Tensile Properties of Robotic 3D-Printed Parts with Varying Contour Layer Thickness

Abstract

Additive manufacturing (AM) technology is rapidly advancing across diverse fields. For instance, the use of robotic arms in various AM processes has led to significant gains in printing flexibility and manufacturing scalability. However, despite these advancements, there remains a notable research gap concerning the mechanical properties of parts 3D-printed with robotic arms. This study focuses on developing a robotic fused filament fabrication (FFF) 3D-printing process with a layer resolution of $50$ to $200 μ m$ . The impact of the robotic printing process on the mechanical properties of printed parts is investigated and benchmarked against a commercial FFF 3D printer. In addition, we propose a novel tool path that can vary contour layer thickness within an infill layer to improve mechanical strength by minimizing air gaps between contours. SEM images suggest this new tool path strategy leads to a meaningful reduction in void area fraction within contours, confirmed by a nearly $6 %$ increase in ultimate tensile strength.

Keywords

fused filament fabrication (FFF)robotic 3D printing mechanical properties polylactic acid (PLA)layer thickness

Introduction

The field of additive manufacturing (AM) has witnessed exceptional growth and countless technological innovations in recent years, fueled in part by progress in materials science, polymer chemistry, process development, and mechanical design.¹ AM technologies can be broadly classified into seven types: vat photopolymerization, material extrusion, material jetting, powder bed fusion, binder jetting, sheet lamination, and direct energy deposition.^2–4 Although these seven classes of AM are diverse in their underlying methodology, a common feature is the sequential layer-by-layer fabrication of a three-dimensional part from a digital model.⁵ The utilization of planar layers in additive manufacturing enables the realization of complex geometries that are not achievable through conventional manufacturing techniques such as casting, forming, molding, and subtractive manufacturing.⁶

Fused filament fabrication (FFF) is a popular material extrusion-based AM process for thermoplastic polymers. In FFF, polymer filament is extruded through a heated nozzle and systematically deposited to build a three-dimensional part layer-by-layer. FFF has seen widespread adoption because of its extensive selection of thermoplastic materials and low cost. Despite their many advantages, FFF processes have some challenges.^7,8 For instance, the uniform slicing of curvilinear geometries results in the “staircase effect,” that is., unwanted surface roughness that is highly dependent on the layer thickness.⁹ To minimize the staircase effect, adaptive slicing algorithms that enforce cusp height requirements have been introduced.^10–12 The surface roughness can also be improved by adopting non-uniform layers along with uniform layers.^13,14

In many cases, 3D printers utilize a gantry mechanism only capable of rectilinear motion, and support structures are necessary for parts with overhanging features. These supports contribute to increased printing and post-processing times and also contribute to degradation in surface quality.¹⁵ New mechanisms have been adopted to minimize support structures and improve fabrication time, such as five-axis printing by adding two degrees of freedom to the print bed,^16–18 and six-degree-of-freedom (6-DOF) robotic arms.^19,20 The flexibility of industrial robotic arms is exhibited, for instance, by hybrid manufacturing processes that simultaneously encompass additive, subtractive, and formative fabrication.^21,22 Furthermore, the utilization of one or more 6-DOF robotic arms enables large-scale printing with a smaller footprint.²³ Robotic arms are renowned for their exceptional precision and repeatability. Thus, the utilization of these robotic systems can greatly enhance additive manufacturing.

The mechanical properties of FFF 3D-printed parts differ from those manufactured using other processes. For instance, 3D-printed parts generally exhibit anisotropic mechanical properties as a result of layer-based manufacturing and directional infill.²⁴ The inter-layer bonding of 3D-printed materials is primarily influenced by layer thickness. Smaller layer thicknesses not only enhance interlayer bonding but also improve the strength of 3D-printed parts.^25–27 In addition to the layer thickness, infill pattern, and percentage, they also have a strong influence on the mechanical properties of FFF 3D-printed materials.^28,29 However, choosing the lowest layer thickness and 100% infill will result in a significant, and perhaps practically prohibitive, increase in the overall build time.

Continuously varied infill pattern (ConVIP) was introduced by Kim et al.³⁰ to enhance mechanical properties and reduce printing time. The authors demonstrated that with a $30 %$ rectilinear ConVIP infill, there was a $15.2 %$ increase in strength, accompanied by an approximately $9 %$ reduction in print time, compared to the conventional infill pattern. A study by Le et al.³¹ focused on optimizing process parameters to achieve a faster print time while minimizing the impact on mechanical properties. Utilizing a design of experiments, the authors found this can be achieved by a large nozzle diameter, increasing the number of outer shells, and reducing the infill percentage. Moreover, minimizing the air gaps between filament paths led to improved strength because of enhanced bonding between the material layers and paths.²⁴

In this article, we develop a robotic FFF 3D-printing process that utilizes a Yaskawa Motoman robotic arm, enabling printing at a layer resolution ranging from $50$ to $200 μ m$ . As with any AM technology, print time, mechanical properties, and surface quality are critical factors. Parameters like layer thickness, air gaps, and infill percentage significantly influence the mechanical properties of FFF 3D-printed parts.³² Thus, we propose a novel tool path that can vary contour layer thickness within an infill layer. This approach aims to minimize air gaps within contours and improve strength by ensuring fine layer resolution on outer surfaces. To assess the effectiveness of the proposed method, statistical analysis is employed to compare the mechanical properties of parts printed with a robot using both uniform contour layer thickness (UT contours) and varying contour layer thickness (VT contours). Furthermore, the mechanical properties of robot-printed parts with UT contours are benchmarked against a commercial printer. Overall, the study presents some new insights into the mechanical properties of robotic 3D-printed parts and explores the potential benefits of utilizing varying contour layer thickness in the FFF printing process.

Material and Methods

System design and printing parameters

The robotic printing configuration consists of two subsystems. The first is the extrusion system, comprised of an MK8 hot end, a filament feeder equipped with a closed-loop stepper motor, and a stationary print bed platform. The second subsystem consists of a Yaskawa Motoman GP7 manipulator with a YRC1000 controller.³³ The extrusion mechanism is affixed to the robot’s end-effector (Fig. 1). An inductive proximity sensor was used to calibrate the extruder’s nozzle tip with the print bed. A Prusa i3 MK3S+, a commercial 3D printer, was used for comparing mechanical properties.³⁴ All parts were 3D-printed with $1.75 mm$ diameter PLA filament (Merlot Red PRO Series, MatterHackers).³⁵ The printing parameters are listed in Table 1. PrusaSlicer software was used for both the Prusa i3 and robot 3D printing.

FIG. 1.

Robotic 3D printing station and parts printed: (A) robot printing setup; (B) hexagon planter CAD model ( $45 \times 50 \times 35 mm$ ); (C) 3D-printed hexagon planter; (D) rocket nozzle and fins CAD model ( $75 \times 75 \times 40 mm$ ); (E) 3D-printed rocket nozzle and fins.

Table 1.

Printing Parameters

Parameters	Values
Nozzle Diameter	0.4 mm
Layer Thickness	0.2 mm
Extrusion Width	0.45 mm
Nozzle Temperature	210°C
Print Bed Temperature	60°C
Max Print Speed	40 mm/s
Contours	3
Infill Pattern	Rectilinear
Infill Angle	45°
Infill Percentage	100%

Robotic printing process

Most FFF printers are equipped with at least four stepper motors. Based on the size of the print area, the X, Y, and Z movements can be achieved by one stepper motor for each axis and one stepper motor to feed the solid filament into the hot end of the extruder. A microcontroller runs 3D printing firmware to control the stepper motors synchronously, with a speed lookup table for uniform polymer extrusion. In contrast, robotic 3D printing consists of two sub-systems that must work synchronously: the robotic arm for rectilinear motion and the extrusion system for filament deposition. Since these two sub-systems have their own independent controllers, achieving consistent extrusion and deposition of the material, and thus high-resolution prints, is challenging. To address this, we have developed two robotic printing architectures that are discussed in what follows.

MATLAB-based Printing Process (MPP): For robotic 3D printing, the G-code commands are converted to robot commands using interfacing software or a G-code Parser. This is the most common architecture used for robot printing.³⁶ So, we developed a process with a similar architecture using MATLAB and MotoCom as interfacing software to communicate tool motion commands with the robot’s controller and printing commands with the microcontroller (Fig. 2a). Open-source Marlin firmware controls the extrusion system. To assess surface quality, an ASTM D638 Type IV specimen was printed (Fig. 2c).

FIG. 2.

Robotic printing process and printed parts: (A) MATLAB-based printing process (MPP); (B) controller-based printing process (CPP); (C) tensile specimen ( $115 \times 19 \times 3.2 mm$ ) printed with MPP; (D) tensile specimen ( $115 \times 19 \times 3.2 mm)$ printed with CPP.

Controller-based Printing Process (CPP): Continuous motion of the robot can be achieved by using JBI files (i.e., job files), which are created specifically for Yaskawa Motoman robots and contain instructions for robotic motion and actuation. The advantage of these files is that they can be executed directly from the controller and can also be used to validate the printing in MotoSim simulation software. So, we developed a two-step process (Fig. 2b). In the first step, we use MATLAB to read the G-code file and generate multiple job files based on the number of printing points. These multiple job files can be executed from the master job file. In the second step, generated job files are loaded onto the controller’s memory for printing. An ASTM D638 Type IV specimen was 3D printed using the CPP architecture (Fig. 2d).

The extruder feed rate (EFR) is calculated from the volumetric flow rate (VFR), which depends on the parameters set in the slicer, such as layer height (LH), extrusion width (EW), print speed (PS), and filament cross-sectional area (FCA). The VFR for a specific print speed is calculated using:

\begin{array}{l} VFR = LH × EW × PS \\ EFR = VFR/FCA \end{array}

(1)

The EFR was calibrated by printing a single contour $20 \times 20 \times 5 mm$ cuboid geometry with a spiral print path (Fig. 3). For these prints, a layer height of $0.2 mm$ extrusion width of $0.45 mm$ , and a print speed of $20 mm / s$ were specified in the slicer software, leading to an extruder feed rate (EFR) of $0.748 mm / s$ (recall that the filament diameter is $1.75 mm$ ). As-printed extrusion width was then measured using digital calipers and found to be $0.54 mm$ , which is $20 %$ more than the targeted extrusion width of $0.45 mm$ . To compensate, the EFR was subsequently scaled down by $20 %,$ which translates to a corresponding reduction in steps per second for the stepper motor. This process was repeated until the measured extrusion width was within an acceptable tolerance $(\pm 0.01 mm$ ) of $0.45 mm$ . This calibration process was executed for different layer heights and print speeds with the same extrusion width.

FIG. 3.

Extruder feed rate calibration: (A) flowchart describing calibration process; (B) spiral print samples with different layer heights ( $0.2 mm and 0.05 mm$ ).

The extrusion flow is controlled directly by the robot controller via digital input–output signals. G-code instructions are categorized into six commands: retract commands, un-retract commands, print commands, travel commands, z-lift commands, and speed change commands. The respective digital input-output signals are turned on based on the command type. Therefore, job files comprise both robot instructions and digital input-output instructions for the extruder feed rate control.

Modified tool path planning

When a polymer emerges from the extruder, it is cylindrical in shape because of the circular nozzle orifice. But when the polymer is deposited on a surface, its cross section is elliptical. A part printed with uniform-thickness contours and a low print resolution (i.e., coarse layer height) can result in air gaps or voids between the elliptical contours (Fig. 4). Because of the air gaps, printed parts are more prone to premature failure when they are loaded mechanically. In contrast, parts printed with finer layer heights tend to minimize voids and improve the strength of the parts, but in general, the overall print time is significantly increased. Therefore, this strategy may be practically infeasible for large-scale prints. The goal of the variable contour layer thickness is to minimize the air gaps between contours to improve strength and surface finish, with only modest increases in print time.

FIG. 4.

Cross-sectional view of a part with three contours and infill: (A) uniform contour layer thickness tool path; (B) variable contour layer thickness tool path.

Three contours were considered for both UT contour and VT contour printing (Fig. 4). When printing with UT contours, the external contour can be printed before the internal contour or vice versa because subsequent material deposition does not disrupt the previously deposited contour (Fig. 4a). However, in the case of the VT contours, a specific tool path is needed to avoid disturbing the previously deposited contours. Upper and lower bounds for the contour layer thickness are based on the volumetric limit of the $400 μ m$ diameter nozzle (upper bound) and the robot repeatability of $\pm 10 μ m$ (lower bound), respectively. The external contour was set to $50 μ m$ layer thickness to minimize the staircase effect for sloped surfaces and ensure smooth outer surfaces. The middle contour was set to 100 $μ m$ , twice the layer thickness of the external contour. Similarly, the inner contour layer thickness and infill were set to $200 μ m$ . In the case of printing with three contours, the internal contour layer height was four times the external contour layer height while keeping the extrusion width the same for all contours. Depending on the volumetric limit and print speed, different layer heights can be considered for faster printing.

For proper material deposition of VT contours, starting from the external contour, a specific order must be followed for each layer (Fig. 4b). The proposed tool path is accomplished using the G-code generated by PrusaSlicer. The pseudo-code representing this process is depicted in Algorithm 1.

Algorithm 1 Varying Contour Layer Thickness Require: G-code file [sliced with a layer thickness (lh) of 200 μm, three contours, and 100% infill] 1: procedure VCLT(Gcode, lh) 2: if n == 1 then ▷ Number of layers ‘n’ 3: MGcode ← x, y, z 4: else 5: for i ← 2, n do 6: switch f do ▷ 4 features 7: case 1 ▷ 1st contour, m points 8: δz = (lh/4) × [3 2 1 0] 9: for j ← 1,4 do 10: for k ← 1, m do 11: P_j(k) = x, y, z − δz(j) 12: end for 13: end for 14: case 2 ▷ 2^nd contour, m points 15: δv = (lh/2) × [1 0] 16: for l ← 1,2 do 17: for k ← 1, m do 18: Q_j(k) = x, y, z − δv(l) 19: end for 20: end for 21: case 3 ▷ 3^rd contour, m points 22: for k ← 1, m do 23: R(k) = x, y, z 24: end for 25: case 4 ▷ infill, n points 26: for k ← 1, n do 27: S(k) = x, y, z 28: end for 29: Add the xyz datasets in the order P₁, P₂, Q₁, P₃, P₄, Q₂, R, S to MGcode 30: end for 31: end if 32: end procedure

Tensile testing and mechanical properties

In this section, we investigate the tensile properties (ultimate tensile strength, Young’s modulus, and elongation at break) of robot-printed materials with UT and VT contours, as well as commercially printed materials with UT contours. Quasi-static uniaxial tension testing was performed on an Instron 3365 universal testing machine according to ASTM D638.³⁷ Type I specimens with a nominal thickness of $2.8 mm$ were used in this study (Fig. 5a). Specimens were pneumatically gripped and tested under displacement control, with a constant crosshead speed of $\dot{u} = 5 mm / \min$ (nominal strain rate of $\dot{ϵ} = \dot{u} / L 0 \approx 0.0007 / s,$ where $L 0 = 115 mm$ is the initial grip-to-grip distance). The axial force was measured using a $5 KN$ Instron load cell. Axial displacement of the crosshead was measured using a built-in Instron LVDT.

FIG. 5.

Specimen dimensions and engineering stress–strain curves: (A) ASTM D638 Type I specimen dimensions ( $165 \times 19 \times 2.8 mm$ ); (B) Prusa with UT contours; (C) robot with UT contours; (D) robot with VT contours. UT contour, uniform contour layer thickness; VT contour, varying contour layer thickness.

FIG. 5.

Continued.

Ten specimens were 3D printed for each contour type to ensure experimental repeatability and statistically meaningful sample sizes. As-printed dimensions of each specimen were measured using digital calipers. Engineering stress–strain curves were generated for each specimen tested, with engineering stress being the measured axial force divided by the undeformed cross-sectional area, and engineering strain being the measured axial displacement divided by the undeformed grip-to-grip distance. A one-way analysis of variance (ANOVA) was performed to determine the statistical significance of the differences in the sample means for each of the three properties. Tukey–Kramer honest significant difference (HSD) analysis was also employed to determine if specific pairs of properties differed significantly.³⁸

The microstructures of the fractured specimen cross-sectional surfaces were examined using a Hitachi TM3000 Tabletop Scanning Electron Microscope (SEM). Prior to the SEM analysis, small samples were extracted at the specimen fracture surface and then sputter coated (Desk II, Denton Vacuum) with a gold-palladium alloy to ensure an electrically conductive surface for effective SEM imaging. The SEM images were processed using ImageJ and MATLAB.

Results and Discussion

Printing process comparison

Specimens printed with the MPP process exhibited over-extrusion at sharp corners as well as rough, curved surfaces. This is attributed to the discontinuous motion because of the MATLAB processing time delay for each instruction. Under-extruded regions highlighted with yellow boxes (Fig. 2c) are attributed to a mismatch between the trapezoidal speed profiles of the extrusion system and the robot. Hence, the process architecture influences both print time and surface quality. Therefore, as discussed in Section Robotic printing process, we developed a CPP printing process to execute instructions in real-time and with proper extrusion feed rate control. The CPP process printed an ASTM D638 Type I specimen in $54 mins$ , $41.3 %$ faster than the MPP process ( $92 mins$ ). The CPP process also yielded prints with smoother contour surfaces (Fig. 2d). In both cases, the actual print time exceeded the slicer’s estimated print time, which we attribute to the execution of supplementary input-output instructions.

Mechanical properties

A total of $30$ samples were printed and tested, with a Prusa i3 (UT contours) and a Yaskawa robot with CPP architecture (UT and VT contours). Each printing method contributes $10$ samples. For the sake of simplicity, the methods of printing with Prusa UT contours, robot UT contours, and robot VT contours are hereafter referred to as groups A, B, and C, respectively.

The key features of the engineering stress–strain curves (Fig. 5) are typical for a thermoplastic polymer: linear elastic deformation (until about $2 %$ strain), followed by yield near the proportional limit (linear-nonlinear transition), plastic deformation, and ultimately fracture. A notable and repeatable non-linear signature is observed near $0.5 %$ strain in the stress–strain traces, perhaps attributable to nozzle temperature and/or an inter-layer mesoscale deformation mechanism.^39–41 Therefore, we calculated Young’s modulus before and after $0.5 %$ strain, named early modulus and late modulus, respectively. The early modulus was calculated as the average slope of three strain values between $0.2 and 0.25 %,$ and the late modulus was the average slope of three strain values between $1 and 1.55 %$ . Means and standard deviations of the calculated mechanical properties (Young’s modulus, tensile strength, and elongation at fracture) were compiled (Table 2) and illustrated graphically in box plots (Fig. 6).

FIG. 6.

Box plots showing means and standard deviations of (A) ultimate tensile strength; (B) percent elongation; (C) early modulus; (D) late modulus.

Table 2.

Mean and Standard Deviation of Tensile Properties

Properties	A	B	C
Properties	Mean	Standard deviation	Mean	Standard deviation	Mean	Standard deviation
UTS (MPa)	50.4	0.97	50.8	2.29	53.7	1.49
Early modulus (GPa)	2.09	0.18	1.96	0.24	2.06	0.10
Late modulus (GPa)	2.01	0.11	2.03	0.09	2.04	0.08
Elongation (%)	5.04	0.24	3.7	0.19	3.97	0.26

ANOVA for the ultimate tensile strengths: The analysis revealed a significant difference between the means of the groups (Table 3a). Additional insight into the specific group means (Table 3b) was obtained through post-hoc pairwise comparison. When comparing the tensile strength of groups A and B, it was determined that the difference between the groups is not statistically significant. This suggests that the tensile strength of the parts printed with uniform contour layer thickness on the robot is equivalent to that of the parts printed on the Prusa. However, group C exhibited a tensile strength that was significantly different from that of groups A and B, with mean differences of $3.25 MPa$ and $2.85 MPa$ , respectively. This suggests parts printed with variable contour layer thickness can be expected to exhibit higher ultimate tensile strength in comparison to those with uniform contour layer thickness.

Table 3.

Statistical Comparison of Tensile Strength Means ( $MPa$ )

Source	SS	df	MS	F	p value
(a) ANOVA for tensile strength.
Columns	63.05	2	31.52	11.15	0.0002
Error	76.31	27	2.82
Total	139.36	29
Group 1	Group 2	Lower Limit	Difference	Upper Limit	p value
(b) Tukey–Kramer pairwise comparison of tensile strength means.
A	B	−2.26	−0.40	1.45	0.8533
A	C	−5.12	−3.25	−1.39	0.0005
B	C	−4.71	−2.85	−0.98	0.0021

ANOVA, SS, df, MS, F.

ANOVA for Young’s modulus: Independent analyses were conducted for both early modulus and late modulus (Table 4). The ANOVA test indicates that there is no significant difference in early modulus means between groups A, B, and C. This is supported by a p value of $0.6$ , exceeding the standard significance level of $α = 0.05$ . Also, pairwise comparisons using the post-hoc test show no significant difference between group means, supporting ANOVA results. Similarly, ANOVA findings for the late modulus suggest no significant difference in averages among the groups, with a p value of $0.76$ exceeding the level of significance. The post-hoc test confirms these findings, with all pairs having p-values much greater than $α = 0.05$ and no rejection of the null hypothesis of equal group means. Therefore, the analysis suggests that the printing method does not have a significant impact on the material’s elastic modulus.

Table 4.

Statistical Comparison of Young’s Modulus ( $GPa$ )

Source	SS	df	MS	F	p value
(a) ANOVA for early Young’s modulus.
Columns	0.10	2	0.05	1.33	0.2805
Error	0.96	27	0.04	—	—
Total	1.06	29	—	—	—
Group 1	Group 2	Lower Limit	Difference	Upper Limit	p value
(b) Tukey–Kramer pairwise comparison of early Young’s modulus means.
A	B	−0.08	0.13	0.34	0.2922
A	C	−0.19	0.02	0.23	0.9585
B	C	−0.32	−0.11	0.10	0.4323
Source	SS	df	MS	F	p value
(c) ANOVA for late Young’s modulus.
Columns	0.01	2	0.002	0.28	0.76
Error	0.26	27	0.01	—	—
Total	0.26	29	—	—	—
Group 1	Group 2	Lower Limit	Difference	Upper Limit	p value
(d) Tukey–Kramer pairwise comparison of late Young’s modulus means.
A	B	−0.13	−0.02	0.09	0.9013
A	C	−0.14	−0.03	0.08	0.743
B	C	−0.12	−0.01	0.10	0.9505

ANOVA for percent elongation at break: The outcome shows a p value of $1.47 \times 10^{- 12}$ , a value that is significantly below the standard level of significance (Table 5). This implies that there exists a statistically significant difference in the means of groups A, B, and C. The pairwise comparisons of groups A and B, and groups A and C, yielded a significantly lower p value than groups B and C (p value of $0.04$ , just below the standard level of significance of $0.05$ ). This outcome suggests that although percent elongation is significantly different among the three printing methods, this significance is weaker between the robot-printed parts. Examination of the engineering stress–strain curves indicates that differences in percent elongation can be attributed to the amount of plastic deformation each printing method exhibits post-yield (Fig. 5). For instance, the Prusa-printed materials exhibit notable inelastic deformation $(> 2 %)$ before fracture relative to their robot-printed counterparts: $(< 1 %)$ for UT contours and ( $< 1.5 %$ ) for VT contours. We suspect that these differences are not because of void area fraction, but rather differing thermal histories in the Prusa-printed versus robotic-printed parts, driven by hardware differences that influence polymer liquefication and, ultimately, the mechanical properties of the solidified polymer. Similar differences in percent elongation have been observed in previous investigations and were also attributed to differences in printer hardware.⁴²

Table 5.

Statistical Comparison of Percent Elongation Means

Source	SS	df	MS	F	p value
(a) ANOVA for percent elongation.
Columns	9.98	2	4.99	88.10	1.47 × 10⁻¹²
Error	1.53	27	0.06	—	—
Total	11.51	29	—	—	—
Group 1	Group 2	Lower Limit	Difference	Upper Limit	p value
(b) Tukey–Kramer pairwise comparison of percent elongation means.
A	B	1.07	1.34	1.60	9.64 × 10⁻¹¹
A	C	0.80	1.07	1.33	1.36 × 10⁻⁰⁹
B	C	−0.53	−0.27	−0.01	0.04

SEM images of specimen fracture surfaces

Analysis of the tensile test data revealed significant differences in ultimate tensile strength among the printing methods. To further analyze these results, representative specimens produced by each printing method were considered for SEM imaging. A montage comprising six images focusing just on the contours and contour-infill interface is shown in Figure 7. Top-row images are cross-sectional views of contours on one half, and bottom-row images are the opposite half of the same fractured specimen. The ImageJ tool is used to identify and differentiate voids that form between solidified filament paths within a layer and adjacent layers. Gaps between the contours are highlighted in red, whereas the gaps between the infill and contours are marked in yellow.

FIG. 7.

SEM images focused on the contours (voids between contours are marked with red, and voids between contour and infill are marked with yellow): (A, B) Prusa with UT contours; (C, D) robot with UT contours; (E, F) robot with VT contours. SEM, scanning electron microscope; UT contour, uniform contour layer thickness; VT contour, varying contour layer thickness.

In the case of uniform contour layer thickness (UT) printing with the Prusa and the robot (Fig. 7A–D), consistent triangular voids between the contours are observed. These voids result from the elliptical shape of the deposited filaments, leading to imperfect inter-filament contact (both within and between layers). The approximate void area within the contours is $0.118 {mm}^{2}$ for the Prusa and ${0.244 mm}^{2}$ for the robot. Additionally, the measured void areas between the infill and contours are ${0.284 mm}^{2}$ for the Prusa and ${0.175 mm}^{2}$ for the robot. Thus, the difference in total void area between the Prusa and the robot (with UT contours) is not significant (the difference is less than ${0.02 mm}^{2}$ ). This is consistent with our finding that differences in the tensile strength of Prusa-printed and robot-printed parts (with UT contours) are statistically insignificant.

Comparing the robot-printed VT contours to the robot-printed UT contours (Fig. 7C-7F), it was observed that the overall void area and ultimate tensile strength for the VT contours were ${0.066 mm}^{2}$ and $51.14 MPa$ , respectively, whereas for the UT contours they were ${0.419 mm}^{2}$ and $47.3 MPa$ , respectively. These results suggest an inverse correlation between the overall void area of the contour and the strength of the printed parts. Air gaps between the deposited paths and voids along the inter-layer bonds (both within the contours and at the contour-infill interface) were the two primary contributions to the overall void area. Notably, inter-layer voids are more prominent in robot-printed UT contours (Fig. 8B and C), leading to higher standard deviations in tensile strength compared to VT contours (Table 2). The presence of voids both within and between the infill layer (Fig. 8), which are a result of the ${\pm 45}^{°}$ zig-zag infill pattern, likely impacts the mechanical properties of the printed parts.

FIG. 8.

SEM images focused on the contours and infill (weak inter-layer bonding is marked with red, and rectilinear infill pattern voids are marked with yellow): (A) Prusa with UT contours; (B) robot with UT contours; (C) robot with VT contours. SEM, scanning electron microscope; UT contour, uniform contour layer thickness; VT contour, varying contour layer thickness.

Conclusion

This study investigated the impact of the robotic FFF 3D-printing process on mechanical properties, using a commercial FFF 3D printer as a benchmark. Additionally, a novel tool path plan (coined VT contours) is introduced to minimize voids within the contours and enhance tensile strength. The key findings of this research are summarized as follows:

Parts produced using a MPP exhibited instances of under-extrusion and over-extrusion of the polymer. So, a controller-based printing process was developed for more uniform filament extrusion. This resulted in improved surface quality and a reduction in overall print time by more than 40%.

Young’s modulus is insensitive to the printing method, with no statistically significant differences observed between VT and UT contours, nor Prusa prints versus robot prints.

Ultimate tensile strength is $5.7 %$ higher in robot prints with VT contours than robot prints with UT contours, despite only a $25 %$ increase in print time ( $67.5 \min$ for VT vs. $54 \min$ for UT with $200 μ m$ layer thickness). These increases in strength are supported by a meaningful $($ approximately $84.2 %)$ reduction in the void area between contours confirmed through SEM imaging. Interestingly, no statistically significant difference in tensile strength was observed between the Prusa and robot prints with UT contours.

The VT contour approach offers an effective alternative to a uniform reduction of layer height to increase strength. For example, reducing UT contour layer resolution from $200 μ m$ to $50 μ m$ leads to an estimated print time of $80.5 \min$ an approximately $20 %$ increase over the VT strategy.

Percent elongation at break differs among the three printing methods, arising from differences in post-yield plastic deformation before fracture. Although pairwise comparisons reveal statistically significant differences in the means of all three printing methods, the significance between the UT and VT robot prints is relatively small. Consequently, we suspect the observed differences in post-yield response are likely not caused by void area fraction, but rather by hardware differences between the Prusa and robotic 3D printers.

Authors’ Contributions

A.V.: Conceptualization, software, formal analysis, investigation, visualization, writing—original draft, writing—review and editing. T.K.: Software, supervision, writing—review and editing. R.L.: Methodology, formal analysis, validation, resources, writing—review and editing. A.N.-D.: Resources, visualization, funding acquisition, writing—review and editing. R.O.: Conceptualization, formal analysis, project administration, resources, funding acquisition, supervision, writing—review and editing. All authors have read and approved the final article.

Footnotes

Acknowledgments

The authors would like to thank Mr. Wade Hickle of Yaskawa Motoman Robotics, Inc. for providing the robots used in this study and robotic software engineer Mr. Tim Luneke for guidance in programming the controller. This work has been supported in part by the University of Dayton Office for Graduate Academic Affairs through the Graduate Student Summer Fellowship Program.

Author Disclosure Statement

The authors declared no potential conflicts of interest with respect to the research of this article.

Funding Information

The authors wish to highlight the financial contributions made by Dr. Raúl Ordóñez and Dr. Amy T. Neidhard-Doll, who are also co-authors of this study. Their support was integral to this research.

References

Ngo

, Kashani

, Imbalzano

, et al. Additive manufacturing (3D print- ing): A review of materials, methods, applications and challenges. Compos B Eng, 2018; 143:172–196; doi: 10.1016/j.compositesb.2018.02.012

Campbell

, Bourell

, Gibson

. Additive manufacturing: Rapid prototyping comes of age. Rapid Prototyp J, 2012; 18(4):255–258; doi: 10.1108/13552541211231563

Alexander

, Wake

, Chepelev

, et al. A guideline for 3D printing terminology in biomedical research utilizing ISO/ASTM standards. 3D Print Med, 2021; 7(1):8; doi: 10.1186/s41205-021-00098-5

F42 Committee. ISO/ASTM52900 additive manufacturing—general principles - fundamentals and vocabulary. ASTM International: West Conshohocken, PA; 2021. Available from: https://www.astm.org/f3177-21.html

Chua

, Leong

. 3d Printing and Additive Manufacturing: Principles and Ap- plications—Fifth Edition of Rapid Prototyping. 5th ed. World Scientific Publishing: Singapore; 2017.

Pereira

, Kennedy

, Potgieter

. A comparison of traditional manufacturing vs additive manufacturing, the best method for the job. Procedia Manuf, 2019; 30:11–18; doi: 10.1016/j.promfg.2019.02.003

Oropallo

, Piegl

. Ten challenges in 3D printing. Eng Comput, 2016; 32(1):135–148; doi: 10.1007/s00366-015-0407-0

Leung

Y-S

, Kwok

T-H

, Li

, et al. Challenges and status on design and compu- tation for emerging additive manufacturing technologies. J Comput Inf Sci Eng, 2019; 19(2):21013; doi: 10.1115/1.4041913

Anitha

, Arunachalam

, Radhakrishnan

. Critical parameters influencing the quality of prototypes in fused deposition modelling. J Mater Process Technol, 2001; 118(1–3):385–388; doi: 10.1016/s0924-0136(01)00980-3

10.

Dolenc

, Mäkelä

. Slicing procedures for layered manufacturing techniques. Com- Put Aided Des, 1994; 26(2):119–126; doi: 10.1016/0010-4485(94)90032-9

11.

Sabourin

, Houser

, Helge Bøhn

. Adaptive slicing using stepwise uniform refinement. Rapid Prototyp J, 1996; 2(4):20–26; doi: 10.1108/13552549610153370

12.

Tyberg

, Helge Bøhn

. Local adaptive slicing. Rapid Prototyp J, 1998; 4(3):118–127; doi: 10.1108/13552549810222993

13.

Huang

, Singamneni

. Curved layer fused deposition modeling with varying raster orientations. AMM, 2013; 446-447:263–269; doi: 10.4028/www.scientific.net/amm.446-447.263

14.

Allen

RJA

, Trask

. An experimental demonstration of effective curved layer fused filament fabrication utilising a parallel deposition robot. Addit Manuf, 2015; 8:78–87; doi: 10.1016/j.addma.2015.09.001

15.

Jiang

, Lou

, Hu

. Effect of support on printed properties in fused de- position modelling processes. Virtual Phys Prototyp, 2019; 14(4):308–315; doi: 10.1080/17452759.2019.1568835

16.

Grutle

ØK

. 5-Axis 3D Printer. Master’s Thesis. University of Oslo; 2015.

17.

Luo

, Hsu

, Hsiao

, et al. 3D digital manufacturing via synchronous 5-axes printing for strengthening printing parts. IEEE Access, 2020; 8:126083–126091; doi: 10.1109/access.2020.3007772

18.

William

, Barnwell

. Strengthening ABS, nylon, and polyester 3D printed parts by stress tensor aligned deposition paths and five-axis printing. Solid Freeform Fab- rication Symposium – An Additive Manufacturing Conference; 2016.

19.

, Dai

, Fang

, et al. RoboFDM: A robotic system for support-free fab- rication using FDM. In: 2017 IEEE International Conference on Robotics and Au- Tomation (ICRA). IEEE; 2017.

20.

, Dai

, Fang

, et al. General support-effective decomposition for multi- directional 3-D printing. IEEE Trans Automat Sci Eng, 2020; 17(2):599–610; doi: 10.1109/tase.2019.2938219

21.

, Haghighi

, Yang

. A novel 6-axis hybrid additive-subtractive manufacturing process: Design and case studies. J Manuf Process, 2018; 33:150–160; doi: 10.1016/j.jmapro.2018.05.008

22.

Keating

, Oxman

. Compound fabrication: A multi-functional robotic platform for digital design and fabrication. Robot Comput Integr Manuf, 2013; 29(6):439–448; doi: 10.1016/j.rcim.2013.05.001

23.

Tiryaki

, Zhang

, Pham

Q-C

. Printing-while-moving: A new paradigm for large-scale robotic 3D printing. In: 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) IEEE; 2019.

24.

Popescu

, Zapciu

, Amza

, et al. FDM process parameters influence over the mechanical properties of polymer specimens: A review. Polym Test, 2018; 69:157–166; doi: 10.1016/j.polymertesting.2018.05.020

25.

Wang

, Ma

, Deng

, et al. Effects of fused deposition modeling process parameters on tensile, dynamic mechanical properties of 3D printed polylactic acid materials. Polym Test, 2020; 86:106483; doi: 10.1016/j.polymertesting.2020.106483

26.

Yao

, Deng

, Zhang

, et al. A method to predict the ultimate tensile strength of 3D printing polylactic acid (PLA) materials with different printing orientations. Compos B Eng, 2019; 163:393–402; doi: 10.1016/j.compositesb.2019.01.025

27.

Chaćon

, Caminero

, Garćıa-Plaza

, et al. Additive manufacturing of PLA structures using fused deposition modelling: Effect of process parameters on me- chanical properties and their optimal selection. Mater Des, 2017; 124:143–157; doi: 10.1016/j.matdes.2017.03.065

28.

Birosz

, Ledenyák

, Andó

. Effect of FDM infill patterns on mechanical properties. Polym Test, 2022; 113:107654; doi: 10.1016/j.polymertesting.2022.107654

29.

Qamar Tanveer

, Mishra

, et al. Effect of infill pattern and infill density on mechanical behaviour of FDM 3D printed Parts- a current review. Mater Today, 2022; 62:100–108; doi: 10.1016/j.matpr.2022.02.310

30.

Kim

, Andreu

, Kim

, et al. Continuously varied infill pattern (Con- VIP): Improvement of mechanical properties and printing speed of fused filament fabrication (FFF) 3D printing. J Mater Res Technol, 2022; 18:1055–1069; doi: 10.1016/j.jmrt.2022.02.133

31.

, Rabsatt

, Eisazadeh

, et al. Reducing print time while minimizing loss in mechanical properties in consumer FDM parts. Int J Lightweight Mater Manuf, 2022; 5(2):197–212; doi: 10.1016/j.ijlmm.2022.01.003

32.

Zohdi

, Yang

. Material anisotropy in additively manufactured polymers and polymer composites: A review. Polymers (Basel), 2021; 13(19):3368; doi: 10.3390/polym13193368

33.

Yaskawa Motoman GP7. n.d. Available from: https://www.motoman.com/en-us/products/robots/industrial/assembly-handling/gp-series/gp7 [Last accessed: August 3, 2023].

34.

Original Prusa I3 MK3S+ 3D Printer. n.d. Available from: https://www.prusa3d.com/product/original-prusa-i3-mk3s-3d-printer-3/ [Last accessed: August 3, 2023].

35.

Matterhackers

. Merlot red PRO series PLA filament—1.75mm (1kg). n.d. Available from: https://www.matterhackers.com/store/l/pro-series-pla/sk/MLUXZSRZ [Last accessed: November 20, 2022].

36.

Urhal

, Weightman

, Diver

, et al. Robot assisted additive manufacturing: A review. Robot Comput Integr Manuf, 2019; 59:335–345; doi: 10.1016/j.rcim.2019.05.005

37.

ASTM D638. Standard test method for tensile properties of plastics. ASTM International: West Conshohocken, PA; 2022; doi: 10.1520/d0638-22

38.

Barnes

. Statistical Analysis for Engineers and Scientists. A Computer-Based Ap- proach. McGraw Hill: New York; 1994.

39.

Wittbrodt

, Pearce

. The effects of PLA color on material prop- erties of 3–D printed components. Addit Manuf, 2015; 8:110–116; doi: 10.1016/j.addma.2015.09.006

40.

Rodŕıguez-Panes

, Claver

, Camacho

. The influence of manufacturing parameters on the mechanical behaviour of PLA and ABS pieces manufactured by FDM: A comparative analysis. Materials (Basel), 2018; 11(8); doi: 10.3390/ma11081333

41.

, Wang

, Yi

, et al. Mechanical properties and a constitutive model of 3D-printed copper powder-filled PLA material. Polymers (Basel), 2021; 13(20); doi: 10.3390/polym13203605

42.

Khabia

, Jain

. Comparison of mechanical properties of components 3D printed from different brand ABS filament on different FDM printers. Mater to- Day, 2020; 26:2907–2914; doi: 10.1016/j.matpr.2020.02.600