Effects of 3D Printing Build Parameters on Tensile Strength
This paper was written by David Edelen III while conducting independent research to support the University of Maryland’s engineering program. Additionally, there were several individuals who helped during the development of this paper and independent research effort. Those individuals are listed below and I would like to thank them for providing thier time and resources into this project.
Dr. Hugh Bruck, Mentoring Professor
Dr. David Barrett
Nick Lanham, Aaron Louis Technologies
Riley Reese, Arevo Labs
Alan Smith, Lab Supervisor
Due to the recent increase of additive manufacturing processes, 3D printing manufacturing methods have rapidly evolved. Throughout this evolution, the “filament” materials used within domestic 3D printing machines have remained fairly limited to the low performance thermoplastics. However, Arevo Labs , located in San Francisco, California, is developing several new composite filament materials. The most impressive material is comprised of the high performance thermoplastic PEEK with carbon fiber reinforcement. This is a break from the current filament offerings that rely on a mixture of thermoplastics to allow for 3D printing using common extruder systems that are not designed for use with the high performance thermoplastics.
The proposed test plan focuses on characterizing the effect of build parameters in relationship to the mechanical properties of common manufactured shapes. In addition, this test procedure will establish an authoritative baseline for which all 3D printer users could reference when estimating optimal build parameters (i.e., fill patterns, fill densities, print speed, nozzle size, layer height, raster angle, grid spacing multiplier, etc.) and selecting the most effective build material. Considering the significant increase in 3D printing technology within the industrial and hobby markets, this test procedure will serve as a frequently referenced filament material and printing process characteristic evolution within the rapidly evolving international 3D printing market.
In order to determine optimal 3D printer build parameters that are explicit to a specified material, the proposal focuses on developing a family of test coupons. These will be grouped by material and similar build parameters, in order to determine the optimal printer settings. Each family of coupons will include a different fill pattern and density that sequentially increases towards a coupon that includes a 100% fill density, or a solid test piece. Once the coupons have been printed, a separate test procedure developed in order to determine mechanical properties will then be utilized to isolate each material’s optimal filament fill patterns, fill densities, print speed, nozzle size, layer height, raster angle, and grid spacing multiplier.
This test proposal will provide a significant increase in the capabilities and effectiveness of most 3D printing methods due to an increased comprehension of optimal build parameters coupled with material performance. In addition, this test proposal will also establish the basis for more advanced additive manufacturing processes and more intelligent 3D printing estimation methodologies. Due to this test proposal’s design flexibility, all available printer filaments have the ability to undergo the same test procedure and would result in a corresponding set of optimal build parameters and material performance specifications.
Even though the majority of 3D printing software packages include the ability to modify build parameters, little is formally documented in regard to deriving optimal build settings’ specific to a given material. As a result, this test proposal has been specifically designed to isolate optimal build parameters and document the performance characteristics of each chosen build parameter in order to establish a documented, formally-tested, and realistic set of build parameters that have been developed using formalized engineering concepts taught at the University of Maryland campus.
Before any testing could be completed, it is essential to understand what research has been done and what information is currently known. After completing preliminary research, a second perspective was developed to supplement and expand upon the initial proposal. This modified the original proposal to make a more robust test plan; building off of what is well established, rather than repeating known findings.
Additive manufacturing is becoming increasingly popular due to the lack of tooling needed and the design flexibility. Last year, the multi-billion dollar industry grew 35.2%, an average growth of 33.8% since 2012. With advancing technology, 34.7% of additive manufactured parts ended up in final products. Over the years, several different processes have emerged for creating the parts. The main types of additive manufacturing are sterolithography, fused deposition modeling, and selective laser sintering (Ziemian et al., 2011). The process that was used in this report is known as fused deposition modeling, abbreviated as FDM. The process of additive manufacturing is unique due to the ability of a three-dimensional part being created by stacking laminated composite layers from a continuous flow of thermoplastics (Sood et al., 2011). Also unique to the layered manufacturing process is the fact that the mechanical properties of the part are not exclusively related to the material, but also the building process as well (Ziemian et al., 2011). Until recently, this process was mostly used to support design representation, as rapid prototyping capabilities have shown to reduce the design life cycle and, therefore, the overall cost (Nancharaiah et al., 2010).
After analyzing several publications applicable to the testing efforts illustrated within this paper, it became apparent that a wide variety of methodologies where used to analyse the resulting data. However, as one would suspect from testing done with any validity, all test were done in accordance with International Organization for Standardization (ISO) or American Society for Testing and Materials (ASTM) regulations. Two of the reports reviewed used advanced algorithms to develop a relationship between the tested build parameters and the result produced.
These advanced algorithms included artificial neural networks and bacterial foraging, which both mimic biological processes (Panda et al., 2009). The other researchers tended to use some form of statistical analysis that would try to develop a relationship between the parameters and the results, such as design of experiments (DOE), central composite design (CCD), or analysis of variance (ANOVA). For me, this indicated that it was likely that the exact relationship between the build parameters and the mechanical properties of the part was not fully understood. Because the bond between filament strings is thermally driven diffusion welding, the variation of the parameter, such as nozzle size and air gap, not only affects the physical make up of the part, but also several of the thermally driven processes (Sood et al., 2011). While nearly every researcher performed tensile testing on printed specimens, many also researched the same parameters affect with compressive tests, three-point bending, impact testing, surface roughness, and dimensional stability. This provided insight on how complicated the relationship between these parameters really is, one parameter led to a different overall effect.
As one would expect, the printed parts where found to have anisotropic material characteristics (Fly et al., 2013). Due to the nature of the FDM process, the continuous flow of thermoplastics tended to have the molecules align with the direction of the extrusion flow (Sood et al., 2011). When the raster angle parameter (e.g. the direction in which the extrusion flows) was altered, it was determined that the zero degree angle produced the highest tensile strength, while the ninety degree angle produced the lowest tensile strength (Ziemian et al., 2011). Positive air gap parts resulted in lower tensile strength performance, while the negative air gap parts produced increased performance (Ahn et al., 2002).
Parts built in the x-axis orientation and y-axis orientation performed similarly, while the z-axis orientation built parts had large voids and failed prematurely (Bagsik et al., 2010). Differing heating and cooling cycles throughout the part result in high residual stress build up and warping (Sood et al., 2011). Detailed temperature profiles show that the bottom layer tends to rise above the glass transition phase and the temperature gradient among layers increases as the number of layers increase (Panda et al., 2011). Lower layer thickness parts tended to have stronger bonds between filament strings and layers (Panda et al., 2011).
As a result of the variety of tests performed thus far, several general behaviors have been recorded however there still is no clear definition of how the parameters are related. The relationship of the complexity that artificial neural networks and particle swarm algorithms must be used to attempt to predict the outcomes of various parameter sets. These results where derived from several different standardized tests and a variety of thermoplastics. This allows for an overall accurate representation of what is known and what is unclear. Numerous tests have shown that the parts behave with anisotropic characteristics, and that the heat transfer relationship is just as much a factor in the material properties as the build parameters.
After reviewing the findings of other researchers’ work that was applicable to the proposed testing, it was clear that some revisions needed to be made. Each build parameter that was originally proposed was reviewed, and ultimately passed on to the testing phase, or vetoed.
Line, rectilinear, and honeycomb are the most popular infill patterns available. There are other patterns that can be used, such as Hilbert curve, but these are slow and do not add beneficial features to the part. As such, they are generally not used. The infill pattern would have the most significant effect on the stiffness of the part, making a three-point bend test the most robust method of testing. Setting this as a tensile test parameter would probably yield nearly identical results given that the skin of the part would take the majority of the load. Although an interesting parameter to test, this was eliminated from the research due to the time constraints of the semester. The three-point bend test is a more complicated test to perform, requiring a sufficient amount of time to be done properly.
The percentage of fill within the part is referred to as the infill density, and is a value between zero and one, with one being a solid fill of the part. As with the infill pattern, this would be an interesting parameter to test, however, it would have little effect on the parts tensile strength, but would modify the stiffness. This would also require a three-point bend test to measure its effect on the performance and for the same reason was eliminated from this research project due to time constraints as a necessary alternative test method.
Number of Perimeters/Shells
Shells, or perimeters, are used interchangeably, and refer to the walls formed around the boundary of the part. The number of perimeters changes the thickness between the outside surface and boundary of the infill pattern. This parameter would most significantly alter the stiffness performance and have little effect on the tensile strength. There is the possibility that it would provide better adhesion of the skin to the core as the number was increased, however, it would probably take a large number to make a significant difference, to the point where it is no longer practical. This parameter would be best suited for testing over a three-point bend test. As the focus of this research project is the tensile strength, this build parameter has also been eliminated.
The speed at which the part is printed is a vast factor in rapid prototyping: the faster the part can be produced, the faster design iterations can be preformed. The slower a part is built, the higher quality the model tends to result in, however this has little affect on the tensile properties, but is directly related to the surface finish and dimensional stability. It would be hard to quantify what a “fast” print and “slow” speed are because, as parameters like the infill pattern and infill density, this would significantly affect the length of time it would take to produce the same part while there is the actual speed the stepper motors are set to. A part could run at the exact same motor speed but have different infill patterns and take twice as long. To make the data collection more simplistic, this parameter was eliminated so that a more dominate build parameter could be the focus.
Typically printers use 0.35 mm, 0.4 mm, and 0.5 mm nozzle sizes. By changing the size of the nozzle, one changes the diameter at which the plastic extrudes. This has been tested by other people, however, it is a complicated parameter due to its relation to the temperature of the extrusion and time to solidification. This will be different for every material and the material I will be using has not yet been researched. Also, the size of the nozzle will allow different levels of resolution, so knowing that one would need a specific resolution, one could see how that would change the tensile strength of the part.
The space between roads of extruded plastic is known as the air gap. A positive air gap means that there is open space between roads, while a negative air gap value packs the roads more tightly. Using a negative air gap is known to increase the bond strength between roads, however the exact best value is not known, so test several negative values to find an optimal would be beneficial to the production of any part. Delamination is a known problem, so anything that can combat that occurrence should be investigated in more detail.
When using the line infill pattern, the angle the part is filled in at relative to the coordinate axis can be controlled with many of the popular slicing software providers. By altering this parameter, it allows for more design optimization because the part can be tailored to any case on a layer-by-layer basis. It is this build parameter that gives some control over the anisotropic characteristics and the specifics of how it performs at different angles is extremely important. This parameter will be tested in depth to attempt to develop a strong relationship between it and the tensile strength.
The thickness of each layer in the vertical direction is called the layer height. Altering this parameter is similar to altering the nozzle size, as they are both related to each other. A lower layer height causes the plastic to spread out further as it is extruded. By changing the value of the layer height, it would directly affect the resolution of the print, just as the nozzle size does. If this parameter is too far off in relation to the nozzle size, it will negatively affect the visual aspect of the prints with high surface roughness and poor dimensional accuracy. For this reason, this parameter will removed.
Originally, I wanted to run the same test procedure on several different materials to develop general performance characteristics that could be scaled by some appropriate material parameter. However, due to constraints on time and funding, only one material can be tested. This material has not had testing like this performed before, so it will be new information and will not repeat anything that has been previously documented.
The axis to which the surface normal to the build platform is defines what the build orientation is. It is standard practice to have the parts x-axis and y-axis touching the platform and the layers stack up on the z-axis. Testing of this has been done and it was proven for a standard tensile test specimen that there was little difference between the x and y orientation build. The z orientation was considerably worse due to the creation of large voids. This is an important parameter to keep in mind when printing a part because depending on the shape; an alternate orientation could be more successful. However, for the tensile coupon, it is known how the parameter affects tensile strength, so it will not be apart of this research testing.
Work Accomplished at Midterm
Preliminary Research: As outlined above, background research was conducted to establish a baseline. Applicable research papers were kept for further review, while reports that did not directly relate to additive manufacturing thermoplastic tensile testing were discarded. For example, several reports on FDM fiber-reinforced thermoplastic fatigue life were found, but since this is not applicable to tensile testing, were not kept for further analysis.
Procurement: In order to make this research project possible, certain resources had to be tracked, purchased, and arranged for use. I already had a (3-D pick one form throughout) three-dimensional printer, however I did not have any material worth testing. Arevo Labs donated a spool of fiberglass-reinforced Ultem, which is a high-performing thermoplastic. Ultem has a crystal structure and is sensitive to moisture. For this reason, the filament must be baked for four hours before printing to eliminate all moisture. For this reason, an oven was acquired. Also, because this is a higher performing thermoplastic, it has a higher extruding temperature. This means that a larger heater cartridge had to be installed on the printer, as well as a higher temperature hot bed. The stock printer software cannot handle these higher temperatures; so external PID controllers were installed to control the temperature of the extruder and hot bed. To perform the tensile test, lab time sessions were arranged with the lab supervisor on a regular basis.
As the project has progressed, some problems have been encountered along the way. While some of them were anticipated, and attempted to be prevented, other issues were encountered that were unforeseen. Leading up to the proposal of this research project, I was in close contact with Arevo Labs deciding what types of test would be most beneficial for me to investigate and after several possibilities, this project was decided upon.
It was also perceived that with a few modifications, this material could be printed efficiently and consistently. The modifications that I was told that needed to be made were a larger header to be able to extrude the higher performance thermoplastic and a hot bed that was able to go to higher temperatures needed to print this material. This was found to be easily attainable and could be done in a timely manner. However, the reality of the situation was that severe modifications would need to be made for this material to be able to be printed. The following section outlines the challenges faced and the found solution, or proposed redesign to fix the problem. These challenges are presented chronologically in the order that they happened in.
Problem: The printer that was used to do this research is a Makerbot Replicator 2 clone made by CTC. This printer uses ReplicatorG (also called RepG for short) software and firmware to control the printer and slice the models that will be printed. The default slicing software used within RepG is called Skeinforge. For general modeling applications, such as printing out small turtles in the library, this software and firmware are acceptable. However, when you wish to have more control, such as attempting to print structural parts, this software does not give the design flexibility necessary to complete the task. RepG has a few basic settings that can be changed, such as infill density and layer height, but it mostly completes prints with one general setting. For doing the proposed research, RepG would not be able to provide the variability preferred to complete the testing.
Solution: An open source slicing program called Slic3r gives full design flexibility. This program allows the user to change and tweak virtually every parameter evolved with the print; making it ideal for producing the robust designs necessary for this research. In fact, the program has so much flexibility that it has 3 default modes: 1) simple, 2) intermediate, and 3) expert; allowing users who are not as advanced to use the program without getting overwhelmed.
The big problem with this program is that it is not directly compatible with the printer’s firmware. As of right now, there are several major companies producing printers, all of which are setup on their own configurations. This means that there is no industry standard set on important parameters such as the file type read by the printer. All printers use an STL file to get the model, but what type of file that is converted into for the printer to read varies greatly. To get around this, the printer’s firmware and software must be wiped and replaced with a more tolerable open source operating system called Sailfish. Slic3r has an output setting that will put all of the g-code in the correct format for a printer using Sailfish. This is a tedious process because the machine is set up for particular offsets, which the native software has stored. When converting to the new system, you must take care that those are kept track of for the machine to remain accurate. The procedure is outlined below, taken from a Google Group on the topic (Jetguy, 2014).
- Download the 32bit version 2.2.1 of PyPy at http://pypy.org/download.html
- Download the 32bit version 2.7.6 of Python at http://www.python.org
- You MUST use Sailfish version of Replicator-G to update and configure Sailfish firmware. Download the version 0040r23 of RepG at http://www.thingiverse.com/thing:32084/#files. MAKE SURE YOU HAVE VERSION 40R23.
- Unzip the Replicator-G file to a folder. Leave all the RepG files in their own folder; don’t move them around (don’t move the EXE). Open RepG (click on the ReplicatorG.exe) MAKE SURE YOU HAVE VERSION 0040r23 running. It says the version in the top bar of the window.
- Make sure the printer is NOT plugged into the computer via USB yet.
- Open RepG Preferences> Advanced. Select the correct file directory for the python and PyPy interpreters.
- From the RepG menu, select “Preferences”. Click the “Advanced” button. Change the “Firmware update URL” field to: http://jettyfirmware.yolasite.com/resources/release/firmware.xml
- Then click “Close”. Wait until the RepG has finished doing its thing! You should see some stuff happening at the bottom command line.
- Connect the printer’s USB cable to both the printer and the computer you will be performing the upgrade from. Next, power your printer on. But DO NOT connect from RepG to the printer. Check the “Connection (Serial Port)” sub-menu of ReplicatorG’s “Machine” menu. Make sure that you see the USB port for your printer is listed. If it does not appear, then select the “Rescan serial ports” item of that sub-menu. You cannot proceed until your printer’s serial port appears (i.e. COM 1). Again, DO NOT actually connect from RepG to the printer: the firmware upload process itself will automatically perform that step.
- From ReplicatorG select the “Upload New Firmware” sub-menu of the “Machine” menu. SELECT “Replicator Dual Extruder”
- Select the firmware to install (r1195), select the COM port, click Next, then Upload
- Once you are done shut the printer down and shut down RepG. Do not connect the USB to the printer yet….
- Open RepG make sure this is selected then click Machine > Machine Type > The Replicator Dual (Sailfish)” or (“The Replicator Dual (Sailfish)” if you have a dual extruder)
- Make sure the USB is plugged into the computer and the printer and turn it back on, open up RepG. If it does not connect then click the connect button and get a connection
- On RepG click Machine > Onboard Preferences. In the lower left hand corner of the window is a button to “reset motherboard to factory settings”. Do that now!
- Give the machine a little time to do the above step, then shut down the printer and RepG and then load them both back up and connect them.
- On RepG click Machine > Onboard Preferences. Click on the “Homing/VREFs”. Ensure that the X and Y Home Offsets are somewhat near 151.998 and 71.998 or 74.998.
- Validate that the tool head offsets (different than homing offsets) are either values less than 1.0mm (AKA 0.85 or something) or is a value less than 33.0mm In other words, if it’s over 34mm you are going to have a problem. Reset to 0 if the number is too large. Again, valid values are less than 1mm (0.88), less than 34mm (33.35) or (32.87). The real nozzle spacing is 33mm and the offset is the adjustment from that value. In the case of then entire measured number being given, it just uses that. If you see anything larger than 34mm reset to 0!
- On RepG click Machine > Onboard Preferences. Click Acceleration (Misc.) and set right and left deprime values to 0.
Problem: The Ultem is a high performance thermoplastic that is able to withstand much harsher conditions then normal thermoplastics today. As a result, this means that the material itself demands a higher extrusion temperature. The factory extruder is capable of going up to 280°C, which is insufficient for printing this material.
Solution: In order to achieve the desired temperature range needed for the Ultem, the extruder’s heater was upgraded. The stock heater cartridge has a 10W output. The upgraded heater cartridge has a 40W output, providing the desired temperature range easily. This cartridge is the same configuration and dimensions as the stock one, so installation was done after soldering two wires and tightening a setscrew.
Figure 6 1: Stock and Upgraded Extruder Heaters
Heated Build Plate
Problem: The factory hot bed is capable of reaching a temperature of 130°C, however Arevo Labs recommends using a build plate temperature of 220°C. This is something that must be modified because it is directly related to the quality of the print. The heat from the bed not only aids in first layer adhesion, but also helps the part to cool more gradually as it is printed.
Solution: There is not a readily available hot bed that can hit the temperature range that is needed to print this material; so one must be custom fabricated. The hot bed is 6”x9”, so it will require a pretty high output heater to maintain the desired temperature. After looking at various options, the fastest and most cost effective option was to use a heating element from a hot plate to heat the printer’s build plate.
The hot plate that was found was 6” in diameter and had an output of 1500W for $29.95, which meets all the required criteria. A new aluminum build plate was cut, and the heating element was attached to the plate. Thermal paste was added to increase the heat transfer. Because the hot bed would be putting off much more heat then normal, the plastic support arms where wrapped in a thermal insulated wrap. This wrap is rated for 1800°F direct contact, which is believed to be sufficient.
Problem: Although the operating system was just converted to an open source version, this does not change any of the native default settings built into the software; it only makes it compatible with other slicing engines allowing you to use something other then the native slicing engine. The Ultem filament needs to be extruded at 340°C onto a build plate that is 220°C. The upgraded extruder heater cartridge and hot bed are now able to hit those target temperatures, however the operating system has built in safety features that will not allow the parts to go to these temperatures. Under normal circumstances, it would likely be catastrophic for one of the parts to hit this temperature range unintentionally; nevertheless for this material the higher temperatures are a must.
Solution: To over ride this feature, external PID controllers where installed for the extruder and heated build plate. This allows the temperature of each device to be controlled completely independent of any other system. This meant that the printer would need to be outfitted with a control panel to house the PID controllers and their accompanying solid-state relays. Figure 6-2 shows the PID controllers added to the side of the printer.
Figure 6 2: Printer With External PID Controllers
Upon installing the PID controller, another problem was encountered. The PID controller operates on 120VAC, so to power the PID; the main power supply for the printer was tied into. There is a power cord that goes directly to the wall outlet, and then branches off to power the various components and power supplies. However when the machine first starts up, a large amount of current gets drawn from the various motors and heating elements, causing a surge in amperage. This overloaded the power cord causing it to short out, and ultimately damage the printer’s power supply. After replacing the power supply, a separate power cord was installed for the PID controllers to prevent this from happening again. Under normal operating conditions the power draw is no where near the limit of the cord, however you must remember that there is an initial surge on start up that can be harmful to the system.
Problem: Now that the printer had the ability to extruder the material and hit the target temperatures necessary while remaining under control, it was now time to begin attempting to print. In order to be able to make the changes necessary for this research, Slic3r would need to be set to expert mode. This meant that many more details would need to be configured for everything to run smoothly. Although Slic3r was output the correct type of file, and Sailfish was loading it to the printer, the g-code commands were not quite correct. The printer moved, but it was off center and sometimes tried to go too far in the wrong direction. The extruder orientation was also backwards pushing the filament out instead of pulling it through.
Solution: After many trial and error attempts, the problem was tracked down and fixed. It rooted from three main areas. The first problem was that Slic3r was not including the necessary starting script to configure the printer’s positioning correctly before printing. Typically upon starting a build, the printer homes itself, then goes to its maximum position, known as the waiting position, the z-axis goes to the first layer level, then the build starts. The script that tells the printer those commands was not there, so it was trying to start from where it was, which is incorrect. The default Slic3r starting script and the final working script can be seen below in Figures 6-3 and 6-4.
Another issue was that the settings where configured using relative estop. Some printers, but not all use this, which is why it is an option for the generic slicing engine Slic3r. When this setting is used, the extrusion distance is calculated based on mass flow rate, rather then a linear distance. The linear distance method does not require this setting to be activated, and is what this system required. This does not have a dramatic effect on the g-code visually, however it makes a huge difference in how the calculations for the tool path are done.
The last issue was with the coordinate system that was set. Some printers use a relative coordinate system, while others use an absolute coordinate system. The relative system sets the center of the bed as the origin and all the moves are based off of that location. The absolute system uses the home position as the origin and bases all of the moves from there. Looking the g-code you can tell which system is in place by the values. For example, if the values range from -80 to 80, then a relative system is used. If the values range from 0 to 160, then an absolute system is being used.
Originally, Slic3r was outputting the tool path based on the absolute system and the printer was configured for the relative system, causing jams and crashes. This same issue was found for the end script and the issue was resolved using the same technique. The two versions of the end script can be seen in Figures 6-5 and 6-6 (Note: The scripts where to long to see in Slic3r so they where copied into a text document).
Figure 6 3: Default Slic3r Start Script
Figure 6 4: Modified Slic3r Start Script
Figure 6 5: Default Slic3r End Script
Figure 6 6: Modified Slic3r End Script
Extruder Feed Tube
Problem: In between the extruder’s nozzle and drive gear, there is a tube known as the feed tube. This tube keeps the filament aligned with the nozzle after passing through the drive gear. The end of the feed tube is tapered towards the thru hole, allowing the filament to easily correct itself if it starts out misaligned. Inside of the feed tube is a small Teflon tube that reduced the friction from the filament sliding through the feed tube. Under normal operating conditions, this tube works great, but when you go above 300°C, this tube will melt and jam the extruder.
Solution: Knowing that this tube was subjected to high temperatures, the best modification appeared to be adding a tube made out of a material that would withstand the heat, but also keep the same dimensions as before. The Teflon tube was first replaced with a copper tube of the same dimensions. After installing it, the filament jammed as soon as it entered the tube. Because the tube was now metal, it conducted the heat put off by the nozzle. So brass and aluminum tubes were tried, thinking that the extruder jammed because the plastic melted and stuck to the tubes walls. Both of these materials also provided a jam as soon as the filament entered the tube.
After looking at the extruder’s design more closely to see what was causing it to jam so easily; I found that the thin metal tube, while able to survive the heat, was also conducting the heat from the nozzle too well, causing the plastic to become molten far too early. The feed tube and drive gear are covered by a heat sink and fan, as seen in Figure 6-7. This heat sink was suppose to be drawing the heat away from the feed tube so the filament did not turn into a liquid until the last little section before the nozzle.
Figure 6 7: MK7 Extruder (Courtesy of Makerbot)
By increasing the temperature of the extruder, it not only made this system less effective, it also exceeded the operating temperature of a key component. In order for this system to work effectively, a new feed tube would need to be made. The inside dimensions would be that of the Teflon tube’s ID, so that the same volume would be maintained. The outside of the feed tube should have concentric fins added, with the fan blowing over them, making the tube able to dissipate heat more quickly. A conceptual representation of the fins can be seen in Figure 6-8. The original tube was made out of stainless steel and it seems reasonable to keep it made out of that material.
Figure 6 8: Concentric Heat Sink Fin Design
After removing the metal and Teflon tube designs in the feed tube, the extruder worked without jamming. This caused there to be an increase in volume, so the extruder took longer to start extruding and stop extruding. While this is a hassle, it would have to make due for this project, as there was not enough time to implement the redesigned feed tube.
Extruder Drive Gear
Problem: Now that the code could be run in the software and the extruder was able to push plastic through the system easily, the prints could be attempted. After a few hours of the printer being on and used, the filament would stop feeding unexpectedly, then start to make a clicking noise. The motor was turning, the filament was not jammed, and it just would not feed.
Solution: After troubleshooting the system, it was determined that the extruder drive gear was slipping. After disassembling the extruder, it was obvious that the setscrew had worked loose because the gear fell right off the motor shaft. The setscrew was tightened, then put back together, only for the same issue to happen shortly after. The higher then normal heat range was affecting every part, and this filament being stiffer then normal materials, was also adding additional strain on the drive gear. To help keep this gear from coming undone, I heated the gear and applied high temperature thread locker to the setscrew before tightening it up. After this was done, no more issues arose with the drive gear coming loose.
Problem: Now that the extruder was working properly, the printer could be calibrated. However, the higher operating temperatures had begun to take their toll on the hot bed support arms. These arms are made out of plastic, so the heat was causing them to sag. This issue was anticipated, so the arms where wrapped in an insulating wrap, however it proved to not be sufficient enough insulation to keep the plastic from softening. The arms are believed to be made out of polycarbonate, which has a glass transition at around 197°C. The hot bed was set to be at 220°C, exceeding that temperature and likely causing softening. This gradual sagging resulted in the printer being very hard to calibrate, and keep calibrated. The distance between the build plate and nozzle is critical to the way the print will start. A rule of thumb is a business card’s thickness between the nozzles and build plate. Any closer and the nozzle could jam or drag, and if further from the build plate, the first layer could not stick or the filament could curl up and never touch the build plate. Although they do not support a heavy load, gravity is enough to gradually cause the arms to sag further and further over time.
Solution: the only way to ensure consistent accuracy is to make the hot bed’s support arms out of metal. After some research, it was discovered that a company was selling aluminum arms for this type of printer. This company was contacted to get a set, however they had a lead-time of nearly three months. I could fabricate this part, however that is a very time consuming task to complete during the semester. For this reason, it is suggested that this modification be made, but due to time restraints and the lack of availability, I would have to make do with what I had.
Leveling Build Plate
Problem: The excessive heat from the hot bed is not only casing the support arms to sag, but also the aluminum build plate was beginning to warp. The cyclic heating and cooling of the aluminum plate was causing the edges to curl, as seen in Figure 6-10. This was making it very difficult to level the build plate, and keep it level. The build plate must be level in order to produce even and accurate prints. In Figure 6-10, you can see that one arm is lower then the other from sagging, and the edges of the build plate show deflection.
Figure 6 10: Hot Bed Support Arm Sag and Plate Warping
Solution: The stock build plate is bolted in each corner, with a spring under the plate so that the plate can be adjusted to a level that is equal across the entire plate. With the arms sagging and build plate warping, the stock springs did not offer enough range to be able to compensate for these events. To fix this, taller springs where added to give the build plate a larger range of adjustability. In the future, a thicker build plate is recommended to minimize the amount of warping that can occur. The results of the hot bed would also come out better if a heating element that evenly heats the build plate were used. This will not only prevent the plate from trying to warp, but it will also provide a more uniform build plate temperature for the print, helping minimize the warping that occurs as it cools. The stock springs (left) and modified springs (right) are shown in Figure 6-11.
Figure 6 11: Build Plate Springs
First Layer Adhesion
Problem: Despite the difficulties with calibrating the machine, it was able to be calibrated within a reasonable amount of time, but did require it to be regularly calibrated to try to maintain consistency. The printer was now attempting to put down the first layer of extrusion, but it was not sticking to the build plate. The filament would come out of the extruder, touch the build plate, then curl up, or drag along with the extruder. Arevo Labs was contacted regarding this issue, and they informed me that they use a proprietary paste to get their prints to stick, which could not be distributed as this time.
Solution: Because I would not be able to get any of the paste Arevo labs had developed for use with this product, I would have to try to come up with my own solution. Although not many people have attempted to print this material, the issue of the first layer not sticking is common in the 3D printing world.
After researching this topic on 3D printing forums and similar information sources, I determined that the first layer’s ability to stick was based upon several main factors. These factors include first layer height, gap between the nozzle and build plate, build plate temperature, extruder temperature, and build plate coating. Although an optimal setting might not have been reached, the layer height, print speed, nozzle gap, extruder temperature, and build plate temperature were all in recommended settings; so the first focus was on build plate coatings. There are three main methods that are commonly used. Some people would coat the build plate in glues cut with water, spray the build plate with spray adhesive or hair spray, and lastly, others preferred to put blue painters tape on the top of the build plate. When printing ABS or PLA, I typically used the blue painters tape, however at this temperature range the tape would just char as the temperature increased, so other methods would be required. After purchasing a wide array of glues and sprays, seen in Figure 6-12, I began trying each one and even various combinations of them together.
Figure 6 12: Assortment of Adhesives Tested
With each trial of adhesive, or combination of adhesives, slight modifications were made to the other mentioned parameters in order to try to find a sweet spot. However, time after time, the first layer was still not sticking. At this high temperature range, most of the glues would burn off before the hot bed even reached temperature. After research on additional methods used by other people, but aside from Arevo Labs, no one was able to provide any ideas that would hold up to the heat. After countless attempts, I determined that a chemical bond would be difficult to achieve at this temperature range, so I set out to achieve a first layer adhesion from a method of my own.
The adhesion of the first layer is absolutely critical, because it is literally the foundation of the print. If there are any flaws in the first layer, the rest of the print is only going to end up worst, so nothing less than perfect would be tolerable. With the chemical adhesion failing, I brainstormed other methods that would not be dependent upon heat or chemicals, arriving at a more mechanical adhesion solution. After trying all of the adhesives, I thought that using a screen would provide a mechanical latching of the first layer to the build plate, and thus allow it to adhere. The screen, seen in Figure 6-13, was made from 120 grit drywall sanding screens, which were then attached to the build plate with binder clips to keep it in place. These screens where able to withstand the temperature, while providing a surface the first layer could bite into.
Figure 6 13: Sanding Screen
With the first layer speed lowered, and the nozzle gap set so that the nozzle was barely dragging the screen, the extruder would literally force the filament into to screen providing excellent adhesion of the first layer. This method was cheap, reproducible, and very consistent. The result of the first layer using the screen can be seen in the figure below.
Figure 6 14: First Layer Adhering To Screen
Problem: The method of using the screen to give the build plate the adhesion necessary for the first layer to stick and work tremendously well, however an unforeseen issue with this method came from the nozzle barely dragging against the gritty screen. These nozzles where not designed to come into contact with any other surfaces, so the slight dragging of the nozzle quickly wears down the tip of the nozzle, causing the tip to lose its inside dimension. When the slicing software does the tool path calculation, it expects the nozzle size to be what it is set at, but the dragging of the nozzle slowly increased this size, throwing off the calculations and resulting in an insufficient print.
Solution: Because the tip was dragging through and was not particularly thick to start with, I attempted to braze the end of the tip in an effort to make it thicker and harder. This design alternative failed due to the nozzle collapsing as the heat was applied to perform the brazing, leaving the nozzle extremely deformed. The only alternative solution was to just start with a new nozzle whenever the nozzle started to loose dimensions. This was not optimal because it is expensive to do, however given the tight timeline; this had to be done to keep things progressing.
Problem: After a very long and bumpy road, the printer was now actually producing prints that were close to acceptable quality to be tested. The screen was keeping the printed locked in place so issues such as warping were not a problem at all. After some small calibrating, the printer was really putting out nice looking test coupons. The only downfall to using the screen and getting this excellent adhesion, is that is was sticking a little too well. After the print was finished, the screen would be entangled in the bottom of the print, leaving a rough surface and screen remnants, as seen in Figure 6-15.
Figure 6 15: Screen Remains In Print
Solution: Now that it is known that the screen sticks into the plastic, an extra piece could be added to the model to make it slightly thicker. This would allow for the screen layer to be sanded off after the print is finished. There is also a setting called a raft that will put down a sacrificial first layer to help prevent warping of the part while printing and cooling. This method seems practical, however has not been tested because it uses significantly amount of material to create the raft, and I am working with a limited amount that Arevo Labs donated for this project.
Thermal Management System
Problem: After all the bugs had been worked out and I was able to print consistently after following a list of procedures, I began the production of my test coupons. I had to allow everything to get up to temperature and stabilize, level the build plate, print a coupon, level the build plate, print etc. until I had all my test pieces. Each print with the 0°/90° infill pattern took 54 minutes to complete, so after the second print, including a few accidents, the printer had been going for at least 5 hours. After being run at these extensive conditions for such a long time, it started to give out. One of the extruder fans flung off several fins, then the heated build plate started becoming unable to maintain a temperature setting, and lastly a fan that cools the motherboard shredded a few fins. It was at this point that I decided to stop from proceeding further. This higher temperature operating range was clearly taking its toll on the entire system and pushing it to its limit. A few fans are cheap to replace, but the next component to go, like the insufficiently cooled motherboard, would not be.
Solution: Although it is unfortunate that I was only able to get two samples before the printer began failing, this does illustrate how changing one small component in a system can make a huge impact on the rest of the system. To be able to successfully print this material, all of the design recommendations would need to be done in order to have a printer that can resist the much harsher conditions.
Having only 2 samples, I was not going to have enough data to prove anything advantageous, however it would still be beneficial to see how the samples turned out. The lab where the test were performed was 24°C and had a barometric pressure of 1012 hPa. The tensile testing machine was set to pull at a rate of 5mm/sec. The results of the testing can be seen in the figure below.
|±45 Honeycomb (default)||83mm||85mm||742.5 N|
|0°/90° Rectilinear||87mm||91mm||1539.17 N|
|0°/90° Rectilinear||87mm||91mm||1386.67 N|
Table 7 1: Tensile Results
The ±45 Honeycomb print was a starting print using default settings to make sure everything was working correctly. It was not a neat, consistent print, however it was added to help illustrate the behavior of the material when compared to the 0°/90° Rectilinear print. The ultimate tensile strength was calculated using: σ=P/A
The dimensions of the cross sectional area is 13mm by 3mm, giving an area of 0.039 m2. The ultimate tensile strength was then calculated to be 19.04 kPa for the ±45 Honeycomb and an average of 37.51 kPa. This material is likely to have failed before it should have because of the stress concentrations created by the plastic sticking in between the screen.
Arevo Labs reports an ultimate tensile strength of 97 MPa for this material, however it is important to note that this could be the value based on an injection-molded sample, or using the most optimal parameters possible.
The overall behavior of the material is analogous to what you would expect from an anisotropic material. The break patterns are also consistent with what was seen in the background research. Although I would like to go into further detail, I do not have sufficient data to back up any claims.
After working through the entire semester in an effort to print this high performance material, I can say with great confidence that this is not an easy material to work with. The material is very strong and is able to perform in a wide array of conditions, however it is equally as difficult to work with. It requires 4 hours in an oven to dry the moisture out of the material for printing, so it is also extremely time consuming to just get a single print done. Originally, I was determined to test various build parameters in order to fin the optimal settings. However, I was under the impression that this material would be printable with some slight modifications. The reality is that the machine needs to be severely redesigned to be able to handle the more extreme conditions. The stock machines are designed to work for a few hours a day in the 100°C to 250°C temperature range, and you are now trying to run it all day everyday in the 220°C to 350°C temperature range. It simply just was not designed to handle this.
Through out the semester I have been in close contact with Arevo Labs, as they have been very helpful in giving me tips and trying to steer me in the right direction. This is a very new state of the art printing filament, so while they are very supportive of research involving their product, there is also a gray area between public knowledge and trade secrets. Keeping that in mind, you have to understand why things like the build plate paste are not given to the public. It would have been helpful to have, but at the same time it is in the best interest of their company to not give it to me at this time. Hopefully in the future, the relationship between them and myself is only strengthened, so eventually I will be trusted with such information. This is sensitive information, and is completely reasonable to try to protect it. From the research that I have done on the material, very few people have access to it, and the few that do have a very difficult time getting it to print, and I was no exception. This material is very temperamental to print, and requires not only a special printer, but a wealth of knowledge on how the process works. As of right now, it seems that Arevo Labs is the only group to achieve both of those feats.
There is a strong lack of supporting data to making any claims about the materials performance, however based on the few samples that where tested, the material did show behavior similar to anisotropic materials. From the background research, we know that this is how printed materials should act. This is a indicator that the material was correctly used.
Having gone through the process, I have become skeptical if some of the information presented in the other studies are even realistic. Several of the published documents show simulated data trends in order to come up with optimal build settings. While mathematically this may make sense, it might not in reality. In order for the material to print, everything has to be just right when it comes to parameters such as the layer height. The model could show that 0.19mm is the optimal layer height, when in reality that height with a 0.4mm nozzle will cause the extruder to jam. In order to really develop optimal settings, test would need to be run with the various settings altered and physically tested.
In the future, it would be beneficial to the field to build a printer that is capable of handling the conditions necessary to print this material, and run the test that I originally proposed to do. This would give hard data, allowing a relationship to be found. This relationship is a complicated one so a method such as multiple variable regression analysis may have to be used, but it is a starting point until more is understood about the process.
Ahn, S., Montero, M., Odell, D., Roundy, S., & Wright, P. (2002). Anisotropic material properties of fused deposition modeling ABS. Rapid Prototyping, 8(4), 248-257. Retrieved April 1, 2015, from http://www.emeraldinsight.com/1355-2546.htm
Bagsik, A., Schuppner, V., & Klemp, E. (2010). FDM Part Quality Manufactured with Ultem*9085. Polymetric Materials 2010, 1-8.
Fly, D., & Acan, M. (2013). 3D Printed Internal Structure: Influence on Tensile Strength. ASEE.
Jetguy. (2014, April 1). UPDATED SAILFISH installation instructions. Retrieved May 20, 205, from https://groups.google.com/forum/#!topic/wanhao-printer-3d/-0pryBjsd1g%5B1-25%5D
Nancharaiah, T., Ranga Raju, D., & Ramachandra, V. (2010). An experimental investigation on surface quality and dimensional accuracy of FDM components. International Journal on Emerging Technologies, 1(2), 106-111.
Panda, S., Padhee, S., Sood, A., & Mahapatra, S. (2009). Optimization of Fused Deposition Modeling (FDM) Process Parameters Using Bacterial Foraging Technique. Intelligent Information Management, 1, 89-97.
Production of Parts for Final Products is Now 34.7% of the Market for Additive Manufacturing. (2014, October 15). Retrieved April 16, 2015.
Sood. A., Ohdar, R., & Mahapatra, S. (2011). Experimental investigation and empirical modeling of FDM process for compressive strength improvement. Journal of Advanced Research, 3, 81-90
Wohlers Associates. (2015, April 6). Retrieved April 14, 2015, from http://wohlersassociates.com/press69.html
Ziemian, C., Sharma, M., & Ziemian, S. (2011). Anisotropic Mechanical Properties of ABS Parts Fabricated by Fused Deposition Modeling (D. Gokcek, Ed.). Mechanical Engineering, 159-180. Retrieved April 1, 2015, from www.intechopen.com/books/mechanical-engineering/anisotrophic-mechanical-properties-of-abs-parts-fabricated-by-fused-deposition-modeling-
Figure 6-7 Page 18
Makerbot Stepstruder MK7. (2011, September 20). Retrived May 18, 2015, from http://www.thingiverse.com/thing:11734
Figure 6-8 Page 19
Bergman, T. L., Dewit, D. P., Lavine, & A. S., Incropera, F. P. (2011). One-Dimensional, Steady-State Conduction. In Fundamentals of heat and mass trasnfer (7th ed., p. 170). New Jersey: John Wiley & Sons.