1 Introduction

Composites, due to their wide field of applications, have been thoroughly studied for their mechanical properties [1,2,3,4]. ABS is an engineering thermoplastic used in household equipment, electric and electronic devices [5, 6], automobile parts [7], and membranes [8]. Composites (carbon based [5, 6, 9], or others) have been developed to enhance ABS performance for specific applications. In MEX 3DP, ABS is the second most used polymer [10]. Therefore, research on its performance in AM is extensive, with the ABS polymer tested in pure form or as a matrix material in composites [11,12,13,14,15,16,17,18,19]. Its mechanical properties have been studied for tensile [20,21,22,23,24,25], flexural [22, 26, 27], creep [28], fatigue [29,30,31,32,33], dynamic loading [34], failure analysis [35], and impact tests [36,37,38,39]. The response of the ABS polymer in MEX 3D printing under different strain rates in the tensile test has also been reported [40, 41]. The effect of the 3D printing settings on the quality (surface roughness, dimensional accuracy, and porosity) of the parts built with the MEX process has been investigated, with the experimental results analyzed with statistical modeling tools [42,43,44,45,46]. The response of the ABS polymer in MEX 3D printing after successive recycling processes has been reported, showing that the material can withstand up to six repetitions of the process without serious compromise of its mechanical properties [47,48,49,50]. It has also been used in hybrid AM processes (AM with laser cutting for surface quality improvement [51], injection molding [52], shot peening [53], rotary friction welding [31], and friction stir welding for manufacturing of large parts [54]) to further expand its fields of application.

The effect of the 3DP parameters on the compression performance has been studied, focusing mainly on a limited number of parameters (one to three) [48, 55,56,57,58,59,60,61,62,63,64,65,66]. Still, research indicates the importance of compressive loading and how 3D-printed parts behave under such loading [67]. Modeling tools have been employed to analyze and optimize the studies’ results [18, 68,69,70]. Such an approach is also applied for the investigation of the effect of the process parameters on the mechanical properties of MEX-printed parts for other polymeric materials as well [71, 72], indicating that this is a common and reliable approach.

Apart from the mechanical performance, a critical aspect of 3DP is the energy consumption due to its environmental effect [73,74,75]. For the assessment of this critical parameter in 3DP, models have been presented [73, 76] based on machine learning [75], life cycle analysis [74], and statistical modeling tools [10]. A very limited number of works are focusing on the energy consumption of the ABS material when 3DP and the effect of the 3D printing parameters [10, 77]. Most works either provide a holistic approach or study a limited number of parameters.

This work introduces for the first time inclusive compressive test results for ABS 3DP, which are now missing in the literature. Additionally, it thoroughly studies the effect of 3DP parameters on the energy performance of the MEX process. Energy performance and sustainability of manufacturing processes are nowadays hot issues with industrial and social interest. ABS was chosen to be studied, as it is the second most popular thermoplastic in MEX AM [47] and it is a popular material for different types of applications as mentioned above. A thorough study, considering simultaneously such a high number (seven) of 3D printing settings, on the response in compressive loading of parts 3D printed with the ABS thermoplastic is missing from the literature. For the energy consumption in MEX 3D printing, literature is still marginal, although it affects sustainability [76], which is a key aspect nowadays for AM technology [78, 79].

Seven control parameters (ORA, RDA, ID, LT, NT, PS, and BT) are studied herein, with three levels each. All are generic parameters, i.e., they do not depend on the specific 3D printer used. The response parameters of the study are related to both the mechanical properties of the 3DP ABS parts under compression loading and the consumption of energy during the MEX process (PT, weight, EPC, SPE, SPP, sB, E, and toughness). For the analysis and the optimization of the experimental results, statistical modeling tools were used. This work provides a roadmap for the seven 3DP parameters studied. The results presented indicate which set of parameter values is better in achieving the expected outcome in each case according to the specifications set (better strength or more eco-friendly energy consumption).

2 Materials and methods

2.1 Preparation of the samples

Figure 1 shows snapshots from the sequence of the experimental processes implemented in the work. ABS was procured from INEOS Styrolution (Frankfurt, Germany). The grade was terluran hi-10,and it has the following properties: B 38 MPa, E 1900 MPa, and density 1030 kg/m3 (https://www.ineos-styrolution.com/Product/Terluran_Terluran-HI-10_SKU300600120831_lang_en_GB.html, accessed 01/12/2022). It was sourced in powder form. It was dried at 80 °C for 2 h following the manufacturer’s specification, and then it was formed into a 1.75-mm diameter filament compatible with the MEX 3DP process (3devo precision, Utrecht, the Netherlands, nozzle 220 °C, mid chamber 230 °C, hopper 240 °C, 3.5 rpm, and 55% fan speed). The filament diameter was monitored with the built-in sensor of the 3devo extruder, which operates in a closed-loop and adjusts the extruding parameters to achieve as accurate a diameter as possible during the process.

Fig. 1
figure 1

Successive experimental steps followed in the work

Full size image

After a repetition of the drying process, compression samples were 3DP with various control parameter values. The 3D-printing settings used are shown in Fig. 2, and they are determined according to the literature review and preliminary tests. Five replicas were manufactured per case. An Intamsys Funmat HT (Shanghai, China) 3D printer was used, and the required G-code for the 3D printing process was prepared on the Intamsuite (Shanghai, China) software tool. Specimens were manufactured following the ASTM D695-02a international standard (samples of 50.8 × 12.7 × 12.7 mm3, Fig. 2). Specimens were manufactured for each set of 3D printing parameters in five replicas.

Fig. 2
figure 2

3D-printing parameters used in the work and control parameter values. The geometry of the compression specimens is shown. a The material degradation graph, as acquired in the TGA, and b the graph acquired in the DSC (TGA and DSC: cycle 30 and 550 °C, step 10 °C/min)

Full size image

2.2 Energy consumption and metrics

Specimens were weighted in a high-accuracy electronic weight, while the consumed energy during the MEX 3DP process was recorded (Rigol DM3058E). The voltage (V) was considered constant during the process. It was measured at the beginning and the end of the 3D printing process to verify the statement. The electrical amplitude (A) was recorded during the 3D printing of the parts (sampling of 20 measurements per second), along with the elapsed time (stopwatch method [80]). From these measurements, the consumed energy was calculated in kWh and then converted to MJ for the determination of the EPC. The consumption of the energy is resolved in three (3) distinct stages: \((a)\) 3D printer start-up, \((b)\) 3D printing process, and \((c)\) 3D printer shutdown, and can be yielded by:

$${E}_{\mathrm{total}}={E}_{\mathrm{thermal}}+{E}_{\mathrm{motion}}+{E}_{\mathrm{auxiliary}}$$
(1)

where

$${E}_{\mathrm{thermal}}={E}_{\mathrm{heating}}+{E}_{\mathrm{cooling}}$$
(2)

\({E}_{\mathrm{motion}}\) is the absorbed energy by the motors of the machine, whereas:

$${E}_{\mathrm{auxiliary}}={E}_{\mathrm{startup}}+{E}_{\mathrm{steadystate}}+{E}_{\mathrm{shutdown}}$$
(3)

Is the energy consumed by the 3D printer’s electronics and other parts.

EPC, as it was recorded with the device, is the energy consumed from the beginning to the end of the MEX 3DP process. The energy required for the machine startup and shutdown is constant and independent of the material used for 3DP. The SPE index is calculated from the following equation (indicating the energy consumed per mass produced):

$$SPE=\frac{EPC}{w}\left[MJ/g\right]$$
(4)

where EPC represents the energy used by the 3D printer (\({E}_{\mathrm{total}}\)), and w is the actual weight of each specimen.

The SPP index is calculated from the following equation (indicating the required power per mass produced):

$$SPP=\frac{EPC}{PT\times w}\bullet {10}^{3}\left[\mathrm{kW}/\mathrm{g}\right]$$
(5)

where PT is the actual printing time for each experimental run.

To ensure that the temperatures used during the filament extrusion and the 3DP process were not affecting the thermal stability of the samples, TGA measurements were taken on the produced material, verifying this hypothesis (Fig. 2). DSC measurements were also taken to verify that the temperatures used in the work are not in areas in which the crystallinity of the material changes (Fig. 2).

2.3 Compression tests

Samples were tested for their compressive response on an Instron KN1200 (MA, USA) at 1.3 mm/min per the standard. Samples were examined for their morphological characteristics with optical microscopy (Kern OKO 1, with a 5MP type ODC 832 camera, Balingen, Germany) and a field emission SEM (Jeol JSM-IT700HR; Tokyo, Japan, Au sputtered samples, high vacuum mode, 20 kV acceleration voltage).

2.4 Design of experiment and ANOVA

Control parameters (ORA, degrees; RDA, degrees; ID, %; LT, mm; NT, °C; PS, mm/s; BT, °C) were selected according to the literature review, and they are satisfying two additional criteria to be generic, i.e., independent of the 3D printer used and to be continuous, i.e., without categorical control such as the infill pattern parameter, for example. Their levels were selected according to the literature [42, 43, 47], as well as the ABS material specs (NT, PS, and BT), and preliminary tests. A Taguchi L27 array was compiled with five replicas each [81]. That is 135 experiments for the modeling process and the analysis of the compression strength results and the energy consumed for the 3DP of the samples. Full factorial modeling would require 5 × 37experimental runs. Results were validated with verification experiments. The five replicas per case evaluate the deviation in the experiments, the validation experiments, and the reliability of the results.

The response parameters were the PT (s) and w (g), which are also used for the calculation of the energy-related outputs, i.e., EPC (in MJ), SPE (in MJ/g), and SPP (in KW/g). Compression strength response parameters were the sB (MPa), E (MPa), and the toughness (MJ/m3). Regression analysis evaluated the reliability of the modeling analysis and provided equations as functions of the process parameters for the prediction of the responses. The regression analysis is provided in the supplementary material of the study.

3 Results

3.1 Examination of the morphological characteristics of the samples

Figure 3 presents stereoscopic images from the top surface of randomly selected samples from different experimental runs. The differences in the 3DP parameters used in each different run, i.e., ORA, RDA, LT, and ID, can be easily distinguished in the images. In all images, a perfect 3DP structure is presented without defects or voids.

Fig. 3
figure 3

Typical 3D-printed samples with the various control parameters’ values

Full size image

Figure 4a shows a graphical representation of the compression experiment, indicating how the 3DP structure is expected to behave and affects the results. Depending on the specimen’s printed structure, shear failure begins as the specimen is subjected to a compressive load and gradually deforms. The specimen is subjected to the compressive load until it fully fails. The shattered surfaces are microscopy-inspected to investigate the failure mechanism. Figure 4b–e shows SEM images at 70 and × 300 from the fracture surface of two randomly selected samples built with ORA 45°, RDA 0°, LT 0.2 mm, and ID 100. The sample presented in Fig. 4b and c has been built with PS 20 mm/s, NT 230 °C, and BT 120 °C, while the sample presented in Fig. 4d and e has been built with PS 40 mm/s, NT 250 °C, and BT 80 °C. The left sample shows higher deformation on the fracture surface than the right sample. The left sample has a more solid surface, while a crack has been formed in the fracture area of the right sample. In all cases studied, samples were fully fractured (divided into two pieces after the compression test) except for the case of samples built with ORA of 45°, in which the samples failed in the test without a full fracture occurring. It should be mentioned also that no buckling was observed in the samples during the tests.

Fig. 4
figure 4

a Graphical representation of a 3D printed sample compression test, SEM images of the fracture surface of samples built with ORA 45°, RDA 0°, LT 0.2 mm, ID 100 and PS 20 mm/s, NT 230 °C, and BT 120 °C at a magnification of b × 70 and c × 300, and PS 40 mm/s, NT 250 °C, and BT 80 °C at a magnification of d × 70 and e × 300.

Full size image

3.2 Experimental results and statistical analysis

Table 1 presents the printing time (s) and weight (g) response parameter values with their deviation for the different control parameter levels. The corresponding MEPs that complied with the average response parameters’ values are presented in Fig. 5. The two response parameters follow the same trend regarding the ORA, similar trends for the RDA, ID, NT, and BT, and different trends for LT and PS. RDA, LT, PS, NT, and BT do not significantly affect the weight of the 3DP parts, which is affected by the ORA and the ID control parameters. ORA of 0 and 90° reduces the weight, and 45° ORA increases it. The increase in the ID also increases the weight. Regarding the printing time, ORA of 0 and 90° reduces it, and 45° ORA increases it. A mild increase in the PT is presented, with an increase in the RDA. The increase in LT radically decreases the PT and is the number 1 ranked control parameter. The increase in the ID increases the PT, while, as expected, the increase in the PS reduces the PT, and PS is ranked as the no. 2 in importance of control parameters regarding the PT. NT and BT also have mild effects on the PT compared to other parameters and are ranked as the two least important control parameters for PT. The rank of all the control parameters for both PT and w is presented in Fig. 5. Table A in the supplementary data provides analytical results for each one of the experimental runs and replicas for the PT and weight response parameters.

Table 1 Taguchi L27 design: control parameters, levels, mean average values, and standard deviations of measured responses for printing time and specimen weight
Full size table
Fig. 5
figure 5

MEP of printing time (s) and weight (g) for the different control parameters

Full size image

Table 2 presents the mean average values with their deviation for the energy metrics and the compression mechanical properties related to response parameters per run. For the two main response parameters of EPC and sB, MEPs have been compiled and are presented in Fig. 6. From the produced graph, no clear relation between the compressive strength and the consumed energy can be derived. The two response parameters follow a similar trend only for the LT control parameters; still, in the sB, high values improve the optimization, while in the EPC, low values are better. Both response parameters are not significantly affected by the NT and BT control parameters. PS does not significantly affect the sB, while the increase in PS reduces EPC and is the rank 2 control parameter for the EPC; 0° ORA provides high sB values, with low EPC. The highest EPC with the lowest sB is observed at 45° ORA; 90° ORA gives better results than 45° ORA regarding the sB, with low EPC. The increase in RDA reduces sB, with a significant decrease presented at 90°. At the same time, the increase in RDA increases the EPC. The 0° RDA gives the highest sB value, with low EPC. EPC is drastically reduced with the increase in LT, while the increase in LT decreases sB. Finally, the increase in ID significantly increases sB. The lowest EPC is reported for low ID values, and it increases with the increase in ID while maintaining the same level of values at the highest ID studied. Table B in the supplementary data provides analytical results for each one of the experimental runs and replicas for the EPC, SPE, SPP, sB, E, and toughness response parameters.

Table 2 Mean average values and standard deviations of measured responses for EPC, SPE, SPP, compression strength, compression modulus of elasticity, and compression toughness
Full size table
Fig. 6
figure 6

MEP of compressive strength (MPa) and energy (MJ) for the different control parameters

Full size image

To further analyze the process mechanism, interaction plots are formed for sB and EPC for the control parameters studied (Fig. 7). For sB, ORA acts synergistically with ID, PS, NT, and BT and antagonistically with RDA and LT. RDA acts synergistically with PS, NT, and BT and antagonistically with the remaining control parameters. LT acts synergistically with PS, NT, and BT. ID acts synergically with PS, NT, and BT. PS acts synergistically with ID, LT, RDA, and ORA. NT acts synergistically with ID, LT, RDA, and ORA. BT acts synergistically with ID, LT, RDA, and ORA. Regarding the EPC, mostly antagonistic relations can be observed. ORA, RDA, and LT act synergistically with LT, and RDA with PS, NT, and BT. ID acts synergistically with PS, NT with LT and RDA, and BT with LT. Figure 8 presents the sB and EPC response parameters as surface graphs for the control parameters with the highest ranks.

Fig. 7
figure 7

Interaction plots of compressive strength (MPa) and energy (MJ) for the different control parameters

Full size image
Fig. 8
figure 8

Surface graphs of compressive strength (MPa) and energy (MJ) for the different control parameters

Full size image

4 Discussion

Herein, the energy consumption, which is a critical parameter for the sustainability of a process, was quantified for 3D printing parts with the ABS polymer using the MEX process. At the same time, an attempt was made to optimize the mechanical properties of the 3D-printed parts under compression loading. It was not possible to optimize both the energy consumption and the compressive strength simultaneously, employing one common set of 3D printing settings. Still, it was found that parts with improved mechanical strength can be built using moderate energy amounts. On the other hand, if energy consumption is the priority, it is possible to build parts with sufficient strength under compression loads, with reduced energy demands. Seven critical 3D printing parameters were studied, and it was found that the selection of appropriate 3D printing settings can significantly affect the performance of the built parts. For example, the compression strength of parts built with different values of the ORA setting can differ up to more than 40%, showing the significance of the 3D printing setting selection. The ID was the dominant parameter affecting the compressive strength of the parts, with the difference in the compressive strength of parts build with different ID exceeding 100%. So, selecting appropriate 3D settings is the first step toward the production of 3D-printed parts with improved mechanical properties. Additionally, additives can further improve and reinforce the mechanical properties of the 3D-printed parts [15]. Such an investigation was outside the scope of the work.

Two works related to the optimization of the mechanical properties of 3D-printed parts made of ABS [10] and PLA [82] follow a similar approach to the current work but with fewer control parameters. Also, different types of mechanical properties were studied. A parameter that was not considered in these previous works was the ORA. This parameter was proven critical in the current study, showing the need for additional 3D printing setting evaluation as it is not obvious which parameter might affect the response metrics. Regarding the compressive strength of ABS parts in MEX 3D printing, different studies have been conducted, but none is studying so many parameters at the same time or employed modeling tools for the analysis of the experimental results [83,84,85,86,87]. Still, the compressive strength results are comparable and summarized in the following Table 3. Any differences can be attributed to the different ABS grades used, the experimental conditions, and mainly to the different 3D printing settings used and the lack of optimization in the existing studies. This highlights the importance of selecting appropriate 3D printing settings as their effect on the mechanical properties was here verified as well. It also highlights the need for optimization due to the wide range of the reported values herein and in the literature. The current study investigates more 3D printing parameters (seven) than similar works in the literature; at the same time, it reports higher compressive strength values than the existing literature. Therefore, it could be assumed that increasing the number of 3D printing parameters that are optimized leads to a more improved performance from the built parts.

Table 3 Experimental results of the current study in comparison to the literature
Full size table

Regarding energy consumption, the increase in the LT reduces the required energy for the 3D printing of the parts to 50% of the energy required at lower LT values. A similar outcome was found with the increase of the PS, which reduces the energy required to about 55% of the energy required at lower PS values. Such outcomes are indicating the importance of the 3D printing parameters in energy consumption as well. It should also be noted that the increase in the temperature parameters (NT and BT) does not significantly affect both the energy consumption and the compressive strength. On the other hand, energy consumption was found to be hugely affected by the printing time, which was also a response indicator in the study. The MEP pattern of the two response indicators is almost the same, showing a direct connection between the two metrics. In a work studying the energy consumption of ABS parts built with the MEX process, LT and PS were also the dominant parameters affecting the energy consumption, same as in the findings of the current work (higher values decreased EPC) [10]. A similar finding is also reported for the energy consumption of PC parts made with the MEX 3D printing process [67].

In the study, the Taguchi plan was followed. Different modeling tools could have been implemented, and there are several other prospect modeling methods. Herein, it was preferred to use this approach since two different research areas were investigated, i.e., energy consumption and the mechanical properties of the MEX ABS 3D-printed parts under compression loading. The analysis showed that each different area suggested a different set of significant parameters. If only the mechanical properties were investigated, approaches such as the fractional factorial plan would probably be more suitable. But it was required to have a comparable experimental effort in an attempt to optimize both research areas simultaneously with a common set of parameters if this was feasible. The fractional factorial plan would require two different and independent sets of experiments and two different optimization efforts. The number of levels was selected to be three, as a large number of experiments were already conducted in this experimental effort for the specific study.

5 Conclusions

This work provided inclusive and comprehensive results on the behavior under compressive loading of ABS MEX 3DP parts. For the time being, seven 3DP parameters are investigated for the effect on both the compressive properties of the parts and the energy consumed for their fabrication. An insight into the significance of each 3DP parameter is provided through the statistical analysis followed. LT and PS were the dominant parameters regarding the EPC, while NT and RDA were the least important settings. For the compression strength, ID and ORA were the dominant parameters and PS and NT were the least important parameters. Regarding the cons of the work, the experimental results presented are not generic for every polymer. Each polymer has different thermomechanical properties, different strand fusion mechanisms, etc., so corresponding processes are required for each polymer. The experimental data provided can be processed with various modeling tools such as artificial neural networks, among others, as future work. Additionally, although a large number of seven 3D printing parameters have been ranked and analyzed in the work for their statistical importance, the experimental process can be further expanded with additional parameters and levels to extend the applicability of the current research results.