**Investigation of a Short Carbon Fibre-Reinforced Polyamide and Comparison of Two Manufacturing Processes: Fused Deposition Modelling (FDM) and Polymer Injection Moulding (PIM)**

**Elena Verdejo de Toro 1, Juana Coello Sobrino 1,2,\*, Alberto Martínez Martínez 2, Valentín Miguel Eguía 1,2 and Jorge Ayllón Pérez <sup>2</sup>**


Received: 31 October 2019; Accepted: 31 January 2020; Published: 3 February 2020

**Abstract:** New technologies are offering progressively more effective alternatives to traditional ones. Additive Manufacturing (AM) is gaining importance in fields related to design, manufacturing, engineering and medicine, especially in applications which require complex geometries. Fused Deposition Modelling (FDM) is framed within AM as a technology in which, due to their layer-by-layer deposition, thermoplastic polymers are used for manufacturing parts with a high degree of accuracy and minimum material waste during the process. The traditional technology corresponding to FDM is Polymer Injection Moulding, in which polymeric pellets are injected by pressure into a mould using the required geometry. The increasing use of PA6 in Additive Manufacturing makes it necessary to study the possibility of replacing certain parts manufactured by injection moulding with those created using FDM. In this work, PA6 was selected due to its higher mechanical properties in comparison with PA12. Moreover, its higher melting point has been a limitation for 3D printing technology, and a further study of composites made of PA6 using 3D printing processes is needed. Nevertheless, analysis of the mechanical response of standardised samples and the influence of the manufacturing process on the polyamide's mechanical properties needs to be carried out. In this work, a comparative study between the two processes was conducted, and conclusions were drawn from an engineering perspective.

**Keywords:** additive manufacturing; fused deposition modelling; composites; 3D printing; polymer injection moulding; polyamide; CFRP

#### **1. Introduction**

Fused Deposition Modelling (FDM) is a promising technology framed within Additive Manufacturing (AM) and has been implemented in the manufacturing processes of functional structures with complex geometries, showing interesting responses in different fields [1]. Medicine, biotechnology, automotive industry or aerospace are domains that use such technologies to obtain parts made of polymers or polymer–matrix composites with high stiffness and low weight values, also reducing material waste during the process [2]. Among the different technologies that are part of AM, those expected to be most suited to the manufacturing industry are Selective Laser Sintering (SLS), Selective Laser Melting (SLM) and FDM [3]. The last of these technologies is considered to be more developed than the other two [4]. FDM technology consists of the layer-by-layer deposition of a thermoplastic polymer previously driven into a semi-liquid state through the nozzle that allows the

deposition of the material. The most widely processed polymers are acrylonitrile butadiene styrene (ABS), polylactic acid (PLA), polyvinyl alcohol (PVA), polyamides (PA) or polyether ether ketone (PEEK) [5]. Three-dimensional printing is also considered a sustainable production technology, which means that material waste is reduced in the manufacturing process, or even avoided [6]. In addition, the high geometrical accuracy in the dimensions of the printed parts is also a determining factor [7]. Design flexibility is considered an advantage of the FDM process as it takes into account not only the idea of modifying the prototype as much as desired, but also considers the wide variety of parameters that this technology allows to be used [8]. Infill density, infill pattern, layer height or printing velocity are the parameters traditionally considered [9]. Nevertheless, other aspects that are sometimes neglected might have an influence on the part being manufactured [10]. For example, Chacón et al. demonstrated the influence of the raster angle and the placement of the part and established the optimum location to obtain the best properties, studying the behaviour of parts under tensile and bending loads [11]. It is necessary to study the influence of these parameters on the mechanical response of printed parts to create a summary that finds the best combination for each application, depending on different stress types [12]. Some authors have also studied the anisotropy of AM parts, determining the significant effect of the placement of the part on the build plate [13].

The main disadvantages of printed parts that need to be addressed include the poor mechanical properties limited by the polymeric matrix and the poor adhesion between consecutive layers during deposition [14]. Some works have studied the influence of parameters on polymers, such as PA12 (ultimate tensile strength (UTS) = 33 MPa) or reinforced PA12 [15]. However, the mechanical properties of PA12 are lower than PA6 (UTS higher than 50 MPa), which has been considered a more suitable choice for replacing metal parts [16]. Higher stiffness and strength values are obtained in PA6 than in PA12, which is more widely used in FDM research, considering the limitations on the melting temperature according to the thermal resistance of printed parts [17]. Other materials, such as ABS and PLA, have been compared to PA, but have yielded worse results [18]. Nylon 6 showed a tensile strength of ~80 MPa vs. ~50 MPa for PLA and ~65 MPa for ABS considering injected parts. In the case of AM manufactured parts, lower results were obtained: a tensile strength of ~50 MPa for nylon 6, ~45 MPa for PLA and ~25 MPa for ABS [18].

Alone, matrices do not reach the expected requirements in mechanical behaviour. To solve the low strength and stiffness problem, the use of reinforcement has been considered the best alternative [19]. Carbon and glass fibres (short or continuous) and nanoparticles, such as nanospheres, are used to reinforce the polymeric matrix [20]. For reinforced polyamide, Chabaud et al. obtained mechanical properties similar to aluminium alloys [21]. Carbon fibres have been chosen as a favourite reinforcement due to their high stiffness and strength, and low weight [9,21].

Despite having shown higher mechanical properties as a matrix in composites, printed parts made of short carbon fibre-reinforced PA6 have not been fully studied. Furthermore, previous research has focused on evaluating the tensile or dynamic properties but has not considered the compressive properties. Moreover, knowledge is needed about which manufacturing technique fits better with the requirements of parts. In addition, some new technologies, such as FDM, must be compared with traditional ones, in this case polymer injection moulding, to determine whether the latter might offer a more effective solution. Taking into account all the above-mentioned considerations, this experimental work focuses on evaluating and comparing the response of printed and injected samples, while also considering the possibility of dealing with compressive loads. The combination, or even the replacement of traditional technologies with new ones, will be essential for the manufacturing process of different functional parts in forthcoming years. This work presents this comparison and provides results for different mechanical tests summarised from an engineering perspective.

#### **2. Materials and Methods**

#### *2.1. Raw Material and Equipment*

The short carbon-fibre-reinforced material used as feedstock in the printing process was CarbonXTM CRF-Nylon (3DX Tech, Grand Rapids, MI, USA), a filament of PA6 with a diameter of 2.85 mm. Fibre content was 20 weight percent (wt %). Furthermore, an Olsson Ruby nozzle of 0.4 mm replaced the printer's default brass nozzle as it has previously been proven that fibre damages brass after a few printed samples. The printer was an Ultimaker 2 Extended + (Ultimaker, Utrecht, Netherlands). Samples were designed using Autodesk Inventor (2016 version, Autodesk, St. Raphael CA, USA), and the slicing program was Cura 3.5.1 (Ultimaker, Utrecht, The Netherlands). In the case of the injection moulding, the same material was used but in granulated form. Although the granulate could have been obtained directly from the filament, we decided to use the granulated composite from the same supplier in order to work with a homogeneous size of grain.

#### *2.2. Parameters for Manufacturing*

As our aim was to study the differences in the behaviour of samples manufactured by injection moulding and 3D-printing technologies, we needed to distinguish and set different parameter types corresponding to each manufacturing technique.

Three-dimensional-printing parameters were set following the recommendations of the feedstock manufacturer. The nozzle temperature was 260 ◦C, the build plate temperature was 80 ◦C, the printing speed was 50 mm/s and the layer height and nozzle diameter were 0.1 mm and 0.4 mm, respectively. In order to compare the samples manufactured by injection moulding, we manufactured samples with 100% infill density using the longitudinal pattern. The 60% infill density parts were manufactured by employing different patterns (triangles, lines ±45◦ and longitudinal; see Figure 1) to compare them in bending, compression and tensile tests.

**Figure 1.** Stereomicroscope images (×1.25) of the appearance of the injected and different patterned printed samples.

The mould temperature was set at 80 ◦C, whereas the polymer temperature during injection was 260 ◦C. The pressure in the process was 8.5 bar. Injection moulding parameters were chosen to be as close as possible to those used in the printing process.

#### *2.3. Measurement Techniques*

The influence of both processes on fibre length was first investigated. The specimens manufactured by both 3D printing and injection moulding were extracted and burnt in Thermogravimetry-Differential Thermal Analysis (TG-DTA) equipment (EXSTAR 6200, Seiko Instruments, Chiba, Japan). The process started at 23 ◦C and went to 900 ◦C in an inert nitrogen atmosphere to avoid fibre degradation. The heating rate was 20 ◦C/min at an inert gas flow of 200 mL/min. Having obtained fibres, measurements were taken with a microscope. Three specimens of each manufactured sample were measured, obtaining the results shown in Figure 2.

**Figure 2.** Results of the fibre length distribution in the raw material, injected and printed samples (**A**) and measurement of diameters in fibres using 400× with a microscope (**B**).

In many experimental studies, the printer's thermal environment affects the degree of crystallinity of the composite's polymeric matrix, which strongly influences its mechanical properties. In general, the higher the degree of crystallinity, the greater is the strength and the less is the deformation generated in mechanical tests. To obtain reliable results in mechanical tests, a previous analysis of the thermal environment in the printing process was carried out. To this end, three build plate temperatures were chosen in order to determine whether the thermal environment had an impact on the mechanical response of printed parts: 110 ◦C, 60 ◦C and 25 ◦C, which was considered representative of the room condition. DSC tests of samples after printing were conducted and bending tests were carried out to study whether crystallinity appears in the material and the response of the printed parts to mechanical stresses. Although in all three cases, the build plate temperature was lower than the initial characteristic crystallisation temperature determined for the material, the gradient of temperature from the printing temperature (260 ◦C) to the temperature set for the build plate was found to promote crystallinity.

All mechanical tests were conducted under room conditions (23 ◦C, HR 50%). The tensile properties of the samples were obtained according to UNE EN ISO 527-2, with a Zwick Z010 TN2S using an extensometer also produced by Zwick with a calibrated length (L0) of 25 mm. Data were collected and processed with the testXpert machine software (8.1 version, ZwickRoell, Ulm, Germany). The distance between grips was 115 mm, and test velocity was set at 1 mm/min.

A compression test was conducted as standard, but recommendations were taken from UNE EN ISO 14126:2001. The cylindrical samples (φs = 12 mm, Ls = 18 mm) were manufactured by both injection moulding and 3D-printing. To do so, a new steel mould with the sample's chosen dimensions had to be manufactured according to the requirements of the above-mentioned injection equipment. Compression velocity was also set at 1 mm/min. Two flat carbon steel discs (φ = 16 mm) were used to compress the sample. The same Servosis machine as in the bending test was used to carry out the test, following UNE EN ISO 178:2011. Molybdenum (IV) sulphide was placed on the surface of the flat discs to reduce the friction between samples and discs, and to prevent the sample from becoming barrel-shaped during the test.

The moulds for injection were designed with only one 6 mm gate; no runners were used in order to obtain a direct material flow into the cavity. The gate was placed at one end of the part to be obtained, which in the worst case was 160 mm (tensile specimens). These designs were elaborated to make the flow of the material easier in spite of the maximum available pressure of the equipment.

#### **3. Results and Discussion**

#### *3.1. Fibre Length Analysis*

After the calcination process in an inert atmosphere, a hundred fibres of each specimen were obtained and measured. The results of the histograms of the lengths of the initial filament (raw material), the 3D printed and the injected material, are shown in Figure 2A. Both processes had a negative influence on fibre length. The average initial feedstock filament length was *L* = 83.16 μm. However, after printing and injection moulding, fibre length decreased by up to 23.27%.

This may have a number of consequences on the behaviour of the manufactured samples if the critical length (*Lc*) is greater than the fibre length of the processed parts, since the reinforcing effect is considered to be less effective. In order to obtain the critical fibre length, it was necessary to determine the tensile strength of the fibre (σ*f*), its diameter (φ*f*) and the interfacial shear strength (IFSS) fibre-matrix. This last property can only be found by performing a pull-out test of the composite material. Once the necessary parameters were identified, the critical length was obtained by applying Equation (1) [16].

$$L\_{\mathfrak{C}} = \frac{\sigma\_f \cdot \phi\_f}{2 \cdot IFSS} \tag{1}$$

The value of <sup>σ</sup>*<sup>f</sup>* = 2.2 GN·m−<sup>2</sup> was taken from the literature [16]. The experimental measurement of the diameter of the fibre was found to be φ*<sup>f</sup>* = 10.12 ± 0.94 μm (Figure 2B). The shear stress of the interfacial bonding between the carbon fibre and the matrix (*IFSS* = 44 MPa) was taken from experimental results reported in different research works [1,22,23] because it was not possible to run the pull-out test in this experimental study. Carbon fibres were considered to be unsized.

The critical length obtained by Equation (1) was *Lc* = 253 μm. Therefore, as the length of the fibres inside the matrix was shorter than the critical theoretical value, the reinforcing effect would be lower than for a fibre length longer than the critical one, especially in the tensile test. Nevertheless, fibres with a longer length than the critical one were observed after both manufacturing processes; that is, 3D-printing and injection moulding. As a result, the reinforcing effect would take place, but to a lower extent and only as a result of the longest fibres. Moreover, some researchers [9] have found that the 3D-printing process reveals a greater influence on the mechanical behaviour of the part than the reinforcement effect itself. In addition, some studies report that the highest reinforcing volumes can only be reached by using short fibres and taking into account that in AM, the mixing–printing process itself tends to result in a length shorter than the critical one [17].

#### *3.2. Study of Variation in Crystallinity in the Printed Samples*

The DSC tests were carried out with the three printed samples of each type (at different build plate temperatures). The results showed no variation in crystallisation peak, starting point or ending point. The samples behaved the same in the thermal analysis and the results obtained were satisfactory, as shown in Figure 3B,C.

DSC analysis of the raw material, i.e., the CarbonXTM CRF-Nylon wire, showed a crystallisation peak (Figure 3A), whereas this was not seen in the DSC analysis carried out for 3D printed parts (Figure 3B,C) independently. For this reason, it can be concluded that the AM manufacturing technique promotes the crystallinity of the PA6 matrix.

There was no variation between the top and bottom parts of the same sample, which were those farthest from and closest to the build plate, respectively (Figure 4). This result is also supported by Figure 3 showing that the DSC curves corresponding to bottom and top surfaces are identical; the initial parts of the curves are typically different according to the initial humidity conditions of the samples and/or the stabilisation time of the furnace chamber. In the bending test, the top and bottom parts in the same samples were tested with bending stresses as a fracture appeared in the stretched part. Both parts were tested to check the influence of the thermal gradient during the printing process.

**Figure 3.** The DSC analysis of the samples printed at different built plate temperatures to analyse the influence of the thermal environment on the degree of crystallinity. The DSC analysis of raw material is shown in (**A**), the top parts are the analyses shown in (**B**) and the bottom parts are the analyses shown in (**C**).

**Figure 4.** Positioning samples in the bending test, labelled (top and bottom) according to printing placement.

The results obtained for the maximum flexural strength (σ*fM*), Young's Modulus (*Ef*) and yield strength (*Yf*) showed a negligible variation between tested specimens (Table 1) and, consequently a negligible influence of the build plate temperature on the flexural behaviour of samples (see Figure 5). The degree of crystallinity was not affected by the thermal gradient, which appeared in the manufacturing process under the set conditions.


**Table 1.** Mechanical properties obtained from the bending tests.

**Figure 5.** Analysis of the mechanical response of the samples manufactured by 3D printing at different build plate temperatures to study if anisotropy was caused by the thermal environment. (**A**) Build plate temperature: 110 ◦C, (**B**) build plate temperature: 60 ◦C, (**C**) build plate temperature: 25 ◦C and (**D**) bending test comparison of (**A**–**C**).

Homogeneous behaviour due to the uniformity of each sample regarding its crystallinity was expected. Therefore, the mechanical properties of parts would be independent of the thermal gradient appearing in the 3D printing process and the build plate temperature would not be considered a variable affecting the part.

#### *3.3. Determination of Properties in the Tensile Test*

The tensile test results showed marked differences between the 60% filled parts and the 100% filled ones. As expected, infill density had a significant impact on the three mechanical parameters under study (see Figure 6). Ultimate tensile strength (UTS) was higher when the completely filled unidirectional pattern was used, which allowed 63% more stress than the 60% filled one. The results obtained for Young's Modulus showed improvements of up to 62% for the completely filled samples. For yield stress, the same occurred as in tensile strength.

**Figure 6.** Tensile test. Results obtained for Young's Modulus, yield strength and tensile strength of the injected and 3D printed samples.

When studying the influence of pattern, the unidirectional pattern behaved better than the triangular one (improvements of 47%) and the linear ±45◦ (up to 37%). For yield stress and Young's Modulus, the same occurred between hollow patterns. Due to the non-deformability of the triangles in the triangular pattern, the samples built using that distribution of filaments displayed greater levels of stiffness, which might be due to a higher Young's Modulus compared to the alternative linear ±45◦ pattern (30% higher in the former). Nevertheless, the best result was obtained with the unidirectional pattern (improvements up to 55%) due to filament orientation according to the preferable distribution of stress in the axial direction in the tensile test. However, the ±45◦ pattern is used in applications where the direction of the stresses is unknown or known but non-axial, whereas the unidirectional pattern might have a worse response to the mechanical stresses loading the part.

Consequently, not only infill density but also the infill pattern was a determining factor in the results obtained.

When comparing the printed and injected parts (Figure 7), it can immediately be seen that the injected samples behaved better under axial stresses. However, differences were not as large as expected. Yield stress was 21% higher, whereas Young's Modulus was 17% greater and tensile strength at the break point was only 20% higher in the injected samples. Comparing FDM and IM, Lay et al. obtained similar results [18], while Blok. et al. tested a printed short carbon fibre-reinforced nylon (6 wt % of fibre weight content) and reported a UTS = 33 MPa and E = 1900 MPa [9]. In the case of the current study, for the unidirectional pattern, UTS = 52 MPa and E = 6191 MPa, but the percentage of fibre inside the polymeric matrix is also higher (20 wt %). Furthermore, greater deviations from the average were obtained with the injected samples, in which more homogeneous behaviour was expected than for the 3D printed samples. It can be concluded that the printed parts would be suitable to replace the injected ones when working under stretching loads.

From the fractographies in Figure 8, a sound structure for the parts obtained by injection moulding (Figure 8A) can be observed. Properly, the fractography is dominated by the matrix and fibres fracture, and to some extent, by the fracture of the matrix–fibre interface. This reveals that the grade of polymerisation obtained in the injection moulding process is reasonable despite the low-pressure injection condition (8.5 bar). The fractures of the samples obtained by 3D printing show that pores and fibres have been separated from the matrix interface in a greater proportion (Figure 8C,D). This is a consequence of the short length fibre used leading to an insufficient interface area with the PA6 matrix. From this viewpoint, the interface is more suited to the injection moulding samples.

**Figure 7.** Comparison of the behaviour of the injected and printed samples with stretching stresses in the tensile test (**A**), and the influence of infill density and printing pattern on the mechanical behaviour of the printed parts (**B**).

**Figure 8.** Fractographies of tested tensile samples: (**A**) Injection moulding 600×; (**B**) injection moulding 1000×; (**C**) 3D printed unidirectional 0◦ 600×; (**D**) 3D printed ±45◦ 600×.

#### *3.4. Determination of Properties in the Compression Test*

The compression test results are shown in Figure 9 and indicate improvements of only 4% in the injected samples in comparison with the 3D printed ones (100% infill). It is worth noting that Young's Modulus is higher in the printed parts (50% improvement with respect to the injected samples). This suggests that when compressive loads are applied, the 3D printing process leads the parts to show higher stiffness than in the IM process. In compression tests, this is due to the direction of the force being opposite to that in the tensile test and the separation of consecutive layers of the printed parts is more difficult. Thus, using high infill density values avoids an early breakage between consecutive layers. Qualitatively, it can be observed that the behaviour of the 3D printed part decreases

as of a determined strain. Furthermore, the decline appears just when the compressive stresses of both samples are equal. This decrease is a consequence of the separation taking place between consecutive layers. This conclusion is supported by the enormous separation observed in Figure 10B corresponding to 60%-filled parts, while this occurred incrementally in completely filled samples. In the 60%-filled non-unidirectional samples (Figure 10C,D), polymeric hardening took place under compression stresses (Figure 11). This can be an advantage for parts that work under compressive stresses with no deformability constraints. Inversely, the unidirectional pattern broke down due to the presence of gaps between consecutive layers. Improvements up to 73% in yield strength and 33% in Young's Modulus were reached when 100% instead of 60% infill density was used, as seen in Figure 9.

**Figure 9.** Compression test. Results obtained for yield strength, tensile strength and Young's Modulus of the injected and 3D printed samples.

**Figure 10.** Fracture images of samples under compressive stress. (**A**) Unidirectional 100%, (**B**) unidirectional 60%, (**C**) triangles 60% and (**D**) linear ±45. (**E**) Comparison between the injected and printed samples in the compression test.

It was possible to see a slight influence of infill pattern for yield strength (unidirectional had a higher yield strength than triangles, and these were higher than the ±45◦ alternate linear one). These smaller differences were due mainly to the stress direction in the test, which tends to compress samples. In this case, the bonding between different layers was not as important as in the tensile test, where the commitment of transmitting load between consecutive layers came into play, and its low failure strength was a determining factor for the properties. However, Young's Modulus was strongly affected

by pattern. Filament orientation was due to pattern choice, which acted as a determining factor to consider the preferential direction of the stresses in each test and, consequently, in each functional application. Both the tensile and compression tests justify this conclusion.

**Figure 11.** Compression test: influence of pattern on the behaviour of sample under compressive stresses.

#### **4. Conclusions**

This work describes a comparative analysis of 3D printing and injection moulding technologies to establish the influence of the process on the behaviour of parts made of short carbon-fibre-reinforced PA6. Both the 3D printing and IM processes led to breakage of fibres, making them shorter by up to 24%. The thermal environment has been shown to have no influence on the behaviour of 3D printed parts.

Tensile tests of printed parts showed worse results than for the IM parts, but differences did not exceed 21% for either yield strength, tensile strength or Young's Modulus. These results are consistent with those obtained by other authors, which validates the IM process used as a reference. A non-linear relationship between infill density and mechanical properties is supported. The unidirectional pattern is the best choice when lower densities are used.

Compared to tensile tests, compression tests revealed a more similar behaviour of 3D printed parts and IM parts (only 4% improvement). Printed samples had higher stiffness values than the IM parts. This phenomenon has not been reported by other researchers. The selection of pattern is a determinant in this case since a hardening effect with the strain appeared for those manufactured using a non-unidirectional pattern, but without reaching the best results.

The comparison between the tensile and compression tests revealed that this reinforced polyamide did not behave the same under compressive and stretching loads, regardless of the manufacturing process. Consequently, as previously reported in the literature, it is crucial to analyse the application for which parts are designed. These outcomes are a novel contribution to the existing literature.

**Author Contributions:** All authors are active members of the research group Science and Engineering of Materials at the Castilla-La Mancha University in Spain and they have actively contributed to the development of the paper. Conceptualization, E.V.d.T., V.M.E., J.C.S. and A.M.M.; data curation, E.V.d.T. and J.C.S.; formal analysis, E.V.d.T., V.M.E., J.C.S., A.M.M. and J.A.P.; funding acquisition, V.M.E., J.C.S. and A.M.M.; investigation, E.V.d.T., V.M.E., A.M.M. and J.A.P.; methodology, E.V.d.T., A.M.M. and J.A.P.; project administration, V.M.E.; resources, V.M.E., J.C.S. and A.M.M.; software, A.M.M. and J.A.P.; supervision, J.C.S., V.M.E.; validation, V.M.E. and J.C.S.; visualization, A.M.M.; writing—original draft, E.V.d.T. and V.M.E.; writing—review and editing, V.M.E., E.V.d.T. and J.C.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Build Time Estimation for Fused Filament Fabrication via Average Printing Speed**

#### **Gustavo Medina-Sanchez, Rubén Dorado-Vicente \*, Eloísa Torres-Jiménez and Rafael López-García**

Department of Mechanical and Mining Engineering, University of Jaén, EPS de Jaén, Campus Las Lagunillas s/n, 23071 Jaén, Spain; gmedina@ujaen.es (G.M.-S.); etorres@ujaen.es (E.T.-J.); rlgarcia@ujaen.es (R.L.-G.) **\*** Correspondence: rdorado@ujaen.es; Tel.: +34-953-212-439

Received: 25 October 2019; Accepted: 29 November 2019; Published: 1 December 2019

**Abstract:** Build time is a key issue in additive manufacturing, but even nowadays, its accurate estimation is challenging. This work proposes a build time estimation method for fused filament fabrication (FFF) based on an average printing speed model. It captures the printer kinematics by fitting printing speed measurements for different interpolation segment lengths and changes of direction along the printing path. Unlike analytical approaches, printer users do not need to know the printer kinematics parameters such as maximum speed and acceleration or how the printer movement is programmed to obtain an accurate estimation. To build the proposed model, few measurements are needed. Two approaches are proposed: a fitting procedure via linear and power approximations, and a Coons patch. The procedure was applied to three desktop FFF printers, and different infill patterns and part shapes were tested. The proposed method provides a robust and accurate estimation with a maximum relative error below 8.5%.

**Keywords:** 3D printing; rapid prototyping; efficiency; printing time; experimental model

#### **1. Introduction**

#### *1.1. About Additive Manufacturing*

Since the first 3D printer was developed in the early 80s, the number of additive manufacturing (AM) solutions, often called 3D printing methods in a non-technical context, and their applications do not stop increasing. It is noteworthy the potential of AM in different applications such as bio-printing [1,2], replicating broken objects or custom parts [3], experimental and educational demonstrators [4], rapid tooling [5] and so on.

According to the standard ISO/ASTM 52900-15 [6], AM solutions produce objects by joining materials, usually layer by layer, from 3D models. This standard classifies the existing solutions in seven types of processes considering how materials are deposited and bonded: material extrusion, binder jetting, material jetting, directed energy deposition, vat photopolymerization, powder bed fusion, and sheet lamination. The main AM advantages over traditional methods are low product development time, material savings, and capability to produce objects with complex shapes, enhanced density, and interior structures [7].

Fused filament fabrication (FFF) based on material extrusion processes is the most widespread AM technique [8], and low price models dominate the shipment numbers [9]. FFF consists of heating a thermoplastic filament, extruding the resulting melt and filling layer by layer a part following 2D paths while the plastic solidifies. Although less accurate than other AM technologies, FFF printers are broadly used because of their price [10], the wide range of plastics that can be printed, and the strength of the obtained parts [1,11].

#### *1.2. Build Time*

AM processes have several issues that limit their potential applications. AM issues in the spotlight are: development of new compatible materials [12], dimensional accuracy and surface roughness [13–15], mechanical properties of printed parts, discretization of CAD model and printed object, printer capabilities, maintenance, optimization of shape and part orientation, and time to manufacture a part (build time) [16,17].

Regarding build time, an accurate estimation can help firms to enhance their processes planning, and to compare AM solutions [18]. Moreover, these estimations can lead to more meaningful results in works dealing with the optimization of process parameters with the target to reduce the operation time [19], as well as those studies that consider the influence of printing speed and, therefore, time in the printed part appearance [20] or strength [21]. Finally, the costs of AM machine-hours depends on build time [22], and it stands to reason that, its contribution to the overall costs will increase the price of the used AM technology. Despite the previous arguments, build time has received less attention than other issues, such as dimensional accuracy and mechanical properties [23].

Time estimation is not a simple task because it depends on the printer and its control characteristics, as well as the printing parameters and the machine path planning. Moreover, time prediction is, in general, not accurate [24,25]. The simplest build time estimation is calculated as the total motion path length divided by the programmed printing speed and, in some cases, can differ more than 30% from the actual build time.

During the last decade, several researchers are concerned about build time estimation in additive manufacturing. According to the detailed work of Zhang et al. [26], there are three main strategies to determine the build time:


Experimental solutions are more accurate than parameter ones and simpler than analytical ones, but there are no rules for data selection neither for the fitting strategy and, therefore, the repeatability is low. A change in the printing parameters forces to construct a new response function, therefore, experimental methods are not flexible.

The method developed by Zhang et al. [26], which is based on Grey theory, is an interesting improvement of the experimental approaches. The authors claim that their estimation error has an average value of 10% and it is better than other existing approaches. On the other hand, it is not well established how to select the input factors and many shape parameters are needed (part volume, support volume, part surface, part height, support height, and part projected area).

Later works to those reported in Zhang's paper determine the build time according to the aforementioned strategies. Zhu et al. [27] developed both, an experimental and a parametric model. The last one is based on a reduced number of printing parameters (volume, height, and density) selected depending on their influence on the build time.

Different strategies are used to build experimental models. For example, the experimental solution explained by Zhu et al. [27] proposes a multi-factor regression, and Mohamed et al. [28] used a Q-optimal response surface methodology.

Examples of analytical methods are the estimations proposed by Habib and Khoda [29] or Komineas et al. [30]. The model developed by Komineas et al. [30] for material extrusion processes is based on a trapezoidal speed profile. It considers that tangential acceleration and deceleration are equal and it does not take into account the influence of normal acceleration limit (direction changes) in the printer speed. On the other hand, Habib and Khoda [29] also propose a simple trapezoidal speed profile model to estimate build time, nevertheless, the goal is not to make an accurate estimation of build time, but to use it to optimize the deposition direction. Moreover, some current computer applications, such as Pronterface [31], provide an analytical time estimation for FFF machines based on the printer characteristics, a trapezoidal speed profile and a cornering algorithm.

#### *1.3. A Blended Solution*

This work explains a new build time estimation model for FFF machines, which combines the analytical and experimental approaches. The proposed model takes into account the kinematics of the problem (as the analytical strategies) and defines it by fitting a low number of printing speed experimental observations. The solution is simpler than that of the analytical models and does not need to know the machine nominal parameters and how the printer is controlled. In contrast to parameter and experimental-approaches, the proposed solution requires only two parameters and few experimental tests to provide an accurate time prediction (estimation error below 8.5% in the printed examples). The experimental procedure requires a low-cost setup and can be easily accomplished with the explanation included in this paper. Moreover, unlike the experimental solutions, once the printing speed is approximated, the proposed method can be applied regardless of the chosen printing parameters.

Because of the printing path in FFF and tool path in a milling process are similar, our build time approach is obtained by modifying and extending the mechanistic model of Coelho et al. [32]. The proposed solution is based on a printing speed model, which not only considers the interpolation segment length, as the mechanistic model does, but also the path shape via the changes of direction. Path planning is key in the resulting build time of material extrusion processes [33]. Direction changes and small interpolation segments have a noteworthy impact on the actual time. Based on experimental observations, we show two different fitting strategies to determine the aforementioned printing speed model.

Some assumptions are considered. The method provides the time to print a part, without including the setup and heating times. Because of the fact that FFF motions are generally based on linear interpolation, the study is limited to this kind of interpolation scheme.

The remaining paper is organized as follows. Section 2 discusses the estimation procedure and Section 3 describes the experimental methods. Section 4 shows the approximation speed surfaces and validation examples. Finally, the main conclusions are drawn in Section 5.

#### **2. Build Time Estimation Model**

In this work, the way proposed to obtain the build time of actual parts is via the estimation of the actual printing speed. Through time measurements of known paths, we obtain information about how path definition (interpolation segment lengths and direction changes), machine characteristics, and its control influence the actual speed between consecutive interpolation points of the printing path. Known an approximation of the actual printing speed, time estimation is simple: reading the CNC printing code and using the speed approximation to estimate the real time of each path segment.

#### *2.1. Average Printing Speed*

We assume hereafter that the main factors that influence the printing speed *f* are the interpolation segment length s and the direction change α. According to Figure 1a, *s* is the Euclidean distance between the two consecutive interpolation points. The direction change α is the angle between the direction of three consecutive interpolation points. To understand how *s* and α influence the printing speed, we measure the average speed along circular arcs (Figure 1b), which are built via repeating interpolation segments with the same length and direction change (a line with α = 0◦ corresponds to a circle of infinite radius). The speed measurement procedure is explained in Section 3.

**Figure 1.** Examples: (**a**) Printing path with random segment lengths and direction changes, (**b**) printing paths for experimental speed estimation.

The speed-segment length relation evolves from linear to a power law (Figure 2a). The linear relation becomes smaller with increasing α, whereas the slope does not change. On the other hand, we have different power laws for each α; printing speed asymptotically decreases as α increases (Figure 2b).

**Figure 2.** Example of experimental measurements of printing speed *f*. (**a**) Speed vs. *s* curves for different direction changes. (**b**) Speed vs. α for different segment lengths.

Note that, we measure *f* from *s* = 0.1 mm to a maximum value *s*Max. We choose *s*Max so that the circular path defined by interpolation segments with *s*Max and α = 10◦ is the maximum circle within the printer bed.

Above *s*Max it is more difficult to measure the printing speed. For this reason and based on the previous discussion, if *s* > *s*Max we assume a power law. This power model is computed by interpolation of *f*(α, *s*Max) and *f*(0, *b*), where *b* is the printer bed diagonal length.

#### *2.2. Printing Speed Surface*

This section is devoted to determining an approximation surface that provides the printing speed for given values of *s* and α. There are different possibilities to define the printing speed surface *f*(*s*, α). The interpolation of the measurement points by means of a degree 1x1 polynomial spline patch provides a straightforward solution, but this approach requires many measurements to accurately predict time.

In order to reduce the number of needed measurements, we propose two alternatives:


To determine the isocurves α = constant on the LP surface, we approximate the segment length *s*c(α) where linear and power approximations intersect (Figure 2). The procedure consists of the next steps:

Step 1. Approximate the speed measurements at α = 0 and *s* < *s*c(α = 0) = *s*c,0 by a line *f*(0, *s*) = *L*(*s*) = *m*·*s* + *n*. The speed profiles, such as those portrayed in Figure 2a, show that if *s* < *s*c(α), then the linear relation *f*(0, *s*) does not change with α.

Step 2. Fit the speed measurements, within the interval *s*c,0 < *s* ≤ sMax, at *k* angles 0◦ ≤ α*<sup>i</sup>* ≤ 180◦, *i* = [0, 1, ... , k−1] by power curves *Pi*(*s*) = *aisbi* . Although *s*<sup>c</sup> varies with α, the relation sc,0 > sc(αi) is satisfied, so that we always are in the region ruled by the power law, and fewer measurements are needed. Note that, by increasing *k* the approximation improves at the expense of accomplishing more measurements.

Step 3. Compute *s* at the intersection of *L*(*s*) and the power curves *Pi*(*s*). Fitting the resulting data, for example by means of a degree-2 polynomial spline, we obtain the curve *s*c(α).

Step 4. Build the curves *f*(α, *s*c(α)) and *f*(α, *s*Max) and interpolate them using a power function *P*(α,*s*) = *a*(α,*s*)*sb*(α,*s*).

Step 5. Define the speed surface as a piecewise function:

$$f(a,s) \equiv \{ L(s), \ s \le s\_\mathbb{C}(a); \ P(a,s), \ s\_\mathbb{C}(a) < s \le s\_{\text{Max}} \} \tag{1}$$

Regarding the CP, the idea is to define *f*(α, *s*) as a spline of two Coons patches. A Coons patch is a surface determined by its boundary curves, i.e., it is a way of filling the space between the curves.

The easiest Coons construction is a bilinear blend of two ruled surfaces and a bilinear interpolation surface [34]. Let *f*(α0, *s*), *f*(α1, *s*) and *f*(α, *s*0), *f*(α, *s*1) be the four parametric boundary curves, then it is easy to build a linear interpolation surface with each couple of curves:

$$\begin{array}{l} r\_1(\alpha, s) = \left(1 - \frac{a - a\_0}{\alpha\_1 - a\_0}\right) f(\alpha\_0, s) + \frac{a - a\_0}{a\_1 - a\_0} f(\alpha\_1, s) \\\ r\_2(\alpha, s) = \left(1 - \frac{s - s\_0}{s\_1 - s\_0}\right) f(\alpha, s\_0) + \frac{s - s\_0}{s\_1 - s\_0} f(\alpha, s\_1) \end{array} \tag{2}$$

On the other hand, we can compute the bilinear interpolation of the four patch corners:

$$r\_{1,2}(a,s) = \left[1 - \frac{a - a\_0}{a\_1 - a\_0}, \frac{a - a\_0}{a\_1 - a\_0}\right] \left[\begin{matrix} f(a\_0, s\_0) & f(a\_0, s\_1) \\ f(a\_1, s\_0) & f(a\_1, s\_1) \end{matrix}\right] \left[\left(1 - \frac{s - s\_0}{s\_1 - s\_0}\right) \frac{s - s\_0}{s\_1 - s\_0}\right]^t. \tag{3}$$

Finally, the Coons surface is:

$$CP(\alpha, \mathbf{s}) = r\_1(\alpha, \mathbf{s}) + r\_2(\alpha, \mathbf{s}) - r\_{1,2}(\alpha, \mathbf{s}).\tag{4}$$

A unique Coons patch with *f*(0, *s*), *f*(180, *s*) and *f*(α, 0), *f*(α, *s*Max) is unable to adequately reproduce the actual printing speed surface. To overcome this drawback, we define the printing speed *f*(α, *s*) as a spline of two Coons joined at the value of *s*c,0 defined in the same way as in the LP construction.

The following steps summarize the CP procedure:

Step 1. Measure the average speed at α = 0◦ and α = 180◦ for different segment lengths *s*, and at *s* = *s*c,0 and *s* = *s*Max for several α.

Step 2. Fit the experimental data to obtain the boundary curves: *f*(0◦, *s*), *f*(180◦, s), *f*(α, *s*c,0), *f*(α, *s*Max). We approximate the experimental data by B-splines curves.

Step 3. Build two Coons patches: CPA with *f*(0◦, *s*), *f*(180◦, s), *f*(α, 0), *f*(α, *s*c,0), and CPB with *f*(0◦, *s*), *f*(180◦, s), *f*(α, *s*c,0), *f*(α, *s*Max).

Step 4. Compute the speed surface by the following spline function:

$$f(\alpha, \mathbf{s}) \equiv \{ \text{CP}\_A(\alpha, \mathbf{s}), \text{ s} \le \text{s}\_{\mathbb{C}, \mathbf{0}}; \text{ CP}\_{\mathbb{B}}(\alpha, \mathbf{s}), \text{ s}\_{\mathbb{C}, \mathbf{0}} \le \text{ s} \le \text{s}\_{\text{Max}} \}\tag{5}$$

Note that, the approximation improves by increasing the number of Coons, but it requires increasing the number of measures.

Either for LP surface or the CP approximation, for *s* > *s*Max, the idea is to interpolate the curves *f*(α, *s*Max) and *f*(α, *b*) (that we assume equal to *f*(0◦, *b*)) using a power function.

Section 4 portraits the resulting LP and CP approximation surfaces and shows the measurements used in both surfaces: 22 measures for LP and 25 for CP.

#### *2.3. Build Time Estimation from G-Code*

Once we have the printing speed surface, it is possible to determine the build time from the path G-code. Observe that, the time required for heating the filament and the bed, and the time needed by the hot-end to go home (setup time) are not considered.

The computation process consists of:


We take the programmed speed for *z* movements, as well as for the hot-end reposition when motors in *x-y-z* axis work at the same time because only the printing speed of *x-y* motors is measured.

• Finally, if the path has *l* segments, the estimated build time *t* is:

$$t = \sum\_{j=1}^{l} \frac{s\_j}{f\_{\mathbf{a}}}, \quad f\_{\mathbf{a}} = \left\{ f(\mathbf{a}\_{\mathbf{j}}, s\_{\mathbf{j}}), \text{ if } f(\mathbf{a}\_{\mathbf{j}}, s\_{\mathbf{j}}) < f\_{\mathbf{\mathcal{P}}\mathbf{j}'}, f\_{\mathbf{\mathcal{P}}\mathbf{\mathcal{P}}} \text{ Otherwise} \right\}. \tag{6}$$

In order to accomplish the above procedure, we write a Mathematica® function. This function reads a G-code file, distinguishes each layer and detects the programmed printing speed along the layer paths. After that, it obtains the coordinates of the interpolation points and computes segments lengths *s* and, using the dot product, angles α. Finally, the developed function uses Equation (6) to estimate the actual build time.

#### **3. Materials and Methods**

We use the previously described estimation method (Section 2) to obtain the build time in three low-cost FFF machines: BQ Hephestos®, Witbox®, and Airwolf 3D HD® printers. 2D dimensional paths with random lengths and random direction changes, hereafter referred to as "random paths," and actual 3D printed parts are designed to validate our time estimation procedure.

#### *3.1. Printers*

Table 1 shows the main characteristics of the tested 3D printers. Extrusion and movements along *x*, *y*, and *z* axes are powered by standard stepper motors. We use the same travel speed for all printers: 120 mm/s.

#### *3.2. Speed Measurement Procedure*

In order to measure the speed at a specific *s* and α, we drive the printer hot-end through interpolation segments of length *s* and direction change α (circular paths with total length *L* ≈ 240 mm, Figure 1b), measure the build time *t*, and finally compute the average speed as the ratio of *L* to *t*.


#### **Table 1.** Main technical characteristics of tested 3D printers.

A chronometer can lead to inaccurate time measurements, and even more for short paths. Thus, we decided to implement a measurement procedure based on producing a sound at the ends of the printing path, which leads to a clear identification of the build time. For printers with a Marlin firmware, such as those studied in this work, this means to add the following line before the first printing path position and after the last one:

"M300 P200 S440; play a 440 Hz tone during 200 ms."

A speaker to run the previous instruction is required. For a printer without a speaker, it is easy to connect a buzzer to an empty port of its electronic card and use the previous command. Witbox machine has a speaker, meanwhile, the Airwolf and the Hephestos machines need a buzzer.

Sound is recorded by a microphone connected to a PC, which allows distinguishing the time between the start and end. Each test was run three times. To determine the speed measurement uncertainty *u*(*f*), we apply the combined standard uncertainty [35] to the equation *f* = *L*/*t*. The maximum *u*(*f*) obtained was lower than 0.2 mm/s for the three studied machines.

#### *3.3. Experimental Tests*

In addition to the speed observations for building the approximation surfaces discussed in Section 3.2, we design two types of validation tests: random paths and printing examples.

All tests were conducted in the Pronterface application. Pronterface, similar to other 3D printing applications, provides an analytical print time estimation based on the planner functions used by the printer firmware to define the printer kinematics. These functions are a model of the speed, which by default is a trapezoidal profile, and a cornering algorithm that deals with the direction changes, and it

is usually based on a limit jerk equation. The performance of analytical models depends on the nominal parameters of motors and control, whose values can differ from the real ones, and an adequate definition of factors such as the jerk limit.

It is interesting to compare the proposed model with a 3D printing software estimation, as many researchers [36–39] trust on those predictions to conduct costs and process optimization studies.

#### 3.3.1. Random Path Tests

A set of random paths are tested to assess the performance of the proposed estimation procedure in specific *s*-α regions. Six paths, with a similar total length of 2400 mm, composed of segments with random lengths and directions (random paths) are conducted at travel speed without extrusion in each printer. A Mathematica function is implemented to provide the random paths and to write the needed G-code file. This function defines *x*-*y* positions and the reference printing speed (travel speed).

#### 3.3.2. Printing Examples

We design three 12 mm high prisms with simple bases (triangle, pentagon, and star) at two scales (1:1, 2:1), and print them using two printer configurations. The well-known software Ultimaker CURA® (Free and open source LGPLv3 application developed and maintained by David Braam for Ultimaker, a 3D printer manufacturer based in Utrecht, Netherlands) provides the G-code. This software provides a similar time estimation to that of Pronterface, but it does not allow estimating times from G-codes obtained outside the software, such as for the random paths. For this reason, we compare our results to Pronterface predictions instead of Ultimaker CURA estimations.

Table 2 shows the factors and levels considered, and Table 3 summarizes the 12 tests performed corresponding to all possible combinations for the considered factors and levels, and the actual printing time measures. Note that, geometry, size, and printing parameters modify the *s*-α values and therefore, the resulting build time.

**Table 2.** Factors (parameters) and levels (factor values) to define the printing examples.


Travel: 120


**Table 3.** Actual printing time measured for the accomplished tests of Table 2.

#### **4. Results and Discussion**

The present section is organized as follows: Section 4.1 shows the experimental results required to build the two proposed approximation surfaces described in Section 2.2 (LP and CP surfaces), as well as the resulting surface models. Section 4.2 presents several tests for assessing the accuracy of the proposed models. This validation procedure consists of carrying out several paths where the printing time is recorded and compared to that provided by each approximation model. In Section 4.2.1, this procedure is applied to random paths (without extrusion of printing material), with the target of facilitating the variation in direction and segment length of a trajectory and analyzing their influence on printing time estimation. In Section 4.2.2., the validation procedure is also applied to several printed parts with different geometries in order to find out if the results provided by the approximation surfaces are also accurate for actual examples. A discussion regarding a comparison between the proposed methods, as well as between them and some usual methods for estimating the printing time, such as the Pronterface and theoretical estimations, is included at the end of the present section.

#### *4.1. Printing Speed Measurements*

Figures 3 and 4 show the printing speed measures and the proposed speed estimation models: LP and CP surfaces obtained. Black dots depict the measures used for the approximations, and grey dots represent additional measures used to verify the goodness of the approximation.

The mean absolute error (MAE) of all measurements and the determination coefficient R<sup>2</sup> of the approximations are included in Figures 3 and 4. The goodness of the approximations is high for all printers since R<sup>2</sup> is close to 1 and MAE value is low.

Note that, we choose a reduced number of measures to be fitted (similar for LP and CP approximations). The idea is to take measurements close to *s* = 0, *s*<sup>c</sup> and *s*Max at α = 0◦, 180◦, and at *s*c,0 and *s*Max for different angles (black dots in Figures 3 and 4). The linear region, for the tested printers, is always obtained for segment lengths lower than 1 mm so that, to capture this behavior we take measurements every 0.1–0.2 mm from *s* = 0. On the other hand, the power region is wider than the linear one and, therefore, we use steps of 2–5 mm from *s*<sup>c</sup> and *s*Max. Regarding the data at *s*<sup>c</sup> and *s*Max for different angles, it is better to take more measurements between α = 0◦ and α = 60◦ because the greatest variations are registered within this range. Considering the previous suggestions, we obtain similar surfaces when fitting different experimental data.

**Figure 3.** Coons patches (CP) printing speed surfaces for the tested printers (different colors are used to identify each patch of the CP spline surface). Dots represent the experimental measurements, CP surfaces fit black dots, and additional measurements (gray dots) are represented to visualize the goodness of the approximation.

LP and CP surfaces are similar and evolve as it is expected considering a trapezoidal speed profile and the machine acceleration limits. With respect to *s*, the surfaces evolve from linear to power, and that is explained because in each interpolation segment the printer accelerates to reference speed and decelerates up to the segment end (trapezoidal speed profile). On the other hand, the speed decreases with α (mainly between 0◦ and 60◦). It stands to reason that the curvature and printing speed have a quadratic relation, which agrees with the experimental data and with the LP and CP surfaces obtained.

Machine characteristics influence the resulting speed surface. Although the surfaces have similar shapes, the Airwolf surfaces (LP and CP) provide the highest speed values and the Hephestos the lowest values in the considered *s*-α domain. In the Airwolf machine, the linear region grows steeper than in the other printers whereas the power region grows smoother. It stands to reason that the actual average speed depends on the printer acceleration limits (see Table 1) and this explains the above results.

**Figure 4.** Linear-power (LP) printing speed surfaces for the tested printers. Dots represent the experimental measurements, LP surfaces fit black dots, and gray dots are additional measurements to visualize the goodness of the approximation.

#### *4.2. Validation Tests*

#### 4.2.1. Random Paths

Six random paths (Section 3.3.1) were printed using the Pronterface application. Figure 5 shows the resulting actual printing time, the theoretical estimation (sum of ratios of segment length to programmed speed), the Pronterface prediction and the print time estimation provided by the LP and CP surfaces in the tested printers.

Comparing the six examples, the proposed approximations provide the most accurate estimations. In each printer, LP and CP average errors are similar and always lower than 5.5%. This value improves Pronterface and theoretical average errors, which are up to 48% and 59% respectively. The dispersion observed in the error values is a consequence of how close the approximations are to the actual printing speed surfaces at each region *s*-α.

Speed mainly changes within the linear region, in the transition from linear to power and because of direction changes up to 60◦ (see Figures 3 and 4). Thus, Pronterface, LP, and CP errors have maximum values at R1 and R2. Regarding the theoretical error, it does not consider *s* and α variations, which leads to a maximum error in R4. On the other hand, theoretical and Pronterface predictions are quite similar in regions R1 to R4, but Pronterface estimation improves when α increases (regions R5 to R6) because it considers the cornering algorithm used by the printer control.

**Figure 5.** Estimation relative error for six random paths at different regions of the *s*-α domain.

Finally, comparing the LP and CP approximations in the printers, the Airwolf speed surface has a MAE greater than those obtained for the Witbox and Hephestos printers (see Figures 3 and 4), and this leads to the differences observed in the time estimation error values.

#### 4.2.2. Printing Examples

To assess the performance of the proposed estimations in real parts, 12 prisms are printed (see Table 3). Examples in the previous Section 4.2.1 point out that estimation error depends on the *s*-α region, but for actual printed parts, *s* and α values are not concentrated in a specific region and that can reduce the prediction errors. Another fact that contributes to differentiate random paths and printing examples is the programmed printing speed. While for random paths, the programmed speed is 120 mm/s, which is always greater than the experimental maximum speed measured, in the printing examples the programmed speed changes along the printing path and can be beneath the actual maximum speed surface.

Figure 6 portrays the relative errors corresponding to the theoretical estimation (considering the programmed printing speeds), the Pronterface prediction and the estimations provided via our LP and CP surfaces.

According to Figure 6, it is easy to observe the improvement in printing time prediction provided by the proposed approaches in comparison with theoretical and Pronterface estimations. For the 12 printed examples, while theoretical and Pronterface approaches show dispersed error values with maximum values in samples 6 and 12, and minimum values in samples 1 and 4, LP and CP solutions show similar relative errors that never exceed 8.5%.

Comparing the printers, the Hephestos shows the worst results followed by the Witbox, and the Airwolf shows the minimum error. With respect to LP and CP estimations, this result differs from those obtained for the random paths, but it makes sense considering that, for the printing examples, the programmed speed can have a value below the maximum actual speed surface. In this case, the estimations consider a speed equal to the programmed speed (see Equation (6) for LP and CP solutions). Estimation errors decrease with the difference between programmed and average actual speed. Printer acceleration determines that difference, in a manner that, the fastest printers show the lowest estimation errors in the printing examples.

**Figure 6.** Relative error of the proposed method, theoretical estimation and Pronterface prediction for the twelve printed samples presented in Table 3.

#### **5. Conclusions**

The printing speed predictions showed in this paper lead to accurate build time estimations (maximum relative error of 8.5 % in the printed examples). The experimental methodology devised to build a printing speed surface can be straightforwardly applied to any FFF or similar machines, by measurement printing times for different segment lengths and direction changes along linear and circular paths. The proposed fitting procedures, LP and CP approaches, provide good mean printing speed approximations (mean absolute error lower than 2.7 mm/s) even for a reduced number of experimental data (22 measurements for LP and 25 measurements for CP).

The estimation procedure requires to read the G-code that defines the printing path. For each interpolation segment, the method compares and chooses the lowest speed between the programmed and predicted one, and computes the required time to travel the segment length at that speed.

It is noteworthy that the proposed method was successfully applied to three low-cost FFF printers. In the experimental tests accomplished (six random paths and twelve printed prisms), LP and CP estimations provide the minimum errors. In many cases, these errors are well below those provided by theoretical and Pronterface (analytical) estimations. Moreover, for all tested printers, while the theoretical and Pronterface estimation errors show high dispersion, CP and LP errors hardly change. Hence, the infill pattern and the component shape and size do not modify the accuracy of the proposed approach.

LP and CP surfaces are defined for a specific maintenance state of the printers, and it stands to reason that wears or maintenance problems could increase time estimation error. Further effort is required to study this fact, which could help to find out when to start maintenance tasks in a FFF machine.

**Author Contributions:** Conceptualization, G.M.-S., R.D.-V. and E.T.-J; methodology, G.M.-S. and R.D.-V.; software, G.M.-S. and R.D.-V.; validation, G.M.-S. and R.D.-V.; formal analysis, G.M.-S. and R.D.-V.; investigation, G.M.-S., R.D.-V., and E.T.-J.; resources, G.M.-S., R.D.-V., E.T.-J., and R.L.-G.; data curation, G.M.-S. and R.D.-V.; writing—original draft preparation, G.M.-S., R.D.-V., E.T.-J., and R.L.-G.; writing—review and editing, G.M.-S., R.D.-V., and E.T.-J.; visualization, G.M.-S., R.D.-V., and E.T.-J.; supervision, R.D.-V. and R.L.-G.; project administration, R.D.-V. and R.L.-G; funding acquisition, G.M.-S., R.D.-V., E.T.-J., and R.L.-G.

**Funding:** This research received no external funding.

**Acknowledgments:** This work is supported by the Spanish Ministerio de Economía, Industria y Competitividad, under research grant DPI2015-65472-R.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Advances in Orthotic and Prosthetic Manufacturing: A Technology Review**

**Jorge Barrios-Muriel † , Francisco Romero-Sánchez \*,† , Francisco Javier Alonso-Sánchez † and David Rodríguez Salgado †**

Department of Mechanical Engineering, Energy and Materials, University of Extremadura, 06006 Badajoz, Spain; jorgebarrios@unex.es (J.B.-M.); fjas@unex.es (F.J.A.-S.); drs@unex.es (D.R.S.)

**\*** Correspondence: fromsan@unex.es; Tel.: +34-924-289-600

† Current address: Escuela de Ingenierías Industriales, Universidad de Extremadura, Avda. de Elvas s/n, 06006 Badajoz, Spain.

Received: 31 October 2019; Accepted: 31 December 2019 ; Published: 9 January 2020

**Abstract:** In this work, the recent advances for rapid prototyping in the orthoprosthetic industry are presented. Specifically, the manufacturing process of orthoprosthetic aids are analysed, as thier use is widely extended in orthopedic surgery. These devices are devoted to either correct posture or movement (orthosis) or to substitute a body segment (prosthesis) while maintaining functionality. The manufacturing process is traditionally mainly hand-crafted: The subject's morphology is taken by means of plaster molds, and the manufacture is performed individually, by adjusting the prototype over the subject. This industry has incorporated computer aided design (CAD), computed aided engineering (CAE) and computed aided manufacturing (CAM) tools; however, the true revolution is the result of the application of rapid prototyping technologies (RPT). Techniques such as fused deposition modelling (FDM), selective laser sintering (SLS), laminated object manufacturing (LOM), and 3D printing (3DP) are some examples of the available methodologies in the manufacturing industry that, step by step, are being included in the rehabilitation engineering market—an engineering field with growth and prospects in the coming years. In this work we analyse different methodologies for additive manufacturing along with the principal methods for collecting 3D body shapes and their application in the manufacturing of functional devices for rehabilitation purposes such as splints, ankle-foot orthoses, or arm prostheses.

**Keywords:** rapid prototyping; additive manufacturing; orthoses; prostheses; fused deposition modeling; laminated object manufacturing; selective laser sintering

#### **1. Introduction**

Assistive technologies, such as orthotic or prosthetic devices, have existed for many centuries. Orthotic devices have been widely used not only to provide immobilization, support, correction, or protection, but also to treat musculoskeletal injuries or dysfunctions [1]. In the 1970s, new techniques like plastic coating were developed, due to the demand of orthotic devices with a more attractive appearance, by applying a tinted rubber-based plastic film [2], allowing the improvement of orthoses appearance and comfort. In the early 1980s the rise of additive manufacturing technologies (AMT), popularly known as 3D printing technologies in a manufacturing environment, with the introduction of the stereolithography technique, based on the cure of photopolymer resin in thin layers with a UV laser allowed construction of 3D models. In the following years, other AMT were introduced, such as: fused deposition modeling (FDM), laminated object manufacturing (LOM), selective laser sintering (SLS), 3D printing, and variable rapid prototyping (Polyjet Technology), among others.

The AMTs are included in the field of rapid prototyping techniques (RPT), producing fully functional parts directly from a three-dimensional model without a machining process. A radical

change in manufacturing of orthotic devices is already happening due to the exponential growth of RPT over recent decades [3,4]. A quick search in Scopus of 3D printing provides only 122 results before year 2000, 303 results between 2001 and 2005, 756 results from 2006 and 2010, 4521 results for the period 2011–2015, and 22,513 results in the last five years. In the biomedical engineering context, the developments have evolved quickly due to the need for individualized devices able to adapt properly to the patient's anatomical shapes [5–9]. For this reason, RPT may be helpful in the orthoprosthetic industry, as these devices must adapt perfectly to the body, not only to accomplish their rehabilitative function, but to avoid disuse, as many of these devices produce blistering, ulcers, or discomfort [10,11]. These techniques have been already applied to the manufacturing of spinal braces [11,12], exoskeleton parts [13,14], and passive orthoses [15–17], and the application in the medical and dental industry represents one of largest serving industries in the world [18]. Moreover, RPT offer advantages in the design of custom orthotic devices (Figure 1): The orthotic and prosthetic devices are highly customizable, as in Zuniga et al. [19], it is possible to fit the devices to complex geometrical features, with high accuracy, and these devices are manufactured efficiently in terms of cost, lead-time, and product quality [20].

**Figure 1.** Examples of 3D printed orthotics (**a**) Forearm static fixation (courtesy of Fitzpatrick et al. [21]). (**b**) Cyborg beast hand prosthesis—a low-cost 3D-printed prosthetic hand for children licensed under the CC-BY-NC license (courtesy of Zuniga et al. [19]). (**c**) Spinal brace (courtesy of Andiamo company [22]). (**d**) Ankle-foot orthosis (courtesy of Andiamo company [22]).

Currently, most rehabilitation devices are designed and hand-crafted by orthopaedists. Therefore, the quality of the product depends on the specialists' skills and experience [23]. The manufacturing process requires time and depends on the expertise of the specialist to obtain products with functional features that match the unique gait dynamics of each subject. Thus, the need for custom-made products such as orthoses and assistive devices is an explicit need considering the evolution of the technology during the beginning of this century [6,24,25].

Regarding orthoprosthetic manufacturing, the first step is to acquire the morphology of the body segment. In the traditional manufacturing process, the subject's morphology is usually acquired

by means of foam or plaster moulds. A prototype is obtained by using a computerized numerical control (CNC) or a milling machine in the thermosetting polyurethane model obtained by the mould. Lastly, the specialist performs several modifications in the device to adjust it to the subject (Figure 2a). However, CNC and milling present some limitations as they are not able to reproduce complex surface designs or to deal with different thickness and materials.

On the contrary, in the new RPT approach, the manufacturing process (Figure 2b) begins with the acquisition of subject's morphology by means of 3D scanning technologies. Then, computer aided design (CAD)-computed aided engineering (CAE) tools are applied to obtain subject specific designs; whereas, functionality is studied by testing different materials and structures. Lastly, the design is easily exported to an additive manufacturing machine where the prototype is obtained. Manufacturing time may vary between several weeks in the traditional process, to a couple of days in the RPT approach. Therefore, the use of RPT, together with the new 3D acquisition methodologies, represents an alternative in the orthoprosthetic industry.

**Figure 2.** Phases of the manufacturing process of custom-fit orthotic devices. (**A**) Rapid prototyping techniques (RPT) methodology (courtesy of J. Barrios-Muriel). (**B**) Traditional methodology (courtesy of Mavroidis et al. [26] under CC-BY License.) Computer aided design (CAD)-computed aided engineering (CAE), computed tomography (CT).

Many other applications of additive manufacturing (AM) and RPT are in the field of manufacturing of medical instruments [27], drug delivery systems [28], engineered tissues [29,30], scaffolds for bone regeneration [31,32], dental implants [33,34], prosthetic sockets, [35] or surgery [36–38]. In this work, we present a review of the developments in the manufacturing of orthotic devices, especially those related to the use of RPT to improve the quality and manufacture times in the rehabilitation field, as in splints, ankle-foot orthoses, or arm prostheses. This review provides comprehensive coverage on the different methodologies ready to be used in the orthotic and prosthetic industry. New 3D data acquisition techniques and the use of different materials are also referred to. This work is intended to be a reference guide on the techniques in this field for practitioners, but also for experienced readers who are interested in pursuing further research.

#### **2. 3D Anatomical Data Acquisition Technologies**

Applications of RPT combined with different techniques for measuring and modelling the human body are useful to generate new criteria for the orthotic device design. Depending on the data acquisition method used, the data can be expressed as a point cloud, voxels (3D volumetric pixels), or three dimensional coordinates of different anatomical points. Up to date, there is no standardized morphology acquisition procedure, however, there are several acquisition methods to support fabrication using RPT within the field of orthotic devices modelling, including computer tomography, 3D scanning, and different optical motion capture systems.

#### *2.1. Computed Tomography*

Computed tomography (CT) is a powerful technique to facilitate diagnostics and for surgical planning. Traditionally, the recorded images were in the axial or transverse plane. Currently, modern scanners record images along different planes, enabling volumetric reconstructions for 3D representations. Several studies have applied CT for the manufacture of orthotic devices. For example, Tang et al. [39] recently proposed the use of CT combined with AM techniques to manufacture insoles for diabetes. In their work, they studied pressure and tissue strain along the plantar foot to correlate these variables with the therapeutic effect of footwear and custom-made orthotic inserts, being able to reduce peak plantar pressure by 33.67%. Artioli et al. [40] studied the use of different acquisition techniques to manufacture 3D printed silicone ear prostheses, concluding that the use of CT and AM (using polylactic acid or polylactide, PLA, resolution 100 μm) produce differences of 0.1% between the manufactured prosthesis and the objective model. Liacouras et al. [41,42] used CT to acquire the morphology of the patients stump and to develop the strategies for designing transtibial prosthetic sockets. Moreover, the data of CT allowed a finite element analysis of the prosthetic model socket to calculate the structural stresses and strain at the sockets, as well as the contact pressure at the fibula head. The high image resolution between tissues is one of the greatest advantages of CT, along with the capacity to improve contrast and reduce noise. However, several drawbacks are worthy of mentioning. Radiation is the main concern and the exposition is directly proportional to the duration of the scanning. Other drawbacks are the partial pixel effect, leading to a blurred boundary, as the different densities share common pixels [43].

#### *2.2. 3D Scanning*

To capture human topography or the external shape, 3D scanning arise as the most practical and comfortable solution. 3D scanning systems use light based techniques to determine the three-dimensional position in space of the different points that integrate the surface of an object. Computer software is then used to reconstruct surfaces from the point cloud and then, the CAD model is obtained.

Currently, 3D scanners for human measurement are available, including the use of single image for reconstruction, structured light technologies, lasers, and different algorithms for stereo reconstruction [44–46]. The most common technologies used to reconstruct human body shape are laser and structured light technologies [44]. The laser technique uses a projected laser dot or line from a hand- held device. A sensor measures the distance to the surface, typically a charge-coupled device or a position sensitive device. For static objects, data is collected in relation to an internal coordinate system and, for dynamic conditions, the position of the scanner must be determined to correctly define the point cloud [47]. Structured light methods use a projector-camera system with pre-defined light patterns projected on the moving object. However, a drawback of this technology is the inability to capture certain topography sections of human anatomy with intricate creases and folds, such as between fingers when the hand is in a neutral position, the back of the knee when flexed, or the armpits. The gathered information is more precise, however, and noise is reduced. Recent techniques explore the feasibility of 4D acquisition [48], but to the best knowledge of the authors, there is no report on its use for the design of orthotic and prosthetic devices yet.

Processing time is significantly reduced compared to magnetic resonance imaging (MRI) and CT, as well as the size of the data files [44]. The use of MRI and CT is mostly used to reconstruct of internal organs or tumours with high accuracy for surgical guidance [49]. Recording time and resolution may vary between different 3D scanners, ranging from 3–5 min and a tenth of millimetres for high accuracy systems to a couple of minutes [50] and millimetres for low cost systems [51]. Other advantages of 3D scanning methods are affordable hardware and software, minimal training requirement, availability, accessibility, and efficiency [44].

Several authors suggest the use of reverse engineering software to obtain a refined model by repairing the data. In the pioneering work of Chee Kai et al. [52], a 3D scanning method was selected

over the traditional methods for prosthesis modeling, such as plaster-of-Paris impressions, MRI, and CT. Mavroidis et al. [26] used 3D laser scanning to create patient-specific foot orthoses. Surface data of the patient anatomy was manipulated to an optimal form using computer aided design (CAD) software and was fabricated using a rapid prototyping machine. The prototype properly fit the subject's anatomy compared to a commercial ankle foot orthosis. Paterson [53] investigated the 3D anatomical data acquisition methods to establish a clinically valid, standardized method. He concluded that laser scanning appears to be the most suitable method to reduce the acquisition of ambiguous data and with a high performance in terms of cost, resolution, speed, accuracy, patient safety, cost, and overall efficiency. More recent works, such as those of Mali and Vasistha [54] or Agudelo-Ardila et al. [55], present efficient solutions for the manufacturing of lower and upper limb orthoses, respectively, using reverse engineering.

#### *2.3. Optical Motion Capture System*

Recently, techniques to measure the topography of the human body in dynamic movements are receiving attention, as the design of orthotic devices should not be designed only for static conditions, as most of them will be used in dynamic conditions to increase rehabilitation. As previously mentioned, there are many commercial 3D systems that are able to measure 3D shapes with high accuracy; however, most of them cannot acquire human motion. An optical motion capture system is a popular technology to capture human movement. These optical systems use several cameras recording in 2D to reconstruct the 3D position of a set of reflective markers placed in anatomical landmarks. These markers should be seen by two or more cameras calibrated to provide overlapped projections. However, the optical motion capture system has a drawback due to the strong limitations related to the number and density of markers [56]. Although only three markers are needed to reconstruct each body segment as a rigid body (e.g., to perform a kinematic analysis), this number is increased if body shape must be also retrieved. The number usually depends on the camera resolution but it does not exceed 60–80 markers per body segment. The main application of the marker-based acquisition technology is in the assessment of the manufactured devices due to the standardized protocols to acquire kinematics [57].

Research has also been focused into structured light using this method [58]. Unkovskiy et al. [59] used a portable structured light scanner to retrieve the topology of the nose cavity and the face to design and manufacture a prosthetic nose. A portable projector was used to project in the region of interest an arbitrary light pattern with a colour code. The system performs the triangulation between the projection pattern and the camera image and to retrieve the correspondence between images. The advantage of this system is the noise reduction compared to the video capture image. The matching of the stereo projection pattern and the camera recorded image is less affected by noise than multiple stereo matching of camera images. Regretfully, synchronous measurement of the entire body shape using multiple projector-camera systems has not been reported yet using this technology. This is essential for capturing the 3D shape motion and to analyse, in the case of gait for example, foot width, length, circumferences, and arch changes during movement [60]. However, there is no commercial system including this technology and thus, more studies should be performed with different methods to compare accuracy and precision between these technologies. The cutting edge technologies in this field are the markerless systems. Chatzitofis et al. [61] proposed recently a low-cost robust and fast system to acquire body kinetics. Although this work still uses reflective straps it could be combined with 4D scanning systems, such as those proposed by Joo et al. [62], to obtain accurate topology and motion of the body segment to rehabilitate.

#### **3. Rapid Prototyping Technologies for Orthotic Devices**

The combination of CAD and computed aided manufacturing (CAM) is a well-known approach that is receiving increasing attention in the field of orthotic devices to replace the traditional craft practices. As stated by Ciobanu et al. [63], customized manufacturing through RPT requires: 3D scanning of the anatomic surface, 3D surface reconstruction, CAD modeling, conversion to

stereolithography format (STL) and, lastly, machining using a special rapid prototyping machine (i.e., a 3D printer) controlled through computer. RPT offer advantages in manufacturing processes of custom-fit orthotic devices, in terms of greater design freedom, ability to create functional elements, superior accuracy and cost efficiency, shorter delivery time, and better user experience of the final product.

In an RPT manufacturing process, a representative virtual 3D CAD model is formed layer upon layer to form a physical object [9]. In RPT, a virtual model of the part is designed through CAD and is converted to a STL file format, which is the default standard file format for RP systems [64,65]. AMT can be categorized in different ways depending on the nature of the fabrication process, such as laser, printer technology, and extrusion technology [7]. There are many different AM processes. Kruth [66] proposed in the early 90's the use of different methodologies for additive manufacturing, classified according to the material used for the prototype (see Table 1). Nevertheless, Paterson et al. [67] demonstrated that only a few of them could be used to manufacture orthotic and prosthetic devices.

Liquid based process, such as stereolitography (SLA), solid ground curing (SGC), UV light curing (ULC), and ballistic particle manufacturing (BPM); or solid based process, such as laminated object manufacturing (LOM) are methodologies used in the manufacturing of orthoses and prostheses. Nevertheless, the most used methodologies to manufacture orthotic and prosthetic devices are fused deposition modelling (FDM), selective laser sintering (SLS), and powder bed and inkjet head 3D printing (3DP). These methodologies represent an optimized trade-off between cost, delivery time, accuracy, and comfort.


**Table 1.** Rapid prototyping techniques available for orthoprosthetics.

#### *3.1. Fused Deposition Modeling (FDM)*

In the FDM process (see Figure 3a), a semi-molten material is extruded through an extrusion head that traverses in the *X* and *Y* axes to create each two dimensional layer of the piece to be manufactured. Two extrusion nozzles compose the movable extrusion head: one to deposit the build material and the other one that contains the support material [68].

In general, the perimeter of each layer is extruded first and then the delimited zone by the previous extrusion is filled by the extruder head by following a pre-defined pattern [68]. Once the layer is completed, the support platform lowers and another layer is extruded. The process continues layer by layer until the piece is complete.

The most common materials for FDM are polycarbonate (PC) and Acrylonitrile butadiene styrene (ABS) or a mixture of them. These materials have similar properties to thermoplastic material for injection moulding [69]. Other materials such as polymers or nylon-based materials may be used. The main advantage of FDM technology is in the use of low-cost materials. Tan et al. [70] were pioneers in the use of used FDM for tibial prosthesis manufacturing and concluded that the functional characteristics of prosthesis were valid for clinical purposes. On the contrary, manufacturing times are high. Since then, an increasing number of applications have arisen for FDM in the biomedical field for

upper [17] and lower limb orthoses [50,71], hand prostheses [19], facial prosthesis [40,72], and drug delivery systems [73].

**Figure 3.** Comparison of the proposed schemes dor rapid prototyping. (**a**) Fused deposition modeling (FDM). (**b**) Selective laser sintering (SLS). (**c**) 3DP. Image adapted from Wang et al. [74] with permission of Elsevier Ltd.

#### *3.2. Selective Laser Sintering (SLS)*

The DTM Corporation (now a part of 3D Systems) introduced the first SLS system in the 90s [75]. The SLS technique creates three-dimensional solid objects or parts by selectively fusing powdered polymer-based materials such as nylon/polyamide with a CO2 laser, turning powder material into solid objects (Figure 3b). A CO2 laser selectively sinters defined regions by traversing across the powder bed in the *X* and *Y*-axes to form a 2D profile [76]. Once the 2D profile has been completed, the platform lowers, a new layer of powder is distributed and the sintering process is repeated. Subsequently, the process is termed a powder-based fusion process. In general, all materials used are thermoplastics, the most common being polyamide 12 (PA), acrylonitrile butadiene styrene (ABS), and polycarbonate (PC) [77]. These materials lead to a considerable weight reduction improving the usability of the rehabilitation devices.

As an example of the use of this technology in the manufacturing of custom orthoses, Schrank and Stanhope [10] evaluated the accuracy of SLS manufacturing process of ankle foot orthoses (AFO). In this work the discrepancy between the CAD model and final product manufactured with SLS was measured through the Faroarm 3D scanner (accuracy ±25 μm). The results showed values below 1.5 mm (SD = 0.39 mm). Deckers et al. [78] developed and tested an SLS-based AFO, highlighting the need to properly characterize the mechanical characteristics of the AFO such as strength, fatigue, and resistance to impacts. Vasiliauskaite et al. [79] tested a polyamide-based orthoses manufactured with SLS and concluded that the features were similar to a thermoformed polypropylene orthosis, the first one being stiffer than the second but enough for the purpose of rehabilitation.

On the other hand, [80] manufactured a splint for upper extremities using the SLS method. They also concluded that the results of this manufacturing technique were good but did not make

any clinical validation of the device. Another similar application of splints using SLS is the design proposed by Evill [81]. In this work, several aspects such as ventilation, hygiene, and aesthetics were improved through CAD. Although there are not conclusive results that confirm these purposes, the new design parameters considerably improve comfort.

#### *3.3. Powder Bed and Inkjet Head 3D Printing: 3DP*

3DP refers to three dimensional printing, which is understood as the process in which the manufactured product is made by means of powder layers stuck with adhesive. In this process, first, a powder layer is spread on the build platform. Second, a liquid binder is deposited selectively through an inkjet printhead by following a patterned layer in the XY plane. Once the 2D pattern is formed, the platform lowers, the next powder layer is spread and so on. This process is sometimes referred as powder bed and inkjet head 3D printing (or 3DP). It should not be confused with the widespread definition of 3D printing, that involves all additive manufacturing process that result in the manufacturing of tree-dimensional objects. The 3DP process is somewhat similar to SLS (see Figure 3b,c): In 3DP, a printing head places liquid adhesive in the material; whereas, in SLS a CO2 laser is used to fuse the layers. The accuracy of this process is lower than in SLS; nonetheless, this method is preferred due to its low cost and quickness. These qualities have lead 3DP to have a predominant role in the prototyping industry.

The employed materials (mainly thermoplastics as ABS) have the required properties to be used in orthotic and prosthetic applications. Herbert et al. [82] investigated whether this technology was suitable to produce functional prosthesis, and they suggested that, although the manufacturing levels were limited, patients felt more comfortable with prostheses made with 3DP machine (Corporation Z402) than the traditional handmade ones. Regretfully, the resistance was not studied in that work and therefore the durability of the product is unknown. Saijo et al. [83] used this technology to develop patient-specific maxilofacial implants reporting a reduction in operation times. As stated by Ventola [84], 3DP is particularly interesting in tissue engineering and regenerative medicine because of its digital precision, control, and versatility.

To summarize, Table 2 shows a comparison of the different RPT by using the commercial models used in the works described in the bibliography.


**Table 2.** Characteristics of the most used machines for AMT.

#### **4. Variable Property Rapid Prototyping**

Variable rapid prototyping differs from other AMT as it aims at producing objects of varied properties. In this sense, each material used to manufacture the object provides specific values of strength, strain, heat deflection temperature, etc. [85]. Object geometries recently introduced 3D printers that use polyjet MatrixTM technology to allow the generation of composite material prototypes of varying stiffness and dual material prototypes. The Connex500TM 3D printer operates by using inkjet heads with two or more photopolymer materials. The material is extruded in 16 μm thick layers. Each photopolymer layer is cured by UV light immediately after the extrusion [86].

A carpal skin was produced by Oxman [80], exploring the multiple material building capabilities available with this technology. Campbell et al. [87] explored the benefits of multiple materials integrated into a wrist splint compared to traditional custom-fitted wrist splints of qualified and experienced clinicians. The work focused on the attempt to place multiple materials to behave as hinges or cushioned features as opposed to traditional fabrication processes where a similar approach would be very difficult to replicate. A drawback for this technology is that the actual commercial CAD software is not efficient to apply the design potential and few computation tools manage the physical interaction between material properties. Nevertheless, recent advances in the use of additive manufacturing of hybrid composites, as recently presented for dental implants [88], may lead to a substantial revolution in the field of orthotics, were hybrid exoskeletons are currently changing from rigid structures to wearable garments.

#### **5. Material Selection for Orthotic Devices**

The choice of material when designing an orthotic device is vital to its success. Physical properties of the orthotic materials include their elasticity, hardness, density, response to temperature, durability, flexibility, compressibility, and resilience [89]. It should be mentioned that each physical property cannot be used alone as a single factor for assessing materials for orthotic devices. A hard material as well as an incorrect aspect in the design may result in an uncomfortable device or a biomechanically detrimental orthotic device. Thermoplastics, composites, and foams are the main materials used to manufacture orthotic devices through AMT [89].

The most known materials used in rapid prototyping manufacturing are ABS (Acrylonitrile butadiene styrene) and PLA (polylactic acid) [90]. ABS is a polymer commonly used to produce car bumpers due to its toughness and strength. PLA is a biodegradable thermoplastic that has been derived from renewable resources such as starch prepared from the grains of corn. These materials are used for the majority 3D printing machines. Rigid and semi-rigid structures can be manufactured with these materials. Depending on the designed thickness, these materials may have different properties.

Soft parts and some semi-rigid parts of orthotic devices are made with foamed materials, usually with open or closed cell structure. The first type allows the movement of gas between the cells whereas the second encloses the gas within the cell walls allowing for a water-tight material. This is desirable for orthotic manufacture as sweat will not be able to penetrate into the material to cause premature degradation. Paton et al. [91] investigated the physical properties of soft materials used to fabricate orthoses designed for the prevention of neuropathic diabetic foot ulcers. They concluded that the most clinically desirable dampening materials tested were Poron®96 and Poron®4000 (thickness of 6 mm) and the material with the best properties for motion control was ethylene vinyl acetate (EVA). With the evolution of AMT, insoles with variable porous structure and adjustable elastic modulus are being manufactured to adapt to the different needs of patients with diabetes [92].

In the SLS manufacture process, polymer powders and ceramics are mixed to form composites. RilsanTM D80 DuraFormTM PA and DuraFormTM GF are examples of materials used in SLS techniques. Faustini et al. [93] explored the feasibility of using RPT for AFO manufacturing process. The study determined that the optimal SLS material for AFO to store and release elastic energy was RilsanTM D80, considering minimizing energy dissipation through internal friction is a desired material characteristic. A relatively recent work from Walbran et al. [94] compares yield stress of custom made orthoses made by SLS of nylon with carbon fibres and FDM with PLA. Although PLA was chosen in terms of mechanical properties and cost, they concluded that the AMT have a strong potential not only to obtain consistent and repeatable models of the subjects affected limb independently of the operator's skill and restraint of the subject, but to further automate the AFO design process.

Materials used in FDM are those with similar properties to thermoplastic materials for injection moulding. In general polycarbonate (PC) and ABS or their combinations, as mentioned above (PC-ABS, PC-ISO or ABSi), nylon-based materials and other polymers can be used. The main advantage of these materials is the low cost. Finally, in the case of variable properties rapid prototyping base materials, digital materials or composite digital materials can be used. The material properties can be modified by combinations and distribution of different types.

#### **6. Discussion**

The traditional manufacturing processes for orthotic and prosthetic devices is still mostly hand-crafted and requires special abilities from the othopaedist to obtain a quality product. Nevertheless, this manufacturing process, in general, produces discomfort to the patient. The acquisition of the morphology of the subject is not a clean process as the use of plaster is required to obtain the mould. Additionally, the final product may produce blistering on the subject's skin, as the morphology is acquired in static conditions.

Alternatively, RPT and AMT have a strong potential to change not only the way in which orthotic and prosthetic products are designed, but also the manufacturing process and the specialist profile. The use of RPT in the orthoprosthetic industry may suppose a considerable change in the know-how; however, it also leads to important benefits. The goal is to accelerate the reconstruction process of 3D anatomical models and biomedical objects for the design and manufacturing of medical products and simulate 3D body shape to design the most suitable orthoses for the patient. Thus, the use of CAD and RPT facilitate the design and fabrication of custom-fit orthotic products with a number of advantages over traditional methods: the use of new materials, customized designs, virtual testing, etc.

The application of these technologies may lead to a significant improvement in the orthotic manufacturing process as production times are lower, morphology acquisition is faster and more pleasant for the patient, as plaster moulds are suppressed and manufacturing errors are minimized. Considerable effort has been applied in the application of AMT to the medical industry and specifically in the design and manufacturing of orthoses and prostheses for rehabilitation purposes [89], to mitigate the effects of aging [95], in the design of active wearable exoskeletons [96], and also to bring this technology closer to the general public [19,97].

The inclusion of these manufacturing methods requires a high investment in equipment, materials, and training that may cause hesitation from investors or orthotic and prosthetic technicians. Moreover, healthcare specialists show some reservations to change their own work routines [86]. Nevertheless, the eruption of 3D printers into the market and the continuous improvements made in this field will give an impulse to the implementation of this technology in the orthoprosthetic industry. In addition, the experience shown in other medical fields, such as in dental implant manufacturing will make possible the implementation of this technology to produce orthoses and prosthesis to reduce waiting lists.

The mass production of these aids must include a series of key points leading to high quality patterns in the manufacturing process. These key points are: acquisition of the subject's anthropometric data, product design, material selection, manufacturing process planning, and product service, among others. The correct application of these steps reduces the total production time and, therefore, the delivery times. Thus, the inclusion of RPT to produce high quality and short delivery time orthotic and prosthetic devices, that also satisfies functionality and patients comfort is now a reality. Finally, the inclusion of manufacturing technology in a traditional environment such as the orthoprosthetic industry will lead to better products to satisfy the specifications required in rehabilitation.

#### **7. Conclusions**

In this work, a review of the different RPT applied to the orthoprosthetic industry has been presented. Specifically, the manufacturing process to manufacture orthoses and prostheses have been analysed and the main works in this field have also been presented. These techniques have been shown to have an exponential growth in the following years in the biomedical field. The new advances in the subject's morphology acquisition as well as the use of RPT can improve the accuracy of the final device, leading to a better rehabilitation process. RPT will help us to optimize the manufacturing process and improve both the design and functionality of assistive devices. Thus, RPT combined with CAD-CAM tools provide a major control in the design and manufacture processes. Finally, the future lines of development in this field will be based on the design of new structures and materials to improve comfort, which will grant the success of the new orthoprosthetic aids.

**Author Contributions:** Conceptualization, J.B.-M. and F.R.-S.; Formal analysis, J.B.-M.; Funding acquisition, F.J.A.-S. and D.R.S.; Methodology, J.B.-M. and F.R.-S.; Project administration, F.J.A.-S. and D.R.S.; Supervision, F.R.-S., F.J.A.-S. and D.R.S.; Writing—original draft, J.B.-M.; Writing—review & editing, F.R.-S., F.J.A.-S. and D.R.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Consejería de Economía e Infraestructuras de la Junta de Extremadura and the European Regional Development Fund "Una manera de hacer Europa" under project IB18103.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Considerations on the Applicability of Test Methods for Mechanical Characterization of Materials Manufactured by FDM**

#### **Amabel García-Domínguez, Juan Claver, Ana María Camacho \* and Miguel A. Sebastián**

Department of Manufacturing Engineering, Universidad Nacional de Educación a Distancia (UNED), 28040 Madrid, Spain; agarcia5250@alumno.uned.es (A.G.-D.); jclaver@ind.uned.es (J.C.); msebastian@ind.uned.es (M.A.S.)

**\*** Correspondence: amcamacho@ind.uned.es; Tel.: +34-913-988-660

Received: 15 November 2019; Accepted: 17 December 2019; Published: 19 December 2019

**Abstract:** The lack of specific standards for characterization of materials manufactured by Fused Deposition Modelling (FDM) makes the assessment of the applicability of the test methods available and the analysis of their limitations necessary; depending on the definition of the most appropriate specimens on the kind of part we want to produce or the purpose of the data we want to obtain from the tests. In this work, the Spanish standard UNE 116005:2012 and international standard ASTM D638–14:2014 have been used to characterize mechanically FDM samples with solid infill considering two build orientations. Tests performed according to the specific standard for additive manufacturing UNE 116005:2012 present a much better repeatability than the ones according to the general test standard ASTM D638–14, which makes the standard UNE more appropriate for comparison of different materials. Orientation on-edge provides higher strength to the parts obtained by FDM, which is coherent with the arrangement of the filaments in each layer for each orientation. Comparison with non-solid specimens shows that the increase of strength due to the infill is not in the same proportion to the percentage of infill. The values of strain to break for the samples with solid infill presents a much higher deformation before fracture.

**Keywords:** additive manufacturing; FDM; ABS; anisotropy; infill density; layer orientation; ASTM D638–14:2014; ISO 527–2:2012

#### **1. Introduction**

Currently, the characterization of parts obtained by additive manufacturing (AM) technologies is a very prolific research field. This fact is a consequence, not only of the significant increase of the presence and the importance of additive technologies in a wide range of industries, but it is also due to the lack of specific standardization, as claimed in previous works such as the one by Rodríguez-Panes et al. [1] and in the review made by Popescu et al. [2], who specifically demanded that test standards for Fused Deposition Modelling (FDM) parts should be developed, including the definition of printing parameters such as layer thickness, perimeters, and raster dimensions. Thus, the most commonly applied international standards for tensile tests are ISO 527–2 [3] and ASTM D638–14 [4], but, in both cases, the guidelines provided respond to the characteristics of plastic injection parts, which represents a very different start context. Contrary to the continuity and isotropy of injected plastics, the layer by layer structure of the pieces obtained by additive manufacturing is discontinuous and anisotropic, so the polymeric nature of the material is the only common feature for both technologies. It is also worth mentioning that, while ISO 527–2 and ASTM D638–14 are identified as the consultation documents for tensile tests by the main standards focused on additive manufacturing technologies, as ISO 17296–3:2014 [5] and ISO/ASTM 52921:2013 [6] are, the Spanish standard UNE 116005:2012 [7]

has also been considered; this standard is only focused on tensile tests and polymer materials, and it is based on ISO 527–2. Despite both standards being similar, UNE 116005:2012 is developed as a specific standard for additive manufacturing, and it is not a standard recovered from other productive sectors. The national standards must follow the guidelines of the international ones and act as complementary information sources to guide researchers and professionals in specific tasks of applicative character, as in this case, with performing mechanical tests for additive manufacturing purposes.

The experiences developed by researchers when applying these standards in additive contexts have a key role, since the obtained results help to clarify the applicability or not of these standards in different additive scenarios, as well as to provide action guidelines. A significant variety of approaches have been faced in order to determinate the influence of manufacturing parameters on the mechanical response of additive products; for example, ABS specimens for unmanned aerial systems produced by FDM were characterized under tension [8] and under compression [9]; Banjanin et al. [10] evaluated the mechanical response of specimens of polylactide (PLA) and acrylonitrile styrene butadiene (ABS), concluding that the results in ABS showed a higher repeatability under tension than under compression, and vice versa. Chacón et al. [11] assessed the effect of build orientation, layer thickness, and feed rate on the mechanical performance of PLA specimens, stating that, for optimal mechanical performance, low layer thickness and high feed rate values are recommended. In the work of Tanikella et al. [12], a set of seven materials has been tested for determination of tensile strength of FDM specimens fabricated with an open-source 3D printer, concluding that this property highly depends on the mass of the specimen. The influence of the type of infill pattern and density, printing temperature, among other parameters on selected mechanical properties of PLA and ABS samples, were evaluated by Cwikła et al. [ ´ 13], showing that, if the maximum strength is the priority, shell thickness should be increased. In the work of Kuznetsov et al. [14], the influence of geometric process parameters on the strength of the sample was also evaluated proposing a new methodology, and concluding that the best combinations of printing parameters allows for obtaining interlayer cohesion close to the one of the feedstock material. Rajpurohit and Dave found a close relationship between the raster angle and failure mode [15]. In addition, Zaldivar et al. [16] investigated the effect of the build orientation on the mechanical properties of a polyetherimide thermoplastic blend as material, concluding that FDM materials behave more like composite structures than isotropic cast resins and that designs should define the build configurations allowable.

However, sometimes, a complementary and previous question is needed because the kind of part we want to produce or the purpose of the data we want to obtain from the tests can guide the definition of the most appropriate specimens. Figure 1 tries to express this idea graphically. A basic geometry usually obtained as a continuous solid by plastic injection (Figure 1a) can be materialized in very different ways when considering additive manufacturing. This includes:


On the other hand, for more complex geometries that are not possible to manufacture with plastic injection processes but are possible in additive scenarios (Figure 1b), these different approaches could be chosen as alternatives for the materialization of the parts which traditionally would be understood as solid. The differences between the internal material continuity that the selection of one or another of these approaches introduces are significant. Thus, this selection implies different ways of understanding the material since, although in every case mentioned before the polymer used can be the same, the behavior of the piece is highly conditioned by internal morphology of those parts traditionally understood as solid.

**Figure 1.** Applicability of conventional test methods for material characterization of parts obtained by additive manufacturing techniques considering their internal morphology: (**a**) basic geometry suitable both for traditional and additive manufacturing technologies. (**b**) complex geometry suitable for additive manufacturing technologies.

As exposed before, from the point of view of the applicability of the mentioned standards, only a continuous material could be characterized. However, it is interesting to reflect the different levels of applicability that the selection of each approach introduces in the context of additive processes. In that sense, the applicable criterion is the continuity or not of the internal morphology of the piece. Thus, a solid layer by layer structure would represent the closest conditions to the ones considered in these standards. In the case of predefined infill patterns, these standards are not really applicable, but, in practice, many studies use this kind of specimens and analyze their behavior by applying them. Finally, cellular structures are similar to these infill patterns in terms of discontinuity, but this group includes lattice structures and approaches in which the sizing of some parts or elements can change along the piece and those fluctuations provide graded densities.

This way, the design of each case study makes each approach more or less appropriate, and it is essential to take it into account. In that sense, the methodology optimization developed by the authors in a previous work [17] and its data needs can be a good example. The mentioned methodology uses optimized lattice structures for the infill areas where each bar is sized according to the stress it can withstand [18]. In order to calculate the section needed in each case, values of the material strength are requested by the structure of the methodology. Since these bars have diameters between 1 and 2.5 mm, infill patterns or cellular structures cannot be implemented into them, so the layer by layer solid configuration is the only option. In addition, as each bar has a different section as a consequence of the different stress values calculated, the lattice structure cannot be considered as uniform along the piece. Thus, those approaches that estimate the mechanical behavior of cellular structures through the testing of specimens out of the standards and that contain a certain number of unit cells would not be of application in this particular case, as in the works of Chen et al. [19], who developed a finite element mesh based method for optimization of lattice structures for AM; Hussein [20], who studied the development of lightweight cellular structures for additive manufacturing with metal; Mahmoud and Elbestawi [21], where lattice structures were fabricated by additive manufacturing in orthopedic implants; Maliaris and Sarafis [22], where lattice structures were modelled using a generative algorithm; Panda [23] and Weeger et al [24], who worked in the design and development of cellular structures for

AM; or Vannutelli [25], where the mechanical behavior of lattice structures was analyzed. As Figure 1 exposes, a solid layer by layer configuration represents the most consistent alternative with the nature of this particular case study and the data that are needed. This way, the identified standards could be applied, always considering their limitations in additive contexts.

Accordingly, the aim of this work is to analyze the mechanical behavior of solid structures manufactured by FDM (as a typology of interest in different design contexts supported by additive technologies, as the ones described above) compared to non-solid ones, and to determine which standard provides better results for the mechanical characterization of materials manufactured by FDM, giving some guidelines in the application of them due to the lack of specific standards for characterization of materials manufactured by FDM. To achieve this goal, a series of tensile tests are carried out using solid specimens obtained by FDM, and according to the specifications of UNE 116005:2012 (based in ISO 527–2) and ASTM D638–14. The obtained results are interpreted together with the layer structures observed in the specimens after fracture.

#### **2. Materials and Methods**

#### *2.1. Work Methodology*

According to the aim of this work, and in order to have a clear view of the research methodology followed in this work, Figure 2 shows the approach definition based in the different situations presented in Figure 1, and a summary of the main steps of the methodology. Four main steps are established, and then basic information related to the particular context of this work is indicated for each one of these stages. As shown in the work diagram, at the end of the process, the obtained results and the conclusions derived from them must be analyzed in order to assess whether the results fit the objectives of the study. The main aspects of the work developed in these stages are exposed in the following sections. Afterwards, the results are compared with the ones obtained by the authors in a previous work in which the specimens used were fabricated using infill patterns [26].

**Figure 2.** Approach definition based in Figure 1 and sketch of the work methodology.

#### *2.2. Materials and Equipment*

The material used in this work is ABS Pro of the commercial house BCN3D (Castelldefels, Spain), whose main physical and mechanical properties are presented in Table 1. The FDM printer to be used is a dual extruder BCN3D R19 printer (BCN3D, Castelldefels, Spain), presented in Figure 3a, whose main technical characteristics are shown in Table 2. The software Cura (Ultimaker, Geldermalsen, The Netherlands) is typically used to export the three-dimensional models of the specimens to G-code. In this paper, the version BCN3D Cura 2.1.4 adapted by BCN3D (Castelldefels, Spain) has been used. Figure 3b shows the equipment for the mechanical testing, a universal testing machine (model Hoytom HM–D 100 kN, Hoytom, S.L., Leioa, Spain).


**Table 1.** Main mechanical and physical properties of the filaments according to the manufacturer.

**Figure 3.** Equipment for additive manufacturing and testing of specimens: (**a**) a dual extruder BCN3D R19 printer; (**b**) detail of the working area in Hoytom HM-D 100kN Universal testing machine.

**Table 2.** Main technical characteristics of the Fused Deposition Modelling (FDM) printer BCN3D R19.


For analysis of the surface, a digital profile projector TESA-VISIO (TESA SA, Renens, Switzerland) and an optical micro-coordinate measurement machine IF-SL ALICONA (Bruker Alicona, Graz, Austria) are used (see Figure 4).

**Figure 4.** Equipment for surface analysis: (**a**); digital profile projector TESA-VISIO; (**b**) optical micro-coordinate measurement machine IF-SL ALICONA.

*2.3. Printing Parameters and Definition of a Work Plan*

The general manufacturing parameters for the FDM printing of the specimens are presented in Table 3.

**Table 3.** Manufacturing parameters with FDM printer BCN3D R19.


Due to the lack of specific standards for material characterization of FDM parts, as explained before, two different standards are going to be used for mechanical behavior testing: UNE 116005:2012 (based in ISO 527–2) [7] and ASTM D638-14:2014 [4]; the geometry of the specimens according to both standards is shown in Figure 5, where Type 1A and Type 1 are the geometries chosen for both standards, respectively.

**Figure 5.** Geometry of the specimens according to different international standards: (**a**) Type 1A: UNE 116005 [7]; (**b**) Type I: ASTM D638–14 [4].

As explained by Rodríguez-Panes et al. [26], the build orientation of the test specimen is one of the most influencing parameters on the mechanical properties of FDM parts. Figure 6a presents the three possible orientations, orientation 1 (flat) and 2 (on-edge) being the ones chosen in this work since orientation 3 is not appropriate for tensile testing due to the arrangement of the filaments perpendicular to the direction of the load. Figure 6b and c shows the two orientation of the samples used in this work.

**Figure 6.** (**a**) typical build orientation of the samples; (**b**) specimen printed with orientation 1; (**c**) specimen printed with orientation 2.

Five specimens for each orientation (1 and 2) are going to be tested according to either standard UNE 116005:2012 and ASTM D638–14:2014. A summary of the experimental work plan and nomenclature used in experiments is presented in Table 4.


**Table 4.** The experimental work plan and nomenclature of the specimens.

Results from the experimental testing of specimens in Table 4 are going to be compared to those ones obtained ii [26] in specimens with non-solid infill; the material of the experiments was ABS (PrintedDreams) in blue and the test procedure according to standard ASTM D638–14:2014 with type I general usage test specimen. The aim is to observe the influence of the infill in the characterization of materials obtained by FDM and for validation purposes. A summary of the cases for comparison is gathered in Table 5.


**Table 5.** Nomenclature of the specimens and printing parameters by Rodriguez-Panes et al. [26].

#### *2.4. Experimental Procedure for Tensile Testing*

The specimens in Table 4 are printed according to printing parameters in Table 3. Figure 7a shows the group of samples according to ASTM D638–14:2014. Each specimen is fixed by the grips as it is seen in Figure 7b, showing a specimen just before tensile testing.

**Figure 7.** (**a**) the test specimens manufactured according to ASTM D638–14; (**b**) specimen placement before tensile testing.

(**b**)

The tests are performed at a velocity of 5 mm/min until breakage. The mechanical properties to be obtained are: tensile strength at yield (σy), tensile strength at break (σU) and nominal strain at break (εt). Equations (1) to (3) show how these parameters can be obtained, as defined in the standard ASTM D638-14 [4]:

$$
\sigma\_y = \frac{F\_{\text{max}}}{A\_0} \tag{1}
$$

$$
\sigma\_{ll} = \frac{F\_{b\text{break}}}{A\_0} \tag{2}
$$

$$
\varepsilon\_{\rm f}(\%) = \frac{\Delta L}{L\_0} \times 100\tag{3}
$$

being:

*Fmax*: the maximum force sustained by the specimen, *Fbreak*: the force sustained by the specimen at breakage, *L0*: original grip separation, Δ*L*: extension (change in grip separation).

#### **3. Results and Discussion**

#### *3.1. Mechanical Behavior of Solid Specimens*

The specimens after tensile testing are presented in Figure 8. Most of them break in close to the radius of fillet close to the grips, which is common for FDM specimens [27], and it is due to stress concentrations at fillet areas [10], being the kind of fracture brittle, with no plastic deformation observed. As Wu et al. reported [28], craze is the main plastic deformation mechanism of ABS, and a great number of crazes are generated perpendicular to the load direction. This brittle behavior is typical in FDM parts, and it has been explained by other authors due to the presence of voids that help the crack initiation and propagation resulting in abrupt rupture [15]; the presence of voids will be analyzed further in Section 3.2.

**Figure 8.** Specimens after tensile testing: (**a**) according to UNE 116005:2012; (**b**) according to ASTM D638–14:2014.

Nominal stresses and strains are obtained for all the tests performed and presented in Figure 9 as stress–strain curves. Figure 9a presents the results according to the standard UNE 116005:2012 for printing orientations 1 and 2; and Figure 9b according to the standard ASTM D638–14, also for both orientations. A comparison of results with the intermediate values of each series of tests is also shown in Figure 9c.

**Figure 9.** Stress–strain curves. (**a**) according to UNE 116005:2012; (**b**) according to ASTM D638-14; (**c**) comparison of results with the intermediate values of each series of tests.

σ σ

The mechanical properties (nominal strain at break, tensile strength at yield, tensile strength at break,) obtained from the tensile tests are presented in Table 6.


**Table 6.** Mechanical properties and nomenclature of the group of tests.

Tests performed according to standard UNE 116005:2012 present a much better repeatability than the ones according to ASTM D638–14. This is a very important finding since the UNE standard is specifically designed for specimens fabricated by additive manufacturing and proves that the geometry used is more appropriate for characterizing materials obtained by FDM than the one used from international standards such as ASTM D638–14, very often used in the scientific literature for these purposes. Except specimen 3 PROB ABS BK 5, all the specimens experience a brittle breakage without plastic deformation. On the contrary, results for ASTM D638–14 show a higher variability, especially for orientation 1, where some specimens exhibit some plastic deformation before fracture (ASTM\_D638\_PROB1\_2, ASTM\_D638\_PROB1\_3 y ASTM\_D638\_PROB1\_4); elongations are particularly higher than the ones obtained by the standard UNE 116005:2012.

As a general trend, orientation 2 provides higher strength to the parts obtained by FDM according to results for both standards, but the influence of the orientation is more important in the case of the standard ASTM. This behavior is coherent with the orientation of the filaments in each layer for orientations 1 and 2, as shown in Figure 10. In this figure, the fracture surface of two of the specimens tested for each orientation (see designation in Figure 8a) is observed, showing the differences in the inner distribution of the filaments as well. These observations are in consonance with the work of Aliheidari et al. [29], who claimed that the nature of the layered structure, and, particularly, the adhesion between the layers, have a direct impact on the mechanical properties.

Observing the images in detail and placing the specimens in the direction of vertical growth of layer (Figures 11 and 12), it is possible to distinguish in both cases two lateral areas (shell) where the beads deposited during the printing have the same arrangement in all the layers, being parallel to the longitudinal axis of the specimen and a central area, in the middle of them, where the arrangement of the beads is different in alternate layers, being in one layer the longitudinal direction of the specimen, and in the above and below layers, the transverse direction. The layers in the longitudinal direction benefit from the alignment of polymer molecules along the stress axis [30].

**Figure 10.** View of the fracture surface of two specimens tested with orientation 1 (**left**) and orientation 2 (**right**) with digital profile projector TESA-VISIO.

**Figure 11.** Layout of the filaments related to the longitudinal stress direction for specimens with orientation 1. Identification of central area and shell.

**Figure 12.** Layout of the filaments related to the longitudinal stress direction for specimens with orientation 2. Identification of central area and shell.

The areas where the filaments are oriented in the direction of the load (shell) are expected to have a higher strength because the filaments contribute together to the strength of the specimen. On the contrary, the central areas present a different behavior: on one hand, there are layers where the filaments are in the same direction than the load, but there are also layers where the filaments are arranged perpendicular to the stress direction, so the cohesion between filaments is crucial to keeping the integrity of the specimen. This situation is analogous to a composite reinforced by continuous or discontinuous fibres, respectively [31]. In this sense, the percentage of section with higher strength provided by the lateral areas (shell), where the filaments are arranged in the same direction than the load, is clearly higher in specimens with orientation 2; this fact justifies their better performance under tension as presented in Figure 9 and Table 6.

In addition, Figures 11 and 12 allow for appreciating a pattern in the fracture of the specimens with both orientations. Figure 13 shows this situation with more detail. This fibre discontinuity has been reported by other authors [15] as the reason for premature failure of the part, resulting in brittle fracture.

**Figure 13.** Fracture planes in central areas in specimens with both orientations.

Throughout the entire central area of both types of specimens, the loss of cohesion between the layers occurs in the same planes. Thus, as shown in Figure 13, the plane between the bottom of a layer (in which the filaments are in the same direction of the load) and the top of the layer below it (in which the filaments are arranged perpendicular to the load) results in being a weak point of these structures. As it is possible to appreciate in Figure 13, and also in Figures 11 and 12, the cohesion between layers along these planes fails in all cases. In non-solid specimens, the brittle interface fracture leading to delamination between layers in FDM specimens is explained by the weak interlayer bonding or to interlayer porosity, as explained in the work by Ziemian et al. [30], who emphasized that the tensile strength is greatly affected by the fiber to fiber fusion and any air gap between the fibers. Figure 14 tries to illustrate the reason of this recurring behavior in these planes. As shown in Figure 14b, when the deformation of the specimen starts the filaments parallel to the load are the ones that are really affected by the load and so they are the first to be deformed. On the other hand, the bonding between the filaments arranged perpendicular to the load is not strong enough, and it fails easily.

**Figure 14.** Elongations and displacements between layers. (**a**) state before deformation; (**b**) state after deformation.

If the elongation of the filaments oriented in the direction of the load is related to the contact surface between the filaments of adjacent layers, very different situations can be observed. As shown in Figure 14a, before the deformation of the specimen, the contact surface (CS) between two filaments located on adjacent layers is clearly defined. However, when deformation starts (Figure 14b), the filaments in the same direction of the load will suffer certain elongation while the ones arranged perpendicular to the load not. Thus, the dimensions of that initial contact surface become different because the load supported by the filaments of each layer depends on their orientation and the cohesion between filaments of different layers is not strong enough to cause equivalent deformation in alternant layers.

Considering as the initial reference a layer with the filaments oriented in the same direction of the load and given that they are the ones which mainly support the load applied and which are deformed, Figure 14b also shows as this phenomenon is very different if the upper or lower layer is considered. In the first case, the contact surface is significantly smaller, and once the union between filaments of layers oriented perpendicular to the load has failed, these filaments can move away maintaining their punctual unions with the filaments parallel to the load of the adjacent layers, which are being deformed. Thereby, as shown in Figure 14b, these relative displacements between the contact surfaces of filaments located on adjacent layers become relevant when the lower layer is considered, due to the larger initial contact surface and its different evolution in each layer. For this reason, the fracture planes shown in Figure 13 appear. Recent studies combined computational and experimental techniques [32]

to evaluate the cohesive strengths between filaments and seems to be a promising field of research in fracture mechanisms of FDM parts.

#### *3.2. Comparative Analysis with Conventional Samples with Pattern Infill*

Before the comparison is done, it is important to clarify that, although the specimens in this work have been printed as solid parts, the additive manufacturing technique leaves some gaps between layers as presented in Figure 15, so the infill is supposed to be close to 100% infill, but not completely. These triangular air voids are responsible for the decrease of the tensile strength because of a decrease in the cross-sectional area of the specimen [30] compared to bulk materials. Moreover, the lower strength of FDM samples compared with injection-molded samples have been attributed to the gaps between filaments and inner pores within them [28].

**Figure 15.** Visualization of the gaps between layers in a specimen performed by an optical micro-coordinate measurement machine IF-SL ALICONA.

A comparison of the results to those ones obtained by Rodríguez-Panes et al. [26], where a pattern infill (non-solid) was used in the FDM specimens, is presented in this section according to Table 5. Comparison is realized with results from standard ASTM D638–14:2014 (Figure 16), as it was the same standard used in the previous work.

**Figure 16.** Comparison of results with non-solid specimens (Table 5) tested byRodríguez-Panes et al. [26]: (**a**) specimens with orientation 1 and 20 and 50% infill; (**b**) specimen with orientation 2 and 20% infill.

Table 7 presents the data of mechanical properties obtained from graphs of Figure 16.

σ σ


**Table 7.** Comparison of mechanical properties in samples with solid (ASTM–O1, ASTM–O2) and non-solid (S1, S2, S3) infill.

Tensile strengths of ABS parts obtained by FDM have been reported to be in the range of 11–40 MPa [10]; the explanation to this wide range is associated to their anisotropic behavior. Results are in good agreement with previous works such as the one by Tymrak et al. [33], with tensile strengths around 30 MPa, or the work by Banjanin et al. [10], with an average value of tensile stresses for ABS samples of 31 MPa, considering that they used non-solid specimens, so they are expected to have lower values than the ones obtained in our work with solid ones; in fact, these reference values are in very good agreement with the ones obtained in our previous work with non-solid samples [26], used for comparison in this section.

As expected, for the specimens with an infill percentage of 20% (S1 and S3), the mechanical properties are the poorest ones, the effect of the infill being more significant in the case of orientation 2, where the differences in tensile strength at yield and break are particularly high (almost double). On the other hand, the sample with an infill percentage of 50% (S2) shows a higher strength than samples S1 and S3, but lower than the strength provided by the solid infill of samples ASTM-O1. Nevertheless, the increase of strength due to the infill does not seem to be in the same proportion to the percentage of infill; that is, for an increase of the infill percentage of almost 400% (from 20% to almost 100% infill for solid specimens), the increase of tensile strength is only 37.83%, and, for an increase of infill percentage of 150%, the increase of tensile strength is 25.04%. In the case of the strain to break, the values for the samples with solid infill presents a much higher deformation before fracture.

#### **4. Conclusions**

In general, tests performed according to the specific Spanish standard for additive manufacturing UNE 116005:2012 present a much better repeatability than the ones according to the general test standard ASTM D638–14, which proves that the geometry of sample and procedure used in the standard UNE is more appropriate for characterizing materials obtained by FDM and for comparison between different materials. All the specimens according to the standard UNE experience a brittle breakage without plastic deformation. Results for ASTM D638–14 show a higher variability, especially for orientation 1, where some specimens exhibit some plastic deformation before fracture and elongations are particularly higher than the ones obtained by the standard UNE.

As a general trend, orientation 2 provides higher strength to the parts obtained by FDM, but the influence of the orientation is more significant in the case of the standard ASTM. This behavior is coherent with the arrangement of the filaments in each layer for each orientation, the percentage of the section with higher strength provided by the lateral areas (shell) being clearly higher in specimens with orientation 2, which justifies their better performance under tension.

Comparison with non-solid specimens (and different percentage of infill) show that, for samples with an infill percentage of 20% (S1 and S3), the mechanical properties are the poorest ones, the effect of the infill being more significant in the case of orientation 2. The sample with an infill percentage of 50% (S2) shows a higher strength than samples S1 and S3, but it is lower than the strength provided by the solid infill of samples, although the increase of strength due to the infill does not seem to be in the same proportion to the percentage of infill. In the case of the strain to break, the values for the samples with solid infill presents a much higher deformation before fracture.

As a general remark and given that national standards follow the guidelines of the international ones (acting as complementary information sources), positive experiences with national standards as the one presented in this work could be considered in future updates of international standards.

**Author Contributions:** Conceptualization, A.G.-D., J.C., A.M.C., and M.A.S.; methodology, A.G.-D. and J.C.; formal analysis, A.G.-D., J.C., and A.M.C.; investigation, A.G.-D., J.C., and A.M.C.; resources, J.C. and A.M.C.; writing—original draft preparation, A.G.-D., J.C., and A.M.C.; writing—review and editing, A.G.-D., J.C., A.M.C., and M.A.S.; supervision, M.A.S.; project administration, J.C. and A.M.C.; funding acquisition, J.C. and A.M.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Annual Grants Call of the E.T.S.I. Industriales of UNED through the projects of references [2019–ICF04] and [2019–ICF09].

**Acknowledgments:** This work has been developed within the framework of the "Doctorate Program in Industrial Technologies" of the UNED and in the context of the project DPI2016–81943–REDT of the Ministry of Economy, Industry, and Competitiveness. We would like to extend our acknowledgement to the Research Group of the UNED "Industrial Production and Manufacturing Engineering (IPME)" and the technical staff of the Department of Manufacturing Engineering for the given support during the development of this work.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Influence of Print Orientation on Surface Roughness in Fused Deposition Modeling (FDM) Processes**

#### **Irene Buj-Corral \*, Alejandro Domínguez-Fernández and Ramón Durán-Llucià**

School of Engineering of Barcelona (ETSEIB), Department of Mechanical Engineering, Universitat Politècnica de Catalunya (UPC), Avinguda Diagonal, 647, 08028 Barcelona, Spain;

alejandro.dominguez-fernandez@upc.edu (A.D.-F.); ramon.duran@estudiant.upc.edu (R.D.-L.)

**\*** Correspondence: Irene.buj@upc.edu; Tel.: +34-934054015

Received: 22 October 2019; Accepted: 19 November 2019; Published: 21 November 2019

**Abstract:** In the present paper, we address the influence of print orientation angle on surface roughness obtained in lateral walls in fused deposition modelling (FDM) processes. A geometrical model is defined that considers the shape of the filaments after deposition, in order to define a theoretical roughness profile, for a certain print orientation angle. Different angles were considered between 5◦ and 85◦. Simulated arithmetical mean height of the roughness profile, Ra values, were calculated from the simulated profiles. The Ra simulated results were compared to the experimental results, which were carried out with cylindrical PLA (polylactic acid) samples. The simulated Ra values were similar to the experimental values, except for high angles above 80◦, where experimental roughness decreased while simulated roughness was still high. Low print orientation angles show regular profiles with rounded peaks and sharp values. At a print orientation angle of 85◦, the shape of the profile changes with respect to lower angles, showing a gap between adjacent peaks. At 90◦, both simulated and experimental roughness values would be close to zero, because the measurement direction is parallel to the layer orientation. Other roughness parameters were also measured: maximum height of profile, Rz, kurtosis, Rku, skewness, Rsk, and mean width of the profile elements, Rsm. At high print orientation angles, Rz decreases, Rku shifts to positive, Rsk slightly increases, and Rsk decreases, showing the change in the shape of the roughness profiles.

**Keywords:** Fused Deposition Modeling; roughness; Polylactic Acid; print orientation angle; build angle

#### **1. Introduction**

In the fused deposition modelling (FDM) process, a filament is heated and then the material is deposited by a nozzle onto a printing bed. FDM printed parts are used in different applications, for example medical, electrical, aerospace, etc. For example, it allows printing patterns for investment casting of biomedical implants [1]. In addition, highly metallic-filled conductive composites can be prepared by FDM to be used in electromagnetic shielding, sensors, and circuit printing [2]. As for aerospace, carbon fiber reinforced PLA printed composites can be used [3].

FDM allows a wide range of materials, and the printed parts have effective mechanical properties. However, printing speed is low and the layer-by-layer building of parts leads to poor surface roughness due to the stair stepping effect [4,5]. When the lateral walls of a certain workpiece are inclined, the use of printing supports is required. In addition, the inclination of the lateral walls will have an effect on surface roughness, since the wall will not be perpendicular to the layer plane.

Different authors have studied the effects of printing parameters on surface roughness. For example, Pérez et al. [6] considered layer height, printing speed, temperature, printing path, and wall thickness. They found that layer height and wall thickness had the greatest influence on arithmetical mean height, Ra. Reddy et al. [7] used layer thickness, material infill, and printing quality as factors. They also

considered build inclination. Both layer thickness and build inclination turned out to be the most influential factors on roughness. Peng and Yan [8] optimized roughness and energy consumption. They employed layer thickness, printing speed, and infill ratio as factors, with layer height being the most important parameter influencing roughness. Kovan et al. [9] studied the effect of layer height and printing temperature on surface roughness. You [10] studied infill ratio, printing temperature, and printing speed. They found that roughness increases with printing speed and decreases with infill ratio. Altan et al. [11] studied the effect of printing processes on surface roughness and tensile strength, with layer thickness and deposition head velocity being the most influential parameters on roughness. Mohamed et al. [12] investigated the effect of printing parameters on the dynamic mechanical properties of polycarbonate–acrylonitrile butadiene styrene (PC-ABS) printed parts. The main factors were layer height, air gap, and the number of contours. Luis studied Ra and Rq values obtained through experimental tests in FDM processes [13].

Regarding previous geometrical models for roughness in FDM processes, Pandey et al. obtained a semiempirical model for roughness, in which they took into account both layer thickness and build orientation [14]. Ahn et al. considered the filaments to have the shape of elliptical curves which overlap in the vertical direction [15]. Boschetto et al. approximated the roughness profiles of printed parts as a sequence of circumference arcs [16]. Ding et al. obtained roughness profiles from the overlapping of different surfaces representing beads [17]. Kaji and Barari obtained roughness profiles from the cusp geometry of the lateral walls of parts, taking into account both straight lines and degree two polynomial curves [18]. On the other hand, the Slic3r manual considers the shape of the cross section of the deposited filaments to be a rectangle with round ends, in which the initial area of the filament is equal to its final area [19]. A similar approach was employed by Jin et al. However, the length of the rectangle in their cross-section model is calculated based on the volume conservation, taking into account the plastic flow-rate and speed-rate [20]. From the assumptions made in [19], Buj et al. calculated pore size from the nozzle diameter, infill, and layer height of printed samples [21]. Other authors take into account the overlapping among filaments, due to diffusion when printing high melting temperature thermoplastic polymers such as polyether ether ketone (PEEK) [22]. However, this effect is not so important with low melting temperature polymers like polylactic acid (PLA) and acrylonitrile butadiene styrene (ABS).

Regarding print orientation, Bottini and Boschetto investigated the effect of deposition angle and interference grade on the assembly and disassembly forces in the interference fit of FDM printed parts [23]. They found that assembly forces depend on both parameters, while disassembly forces do not depend on deposition angle, as surface morphology is modified as a result of assembly. In addition, different authors have studied the influence of print orientation on the mechanical strength of parts [24]. Domingo-Espín et al. studied six different orientations and determined stiffness and tensile strength of polycarbonate (PC) samples [25]. They recommended that, when the yield strength of a material is exceeded, the parts should be oriented in a way that the greater tensile stresses are aligned with the direction of the longest contours, in order increase their tensile strength. Knoop et al. studied the effect of building orientation on the tensile, flexural, and compressive strength of polyamide (PA) parts [26]. As a general trend, they found higher tensile strength for build orientation X of the tensile test specimens (on its edge), than for build orientation Y (flat lying), or Z (upright). Uddin et al. studied the effect of print orientation on the tensile and compressive strength of ABS parts [27]. They obtained the highest stiffness and failure strength for layer thickness 0.09 mm, printing plane YZ and horizontal print orientation. Chacón et al. [28] studied the influence of print orientation on the tensile and flexural strength of PLA parts. They observed that low layer thickness and high feed rate values improved mechanical performance. Sood et al. [29] investigated the effect of layer thickness, build orientation, raster angle, raster width, and air gap on the compressive strength of parts. They found that an artificial neural network (ANN) model was better for modeling compressive strength than a regression model. The optimal value for layer orientation, giving higher compressive strength, was 0.036◦. McLouth et al. [30] analyzed the influence of print orientation and raster pattern on the

fracture toughness of ABS parts. They concluded that samples with layers that are parallel to the crack plane turned out to have lower fracture toughness than samples with other print orientations. As for the influence of print orientation on roughness, Chaudhari et al. studied the surface finish of ABS parts printed with different layer thickness, infill, orientation, and postprocessing operation. They found that infill and postprocessing had the greatest influence on roughness [31]. Thrimurthulu et al. [32] simultaneously optimized surface roughness and build time, as a function of slice thickness and build deposition orientation. Both parameters influenced roughness. Wang et al. [33] studied the effects of: layer thickness, deposition style, support style, deposition orientation in the Z direction (build angle), deposition orientation in the X direction (raster angle), and build location on the tensile strength, dimensional accuracy, and surface roughness of printed parts. They observed that layer thickness was the most influential parameter.

The aim of the present paper is to define a geometrical model for surface roughness in lateral walls, in FDM printing processes. The model considers the different print orientations, with simulated results being compared to experimental results. To do so, cylindrical samples are printed with different print orientations of between 0◦ and 85◦, in PLA. Roughness is measured along the generatrix of the samples, by means of a contact roughness meter. Then, the results from the model are compared to the experimental results for different print orientation angles.

#### **2. Materials and Methods**

#### *2.1. Geometrical Model*

A geometrical model was defined to calculate roughness in lateral walls, for parts with different print orientations. Two assumptions were made (Figure 1):


**Figure 1.** Schematic of the cross-section of two adjacent deposited filaments with print orientation angle of 0◦ (the horizontal line corresponds to the printing bed).

Considering these assumptions, arithmetical mean height Ra values were calculated for each print orientation, according to the following procedure:

1. The geometry of two deposited filaments (one on top of the other) is drawn for each print orientation studied, using the Solid Woks 2017 software (Dassault Systèmes Solidworks Corporation, Waltham, MA, USA). The tangent line at the edge of the two filaments is determined and the figure is rotated until the tangent line becomes a horizontal line. Figure 2 shows an example for print orientation angle of 45◦.

**Figure 2.** Schematic of the cross-section of two deposited filaments with print orientation angle of 45◦.


**Figure 3.** Profile for print orientation angle of 45◦, with the areas highlighted in grey.

4. The center line of the profiles was found with Solid Works, taking into account the mean value theorem for integrals. The center line divides a profile function into two parts, so that the areas contained by the profile above and below the center line are equal (Figure 3). The first mean value theorem for integrals says that for all continuous functions in the area [a, b] a point c exists within the interval [a, b], which makes the area below the function equal to its image at point c for all the interval length, according to Equation (1).

$$f(b-a) \cdot f(c) = \int\_{a}^{b} f(\mathbf{x}) \, d\mathbf{x}.\tag{1}$$

5. The arithmetical mean height roughness parameter Ra (in μm) was calculated according to Equation (2).

$$Ra = \frac{1}{L} \int\_{0}^{L} \left| f(\mathbf{x}) \right| d\mathbf{x} \tag{2}$$

where L is the measurement length in mm, and *f*(*x*) is the discrete function that defines the roughness profile, in mm.

In order to compare the simulated results of the model with the experimental results obtained with a contact roughness meter, the geometry of the roughness meter tip was added to the ideal geometry of the layers. Its cross-section was assumed to be an isosceles rectangle triangle of 1 mm height, with sharp edges.

Two different cases were found:

(a) For print orientation angles lower or equal to 45◦, the tip leans on two surfaces, and a new profile is obtained which shows shallower valleys than the previous one (Figure 4).

**Figure 4.** Schematic of the printed layers with the roughness tip, for print orientation angles higher than 45◦.

(b) For print orientation angles higher than 45◦, the tip leans on one of the two sides of the profile. Moreover, it is not able to reach the lowest part of the profile (Figure 5). The modified valleys have the same depth as the original ones, but the shape of the profile changes.

**Figure 5.** Representation of the roughness tip with the printed layers, for print orientation angle higher than 45◦.

New simulated Ra values were calculated from the modified profiles.

#### *2.2. Printing Process*

A double extruder Sigma printer from BCN3D Technologies (Barcelona, Spain) was used. Cylindrical PLA samples were printed, of 12.7 mm diameter and 25.4 mm height, according to a height-to-width ratio of 2.

Printing parameters are provided in Table 1 (Appendix A).

**Parameter Values** Layer height (mm) 0.25 Infill ratio (%) 50 Nozzle diameter (mm) 0.4 Printing speed (mm/s) 60 Printing temperature (◦C) 205 Print orientation angle (◦) From 5 to 85

**Table 1.** Printing parameters of the experimental tests.

Layer height is the thickness of each deposited layer. Infill ratio is the amount of solid material within the volume of a printed structure. Infill type was rectangular in all cases, with raster angle 0◦. Air gap is the space between filaments, and depends on the infill ratio used. Shells are the layers that are printed around the infill area. No shell was printed in this case.

Print orientation angle and build angle are complimentary angles. They are shown in Figure 6.

**Figure 6.** Schematic of a printed part with the print orientation angle and the build angle.

#### *2.3. Roughness Measurement*

Roughness was measured in a contact Taylor Hobson Talysurf 2 roughness meter (AMETEK Inc., Berwyn, PA, USA), with two different Gaussian filters of cut-off 8 mm and 2.5 mm respectively. Several roughness parameters were taken into account: arithmetical mean height, Ra, maximum height of the profile, Rz, kurtosis, Rku, skewness, Rsk, and mean width of the profile elements, Rsm.

Measurement direction coincides with one generatrix of the cylinders, specifically the one that is placed opposite the printing supports. As an example, the blue lines in Figure 7 show the measuring direction of two specimens with different print orientation angles.

**Figure 7.** Printed specimens with the measurement direction highlighted in blue.

If a print orientation angle of 0◦ were considered, there would be no need to use printing supports. Thus, roughness would be measured along any generatrix of the specimen.

#### **3. Results**

#### *3.1. Roughness Profiles*

As an example, Figure 8 presents experimental roughness profiles for different print orientation angles. A print orientation angle of 5◦ (Figure 8a) corresponds to a regular profile, with the typical shape obtained in lateral walls when layers have no inclination, in FDM processes. The profile shows rounded peaks and sharp valleys, and the peak width corresponds to the layer height employed. As the angle increases, similar profiles are obtained, for example for a print orientation angle of 55◦ (Figure 8b). For a print orientation angle of 80◦, a sawtooth shape is observed for the profile. For a print orientation angle of 85◦, the profile becomes more irregular, combining high peaks for the filament edges with a transition flat area between consecutive peaks. The distance between peaks increases. At a print orientation angle of 90◦, the layers would be parallel to the direction in which roughness is measured. For this reason, the theoretical roughness value would be zero.

**Figure 8.** Roughness profiles for print orientation angle of: (**a**) 5◦, (**b**) 55◦, (**c**) 80◦, and (**d**) 85◦. Figure 9 shows a picture (plan view) of a sample manufactured with print angle of 85◦.

**Figure 9.** Plan view of a sample with print orientation angle of 85◦.

As print orientation angle increases, the stair-stepping effect becomes more evident. It can be observed that the high inclination of layers leads to a greater distance between crests, with wide plateaus that provide lower roughness values. In addition, the measured roughness profile in this case is more irregular than the rest of the profiles (Figure 8d), causing greater discrepancy between experimental and simulated roughness values.

#### *3.2. Roughness Values*

Figure 10 presents the simulated Ra results, considering the tip geometry or not, as well as the measured roughness with cut-off of either 8 mm or 2.5 mm. According to ISO 4288 standard [31], a cut-off value of 2.5 mm is recommended for Ra values between more than 2 μm and 10 μm, and a cut-off value of 8 mm is recommended for Ra values higher than 10 μm. Error bars correspond to ± standard deviation values.

**Figure 10.** Arithmetical mean height (Ra) vs. print orientation angle.

In all cases, as expected, the roughness results simulated with the tip were lower than those simulated without the tip, since the tip reduces the valley depth of the profile. As a general trend, the experimental values agree with the simulated values with tip up to a print orientation angle of 80◦. The results agree with those of Reddy et al. [4], who found that Ra decreases with build angle, which is the complimentary angle of the print orientation angle. They found maximum Ra values of 50 μm for build angles of 10◦ (printing angle of 80◦). However, in the present work, at 85◦ the experimental roughness decreases significantly with respect to 80◦. Such decrease is more important for the cut-off of 2.5 mm than for the cut-off of 8 mm. This suggests that the abrupt transition from high simulated roughness values at the print orientation angle of 80◦ to the zero simulated roughness value at the print orientation angle of 90◦ is more gradual in the experimental tests.

In order to analyse the shape of the roughness profiles at high print orientation values, Table 2 provides the experimental values of other roughness parameters, Rz, Rsk, Rku, and Rsm, measured with a cut-off of 8 mm.


**Table 2.** Rz, Rsk, Rku, and Rsm values.

Rz increases with print orientation angle, as expected, up to 80◦, and then decreases at print orientation angle of 85◦. Skewness shows negative values up to 70◦, corresponding to higher valleys than peaks (Figure 8b). At 75◦ and 80◦ skewness values are close to zero, corresponding to symmetric profiles (Figure 8c). At 85◦, skewness has a positive value, with higher peaks than valleys (Figure 8d). Kurtosis is lower than 3 in all cases, pointing out that the peaks are sharper than those corresponding to a normal distribution of heights. At print orientation angle of 85◦, the highest Rku value is obtained of 2.312, corresponding to rounder peaks and valleys. Parameter RSm, mean width of the profile elements, increases with print orientation angle, since the effective distance between layers increases. However, at 85◦ the parameter decreases, because small roughness peaks are measured in the gaps between adjacent peaks (Figure 8d).

#### **4. Discussion**

The proposed model allows simulating Ra values to be obtained in lateral walls of FDM printed parts. Unlike other models, which take into account overlapping among adjacent deposited filaments [15,17], the present model makes the assumption that printing temperature is low enough to avoid overlapping. It also assumes that the shape of the cross-section of the deposited filament is rectangular with rounded edges [19,20].

Experimental Ra values are similar to simulated ones at low print orientation angles, and they increase with print orientation angle as reported by Reddy et al. [7]. However, at high angles above 80◦, the experimental roughness values are lower than the simulated ones. This suggests a gradual decrease in the experimental roughness between 80◦ and 90◦. At 90◦, the printing direction would be parallel to the measuring direction and, for this reason, the experimental roughness values would be close to zero.

At low print orientation angles, regular profiles are obtained with round peaks and sharp valleys, which are typical of FDM processes [34]. At a high print orientation angle of 85◦, the distance between

consecutive peaks increases, leading to a flat area or gap. In this case, not only the arithmetical mean height of the profile Ra decreases but the maximum height of profile Rz and the mean width of the profile Rsm. Skewness parameter Rsk becomes positive and kurtosis parameter Rku increases, noting the change in the profile shape [35].

In the future, a similar methodology using the mean value theorem for integrals, can will be applied to calculate simulated Ra in other manufacturing processes, either additive manufacturing processes or subtractive processes, provided that the theoretical geometry of the roughness profile can be obtained.

#### **5. Conclusions**

This paper presents a geometrical model for the simulation of roughness profiles obtained with different print orientation angles in FDM processes, in order to determine the mean height of the roughness profile, Ra. In addition, experimental tests were performed. The main conclusions of the paper are as follows:


**Author Contributions:** conceptualization, I.B.-C. and R.D.-L.; methodology, I.B.-C. and R.D.-L.; software, A.D.-F.; validation, I.B.-C. and A.D.-F.; formal analysis, I.B.-C. and A.D.-F.; investigation, I.B.-C., A.D.-F. and R.D.-L.; resources, I.B.-C.; data curation, I.B.-C., A.D.-F. and R.D.-L.; Writing—original draft preparation, I.B.-C. and R.D.-L.; Writing—review and editing, I.B.-C.; visualization, A.D.-F.; supervision, I.B.-C.; project administration, I.B.-C.; funding acquisition, I.B.-C.

**Funding:** This research was funded by the Spanish Ministry of Industry, Economy and Competitiveness, grant number DPI2016-80345R.

**Acknowledgments:** The authors thank Ramón Casado-López for his help with experimental tests.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Appendix A**

List of printing parameters [profile] layer\_height = 0.25 wall\_thickness = 1.2 retraction\_enable = True solid\_layer\_thickness = 1.2 fill\_density = 50

print\_speed = 60 print\_temperature = 205 print\_temperature2 = 205 print\_temperature3 = 0 print\_temperature4 = 0 print\_temperature5 = 0 print\_bed\_temperature = 65 support = Everywhere platform\_adhesion = Raft support\_dual\_extrusion = First extruder wipe\_tower = False wipe\_tower\_volume = 50 ooze\_shield = False filament\_diameter = 2.85 filament\_diameter2 = 2.85 filament\_diameter3 = 0 filament\_diameter4 = 0 filament\_diameter5 = 0 filament\_flow = 100 nozzle\_size = 0.4 retraction\_speed = 40 retraction\_amount = 6.8 retraction\_dual\_amount = 3 retraction\_min\_travel = 1.5 retraction\_combing = No Skin retraction\_minimal\_extrusion = 0 retraction\_hop = 0.08 bottom\_thickness = 0.2 layer0\_width\_factor = 100 object\_sink = 0 overlap\_dual = 0.15 travel\_speed = 200 bottom\_layer\_speed = 35 infill\_speed = 35 solidarea\_speed = 35

inset0\_speed = 35 insetx\_speed = 35 cool\_min\_layer\_time = 5 fan\_enabled = True skirt\_line\_count = 2 skirt\_gap = 2 skirt\_minimal\_length = 150.0 fan\_full\_height = 0.5 fan\_speed = 85 fan\_speed\_max = 100 cool\_min\_feedrate = 10 cool\_head\_lift = False solid\_top = True solid\_bottom = True fill\_overlap = 15 perimeter\_before\_infill = True support\_type = Lines support\_angle = 20 support\_fill\_rate = 50 support\_xy\_distance = 0.6 support\_z\_distance = 0.15 spiralize = False simple\_mode = False brim\_line\_count = 5 raft\_margin = 3.0 raft\_line\_spacing = 3.0 raft\_base\_thickness = 0.3 raft\_base\_linewidth = 1.0 raft\_interface\_thickness = 0.28 raft\_interface\_linewidth = 0.6 raft\_airgap\_all = 0.0 raft\_airgap = 0.22 raft\_surface\_layers = 2 raft\_surface\_thickness = 0.15 raft\_surface\_linewidth = 0.4

```
fix_horrible_union_all_type_a = True
fix_horrible_union_all_type_b = False
fix_horrible_use_open_bits = False
fix_horrible_extensive_stitching = False
plugin_config = (lp1
       (dp2
       S'params
       p3
       (dp4
       sS
          filename
       p5
       S'RingingRemover.py
       p6
       sa.
object_center_x = −1
```
### **References**

object\_center\_y = −1


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Elastic Asymmetry of PLA Material in FDM-Printed Parts: Considerations Concerning Experimental Characterisation for Use in Numerical Simulations**

#### **Ma-Magdalena Pastor-Artigues 1,\*, Francesc Roure-Fernández 1, Xavier Ayneto-Gubert 1, Jordi Bonada-Bo 1, Elsa Pérez-Guindal <sup>2</sup> and Irene Buj-Corral <sup>3</sup>**


Received: 31 October 2019; Accepted: 15 December 2019; Published: 18 December 2019

**Abstract:** The objective of this research is to characterise the material poly lactic acid (PLA), printed by fused deposition modelling (FDM) technology, under three loading conditions—tension, compression and bending—in order to get data that will allow to simulate structural components. In the absence of specific standards for materials manufactured in FDM technology, characterisation is carried out based on ASTM International standards D638, D695 and D790, respectively. Samples manufactured with the same printing parameters have been built and tested; and the tensile, compressive and flexural properties have been determined. The influences of the cross-sectional shape and the specimen length on the strength and elastic modulus of compression are addressed. By analysing the mechanical properties obtained in this way, the conclusion is that they are different, are not coherent with each other, and do not reflect the bimodular nature (different behaviour of material in tension and compression) of this material. A finite element (FE) model is used to verify these differences, including geometric non-linearity, to realistically reproduce conditions during physical tests. The main conclusion is that the test methods currently used do not guarantee a coherent set of mechanical properties useful for numerical simulation, which highlights the need to define new characterisation methods better adapted to the behaviour of FDM-printed PLA.

**Keywords:** FDM; PLA; mechanical properties; bimodulus materials; standards; finite element analysis (FEA)

#### **1. Introduction**

Additive manufacturing (AM) technologies allow for converting virtual models into physical models in a quick and easy way by means of tool-free processes. Different polymeric materials are being produced for 3D Printing (3DP) with a wider range of properties. 3DP has many applications in sectors such as automotive, electronics or medical [1]. Aliphatic polyesters, in particular poly lactic acid (PLA), are suitable materials for in vivo applications because of their biocompatibility, biodegradability, good mechanical strength and processability. PLA is the most researched and used aliphatic biodegradable polyester. It is a leading biomaterial for numerous applications in both medicine and industry, and the ability to adapt its properties for specific applications makes the market capacity of PLA products

very broad, which has catalysed an extensive and growing amount of research aimed at the use of this material in innovative forms and applications [2–7].

If PLA-printed parts are to be usable as real industrial or biomechanical components, their structural and mechanical reliability has to be proved by means of strength and stiffness verifications. These will be done by classical strength of materials calculations or by finite element (FE) simulations. In either case, it is essential to have a coherent set of mechanical properties of the material under the different service conditions (tension, compression, bending, torsion, etc.). Flexural strength and tensile strength are two of the most commonly used values for comparing plastic materials. Compressive strength gives a good indication of short-term load capacities. Rigidity is expressed by the modulus of elasticity in tension and flexion. Reference data on the mechanical properties of PLA are available in the literature [2,8–14], with the tensile test being the most common of the tests performed to characterise the mechanical behaviour of PLA [15–21].

However, information of the bulk material behaviour is not always useful for carrying out numerical simulations to check the proper in-service behaviour of a fused deposition modelling (FDM)-printed component. To simulate the behaviour of a FDM-printed component, it is previously necessary to characterise the material in the same way that it is in the component. Many works start from the existing standards for polymeric materials to characterise also their FDM-printed versions. The ASTM standards D638 (tensile properties), D695 (compressive properties) and D790 (flexural properties) are widely used for this purpose.

The objective of this work was to obtain the mechanical properties of parts printed on PLA for use in numerical simulations, and to contrast the procedures for such purposes based on the use of ASTM standards. The results obtained considering an isotropic behaviour were compared for the different types of loads in the standards (tension, compression and bending). Nevertheless, when considering the bimodular behaviour of the material [22–25], inconsistencies were observed between the results obtained in the different tests. Differences in compression behaviour were also observed depending on the shape and proportions of the samples.

It is concluded that the currently accepted approach, based on characterisation according to the above standards (or the equivalent ISO standards), is not suitable for this purpose. It is necessary to define new characterisation methods that take into account of the bimodularity of the material and ensure a consistent set of mechanical characteristics for numerical simulation. This will be the next step of this research.

#### **2. Test Methods**

The PLA samples were 3D-printed with FDM technology, in a BCN3D Sigma v17 machine (BCN3D Technologies, Barcelona, Spain). It is an FDM desktop 3D printer with an independent dual extruder.

Specifications and properties of the PLA filament, provided by the manufacturer of the machine, are shown in Table 1.


**Table 1.** Filament specifications and properties

The slicing software was Cura 0.1.5. All the specimens were manufactured with constant printing parameters, which are provided in Table 2.


**Table 2.** Printing parameters.

Anisotropy as well as the influence of different printing parameters, like infill percentage or printing speed, were beyond the scope of this study.

ASTM (American Society for Testing and Materials) and ISO (International Organization for Standardization) mechanical testing standards are widely used to determine mechanical properties of plastics. These test procedures assume the material is continuous and homogeneous, although not necessarily isotropic. They do not include particular considerations for additive manufacturing [26,27]. Forster [28] reviews existing procedures for testing polymers and analyses their feasibility for additive manufacturing processes. It should be mentioned; however, that there are technical committees working on developing new standards for additive manufacturing (ISO/TC261, ASTM F42).

ASTM D638-02a Standard Test Method for Tensile Properties of Plastics [29], ASTM D695-02a Standard Test Method for Compressive Properties of Rigid Plastics [30] and ASTM D790-02 Flexural Properties of Unreinforced and Reinforced Plastics and Electrical Insulating Materials [31] were followed in the tests carried out.

As regards the shape and dimensions of the specimens tested, the recommendations of the standards were followed. For the tensile and bending tests, the options were quite clear; however, in the case of the compression test, two cross-section shapes were possible, and the specimen length depended on the mechanical property to be obtained.

#### *2.1. Specimens*

The shape and dimensions of the specimens were defined in accordance with the standards (Figures 1–3) as above mentioned, and six samples of each series were manufactured: One set of specimens for tensile tests, another set for bending tests, and four sets for compression tests, where two different cross-section shapes and two different specimen lengths were possible.

**Figure 1.** Drawing of the tensile test specimen (dimensions in mm).

**Figure 2.** Drawing of the compression test specimens (dimensions in mm).

**Figure 3.** Drawing of the three-point bending test specimen (dimensions in mm).

As mentioned above, the printing parameters were kept constant (Table 2). The layers were oriented in the direction of the stresses. Figure 4 shows the print raster and the direction of the layers.

**Figure 4.** Structure of specimens with the raster and layers.

Figure 5 shows the 36 PLA 3D-printed samples grouped in six sets.

**Figure 5.** Poly Lactic Acid (PLA) samples for mechanical testing.

All test specimens were weighed on a KERN 400-55N precision scale (KERN & SOHN GmbH, Balingen, Germany), and measures were taken by means of a calliper to obtain the actual dimensions of each one of them. Tensile, compression and three-point bending tests were performed. All of them were carried out by means of an INSTRON machine model 3366 (INSTRON®, MA, USA) with the necessary equipment for each test.

#### *2.2. Experimental Tests*

#### 2.2.1. Tensile Test

According to the ASTM D638 standard [29], the speed of testing was set at 5 mm/min and the load–extension curve of the specimen was recorded. Longitudinal (INSTRON 2630-102) and transverse (INSTRON I3575-250M-ST) strain measuring devices (extensometers) were attached to the specimen in order to determine the Poisson's ratio (Figure 6) (INSTRON®, MA, USA).

**Figure 6.** Tensile test assembly. Tensile test diagram (**a**); the longitudinal (**b**) and transverse (**c**) extensometers fitted on the specimen are shown.

From the data recorded during the test, the values of stress (σ), strain (ε) and modulus of elasticity (*E*) were calculated (Equation (1)).

$$\begin{cases} \sigma = \frac{P}{A\_0} & \varepsilon = \frac{\Rightarrow l}{l\_0} \quad E = \frac{\sigma}{v} \end{cases} \tag{1}$$

where:

*P* = tensile load;

*A0* = initial cross-sectional area;

Δ*l* = increment of distance between gauge marks;

*l0* = initial specimen gauge length.

#### 2.2.2. Compression Test

In the case of the compression test, the ASTM D695 standard [30] specifies that the test specimen shall be in the form of a right cylinder or prism (square), whose length is twice its width or diameter. However, when the modulus of elasticity and offset yield-stress are desired, the test specimen shall be of such dimensions that the slenderness ratio (λ) is in the range from 11 to 16:1. In this case, the preferred specimen sizes were 12.7 (a) by 12.7 (a) by 50.8 mm (L) (prism), or 12.7 in diameter (*D*) by 50.8 mm (L) (cylinder) (Figure 7). In literature consulted concerning this test, the length of the specimens was usually twice its principal width or diameter. Exceptionally [32–35] long-length specimens were tested.

$$\lambda = \frac{L}{\bar{l}} = \frac{L}{\sqrt{\frac{l}{A}}} = \begin{cases} \frac{\frac{L}{L}}{\sqrt{\frac{l}{4D^2}}} & \frac{4L}{D} \in [11, 16] \Rightarrow L \in [2.75D, 4D] \quad \text{for cylinder} \\\\ \frac{L}{\sqrt{\frac{l}{4D^2}}} & \frac{2\sqrt{3}L}{a} \in [11, 16] \Rightarrow L \in [3.2a, 4.6a] \quad \text{for prism (square)} \end{cases} \tag{2}$$

where:

*i* = least radius of gyration;

*I* = moment of inertia;

*A* = area of the cross-section.

**Figure 7.** Dimensions of compression test specimens.

Limiting the slenderness ratio (λ) to the range 11 to 16:1 avoids 1) the influence of the end conditions on the results and 2) the buckling of the sample in the elastic range of the test.

Four series of specimens were built, called short specimens (L = 2a, L = 2D) and long specimens (L = 4a, L = 4D), whose length values were in the range given by Equation (2).

The speed of testing was set at 1.3 mm/min for short specimens (L = 2a, L = 2D) in accordance with paragraph 9 of the standard, and at 2.6 mm/min for long specimens (L = 4a, L = 4D), thus preserving the constant strain rate in all tests.

The test setup is shown in Figure 8.

**Figure 8.** Compression test diagram (**a**); long specimen at the end of the test (**b**).

In a similar way to the tensile test, the values of stress (σ), strain (ε) and modulus of elasticity (E) were calculated (Equation (1)).

#### 2.2.3. Three-Point Bending Test

The three-point bending test was performed in accordance with ASTM D790 [31]. Figure 9 illustrates the test setup. The dimensions of the specimens can be seen in Figure 3.

The machine was set for the rate of crosshead motion (*R*) calculated according to Equation (3) taken from the standard:

$$R = \frac{ZL^2}{6h} = \frac{0.01 \times 52^2}{6 \times 3.2} \approx 1.4 \text{ mm/min},\tag{3}$$

where:

*L* = support span (52 mm);

*h* = depth of beam (3.2 mm);

*Z* = rate of strain of the outer fibre (mm/mm/min), where Z shall be equal to 0.01.

The test ended when the deformation reached 5%, which was equivalent to a vertical displacement (δ) of 7 mm (Procedure A).

**Figure 9.** Bending test diagram (**a**); specimen under loading (**b**).

The flexural stress (σ*f*) and flexural strain (ε*f*) were calculated according to Equation (4). The flexural modulus of elasticity (EB) is calculated from Equation (5).

$$\begin{cases} \sigma\_f = \frac{\beta PL}{2bl^2} & \varepsilon\_f = \frac{6\beta l}{L^2} \end{cases},\tag{4}$$

where:

*P* = load at a given point on the load–deflection curve (N); *b* = width of beam tested (mm)

#### *2.3. Numerical Test*

Finite Element (FE) analyses were carried out previously to reproduce mechanical testing on ABS 3D-printed parts [36,37]. When using experimental data in numerical models, it is important to ascertain under what conditions the mechanical properties were obtained [34].

This section presents a first approximation of a finite element model of the compression test. A three-dimensional model is shown that does not include the internal structure of the material. The software used was ANSYS® Academic Research Mechanical, Release 19.1 [38].


The features of the analysis were:


**Figure 10.** Enlargement of the surface area of a sample, where the porous nature of the material can be seen.

**Figure 11.** Finite element model for the cylindrical (**a**) and prismatic (**b**) short specimens. The purple elements correspond to the steel plates of the testing equipment.

The analysis reproduced the conditions of the compression test. The steel plates of the test setup were also simulated, and a contact between the specimen and the steel was introduced with a coefficient of friction of approximately 0.4 [39]. To define the boundary conditions, the bottom line of the bottom steel plate was fixed with zero displacement in all directions, and a negative vertical displacement was defined in the top steel plate. Symmetry boundary conditions were introduced at the two planes of symmetry.

#### **3. Results**

#### *3.1. Results of Experimental Tests*

#### 3.1.1. Tensile Test Results

Figure 12 shows the stress–strain curves, overlapped for all specimens. Figure 13 illustrates the tensile failure of a sample after the test.

The modulus of elasticity (E), strength (σ), elongation (ε) and Poisson's ratio (ν) were derived from the data recorded during the test and subsequent calculations following the recommendations of the standard (Equation (1)).

**Figure 12.** Tensile stress–strain curves.

**Figure 13.** Aspect of brittle break of a tensile specimen by the minimum cross-section—7 <sup>×</sup>19 (mm2).

The regression line of the initial linear part of the stress–strain curve was plotted, and the slope was taken as an E value (Figure 14).

**Figure 14.** Regression line (dotted line) taken for calculation of modulus of elasticity (E).

The Poisson's ratio (ν) was calculated from the data recorded by the two extensometers (linear and transverse). Figure 15 shows both deformations as a function of the applied load. With the slopes of these two straight lines, ν was calculated by dividing the values by each other.

**Figure 15.** Regression lines (dotted lines) taken to calculate the Poisson's ratio. The continuous lines represent the strain measured by the extensometers as a function of the applied load.

Values of modulus of elasticity (E), tensile strength (σ), elongation at tensile strength (ε) and Poisson's ratio (ν) derived from tensile tests are shown in Table 3, including mean and standard deviation (SD).


**Table 3.** Tensile properties obtained from tensile tests (ASTM D638).

#### 3.1.2. Compression Test Results

Analogously to the methodology followed for the tensile tests, from the force-displacement data recorded during the test, the stress–strain curve was obtained. The graphs were adjusted by doing the toe compensation indicated by the Annex A1 of the ASTM D 695 standard [30]. It consisted of ignoring the initial "toe" region of the stress–strain curve, and obtaining the corrected zero point of the stress axis by intersecting the prolongation of the linear region of the curve with the stress axis.

The modulus of elasticity (E), strength (σ) and elongation (ε) were determined from the data recorded during the test and subsequent calculations (Equation (1)).

It can be seen that, apparently, the strength increased in the plastic range of the material. This was because, as a result of compression, transformations occurred in the material, porosity was reduced, volume decreased and density increased. When the internal structure of the material changed, the load capacity changed. However, this did not affect the calculated parameters.

Values of modulus of elasticity (E), compressive yield point strength (σ) and elongation at compressive yield point strength (ε) are shown in Tables 4–6, including mean and standard deviation (SD). Since there were four series of test specimens in this case, it was deemed convenient to show the results separately (i.e., a table for each magnitude).


**Table 4.** Modulus of elasticity from compression test (ASTM D695).

**Table 5.** Compressive yield point strength from compression test (ASTM D695).


**Table 6.** Elongation at compressive yield point strength from compression test (ASTM D695).


Figure 16a,b shows the compressive response at medium levels of deformation for short specimens. For high levels of deformation, the long specimens failed by global buckling (Figure 16c,d), and the failure pattern of the short specimens was a barrel shape (Figure 17).

**Figure 16.** Compressive stress–strain diagrams: prismatic short shape (**a**), cylindrical short shape (**b**), prismatic long shape (**c**) and cylindrical long shape (**d**).

**Figure 17.** Samples tested until different levels of deformation.

#### 3.1.3. Bending Test Results

Stress–strain curves, overlapped for all specimens, are shown in Figure 18. The maximum value of σ<sup>f</sup> was taken as flexural strength, and ε<sup>f</sup> was the respective deformation.

Values of the tangent modulus of elasticity (EB), flexural stress (σf) and flexural strain (εf) are shown in Table 7, including mean and standard deviation (SD). Flexural stress (σf) and flexural strain (εf) were calculated using Equation (4). According to the standard, the modulus of elasticity in bending (EB) was calculated by means of Equation (5):

$$E\_B = \frac{L^3 \frac{P}{5}}{4bh^3} = \frac{PL^3}{486I'}\tag{5}$$

#### where:

*I* is the moment of inertia of the cross-section (mm4).

**Figure 18.** Flexural stress versus flexural strain curves.


**Table 7.** Flexural properties from three-point bending test (ASTM D790).

#### *3.2. Numerical Testing Results (FE Analysis)*

Figures 19–22 show the normal stress plots for each of the four specimen types: short prismatic (Figure 19), short cylindrical (Figure 20), long prismatic (Figure 21) and long cylindrical (Figure 22). There were no significant differences in the influence of the shape of the cross-section.

It can be perceived that longer samples exhibited a quasi-uniform distribution of stresses over a wide intermediate length, while shorter samples had hardly a uniform distribution in the intermediate cross-section, which means that the boundary conditions at the ends of the specimen had more influence on shorter samples. This effect was more notorious in the prismatic rather than the cylindrical specimens.

**Figure 19.** Distribution of normal compressive stresses (N/mm2) for the middle section of the symmetry plane. Short prismatic specimen.

**Figure 20.** Distribution of normal compressive stresses (N/mm2) for the middle section of the symmetry plane. Short cylindrical specimen.

**Figure 21.** Distribution of normal compressive stresses (N/mm2) for the middle section of the symmetry plane. Long prismatic specimen.

**Figure 22.** Distribution of normal compressive stresses (N/mm2) for the middle section of the symmetry plane. Long cylindrical specimen.

#### **4. Discussion of Results**

#### *4.1. Experimental Tensile, Compression and Bending Tests*

In order to compare the results between the three types of tests, in the case of the compression test, the average of the long specimens was taken, given that the standard recommends this when determining the modulus of elasticity.

The second, third and fourth columns of Table 8 summarize the modulus of elasticity (E), strength (σ) and elongation (ε) found for each type of test, respectively, while the last two columns capture mechanical properties of processed PLA by injection [2,22]. Figures 23–25 show the results in bar graph format.


**Table 8.** Summary table of the mechanical properties of the three tests.

(\*) depending on the strain rate.

It can be seen that the highest modulus of elasticity was the flexural modulus, with a value of 2640 MPa, followed by the tensile modulus with 2390 MPa. This was not a significant difference; however, the value of the modulus of elasticity in compression was 1445 MPa, approximately 40% lower than the previous two. This scenario would be anomalous from the point of view of a bimodular behaviour, as it should be an intermediate value between them. In effect, using the calculation method described in [23], given the elastic modules of 2515 MPa in tension and 1445 MPa in compression, the modulus in three-point bending should be around 1836 MPa (the effect of shear deformation in the specimen was taken into account), considerably lower than the observed value of 2575 MPa.

**Figure 23.** Bar graph with the mean value of the E variable.

This apparent anomaly was investigated by re-evaluating the values of the elastic modules in order to find out the causes of the discrepancy. The following corrections were made to the data sets of the three tests:


Table 9 lists the new elastic modulus values found after these adjustments.


**Table 9.** Summary of re-evaluated modulus of elasticity.

The new values obtained showed less dispersion, both between samples of the same test and between the different types of cross-sections in the compression tests of long specimens.

As can be seen, the values of the tensile and bending modules were now more similar to those shown in Figure 23, although the mean bending value was slightly higher than the mean tensile value.

However, the results obtained from the three standard tests were still inconsistent with bimodular behaviour. This inconsistency is mainly attributed to differences in sample geometry, which also generates process differences, and to different strain rates in the standard tests. Since these are uniaxial tests, and the print layers are very thin, the effect of anisotropy was considered to be minor.

**Figure 24.** Bar graph with the mean value of the σ variable.

The most remarkable thing about graphics related to strength is that the flexural value was considerably higher than the other two, 60% higher than the compressive strength and twice as high as the tensile strength, which is clearly the lowest.

**Figure 25.** Bar graph with the mean value of the ε variable.

Finally, the graphs of the elongation values were compared. As above, the highest value corresponded to the bending test, followed by compression, although they did not show such a significant difference. The elongation for the tensile test was the lowest.

These results are shown here in summary:

*Ebending* ≈ *Etensile Ecompression* σ*bending* σ*compression* > σ*tensile* ε*bending* ε*compression* ε*tensile*

The results obtained from the tests of the four series of compression specimens, two lengths and two cross-section shapes, are analysed in the following Section.

#### *4.2. Specific Analysis of the Compression Test*

As can be seen in Figure 26, the modulus of elasticity of long specimens was considerably higher regardless of the cross-section shape. This result is consistent with [34]: Young's modulus decreases when the diameter:length ratio increases. In the case of cylindrical specimens, the value for long specimens was 40% higher, being 70% higher in prismatic specimens. On the other hand, the cylinder shape also had a higher value within each length. Thus, the highest value was that of long-cylinder specimens.

**Figure 26.** Comparison of modulus of elasticity in compression tests.

In the last row of Table 4, the dispersions of the values of the compressive modulus of elasticity (in %) were calculated for each type of sample. All values are within the acceptable range of variation when the modulus of elasticity is measured, as this is a parameter that tends to present higher dispersion than other ones (e.g., breaking stresses). Some conclusions can be drawn from Table 4:


By combining reasoning a) and b), it is concluded that the most appropriate sample type to determine the modulus E is the L = 4D type.

In addition to the dispersion inherent in a compression test, the influence of two other factors was identified and analysed:

(c) The flexibility of the testing machine caused an increase in the displacement between the compression plates, which implied obtaining an elastic modulus lower than the real modulus of the material. Although the compression test standard ASTM D695-02a did not provide for correction for this effect, the actual modulus E can be calculated from the measured modulus E' by using the following Equation:

$$E = \frac{L}{\frac{L}{E'} - \frac{A}{K\_M}} \text{ } \tag{6}$$

where:

*L* is the length of the specimen;

*A* is the area of the cross-section;

*KM* is the rigidity of the testing machine.

Table 10 shows the values initially determined for each type of specimen together with the values achieved after making the correction for the flexibility of the testing machine, as well as the variation (in %).


**Table 10.** Influence of the flexibility of the testing machine.

It was observed that the effect of the flexibility of the testing machine was less for long specimens (4a and 4D). This is an additional reason for using long specimens to determine the modulus E.

(d) Friction between the sample and the load plates of the machine limited the effect of Poisson at the ends of the sample, and caused a barrel shape, which altered the assumed uniform distribution of stresses and deformations, thus increasing the apparent elastic modulus. The barrel shape was observed in some of the specimens (see Figure 17). To verify the influence of this effect on the Young's modulus, two finite element simulations of the compression tests were performed. In the first one, it was assumed that there was no friction, and in the second one, the friction completely blocked the sliding between both surfaces in contact. The sample stiffness variation (which is proportional to the Young's modulus) is shown in Table 11, where the percentage compression stiffness variation is expressed with regard to the frictionless model.

**Table 11.** Variation of the stiffness with respect to the model without friction, obtained by FEA.


It was noticed that the influence of the coefficient of friction between the specimen and the test machine plates was low, being lower for long specimens (4a and 4D). This is an additional argument for using long specimens to determine the modulus E.

In the values of compressive strength (Figure 27), it is observed that cylindrical samples had higher strength values, regardless of the length of the sample. The differences between them were not as noticeable as in the case of the modulus of elasticity.

**Figure 27.** Comparison of compressive strength.

Finally, the elongation (Figure 28) obtained for cylindrical samples was also slightly higher than that obtained for prismatic samples. In addition, higher values were found in the short samples than in the long ones. However, in the case of elongation, the differences were less meaningful than in the case of the modulus of elasticity and strength.

**Figure 28.** Comparison of elongation in compression tests.

These results are summarized in the following expressions:

*ECylinder* > *EPrism and ELong EShort* σ*Cylinder* > σ*Prism and* σ*Long* ≈ σ*Short* ε*Cylinder* > ε*Prism and* ε*Long* < ε*Short*

#### **5. Conclusions**

The mechanical properties of PLA manufactured by FDM were determined under tensile, compressive and flexural stresses. The results obtained are quite consistent, considering the low dispersion of results within each group of specimens and in comparison with available data.

However, the results obtained in this work show that PLA has a double asymmetry in its tensile and compressive behaviours: On the one hand the asymmetry in strength, and on the other hand asymmetry in the constitutive behaviour, which suggests the need to treat this material by means of a bimodular elasticity model.

Nevertheless, characterisation based on standard tests presents significant difficulties when its purpose goes beyond its application to quality control tasks, such as numerical simulation.

The behaviour of 3D-printed materials is highly sensitive to process factors and, in the case of polymeric materials, also to the effect of different strain rates applied in each test. This makes it difficult to achieve a consistent characterisation of the elastic constants among the various types of standard tests available today, even when taking into account that the different dimensions and shapes of the specimens in each test can cause process differences that affect in the measured properties.

For this reason, it is necessary to define a new characterisation procedure that allows obtaining a consistent set of elastic constants, with the minimum number of tests possible, especially adapted to bimodular materials. This result would be of great interest for carrying out simulations of the structural behaviour of 3D printed parts. Recently, Mazzanti et al. [40] concluded there is no recognised international standard governing the characterisation of the tensile, compressive or flexural properties of 3D-printed materials. The current standards are those used for the characterisation of bulk polymeric materials. In this case, the geometric characteristics are standardized through the concepts of stress and strain, but in the case of 3D printing, this is difficult because the specimen is actually a structure, not a material.

With respect to the compression test, it has been demonstrated that there is less variability of results in cylindrical specimens than in prismatic specimens (probably a result of the manufacturing process, since at the corners of the prismatic shapes the printing head can deposit excess material). In addition, when determining the modulus of elasticity, it is confirmed that, following the recommendation of the ASTM D695 standard, the longer specimens provide results that are more consistent.

The need to perform compression tests to characterise the elastic modulus to compression should be reconsidered. This could be consistently deduced using a bimodular model from the bending test, through the flexural modulus, or also from the shear test, through the transverse modulus of elasticity G, thus avoiding the uncertainties associated with the compression test.

The simulation of the compression test shows that the model is sensitive to the boundary conditions applied at the ends of the specimen (friction, etc.). Despite the introduction of geometric non-linearity in the analysis (GNA), this is not enough to correctly reproduce the actual behaviour of the material in a coherent way.

**Author Contributions:** Conceptualization, M.-M.P.-A., F.R.-F. and X.A.-G.; methodology, M.-M.P.-A and F.R.-F.; software, J.B.-B.; formal analysis, X.A.-G.; resources, E.P.-G. and I.B.-C.; writing—original draft preparation, M.-M.P.-A.; writing—review and editing, F.R.-F., X.A.-G., J.B.-B., I.B.-C. and E.P.-G.; supervision, M.-M.P.-A. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research has been funded by the Spanish Ministry of Economy, Industry and Competitiveness; Grant Number DPI2016-80345-R.

**Acknowledgments:** The authors thank Marina Blasco, Bachelor of Mechanical Engineering, for her important contribution to the work. They also thank Ramón Casado-López and Francesc-Joaquim García-Rabella for their help with experimental tests.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
