Abstract
Additive manufacturing (AM) techniques, such as fused deposition modeling (FDM), are able to fabricate physical components from three-dimensional (3D) digital models through the sequential deposition of material onto a print bed in a layer-by-layer fashion. In FDM and many other AM techniques, it is critical that the part adheres to the bed during printing. After printing, however, excessive bed adhesion can lead to part damage or prevent automated part removal. In this work, we validate a novel testing method that quickly and cheaply evaluates bed adhesion without constraints on part geometry. Using this method, we study the effect of bed temperature on the peak removal force for polylactic acid (PLA) parts printed on bare borosilicate glass and polyimide (PI)-coated beds. In addition to validating conventional wisdom that bed adhesion is maximized between 60 and 70 °C (140 and 158 °F), we observe that cooling the bed below 40 °C (104 °F), as is commonly done to facilitate part removal, has minimal additional benefit. Counterintuitively, we find that heating the bed after printing is often a more efficient process for facile part removal. In addition to introducing a general method for measuring and optimizing bed adhesion via bed temperature modulation, these results can be used to accelerate the production and testing of AM components in printer farms and autonomous research systems.
1 Introduction
Additive manufacturing (AM), the process of building an object layer-by-layer from a three-dimensional (3D) digital model, has experienced tremendous growth for rapid prototyping and producing parts in small volume or with complex geometries [1]. Among the myriad and growing families of AM [2], material extrusion (ME), commonly known as fused deposition modeling (FDM) or fused filament fabrication (FFF), is a popular method for AM because of its low cost and ease of use. In FDM, thermoplastic polymer filament is pushed through a heated nozzle and deposited onto a print bed [3,4]. Despite the simplicity of this approach, selecting process parameters for successful printing can often be subtle, which has sparked numerous studies focusing on process parameters such as extruder temperature, layer thickness, raster width, raster orientation, air gap, bed temperature, print speed, and print orientation [5–11]. An additional and less studied property is bed adhesion, which refers to the adhesion between the part and the printer bed. Bed adhesion is a critical factor to manage since insufficient adhesion can lead to premature detachment of the part from the bed, resulting in a failed print [12–14].
With a focus on avoiding failed prints from low adhesion, several studies have sought to maximize adhesion by modulating bed temperature and bed composition [15–17]. While achieving high adhesion is desirable during printing, high adhesion is undesirable after printing has concluded because the part could be damaged during removal. Further, automated part retrieval has been demonstrated as a method for dramatically accelerating part production and mechanical design, but it requires careful tuning of adhesion to enable robotic retrieval [18,19]. To circumvent the need for automated part removal from the bed, one can automatically replace the entire build plate after each print; however, this approach prevents automated mechanical testing [20]. To minimize adhesion after printing, common practice is to let the bed cool to room temperature [13], but this practice has not been systematically evaluated. In order to accelerate automated part retrieval in AM, a basic understanding of how to efficiently manage adhesion during and after printing is needed.
Here, we study the peak force required to remove a part printed using FDM as a function of bed temperature on both bare and polyimide (PI)-coated glass beds. Initially, we identify a simple and transferable approach to rapidly measure the peak removal force required to remove a part from the printer bed, which can be used to study different part geometries. In order to independently evaluate the effect of adhesion during and after the print, we conduct studies with distinct temperatures during and after printing. Through a systematic study of removal force versus bed temperature, we determine the optimal print bed temperature, the minimum effective removal temperature, and several benefits of heating rather than cooling for part removal. Finally, we evaluate the complete cycle time associated with printing and removing parts and find that, strikingly, heating after printing is more efficient than cooling after printing. In addition to being of fundamental interest to the AM community, this bears important implications for FDM-based autonomous research and high throughput AM. While the discussion here is limited to polylactic acid (PLA) cylinders, the simplicity of this testing approach would allow researchers to rapidly evaluate other chemistries or geometries. Given the dramatic increase in interest in custom filaments for FDM, we anticipate that optimizing adhesion for emerging materials will only increase in importance [21–23].
2 Material and Methods
2.1 Design and Fabrication.
Unless otherwise specified, the hollow cylindrical parts used in this study were designed using openscad to be 20 mm (0.79 in.) in radius and 30 mm (1.2 in.) in height. The G-code for the parts was generated using slic3r and then sent to the FDM 3D printer (Makergear M2) using repetier. The parts were printed using PLA (Makergear) filament and a 0.35 mm (0.014 in.) diameter brass extrusion nozzle. The parts were either printed directly on a borosilicate glass bed (Makergear) or a borosilicate glass bed coated with a 51 mm (2 in.) wide strip of PI tape (Makergear). The PI film was replaced if damage such as tears or bubbles was observed. Although it is common for PI films to be used for many months before replacement, the high removal forces in some of our experiments caused premature damage. Therefore, we inspected the film following each set of experiments to ensure that all samples were printed under identical conditions.
The hollow cylindrical parts were printed with three top layers, three bottom layers, two perimeter lines, a 0.39 mm (0.015 in.) perimeter line width, no infill, and an alternating rectilinear pattern for the top and bottom layers. The layer height was set to 0.3 mm (0.012 in.), with the exception of the first layer height, which was set to 0.35 mm (0.014 in.). No initial layer height adjustments were made to compensate for the thickness of the PI film. The first layer was printed at 30 mm/s (1.2 in./s) while subsequent layers were printed at 60 mm/s (2.4 in./s), with the exception of the outermost perimeter, which was printed at 30 mm/s (1.2 in./s) for all layers. Bed temperature was measured using the Makergear M2’s temperature interface. When a print was started, the bed and nozzle were heated to the required temperatures. The nozzle was set to 220 °C (428 °F) for all prints. The bed print temperature Tp was set to either 60 or 70 °C (140 or 158 °F). This range was chosen as, during our initial tests, bed print temperatures above or below these values commonly resulted in failed prints on the glass bed. The print would start when both the bed and nozzle temperatures reached their set points, as measured by the M2’s built-in temperature interface. The extruder fan was set to 0% for the first three layers, and then to 100% for subsequent layers. After the print ended, the bed temperature was set to the removal temperature Tr. Part removal started immediately upon reaching the target bed temperature, and removal of all specimens on the bed typically took less than a minute. For each production cycle, three cylinders were printed on glass and three were printed on PI-coated glass. Therefore, six cylinders were printed simultaneously. For each set of conditions tested, three production cycles were conducted, providing nine data points for each unique parameter combination.
2.2 Testing Procedure.
During testing, the peak removal force was measured using a digital force gauge (HF-500), which has a 0.1 N (0.02 lbf) resolution and a 5 N (1.1 lbf) minimum detectable force. The gauge was placed horizontally with the tip centered on the top of the cylinder. Increasing pressure was applied laterally until the part completely released from the bed surface or until the part broke. The peak measured force was recorded as the removal force Fr. If the Fr < 5 N (1.1 lbf), then 0 N (0 lbf) was taken as Fr as the sensor could not register forces below its threshold value. If the part broke before removal, the peak force prior to the break was taken as Fr. Each temperature condition was evaluated for nine identically prepared samples for a total of approximately 400 tested cylinders.
3 Results and Discussion
In order to rapidly evaluate part-bed adhesion in FDM, we sought to develop a facile method to measure the force required to remove a part from the bed after printing. Prior methods either relied upon shearing thin strips of printed material that were three layers tall [17] or involved pulling parts vertically away from the print bed [15,16]. While these methods provide valuable insight, they require the development of an expensive testing system and do not directly measure the forces required to remove parts in practice. Thus, we hypothesized that pushing a part laterally with a force gauge while recording the peak force would more closely resemble how humans or robots remove parts from the print bed in practice. Pushing laterally on a part with a force gauge is likely not the most efficient method of part removal in terms of minimizing the force required. For example, a two-sided gripper could impart a controlled torque along with a lateral force and also have superior control over the part after removal. However, the force gauge method used here provides repeatability at low cost, allowing operators to quickly evaluate the ideal print conditions for their part. For this reason, we believe it could be a useful tool for studying adhesion and optimizing print parameters.
We developed a facile process for measuring the adhesion between FDM printed components and the bed (see Table 1 and Sec. 2). In brief hollow cylindrical parts were printed on either bare glass or PI-coated glass at a constant bed temperature Tp and constant nozzle temperature Tnoz (Figs. 1(a) and 1(b)). Subsequently, the bed was heated or cooled to the removal bed temperature Tr and then the force gauge was used to remove the part through an application of force that was subsequently recorded (Fig. 1(c)) with the maximum force being recorded as Fr (Fig. 1(d)). While this force was representative of what is required to remove a specific part, whether this force provides any predictive information that is transferrable to other part geometries was not known.
Radius r | Height h | ||
---|---|---|---|
15 mm (0.59 in.) | 30 mm (1.2 in.) | 45 mm (1.8 in.) | |
16 mm (0.63 in.) | 0 | 9 PI, 9 glass | 0 |
20 mm (0.79 in.) | 9 PI, 9 glass | 9 PI, 9 glass | 9 PI, 9 glass |
24 mm (0.94 in.) | 0 | 9 PI, 9 glass | 0 |
Radius r | Height h | ||
---|---|---|---|
15 mm (0.59 in.) | 30 mm (1.2 in.) | 45 mm (1.8 in.) | |
16 mm (0.63 in.) | 0 | 9 PI, 9 glass | 0 |
20 mm (0.79 in.) | 9 PI, 9 glass | 9 PI, 9 glass | 9 PI, 9 glass |
24 mm (0.94 in.) | 0 | 9 PI, 9 glass | 0 |
Note: In all, 90 samples were printed to test different geometries.
To test whether the force needed to remove a part can be interpreted in a general context, we performed a series of adhesion tests on 90 cylinders with varying geometries. For instance, as the cylinder radius was increased from 16 to 24 mm (0.63 to 0.94 in.), Fr for parts printed on PI was found to increase by 50% (Fig. 2(a)). Interestingly, Fr required for parts printed on glass exhibited a similar increase, albeit with a lower magnitude. Qualitatively, the need for a larger force is consistent with the larger contact area between the part and the build plate, resulting in stronger adhesion. In contrast, as the cylinder height was increased, Fr was observed to decrease (Fig. 2(b)), which we hypothesize was due to a longer moment arm between the applied force and the part-bed interface. While these measurements have large standard deviations, one-way analysis of variance shows that these points have a less than 0.1% chance of reflecting the same distributions. Similarly, the trend for parts printed on bare glass was consistent with that exhibited on PI-coated glass albeit with a lower magnitude. From these results, we conclude that while the absolute Fr depended on the part geometry, the ratio of Fr between glass and PI remained statistically indistinguishable (p-values significantly above 0.05) in all geometry-based experiments (Fig. 2(c)). Specifically, for all 90 specimens, the average ratio was 0.27 with a standard deviation of 0.10, which implies that, at this removal temperature, parts printed on PI required ∼3.5 times more force to remove than parts printed on glass. This shows that, at least in a relative sense, this testing method allows one to measure a quantity that is reflective of the fundamental interaction between the part and the bed. Further, this implies that upon using a standard part geometry, measuring Fr can allow one to study how altering processing parameters affects adhesion.
Having shown that Fr for a specific part is of fundamental value, we sought to determine whether other factors could result in robust control over adhesion. Specifically, it is important to maximize adhesion during printing to prevent print failure, but after printing, low adhesion is preferable to facilitate part removal. Thus, we hypothesized—building off conventional wisdom of the FDM community—that modulating bed temperature could be a rapid path to dynamically tuning adhesion. To explore this, 288 hollow cylinders (see Table 2 and Sec. 2) were printed using various Tp and Tr (Fig. 3). Two major trends were immediately apparent: (1) PI robustly exhibited larger Fr and (2) Fr was maximized when Tr ∼60–70 °C (140–158 °F) with lower adhesion observed at hotter or colder temperatures. Intriguingly, Fr always appeared to be maximized when Tp differed from Tr. Specifically, when printing at 60 °C (140 °F), the force was largest when removing the part at 70 °C (158 °F) and vice versa. While these points exhibited a great deal of variation and these trends could be due to limited statistics, it is interesting to hypothesize mechanisms that might lead to such a trend. For instance, perhaps printing at 70 °C (158 °F) results in more conformal contact between the part and bed, but the work of adhesion is higher at 60 °C (140 °F). This phenomenon warrants further study.
Removal temperature Tr | Print temperature Tp | |
---|---|---|
60 °C (140 °F) | 70 °C (158 °F) | |
30 °C (86 °F) | 9 PI, 9 glass | 9 PI, 9 glass |
40 °C (104 °F) | 9 PI, 9 glass | 9 PI, 9 glass |
50 °C (122 °F) | 9 PI, 9 glass | 9 PI, 9 glass |
60 °C (140 °F) | 9 PI, 9 glass | 9 PI, 9 glass |
70 °C (158 °F) | 9 PI, 9 glass | 9 PI, 9 glass |
80 °C (176 °F) | 9 PI, 9 glass | 9 PI, 9 glass |
90 °C (194 °F) | 9 PI, 9 glass | 9 PI, 9 glass |
100 °C (212 °F) | 9 PI, 9 glass | 9 PI, 9 glass |
Removal temperature Tr | Print temperature Tp | |
---|---|---|
60 °C (140 °F) | 70 °C (158 °F) | |
30 °C (86 °F) | 9 PI, 9 glass | 9 PI, 9 glass |
40 °C (104 °F) | 9 PI, 9 glass | 9 PI, 9 glass |
50 °C (122 °F) | 9 PI, 9 glass | 9 PI, 9 glass |
60 °C (140 °F) | 9 PI, 9 glass | 9 PI, 9 glass |
70 °C (158 °F) | 9 PI, 9 glass | 9 PI, 9 glass |
80 °C (176 °F) | 9 PI, 9 glass | 9 PI, 9 glass |
90 °C (194 °F) | 9 PI, 9 glass | 9 PI, 9 glass |
100 °C (212 °F) | 9 PI, 9 glass | 9 PI, 9 glass |
Note: Bed temperature Tp during printing was set at either 60 °C (140 °F) or 70 °C (158 °F). Bed temperature Tr during removal ranged from 30 to 100 °C (86 to 212 °F) in increments for 10 °C (18 °F).
Interestingly, significantly greater variance of Fr was found at Tr = 60 °C (140 °F). We believe that this is because the glass transition temperature of PLA is ∼60 °C (140 °F) [17], leading to slight variations in bed temperature having an outsized effect on adhesion. This hypothesis could be tested by performing equivalent studies with materials that have a higher or lower glass transition temperature.
The systematic study of Tr provided several important observations with relevance to efficient FDM. (1) In a validation of common practice, cooling the bed after printing made parts significantly easier to release. Interestingly, 40 °C (104 °F) was the most effective choice for Tr as Tr < 40 °C (104 °F) showed no additional reduction in Fr. (2) The consistently higher Fr observed when printing on PI was promising for avoiding adhesion-related print failures, but its substantially higher floor posed a barrier to facile automated, or manual, part retrieval. Thus, counterintuitively, glass beds may likely be superior for high throughput manufacturing. (3) Strikingly, for PI, heating after printing may be more effective than cooling at lowering Fr. In particular, Fr only decreased ∼50% when testing on PI-coated glass when Tr < Tp. However, for Tr > Tp, decreases in Fr of 80–90% were observed. An additional tradeoff observed for heating after printing, however, was that heating the bed to 90–100 °C (194–212 °F) sometimes led to warping or deformation of the bottom of the part, highlighting the need for part-specific testing.
While we did not observe any drawbacks to having high adhesion during printing, care is warranted when minimizing adhesion after printing by elevating temperature as we observed several instances of plastic deformation during removal at temperatures ≥90 °C (194 °F), especially when printing on PI film. Interestingly, while parts printed on PI film sometimes failed during removal, they did so in one of two distinct regimes. When Tr = 60 or 70 °C (140 or 158 °F), parts sometimes failed by interlayer delamination as the removal force overcame interlayer bonding. However, at Tr ≥ 90 °C (194 °F), the heated parts sometimes buckled due to lower structural integrity cause by the higher bed temperature softening the part. Interlayer delamination was also occasionally seen at these higher temperatures, although less often. Importantly, no parts that were printed on glass were damaged during removal. Thus, we do not believe that there is a single ideal removal process for all situations, but rather that heating should be considered as an alternative to cooling to accelerate the process. Importantly, the rapid method for measuring adhesion is an important part of evaluating potential strategies for accelerating part removal.
The observation that heating after printing may be superior to cooling raises the further question of whether this could be faster from the perspective of printing cycle time. Thus, defining the cycle time t to be the time required to heat/cool the bed from Tp to Tr and back, it is clear from the measured trajectories of the bed temperature that heating was faster than cooling for a given temperature swing (Fig. 4(a)). This is because the cooling rate is faster at temperatures further from room temperature. Additionally, the heating rate was faster than the cooling rate as the speed of heating was determined by the strength of the heater while cooling required convective/conductive transport into the surrounding environment. Considering two cycles with equivalent total temperature swings, cycling the bed between 60 °C (140 °F) and 90 °C (194 °F) took t = 12.3 min, while cycling the bed between 60 °C (140 °F) and 30 °C (86 °F) took t = 28.2 min. When these trajectories were combined with measurements of Fr, they allowed us to compute Fr versus t (Figs. 4(c)–4(f)). Critically, these plots depict the tradeoffs inherent to minimizing both Fr and t. For instance, when Tp = 70 °C (158 °F) on either bare glass or PI-coated glass (Figs. 4(d) and 4(f)), heating to 80 °C (176 °F) provided a more than 50% decrease in Fr while taking less than 5 min. Out of all the conditions tested, cooling was only consistently more efficient when Tp = 60 °C (140 °F) on bare glass (Fig. 4(c)). This process would also allow one to evaluate the acceleration afforded by hardware alterations such as the addition of a fan or stronger heater to reduce cycle time.
4 Conclusion
Ultimately, this work represents a roadmap for implementing a testing protocol to quickly determine the optimum conditions for a specific printer both in terms of maximizing adhesion during printing and minimizing cycle time between prints. With the expanding influence of FDM and advent of 3D printing farms and autonomous research systems based on AM, managing part adhesion to efficiently swing between high adhesion during printing and low adhesion during removal is of critical importance. Here, we explored both Tr and bed material to elucidate optimal conditions for managing these disparate constraints. Key takeaways include:
A novel and easy-to-implement testing method was developed that replicates how humans and robots remove parts to quickly and cheaply measure Fr. This testing method was validated by testing Fr for parts of varying geometry to find that PI-coating a glass bed increases Fr by ∼3.5 fold across various print bed temperatures and cylinder shapes.
By exploring Fr versus Tr, we found that not only can heating be more effective at decreasing Fr, but that it may also lead to quicker cycle times. Printing a single object at a time brings significant advantages on an FDM printer, such as eliminating travel time between parts and improving part surface finish. However, job times for single-part prints are significantly shorter, making it critical to minimize the downtime between prints. Thus, these results will have high relevance to both small-batch manufacturing and autonomous research using AM.
Finally, Tr < 40 °C (104 °F) did not further decrease Fr, showing a point of diminishing return in the conventional method of allowing the bed to cool after printing.
Acknowledgment
This work was supported by Google LLC, the Boston University Rafik B. Hariri Institute for Computing and Computational Science & Engineering, the National Science Foundation (CMMI-1661412), and the US Army CCDC Soldier Center (contract W911QY2020002).
Conflict of Interest
There are no conflicts of interest.
Data Availability Statement
The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request. The authors attest that all data for this study are included in the paper. Data provided by a third party are listed in Acknowledgment.