1. Introduction
Additive Manufacturing (AM) has enabled the rapid manufacturing of prototypes, one-off productions, and complex geometries that traditional manufacturing could not achieve effectively. Fused filament fabrication (FFF) has undergone wide spread adoption, enabling printing of thermoplastics such as PLA and ABS [
1]. Metal based additive manufacturing (MBAM) has enabled the use of additive manufacturing for structural components that require the durability and performance of metallic materials [
2,
3]. Many MBAM processes are well established, such as laser-powder bed fusion (L-PBF), which creates 3D geometries by melting metallic powder in an inert environment in a layer wise fashion [
4,
5,
6]. While L-PBF can be enabling for unique geometries and alloys, the equipment, infrastructure, and associated hazards are complex and potentially cost prohibitive for smaller scale research/commercial efforts.
While FFF printing has been historically limited to thermoplastics, metal extrusion additive manufacturing (metal MEX) has enabled methods to create metallic components with FFF hardware and post-processing. Metal MEX originated by re-purposing FFF equipment and experimental methodologies before all-in-one commercial solutions from MarkForged and Desktop Metal came to market which are comparable in cost to some L-PBF machines. Instead, hybrid materials like Ultrafuse 316L (BASF 3D Printing Solutions GmbH, Heidelberg, Germany) metal filament can be utilized to create fully dense metallic components on common FFF printers, with post-processing standardized and outsourced [
7]. Ultrafuse 316L metal filament is comprised of 316L stainless steel particles (approximately 90% by mass) suspended in a polymer binder, allowing the material to be printed in nearly an identical method to thermoplastics [
8]. However, two stages of post-processing are required to produce a fully dense, metallic component. Immediately after printing, the component is referred to as a “green part” and looks visually similar to traditional FFF components since it is mostly composed of polymer binder. First the “green part” undergoes catalylic debinding, which removes a majority of the polymer binder, leaving behind a “brown part”. The remaining binder ensures the encased metallic powder remains the correct geometric shape. Finally, the “brown part” is sintered, causing the particles to combine into a fully dense steel part [
9]. The result is a metallic component; however, this post-processing causes the parts to shrink anisotropically [
10]. This process does not seek to replace traditional MBAM processes, but allows for the creation of metal components at a significantly reduced cost compared to traditional MBAM. While MEX will not likely be utilized to create structural components, there are many applications that will see higher stresses than plastics can withstand. Rapid prototyping and one-off production of components such as levers, instrument panels, or any class B1-B2 structure outlined in NASA’s “Standard for Additively Manufactured Spaceflight Hardware by Laser Powder Bed Fusion in Metals could be produced with MEX” [
11].
A number of studies have been conducted to determine optimal process parameters and properties with Ultrafuse 316L. Thompson et al. [
8] published one of the earliest overviews of the process, using a Prusa i3 Mk2 printer. Several studies have set out to determine optimal print parameters, varying nozzle temperature, infill strategy, build orientation, print speed, extrusion multiplier, line width, and layer height [
12,
13]. The effects of print parameters on shrinkage during processing [
12,
14,
15,
16], geometric accuracy [
16], density [
12,
17,
18], tension tests [
8,
13,
14,
15,
18,
19], optimization of post-processing [
20], fatigue [
21], hardness [
14,
18,
22], and surface roughness [
22] have been investigated, often comparing to samples produced from L-PBF [
19,
22], other MEX processes such as Desktop Metal and Markforged [
21,
22] or samples from conventionally manufactured 316L plate [
8,
14]. Typically industrial or commerical grade FFF printers have been utilized in previous research. Industrial printers such as the Apium P155 (
$10,000–
$25,000+) or commercial grade such as the Creatbot F430 [
16], Lulzbot TAZ6 [
20], Intamsys Funmat HT [
19], Ultimaker S5 [
12], Ultimaker 3 [
13] (
$2000–
$10,000), or consumer grade such as Felix Tec 4 [
15] and Prusa i3 [
8,
22] (
$1000–
$2000) have been utilized. At present, no studies have utilized any of the common hobbyist grade printers available for under
$500.
Previous research has established that printing specimens in the horizontal orientation provides the highest tensile strength [
17]. Furthermore, it has been established that factors such as nozzle temperature have little effect on density [
12]. The process has proved to produce 316L samples with tensile strengths similar to L-PBF or conventionally produced 316L but with a reduced yield strength [
9,
13] and approximately 2–5% porosity [
8,
12,
14,
17,
22]. However, previous literature has highlighted that typical mechanical properties are reduced from those advertised by the manufacturer due to porosity [
19]. Process-induced porosity is a significant issue for metal FFF, with pores naturally forming between parallel print lines during deposition and post-processing. The reported density from previous research varies, with reports of up to 95–98% dense parts [
13,
17], resulting from many different methods being utilized to calculate the volume/density/porosity. Previous research has utilized Archimedes principle [
13], X-ray micro-computed tomogrpahy (micro CT) [
17], gas-pycnometry [
12], and optical micrographs [
8,
17]. Damon et al. [
17] correlated process-induced porosity to mechanical properties using micro CT and micrographs, concluding the formation of pores is strongly connected to build direction with elongated pore channels forming along parallel print paths and turning points between internal hatching and perimeters. Damon et al. [
17] recommends further research focus on reduction of process-dependent porosity through optimization of extrusion parameters. These findings are further supported by Safka et al. [
15] utilizing a 106% extrusion multiplier to increase tensile and yield strength; however, they did not report density or porosity. Developing an understanding of which extrusion parameters have the greatest potential to improve mechanical properties and mitigate porosity requires further investigation. Additionally, there is little understanding of how this may vary across multiple FFF machines and if process parameters are machine agnostic.
Previous research has only utilized one FFF printer within each respective study, often focusing on the effects of print parameters on density/shrinkage [
12] or mechanical properties [
13]. At present, the variation of optimal process parameters between FFF printers is not understood, since control software, slicing/print path, range of allowable parameters, and repeatability varies with each printer. There is little understanding of how equivalent print parameters vary between FFF printers. Few studies have compared multiple density measurement techniques, tension, and hardness testing across replicates within a single design of experiments.
To understand the variation of print parameters across multiple FFF printers, this study compares two FFF printers, by analyzing the effects of varying extrusion process parameters on density and mechanical properties with Ultrafuse 316L (BASF 3D Printing Solutions GmbH, Heidelberg, Germany). A Makerbot Method X, a commercial grade FFF printer is directly compared to a Creality Ender 3 V2, one of the lowest cost and most widely available hobbyist FFF printers. This study has three specific goals: evaluate if budget, hobbyist grade, FFF printers can utilize Ultrafuse 316L to produce metallic components for a total investment of approximately $1000, explore the use of gas-pycnometry and optical micrographs to measure density, and evaluate the impact of extrusion process parameters on density and determine if they are machine agnostic.
3. Results & Discussion
The measured density, relative density, and optical microscopy (OM) density of FFF and L-PBF specimens are presented in
Table 4. Relative density and OM density is both presented relative to the published density of 316L stainless steel of 7.84 g/cm
3 [
21]. OM density was measured from micrographs of three specimens for each process parameter combination. The MBDS samples exhibited the highest OM density at 99.3% for specimens printed one-by-one and four at a time. However, M221 demonstrated a similar OM density of 99.1%. The Ender samples exhibited a maximum OM density of 95.9% for E212, and minimum of 80.8% for E221.
Method X samples demonstrated the highest OM density with little difference in mechanical strength between all parameter combinations. However, a difference in strength and porosity was present in the Ender samples. E122 and E212 performed the best with only a 7–10% reduction in strength as compared to the average Method X sample. E221 and E111 had a significant reduction in mechanical strength with the lowest OM density. The average (
) mechanical response and the standard deviation is presented in
Table 5. A representative stress-strain curve is presented in
Figure 4, where “S” denotes the sample shown on the plot.
Ender samples made with the 120% extrusion multiplier obtained comparable mechanical properties to previous research, but had a reduced yield strength of 135–142 MPa compared to previous values of 153–180 MPa in literature [
13,
14,
19]. All of the Method X prints obtained comparable or improved properties to previous research. The Young’s modulus was comparable to the 151.3–185 GPa [
13,
14,
17] and tensile strength was higher than the 465–498 MPa [
13,
14] reported in previous studies. Due to porosity, only three E212 (0.5 mm line width, 0.1 mm layer height, and 120% extrusion) samples were produced. Furthermore, large irregular pores in E111 meant hardness testing could not be performed. The most dense Ender sample, E212 and least dense, E221, when compared show the same trend of reduced strength and elongation as previous research [
28].
3.1. Gas Pycnometry vs. Optical Microscopy Density (%)
Since gas pycnometry measures enclosed volume, the reported density is of the solid components and only accounts for entrapped porosity, resulting in artificially high density values for these samples due to elongated, open pores. Optical microscopy can account for both open and closed porosity within the sample and produces density values more representative of the bulk component density when open pores are present. Even with an increased porosity, the Ender samples often had open pores forming along print lines, causing the artificial increase in measured density. The Method X and L-PBF samples often had closed pores, resulting in the more representative measured density, which correlates to the respective OM density. While gas-pycnometry can accurately measure density when closed pores are present, OM density or
should be utilized to accurately measure density when open pores are present. Examples of the process-induced porosity is highlighted in
Figure 5.
3.2. Process-Induced Porosity
The distribution, shape, and size of pores varied with combinations of process parameters and between both FFF printers. At the lowest line width, layer height, and extrusion multiplier, porosity is the most exaggerated. E111 featured large irregularly shaped pores/voids (average pore size of 12 μm and circularity of 0.32), alongside small pores distributed throughout. Pores were much smaller and more regularly shaped in the M111 samples, with an average pore size of 1.5 μm and circularity of 0.53. However, both 111 samples show a trend of pores forming in the vertical direction, particularly around the exterior faces, and the interface of the two outer contours shown in
Figure 5a,b.
In general, closed pores were present in Method X samples between print lines and layers throughout the infill regions. Porosity in Ender samples followed two trends: (a) homogeneous porosity across the infill region when the 100% extrusion multiplier was used (
Figure 5a,b) and the formation of pores along the vertical direction when the 120% extrusion multiplier is used, which is exaggerated at the contour-infill interface (
Figure 5c). The spacing between the vertical pores within the infill region of sample E122 (0.57 mm) is slightly larger than the target line width of 0.5 mm for these parts. The differences in closed vs. open porosity between the two FFF printers (
Figure 6) suggest the influence of uncontrolled process variables. Examples could include differences in machine architecture such as filament feed mechanism or chamber temperature control. Though these factors were not accounted for in this study, they are potential contributors to the discrepancy between specimens made on two different systems while the primary printing parameters were held constant.
The aforementioned differences in distribution, shape, and amount of porosity impacted the mechanical response. All Method X specimens had an OM density of over 97%, and the Ender 3 V2 specimens had a maximum OM density of 96%. The small isolated pores in Method X samples did not greatly reduce the elastic modulus, but are likely responsible for the reduction in yield strength without greatly impacting tensile strength. The large irregular pores in the E111 specimens highlight the other extreme of this trend, where the elastic modulus is decreased due to reduction in effective cross-area, and the yield and tensile strengths were reduced due to stress concentrations around larger voids/pores.
3.3. Differences in FFF Printers
As expected, there are a variety of differences between the Method X and Ender 3 V2. The Method X has an automated bed leveling system, with built in calibration to maintain reproduce-ability. The Ender 3 V2 has no such system, requiring the print bed to be leveled by hand. Furthermore, it is up to the user to maintain/calibrate the printer and is susceptible to changes over time. Additionally, the build chamber of the Ender 3 V2 has no heating aside from the print bed to maintain the temperature within the printers enclosure, while the Method X has a heated bed and build chamber. The Ender 3 V2 uses a bowden tube style extruder, as opposed to the direct drive extruder used on the Makerbot Method X. While base capabilities of both systems are similar, the additional features of the Method X reduce human-factor induced variability and provide a larger processability space.
It is worth noting, other than an upgraded nozzle, the Ender 3 V2 was completely stock, as it came out of the box. This was done to evaluate the performance of this FFF printer off the shelf, as it is widely utilized. There are many aftermarket components and modifications that may improve the repeatability and performance of the Ender 3 V2 when printing with Ultrafuse 316L (BASF 3D Printing Solutions GmbH, Heidelberg, Germany). Examples of modifications to mitigate environmental effects and improve print consistency could include an upgraded hot-end, a heated build chamber, and utilizing the auto-bed leveling system. Despite best efforts of setup and maintenance, the Ender does not have the repeatability of commercial grade FFF printers in its stock configuration.
3.4. Specimen Failure
All samples exhibited ductile tensile behavior. The fracture surfaces in
Figure 7 show ductile fracture, although the fracture surface is rather unique for the Ender sample in
Figure 7b. Typical necking was evident in the more dense samples, but the less dense samples exhibited more elongation of the infill “fibers” as opposed to the bulk structure. Failure of the more porous samples occurred along planes perpendicular to the infill paths producing a faceted surface (
Figure 7b) or a 45° plane (
Figure 8b). Good fusion of the contour pass to infill passes as shown in
Figure 7b led to improved mechanical response.
Several samples showed delamination of the infill from the contour paths as in
Figure 8a. The more porous samples typically failed along one of the 45° infill paths. Both
Figure 8a,b show that only a portion of the infill was carrying load, leading to a higher stress on the remaining sections and reducing the bulk strength of the component. The larger, regularly patterned pores with continuous infill lines of E221 showed poorer mechanical strength but higher ductility than E111 which had more seemingly random porosity.
3.5. Effects of Print Strategy
As previously stated, the Ender 3 V2 and Method X utilized different software for slicing. Both machines used 2 outer contours per recommendation of previous research, and different patterns for alternating infill strategy throughout the build height as previously highlighted in
Figure 2 [
9,
17]. The increase in porosity levels in the upper half of the MBDS specimen (
Figure 5d) suggests that infill porosity is dependent on printing strategy. However, the consistent lack of adhesion at the contour-infill interface throughout the build height suggests this defect pattern is extrusion parameter dependent. The differences in porosity throughout the build height between FFF printers suggest further research is needed.
3.6. L-PBF/Plate Comparison
The L-PBF produced samples resulted in the highest elastic modulus, yield strength, and hardness of all tested samples. L-PBF showed an increase of 11%, 62%, and 5% for the elastic modulus, yield strength, and tensile strength compared to the average Method X response. In previous research, a Renishaw AM400 was compared to MEX samples in the same build configuration (and powder supplier) and showed a 32% reduced elastic modulus, but a 12% increase in yield strength and a 6% increase in tensile strength [
19], highlighting the variability present in L-PBF and environmental factors. However, the 316L plate specimens featured the highest ultimate tensile strength at 631 MPa. The hardening behavior of the FFF samples follows a similar trend to that of the 316L plate samples. The produced L-PBF samples showed minimal hardening, consistent with other vertically printed L-PBF 316L components [
19,
29]. Makerbot and Ender FFF parts showed similar yield strengths with approximately 50% and 40% reduction of yield strength compared to plate and L-PBF samples, respectively.
3.7. Future Work
A more accurate specimen density could be achieved by performing micro CT as done in [
17]. This would reduce the effects of irregularities such as open pores during gas pycnometry measurements or location dependence of optical micrographs. Furthermore, micro CT could provide additional insight into the subtle differences between MBDS and MBDS1x1. While there was an insignificant difference in mechanical response of MBDS and MBDS1x1, further testing would need to be performed to fully conclude that there is no significant difference between printing samples on a per-part vs per-layer basis with regards to part density and strength.
Future research should investigate the effects of print strategy on process-induced porosity, and consider the evaluation of printing parameters not within the scope of this work (i.e., layer rotation angle, number of perimeters, perimeter overlap, and further investigation into extrusion parameters). Additionally, a higher sample size of OM density measurements taken from different regions and cross-sectional planes of the sample may help to elucidate some of the unique porosity phenomena occurring in these samples.
4. Conclusions
This research evaluated the use of two FFF printers, the Makerbot Method X and the Creality Ender 3 V2, to print tensile specimens using Ultrafuse 316L filament. Layer height, line width, and extrusion multiplier were varied for each printer, and the resulting tensile response, microhardness, and density were compared. The resulting mechanical properties were comparable to those published in literature for components printed with Ultrafuse 316L filament. The Method X produced 99% dense 316L specimens using the default print settings, and all successful print parameter combinations gave comparable or improved mechanical results compared to previous research. The Ender samples with an increased extrusion multiplier from the default parameters showed comparable strength to the Makerbot produced samples and literature data, but had a reduced OM density of approximately 95%. Porosity reduced the mechanical strength of the Ender samples, shown by process parameter combinations with an extrusion multiplier of 100%.
Additionally, the aforementioned FFF parameter combinations produced comparable elastic modulus, and slightly reduced tensile strengths compared to the L-PBF samples. Both L-PBF and FFF samples showed reduced tensile strengths compared to the plate material. FFF produced samples showed a reduction in yield strength, approximately 50% of 316L plate and 40% of L-PBF produced samples for both the Makerbot and Ender produced parts. This work supports the claim that FFF metal printing is a viable, low-cost option for lower strength requirement metal components. Comparing between FFF printers, the optimal settings for the Ender 3 V2 showed only a 7–8% reduction in yield and tensile strength compared to the average Method X sample. This is substantial, considering the Ender 3 v2 costs 95% less than the Method X. Accordingly, this work demonstrated that users are not limited to printing this material on higher-end FFF printers, but with the right settings, metal parts can be produced using a low-cost, hobby-grade printer with minimal modifications or up-front costs.