly Evaluation
Manufacturing of the integration block was found to be a lengthy process, however three
successful integration blocks were manufactured. The drilling of the vertical fiber channel
proved to be the most difficult operation because it required a depth of 40mm, 1.97mm
hole (pilot hole diameter for 2mm ream). This operation required the 2 flute drill bit shank
to be extended beyond the recommended contact length for a 3 toothed milling chuck. This
issue was overcome by decreasing the turning speed of the drill to minimize vibration
while simultaneously peck drilling a millimeter per plunge to provide proper chip clearing.
While machining the ceramic glass mica for the integration block, it was found that high
speed steel drill bits were not adequate for the drilling operation. High speed steel end mill
with a large thermal mass was found to be adequate for the facing operations. Due to the
low thermal conductivity of the workpiece, the heat transfer from the milling and drilling
operations was largely dissipated into the tool instead of the chip or workpiece. In addition,
the ceramic was extremely abrasive and dulled the high speed steel drill bits. To overcome
this issue, titanium nitride coated bits were used to drill the holes for the component.
The modular assembly of the extruder heads was found to operate as designed. Having the
capability to individually clean different areas of the extrusion process proved to increase
setup speed and simultaneously decrease post testing cleaning time. In addition to
cleaning, the ability to quickly change the nozzle was extremely convenient for testing.
Unit Cost Analysis
This project was an alpha-phase design, and as such the production cost can only be
analyzed from the costs of the first iteration of each print head. Each printer head version
had a different cost to produce, with each successive version being more expensive than
the last. This was due to an increase in complexity for each design. The commingled tow
head was simple and required only a standard E3D hotend modified with a newly designed
and manufactured nozzle. This kept production price low for the first design, costing
$105.90 to produce when a single set of the production tooling was accounted for, and only
$35.44 in parts and raw materials.
The next design, the fiber filament integrator, increased in complexity and required more
complicated machining thus requiring more tooling. It also took more raw materials to
create the integrator block, thermally-isolating glass-ceramic mounting block, as well as the
same nozzles as before. Additionally it required the use of a Pico all-metal hotend to reach
the higher temperatures required to lower the viscosity of our thermoplastic filament.

27
Obviously this led to higher development and production costs, adding up to $327.46
including one set of production tooling, and $263.45 without the tooling.
The final design iteration came out to be the most expensive; it required all the parts used
in the fiber filament integrator with the addition of a stepper motor with a planetary
gearbox, high load bearings, an aluminum rod for the pellet infeed, and an auger screw to
act as a leadscrew. This was by far the most expensive design of the project, costing
$390.46 with a single set of production tooling, and $323.90 without any tooling. If any of
these designs were to be brought into higher levels of production, costs could be minimized
with the use of tooling for multiple parts rather than just the single set produced, by buying
parts and materials in more bulk size and thus reducing the price of procurement, and by
streamlining the design and production processes to minimize the total number of parts
needed.
Project Development Cost
This project was given a budget cap of $2,000, provided by Dr. Radford, to develop 3
iterations of print heads that increased in complexity with each design. Rough guidelines
including required parts and raw materials, were made for each iteration to help decide
how much of the budget could be allocated to each design. The total amount of the budget
spent at the end of the project was $1,659.64 (a total of $340.46 below the overall allocated
budget). Additionally, consumable parts and tooling used by the team in the MERC over the
course of the project were replenished, and some tooling needed by Dr. Radford was
purchased to help repay the lab for consumables used. No more of the budget will be
required from here on out, because all the parts needed to finish the project are already
purchased and made.
Deviations from Original Plan
The composite extruder head development initial project plan called for three different
prototype extruder heads to be produced, with the results from the first prototype
informing the design, of the second and so on. This general project outline was followed by
the extruder team and eventually three different prototype heads, each designed to print
composite from a different form of stock material, were produced. However, there were
deviations from the development plan with specific prototypes, especially when compared
to the plans in the critical design review.
The plan called for the testing of multiple nozzle geometries to determine which provided
the best wet out of fiber and consolidation between layers. Multiple nozzles of varying
shapes, sizes and outlet diameters were manufactured however it became clear early on
that nozzles with the largest flat area and a large outlet diameter were most effective. The
large nozzles were therefore used to print most of the commingled tow beams rather than
an equal distribution of prints between the different nozzle designs.
While there had been discussion about having nozzles manufactured via EDM in the first
semester the team did not believe that the budget was large enough to cover the cost. For
this reason no plans to have complex geometries, which would require EDM, were included
in the project plan. During the second semester the group discovered that a company

28
would be willing to pay for EDM work. This allowed more complicated nozzles, specifically
nozzles with tapered channels, to be manufactured. These tapered, EDM nozzles were not
include in the critical design review or original testing plan, but they resulted in
significantly better results than the hand machined nozzles.
The final fiber-filament integrator that was manufactured by the group was very close to
the drawings shown during the critical design review. The most important difference was
the addition of barrels which threaded into integrator block. These barrels were added
because they could be removed from the block which made cleaning the integrator much
simpler. The plan presented at the critical design review called for integrators with
different channel angles to be produced and tested. The first integrator manufactured used
a 45 degree angle because it was the easiest to machine. With testing it was determined
that integrators wet out issues couldn’t be solved with a simple angle change. This meant
that machining of more integrators would not yield more successful results and therefore
integrators with the two other angles were not created.
The last prototype, the pellet extruder, was still in the concept phase during the critical
design review. The plan, at that point, was to create an extruder which would be mounted
separately to the rest of the extruder head. The extruder would turn thermoplastic pellets
into filament which would then be driven into the fiber-filament integrator. More research
showed that creating filament with a consistent diameter, and being able to feed it
regularly, would be very difficult without a large and expensive lead screw set up which
was well outside the scope, and budget, of the project. Even more research discovered a
possible solution in the form of an auger and heater melting and directly driving pellets
into the integrator block. This was the solution that the team decided was the most feasible
and therefore produced.
Lastly, the testing and evaluation for the prototypes went as expected, with some
exceptions. Because the best nozzle option was determined early on in the testing process
most commingled tow testing revolved around layer printing parameters such as layer
height, feed rate and hot end temperature. While all of the mechanical testing that had been
planned was carried out on the commingled tow samples no resin burnout was attempted.
It was determined that because the comingled tow was already of a fixed fiber volume
fraction there was no need to test for that fraction again. Instead, photo-microscopy was
carried out on the printed samples in order to determine the level of consolidation between
the different layers of the composite.
When the fiber-filament integrator was tested it showed difficulties wetting out fibers
enough to create a useable composite beam. After discussion with the project advisor it
was determined that mechanical testing on a beam would produce no useable results and
was not a prudent use of time and resources. Instead photo-microscopy was used to
determine how much thermoplastic had actually found its way into the bundles of fiber.

29
Conclusions and Recommendations
Hypothesis testing the samples that were produced in both a traditional molding technique and
additive method shows that when printing parameters are optimal, there is no statistically
discernable difference in either the flexural stiffness or interlaminar shear strength in samples
with a 10% higher volume fraction in the extruded samples. These results are extremely
encouraging because it demonstrates that additive manufacturing of continuous fiber
reinforced composites is a realistic alternative to molding without sacrificing structural
integrity a detrimental amount. Upping the volume fraction of a material could be seen as a
small price to pay when it allows for the elimination of expensive tooling, long molding
processes, and costly labor of hand layup.
Based on data from test results there are several key findings that allow for more successful
testing of this system in future design iterations. Clearly there was a drop-off in both short
beam shear strength and flexural stress when the layer height of the print increased past a
threshold value of 0.30 mm. The reason for this phenomena rests mostly in the need for
pressure to be applied to composite parts to consolidate the fibers and the lower the extrusion
point is to the bed, the larger pressure that is applied. However, lowering the layer height
comes with a tradeoff as it causes more matrix to be squeezed out the sides of the deposition
and therefore geometric control suffers. A decision can be made on a part by part basis to
determine if structural integrity or geometric tolerance is the more crucial factor and then an
appropriate layer height can be selected. These findings hold true after inspecting the parts
microscopically as the samples printed with higher layer heights showed larger regions of
thermoplastic dense areas and larger dark areas representing air voids.
Based on optical inspection of parts printed with the fiber-filament integration systems, it is
clear that matrix viscosity is incredibly important to wetting out fibers and therefore the
temperature of the matrix inlet should be as high as possible without degrading the matrix
material or damaging the printing apparatus. Tests with the Pico hot end running at 290 C
yielded the composites with the greatest wetting.
Difficulties arose during the testing of the fiber-filament integration system due to a lack of
quality wetting out of the fibers. One factor making it difficult to achieve better wetting out is
the sizing that is attached to the fibers for handling. Sizing is a chemical formula that allows
fiber to be wound neatly but different sizings exist for different matrix materials. Due to
thermoplastic sizing being a rare material, no companies contacted carried it in stock and lead
times were often too long to be of use for the team’s research. Due to a surplus of fibers sized
for epoxies being available, the integrator was tested with them. This potentially caused the
fibers to act like a more solid section and not allow for the thermoplastic matrix to penetrate
the fibers. To try and mitigate this issues, a small system of dowel pins were attached to the
integrator just above where the fiber infeed channel is located. This allowed for the fibers to
spread out and break up the sizing before being introduced to the matrix material. This did in
fact help the wetting of the fibers but not to a level to achieve adequate wet out.
A particular challenge of this design process was being able to customize the commercial
Zmorph printer to allow for use with the new extruder heads. Due to the lack of community
resources available and the proprietary nature of the printer, small customizations proved time
consuming and exigent. Modifications done to the printer included adjusting the printer

30
configuration file to reach a higher maximum temperature, removing the enclosure face to
allow for material infeed, and mounting a Bowden drive to the base of the printer. It was
determined that originally choosing a printer with a more involved open source community
would have been a better path to take though the printer decision was one made across several
different departments.
One recommendation for the future of this design would be to design in a way for the fibers to
run over a gentle radiused part such as the dowel pins where it is introduced to the matrix
material. As it stands now, the fibers break up and spread but then as they enter the channel
they are constrained back to a 2 mm channel and bunch back up discouraging penetration. The
team has designed several different ways to attempt to mitigate this problem. Such designs
include drilling a wider fiber channel and press fitting dowel pins so that they tangentially
protrude into the sides of the channel. There is a potential difficulty in this method as it creates
a new exit for the molten matrix to leave the integrator which would cause issues with volume
fraction control and repeatability. Another way would be to redesign the integration block to
allow for rollers to be used instead of the fiber and matrix channels. This would allow for the
fibers to spread over a large radius and then the system can hit the fibers with the matrix along
to roller surface. The potential issues with this design deal with getting the wetted out fibers to
release from the roller.
Another redesign feature to the integrator block that could potentially help with printer
efficiency would be to reduce the extruder head’s thermal mass. With the current design a large
amount of heat is needed for the whole system to reach equilibrium which means that a larger
uncertainty with temperature fluctuations is present during testing.
Another way to increase the wetting of fibers would be to better select the materials used in
the integration system. The focus on selecting a thermoplastic matrix was based on its behavior
in 3D printing applications such as self-adherence and maintenance practices. While weighing
those options, ABS was deemed the best matrix material to initialize testing because it cleans
well and sticks to itself fairly well. However, when analyzing composite production methods,
the measurement of how well fibers adhere to the matrix material comes down to the
respective surface energies of each. When viewing the chemical make-ups of both the fiber and
matrix materials, it becomes clear that those with largely different surface energies will attract
to one another. The more similar surface energies are to one another, the less driving force
there is to break the surface bonds unique to each and reattach to create a new surface. E-glass
has an accepted surface energy on the order of 400 mN/m [11]. ABS on the other hand has a
surface energy on the order of 42 mN/m [17]. This is large differences but for perspective, the
surface energy of polypropylene (PP) is 30 mN/m; that creates a 4% larger difference in
surface energies which could allow for slightly better adherence between the constituents.
Polypropylene has its flaws as well when it comes to printing as it struggles to adhere to a build
plate without a PP based foundation such as a PP sheet to print on. Another way to put it, ABS
is a better material for 3D printing but PP is a better one for composites production and a
recommendation for future design iterations would be to try and overcome the printing
difficulties of PP but enjoy its composites production properties instead of the other way
around. Also, if ABS is used in the printing process it is important to note that washing it with
deionized water and storing clean prior to printing helps to bring down the surface energy a
small amount to a reported 38 mN/m, the same value as polyactic acid (PLA), another material
commonly used in 3D printing [17, 15].

31
Appendices
Appendix A: ASTM D2344 – Short Beam Shear Testing Brief

32
Appendix B: ASTM D7264 – Flexural Stiffness Testing Brief

33
Appendix C: Mechanical Testing Results
ABS-GF Molded Composites Testing
(1-1) 98.1 2.588845883
(1-2) 138.7 3.660274455
(1-3) 93.4 2.464813512
(1-4) 124 3.27234342
(2-1) 112 2.955665025
(2-2) 116.1 3.063863476
(3-1) 154 4.064039409
(3-2) 68.2 1.799788881
(3-3) 147.8 3.900422238
(4-1) 126.7 3.343596059
AVG: 117.9 3.111365236
SD: 26.341644 0.69515247
Commingled Tow Testing – Feed Rate Study
CT1-1 108.3 2.807677949
CT1-2 98.2 2.545835407
CT1-3 113.4 2.93989547
CT2-1 110.7 2.869897959
CT2-2 110.3 2.859527958
CT3-1 89.6 2.322880372
CT3-2 95.6 2.478430397
CT3-3 102.3 2.652127924
Average: 103.55 2.68453418
SD: 8.48746976 0.220037689

34
Commingled Tow Testing – Layer Height Study
CT1-1 12.2 19.103
CT1-2 13.6 21.295
CT2-1 13 13.292
CT2-2 13.6 13.905
CT3-1 15.6 15.950
CT3-2 12.3 12.576
CT4-1 10.6 13.090
CT4-2 11 13.584
CT5-1 11.5 12.190
CT5-2 13.8 14.628
Average: 12.72 13.961
SD: 1.51 3.00
Sample 1 Avg: 20.20 LH=0.42
Long Beam Flexural Stress – Layer Height of 0.3 mm with varying Feed Rate
CT1-1 – Feed=200 mm/min 30.56 26.64542
CT2-1 – Feed=600 mm/min 37.37 23.6136
CT3-1 – Feed=1000 mm/min 17.44 26.49978
AVG: 28.45667 25.58627
SD: 10.13012 1.709932

35
Long Beam Flexural Stress – Feed rate of 600 mm/min with varying layer height
CT1-1 12.2 19.10371
CT1-2 13.6 21.29594
CT2-1 13 13.29234
CT2-2 13.6 13.90583
CT3-1 15.6 15.95081
CT3-2 12.3 12.5766
CT4-1 10.6 13.09062
CT4-2 11 13.5846
CT5-1 11.5 12.19008
CT5-2 13.8 14.6281
AVG: 12.72 14.96186
SD: 1.508347 3.000702

36
Appendix D: Hypothesis Testing Results from Minitab Statistical Software

38
Appendix E: Composites Molding SOP

40
Appendix F: Maximum bending stress of glass fibers calculation
Maximum bending Stress of fibers:
𝜎𝑏𝑒𝑛𝑑𝑖𝑛𝑔 =
𝑀 ∗ 𝑦
𝐼
𝜎 𝑚𝑎𝑥 = 𝜀 ∗ 𝐸
𝜀 =
−𝑦
𝑟
𝜎 𝑚𝑎𝑥 =
−𝐸 ∗ 𝑦
𝑟
𝑟 =
𝐸 ∗ 𝑦
𝜎 𝑚𝑎𝑥
Properties of E-Glass Fiber[ref]:
𝐸 = 72 𝐺𝑃𝑎
𝜎 𝑚𝑎𝑥−𝑏𝑒𝑛𝑑𝑖𝑛𝑔 = 1950 𝑀𝑃𝑎
𝑑𝑖𝑎𝑚𝑒𝑡𝑒𝑟 = 10 𝜇𝑚
𝑑 =
𝑑
2
= 5 𝜇𝑚
𝑟 =
(72 𝐺𝑃𝑎)∗ 5 𝜇𝑚
1950 𝑀𝑃𝑎
= 0.18 𝑚𝑚

41
Appendix G: Cambridge Engineering Selector Material Selection Stages of Nozzle

42
Appendix H:Fiber volume fraction calculation
Vf = volume of fiber (mm^3), Vm = volume of matrix (mm^3)
Af = cross sectional area of fiber (mm^2), Am = cross sectional area of matrix (mm^2)
Uf = speed of fiber (mm/s), Um = speed of matrix (mm/s), Up = speed of print (mm/s)
𝑣𝑜𝑙. 𝐹𝑟𝑎𝑐𝑡𝑖𝑜𝑛 =
𝑉𝑓
𝑉𝑚 + 𝑉𝑓
𝑉𝑓 = 𝐴𝑓 ∗ 𝑈𝑓 = 𝐴𝑓 ∗ 𝑈𝑝
𝑉𝑚 = 𝐴𝑚 ∗ 𝑈𝑚
𝑣𝑜𝑙. 𝑓𝑟𝑎𝑐𝑡𝑖𝑜𝑛 =
𝐴𝑓 ∗ 𝑈𝑝
( 𝐴𝑚 ∗ 𝑈𝑚) + (𝐴𝑓 ∗ 𝑈𝑝)
𝑡𝑜𝑡𝑎𝑙 𝑣𝑜𝑙.= (( 𝐴𝑚 ∗ 𝑈𝑚) + ( 𝐴𝑓 ∗ 𝑈𝑝)) ∗ 𝑝𝑟𝑖𝑛𝑡 𝑡𝑖𝑚𝑒

43
Appendix I:Detailing Design ofFiber Filament IntegrationSystem
Cross section view of Fiber-Filament Integrator Head
Exploded side view of Fiber-Filament Integrator Head
Full view of Fiber Filament
Integrator Head

44
Isometric Exploded View of Fiber-Filament Integrator Head
1 Brass Nozzle Tip (C360 Brass)
2 Integrator Blocl (6061-T6)
3 Vertical Channel Mounting Screw (416 Stainless Steel)
4 Thermal Isolation Block (Ceramic Glass Mica)
5 Sizing Breaker Device (3 Hardened Steel Dowel Pins, ABS 3D Printed Frame)
6 Pico Hot End (Stainless Steel)
7 External Hot End Mounting Pin (Brass)
*1 Groovemount For Bowden Drive (Aluminum)
*2 M6X1 Locking Nut (Zink Coated Steel)
*3 M6X1 Locking Nut (Zink Coated Steel)

45
References
[1] al., F. N. (2015). Additive Manufacturing Of Carbon Fiber Reinforced thermoplastic
Composites using Fused Deposition Modeling. Composites: Part B, Engineering, 80,
369-378.
[2] Campbell, F. (2010). Structural Composite Materials. ASM International.
[3] MarkForged Develops 3D Printer For Carbon Fibre. (2015). Reinforced Plastics, 1(59).
[4] Michaeli, W. (2004). Processing Polyethelylene Terephthalate on a Single Screw
Extruder Without Predrying Usin Hopper and Melt Degassing. ANTEC, 296-298.
[5] P., & Mason, K. (n.d.). Sizing Up Fiber Sizings. Retrieved April 26, 2016, from
http://www.compositesworld.com/articles/sizing-up-fiber-sizings
[6] TWI, 'FAQ: How are composites manufactured?', (2015). Available: http://www.twi-
global.com/technical-knowledge/faqs/process-faqs/faq-how-are-composites-
manufactured/. [Accessed: 02- Oct- 2015].
[7] Eichenhofer, Maldonado, Florian, Ermanni, M. (2015). ANALYSIS OF PROCESSING
CONDITIONS FOR A NOVEL 3D-COMPOSITE PRODUCTION TECHNIQUE. 20th
International Conference on Composite Materials, 20th.
[8] Budinski, K. (1979). Engineering Materials: Properties and Selection (9th ed., Vol. 1, p.
773).
Upper Saddle River, New Jersey: Reston Pub.
[9]"TWINTEX® PP Mechanical Properties (non Standard)." Fiberglass Industries, Inc.
Fiber Glass Industries, Inc, 2013. Web. 5 Oct. 2015.
<http://fiberglassindustries.com/twintextechdata.htm>.
[10] Gibson, I., Rosen, D., & Stucker, B. (2010). Additive manufacturing technologies rapid
prototyping to direct digital manufacturing (2nd ed., Vol. 1, p. 487). New York:
Springer New York.
[11] Volf, M. B. (1990). Technical approach to glass. Amsterdam: Elsevier.
[12] Sloan, J. Arevo Labs launches 3D printing platform for composite parts fabrication.
(2015). Composites World.
[13] Dons, K. (2013). Filament Winding of Composite Tubes-Master Thesis. Retrieved
November 8, 2015.
[14] Sung-Hoon Ahn Michael Montero Dan Odell Shad Roundy Paul K. Wright,
(2002),"Anisotropic material properties of fused deposition modeling ABS", Rapid

46
Prototyping Journal, Vol. 8 Iss 4 pp. 248 - 257, from
http://dx.doi.org/10.1108/13552540210441166
[15] Rong, X., & Kief, M. (n.d.). A Study of PLA Printability with Flexography (Tech.).
[16] Mori, K., Maeno, T., & Nakagawa, Y. (2014). Dieless Forming of Carbon Fibre
Reinforced Plastic Parts Using 3D Printer. Procedia Engineering,1595- 1600.
Retrieved November 5, 2015, from www.sciencedirect.com
[17] Analytica, T. (n.d.). Surface Free Energy of ABS Plastic (Rep.).
[18] Sells, E., 2009, “Towards a Self-Manufacturing Rapid Prototyping Machine,” Ph.D.
thesis, University of Bath,Bath, UK.
[19] Hague, R., Campbell, I., & Dickens, P. (2003). Implications on design of Rapid
Manufactruing. SAGE Journals, Special Issue Paper, 281-301. Retrieved
November 10, 2015, from http://pic.sagepub.com/content/217/1/25.full.pdf
html
[20] “Filastruder Kit”. Filastruder.Filastruder (2015).Web. 4 Dec. 2015

47
http://www.filastruder.com/products/filastruder-kit

Composites Extruder Head Development - Final Report (7)

More Related Content

Viewers also liked

Composites Extruder Head Development - Final Report (7)