*Review* **3D Printing of Fibre-Reinforced Thermoplastic Composites Using Fused Filament Fabrication—A Review**

#### **Andrew N. Dickson \*, Hisham M. Abourayana and Denis P. Dowling**

School of Mechanical and Materials Engineering, University College Dublin, Belfield, D04 V1W8 Dublin, Ireland; hisham.abourayana@ucdconnect.ie (H.M.A.); denis.dowling@ucd.ie (D.P.D.)

**\*** Correspondence: andrew.dickson@ucdconnect.ie

Received: 26 August 2020; Accepted: 20 September 2020; Published: 24 September 2020

**Abstract:** Three-dimensional (3D) printing has been successfully applied for the fabrication of polymer components ranging from prototypes to final products. An issue, however, is that the resulting 3D printed parts exhibit inferior mechanical performance to parts fabricated using conventional polymer processing technologies, such as compression moulding. The addition of fibres and other materials into the polymer matrix to form a composite can yield a significant enhancement in the structural strength of printed polymer parts. This review focuses on the enhanced mechanical performance obtained through the printing of fibre-reinforced polymer composites, using the fused filament fabrication (FFF) 3D printing technique. The uses of both short and continuous fibre-reinforced polymer composites are reviewed. Finally, examples of some applications of FFF printed polymer composites using robotic processes are highlighted.

**Keywords:** fused filament fabrication; polymers; fibre reinforcement; mechanical properties

#### **1. Introduction**

Three-dimensional (3D) printing, also known as additive manufacturing (AM), can be used to print a range of metallic, polymer and composite parts with complex geometries and great design flexibility, while minimising processing waste [1,2]. Applications of this processing technology have included parts fabricated for use in the biomedical, automotive and aerospace sectors [3]. Three-dimensional printing was first introduced during the early 1980s, using the process of stereolithography, in which UV lasers are used to cure layers of polymer into 3D shapes [4]. These methods can be used to process materials such as epoxies. For example, a hydrogel combined with a UV curable adhesive to form a composite exhibiting properties similar to organic tissues, such as cartilage, was demonstrated [5]. A range of other polymer 3D printing technologies are also available, including Selective Laser Sintering (SLS), Laminated Object Modelling (LOM), Multi Jet Fusion Printing and Fused Filament Fabrication (FFF) processes [6–8]. The latter technique, which is also known by the trade name Fused Deposition Modelling, is one of the most widely used amongst all the 3D printing techniques, showing great potential for fabricating 3D geometry parts with the capacity to compete with conventional processing techniques [9–11]. In this technique, the polymer filament is extruded through the nozzle that traces the part's cross sectional geometry layer by layer, as shown in Figure 1 [12]. The nozzle contains resistive heaters that keep the polymer at a temperature just above its melting point, so that it flows easily through the nozzle and forms the layer [13]. The extruding apparatus is typically mounted onto an X–Y computer numerical control (CNC) gantry, allowing the printing of complex geometric patterns. Once a pattern is deposited, the build platform is lowered, or the extruding orifice is raised-up, to deposit the next layer [14,15].

**Figure 1.** Schematic of FFF process for the printing of parts using the melted polymer filament.

At present, thermoplastic polymers are the most frequently utilised feed-stock materials for the FFF process, due to their relatively low cost as well as their low melting temperatures [16]. These polymers include acrylonitrile butadiene styrene (ABS), polylactic acid (PLA), polycarbonate (PC), polyether ether ketone (PEEK) and nylon [17]. The resulting pure polymer products, however, can often lack the strength to produce fully functioning engineering parts, which has restricted the wider adoption of this technology [18]. In order to address this issue, reinforcing materials, such as fibres, are added into the polymer matrix during printing, in order to produce a composite structure which typically exhibits improved mechanical properties [19].

Reinforcing fibres in composite materials can be in the form of continuous fibres or discontinuous (short) fibres [20]. Continuous fibres have long aspect ratios and have a preferred orientation, while short fibres have short aspect ratios and generally have a random orientation [21]. Due to the fibre orientation, continuous fibre composites offer higher strength and stiffness qualities than those of discontinuous fibre composite [22]. This paper will initially introduce and discuss short fibre-reinforced polymer composites, before reviewing the rapidly expanding field of 3D printed continuous fibre composites. Commercial developments of FFF, including the use of robotic printing techniques for larger scale printing, are also reviewed.

#### **2. 3D Printing of Short Fibre-Reinforced Composite**

Composites fabricated reinforced using short fibres are attractive because of their ease of fabrication, economy and superior mechanical properties [23]. They are typically produced by extrusion compounding, injection, or compression moulding processes [22]. For FFF processing, the filaments are fabricated in a two-step process; this firstly involves mixing the polymer pellets and fibre in a blender and secondly extruding the compound to create the filament [24]. Typical short fibres used as reinforcements include carbon and glass fibres. Recently, basalt fibre has also received attention [25].

As detailed in Table 1, several authors have investigated the addition of short fibres into a thermoplastic polymer to provide composite filaments used as a feedstock for FDM process. The reported studies included investigations of the effect of fibre content, as well as its orientation and length, on the processability and performance of the resultant fibre-reinforced composites. Some studies involved a comparison between the properties of 3D printed composites and those fabricated by traditional compression moulding techniques.


**Table 1.** A summary of materials used for 3D printing of short fibre-reinforced polymer composites.

Fibre content plays an important role in determining the properties of FFF composite filaments. Generally, tensile strength increases with increasing fibre content. Composite filaments, however, with high fibre content, can be very difficult to print, due to issues with nozzle clogging, in addition to the excessive viscosity of the melted composite filament [34,35]. Therefore, the determination of an appropriate fibre content in the composite used as a filament for FFF is often a compromise between processing difficulty and the performance characteristics of the resulting composites [32].

Ning et al. [36] investigated the effect of carbon fibre content and length on the mechanical properties and porosity of FFF printed ABS/carbon fibre composite parts. The composite filaments were fabricated with different fibre contents (3 to 15 wt %) and different fibre lengths (100 and 150 μm). This study demonstrated that the best performing FFF printed parts were obtained for samples reinforced with 5 wt % carbon fibre, which achieved 22.5% and 30.5% increases in tensile strength and Young's modulus, respectively, compared with the ABS only specimens. A further increase in the fibre content to 10% or higher resulted in a decrease in tensile strength due to the higher porosity. Moreover, the composite specimens reinforced with longer carbon fibres (150 μm) exhibited higher tensile strength and Young's modulus values, and lower toughness as well as ductility, compared with those reinforced with shorter carbon fibres (100 μm).

A study by Tekinalp et al. [34] investigated short fibre (10–40 wt %)-reinforced ABS composites as a feedstock for FFF printing, in order to report on their processability, microstructure and mechanical performance. The additive components were also compared with traditional compression moulded (CM) composites. The results showed that FFF 3D printing yielded samples with very high fibre orientation, lower average fibre length and high porosity levels (16–27%) compared with those obtained using the CM process. Tensile test results demonstrated that tensile strength and modulus were increased, with increasing fibre content for both the FFF and compression moulded samples (Figure 2). However, the improvement in the mechanical properties of FFF 3D printed samples is close to that

of those fabricated with CM process, attributed to the high degree of fibre alignment in FFF 3D printed samples compared to the random orientation of the fibres in moulded samples. The authors compensated for some of the loss of strength due to high porosity and decreased fibre length.

For load bearing applications, the composite filaments used in the FFF process must exhibit adequate mechanical properties, such as strength, stiffness, ductility and flexibility [19,32]. The addition of short fibres into pure thermoplastics polymers, however, while improving the resulting printed parts' tensile and flexural strengths, can be at the cost of the reduced flexibility and handleability of the resulted composite filaments. Authors have addressed this issue through the addition of a small amount of plasticiser and compatibility agents. For example, the processability values of short glass fibre-reinforced ABS composites with three different glass fibre contents (10.2, 13.2 and 18 wt %) were investigated in relation to their use as a feedstock filament for FFF [32]. The ABS was mixed with glass fibre in a twin-screw extruder, and then granulated into small pellets. These pellets were then fed into a single screw extruder and extruded into a filament. The glass fibres were found to reduce the flexibility of the resulting filament and make it impossible to feed into the FFF printer, therefore plasticiser (linear low-density polyethylene) and compatibiliser (hydrogenated Buna-N) were added to improve the ability to process the filaments through the printer nozzle and the properties of the FFF parts. The properties of the resulting composite filaments showed they would work well as a feedstock for FFF processes. This study also demonstrated that adding short glass fibre to ABS polymer resulted in a reduction in adhesive strength between the layers in the resulting FFF 3D printed samples, while the tensile strength was increased with increasing fibre contents. The authors reported that this may be due to enhanced fibre bridging across layers during printing, as fibre content increased.

A study by Sodeifian et al. [33] reported on how the flexibility of glass fibre-reinforced polypropylene composite filaments was enhanced by adding maleic anhydride polyolefin (POE-g-MA) as a modifier. POE-g-MA was added with three different weight percentages, namely 10, 20 and 30 wt %. The filaments were used to produce test specimens using FFF printing. The test specimens were also provided using the CM method, to compare the results with those of the FFF method. This study demonstrated that the tensile strengths of the specimens with 10 wt % POE-g-MA were the same irrespective of the manufacturing method used, but the FFF 3D printed specimens exhibited higher flexibility. The FFF 3D printed specimen with 20 wt % POE-g-MA showed superior mechanical properties, compared with those prepared by using the CM method. Increasing POE-g-MA to 30 wt % yielded an increase in the strength and a decrease in flexibility. X-ray diffraction analysis indicated the higher crystallinity of the specimens prepared by compression moulding, compared with that obtained using 3D printing. Figure 3 helps to demonstrate the interlayer and intralayer adhesion of the 3D printed specimens, as compared to that prepared using the CM method.

**Figure 2.** Effect of fibre content and preparation process on (**a**) tensile strength and (**b**) modulus of ABS/carbon fibre composites [34].

**Figure 3.** Scanning electron microscopy (SEM) images of the specimens manufactured using (**a**) 3D printing and (**b**) CM methods [33]. These images help to illustrate the differences in interlayer adhesion of 3D printed samples compared to that prepared using CM process.

A study by Sang et al. [35] investigated PLA composites reinforced with silane treated basalt fibre at three different fibre weight fractions (5, 10 and 20 wt %). The effects of fibre type, as well as its weight fraction, on the thermal properties, mechanical performance and rheological behaviour of PLA/BF composite filaments were investigated. The study included a comparison with composites fabricated using both compression-moulded and carbon fibre-reinforced counterparts with the same fibre weight fraction. The results of tensile tests indicated that 3D printed specimens exhibit similar tensile strength values to the compression moulded specimens, and this was attributed to the high alignment of fibre orientation in FFF 3D printed samples versus the random orientation of fibres in the moulded samples. A 33% increase in tensile strength was obtained for FFF samples reinforced with 20 wt % basalt fibre. Rheological results, as anticipated, demonstrated that the viscosity of the carbon fibre composites is higher than that of the PLA matrix, while that of the basalt fibre composite counterparts is close to the PLA only material. The high viscosity of the carbon fibre composite results in poor printing flowability and interlayer bonding defects, which causes stress concentration and failures in test samples. This study indicates the ease of processability of basalt compared to carbon fibre for the fabrication of FFF composites.

#### *Applications of FFF Printed Short Fibre-Reinforced Composites*

Some examples of the applications of fibre-reinforced polymer composites include high-temperature inlet guide vanes (IGV) for aerospace applications, printed using polyetherimides-Ultem 1000 mixed with 10% chopped carbon fibre [19,37]. Sang et al. [38] developed promising PLA-PCL/basalt fibre composite filaments, to be used as FFF feedstock for manufacturing honeycomb structures that exhibit elastic deformations with superior energy absorption under compressive loading, and which provide valuable means to obtain an excellent compressive mechanical performance in honeycomb structures (Figure 4).

**Figure 4.** 3D printed circular honeycombs of PLA-PCL/KBF with varying ratios [38]. The PLA-PCL/KBF composite consists of polylactic acid (PLA) as the stiff matrix, polycaprolactone (PCL) as an elastomer phase and silane-treated basalt fibres (KBF) as the reinforcing filler.

#### **3. 3D Printing of Continuous Fibre-Reinforced Composite**

As discussed in Section 2, the incorporation of short fibre reinforcement can usually increase the stiffness of the resulting composite; however, the part strength is often only marginally increased. This is due to reliance on the matrix material to transfer loads between fibres. In contrast, a continuous fibre reinforcement transfers and retains primary loads within unbroken strands of fibre, and this results in a significantly lower load transfer through the polymer and allows for a load-bearing capacity orders of magnitude higher than that which short reinforcement is capable of achieving. In the case of continuous fibre composites, the polymer serves to transfer off-axis loads between fibres, such as shear forces. This protects the fibres, as high modulus reinforcement such as carbon and glass fibre exhibit poor mechanical properties under shear loading [39]. Continuous fibre-reinforced polymers are currently one of the largest areas of focus in 3D printing research [40]; this is due to their potential to match or exceed the mechanical performance of conventional composites. A number of authors have reported on modifications of the standard FFF process for the printing of continuous fibres; these include hardware changes, such as more wear-resistant nozzle materials, fibre cutting mechanisms, dual inlet hot-ends and fibre preheaters [41,42].

Two main categories of continuous fibre printing have been described in the literature, these being "in-situ fusion", and "ex-situ prepreg" [41,42]. The in-situ systems utilise two input materials, typically a dry fibre feedstock (the reinforcing fibre) and a neat polymer (the matrix polymer), which are combined together during the printing process. One of the most widely used techniques is known as "in-nozzle impregnation" [43]. In this process, the dry fibre is typically pre-threaded through the printer nozzle prior to printing, and the fibre is also preheated using a coil heater or IR lamps, so as not to excessively cool the molten polymer during printing. The polymer is fed by a motor-driven hobbed gear into the melt zone of the hot-end, and the preheated fibre and melted polymer converge in this melt zone where they are pushed together by the feeding polymer filament. The polymer continues to flow as long as the motor drives the filament, and the continuous fibre bundle is pulled through the nozzle by traction force as it is anchored to the build plate, as shown in Figure 5a. This method has the advantage of a single manufacturing step, which uses low-cost commercially available feedstocks, such as carbon fibre tow and FFF filaments. It also allows for real-time control over the local volume fraction of the part by altering the flowrate of polymer. This single-step printing approach, while rapid, has been reported to yield relatively poor-quality composite parts [44,45]. The short dwell time within the heated nozzle results in poor polymer infusion into the fibre bundles, and ultimately increases the porosity of the composite [43–45]. Tian et al. [45] reported that increasing the printing temperature led to a marked increase in polymer impregnation and decreased porosity, however a quantitative

analysis was not performed. Despite these issues, these studies have demonstrated that significant strength increases can be achieved by the addition of the continuous fibre to polymer materials. Matsuzaki et al. [46] reported on a 3.4-fold improvement in the strength of carbon fibre-reinforced PLA, versus unreinforced polymer, when only 6.6 volume fibre percent (VF%) was used. Similarly, Bettini et al. [47] observed up to a six-fold improvement in strength with the incorporation of 8.6 VF% of aramid fibre in PLA. This aramid composite was shown to exhibit a lower porosity compared with that obtained for carbon fibre composites. Another "in-situ fusion" method involves first printing polymer parts using standard FFF 3D printing techniques; these are then sandwiched together with a composite reinforcement and bonded with the application of heat and/or pressure [48]. The fibres can be added during the printing process and overprinted or added after printing between layers of the printed polymer, and then placed into an oven to facilitate bonding (Figure 5b). Mori et al. [48] developed a method for overprinting and compared it with the in-nozzle impregnation method discussed earlier. Whilst overprinted carbon fibre/PLA samples led to a 180% increase in load before failure versus the unreinforced PLA, the in-nozzle impregnated composite withstood a force of 500%, which was obtained using PLA only. This superior performance was attributed to the better contact between the matrix and fibre, and as a result significantly reduced porosity. A case study by Brooks et al. [49] used a topologically optimised FFF polymer base structure, onto which large fibre tows were adhesively bonded, and the results were a lighter part that exhibited a 4000% higher strength, compared with the unreinforced equivalent. A potential weakness of these processes is the short amount of time the fibre spends in the molten polymer, as well as the low pressure applied to the polymer, which typically leads to poor infiltration and high porosity. The approach also necessitates that multiple manufacturing steps occur simultaneously, which makes optimisation difficult.

**Figure 5.** Schematics of in-situ fusion techniques: (**a**) in-nozzle impregnation with polymer and coaxial fibre extrusion, and (**b**) embedding of continuous carbon fibre (CCF) after 3D printing in a thermal bonding process. Images from [42].

In contrast with "in-situ fusion", the "ex-situ prepreg" systems separate the manufacturing of the filament and the printing of the composite into two separate steps. This allows for greater control over the individual processes. As with in-situ systems, the method utilises two input materials (a fibre tow and polymer); however, these are combined together prior to printing into a pre-impregnated filament (prepreg), via a separate extrusion process (Figure 6). The prepreg filament is then spooled and transferred to the printing system for deposition. Where in-nozzle impregnation requires a drive motor for the constant extrusion of the polymer, this method only requires a motor to feed the initial few millimetres of filament through the nozzle. After the filament is anchored to the plate, this motor disengages and is pulled through the nozzle by the continuous fibre reinforcement, which remains solid throughout the process. This simplifies the printing process significantly, and as well as allowing for superior fibre impregnation during the filament manufacture stage, the dedicated extrusion apparatus can exert more pressure on the polymer to fully infuse the fibre tow, and allows for higher manufacturing speeds.

**Figure 6.** Schematic of ex-situ prepreg process: (**a**) The extrusion and cooling apparatus for production of the prepreg filament. (**b**) The printing process utilising the prepreg filament requires no drive gear as the fibre is pulled through the nozzle, extruding the polymer as it moves [42].

In 2014, the MIT spin-out company Markforged was the first to commercially offer the ex-situ prepreg composite printing system [50]. Their system could print using carbon, glass and Kevlar fibre, and used a cutting apparatus to deposit the correct amounts of fibre in specific locations in the part. The printer included a second FFF printhead for the printing of unreinforced nylon. Selected regions of a polymer part can be reinforced, rather than strengthening the entirety of the component. In doing so the components' fibre fractions are limited to approximately 34 VF% (slightly lower than the VF% of the prepreg filament used), and large areas of the parts remain unreinforced [41]. Compared with printed polymer only parts, the FFF of the composites exhibited significantly improved mechanical performances. For example, Blok et al. [51] reported tensile strengths as high as 725 MPa for carbon fibre/PA, compared with 84 MPa for the printed PA polymer. It is important to highlight, however, that the results from this study are based upon samples that were modified after printing. Several studies utilise an in-house designed filament or printer with a similar mechanism to that used by Markforged. Hu et al. [42], for example, produced a custom PLA/carbon fibre prepreg filament printed using a modified open source 3D printer, and these composites achieved flexural strengths five times higher than unreinforced PLA. However, the VF% was not provided for cross comparisons to be made. The air void content of these samples is typically lower than in the in-nozzle impregnated equivalents, primarily due to the initial impregnation step. As highlighted by Matsuzaki et al. [52], increasing the fibre count in a tow results in an increased air content, however after deposition this is typically reduced. This study also demonstrated that after printing under pressure (from the printing head) the nozzle serving to push the air out of the filament, a reduction in porosity from 33% to 4% was observed for some samples [52]. Whilst this result was reported for a single line of printed filament, the overall composite's void content increased due to the formation of air pockets between filaments and subsequent layers, as they are placed adjacent to and on top of one another.

Both "in-situ fusion" and "ex-situ prepreg" systems demonstrate that entrained air within the composite matrix is the primary challenge when 3D printing continuous fibre composites. Prepreg systems have the advantage of a dedicated manufacturing step, which can reduce filament air content to nearly 0%. However, some level of porosity remains after printing, and it is therefore evident that the printing process itself induces porosity and still requires optimisation. Goh et al. [41] observed that the overlapping of fibre bundles can reduce this porosity, however it could not be eliminated completely. The reduction in printed part porosity associated with the use of low pressure processing conditions during printing (1 Pa) has been successfully shown to increase the interlaminar shear strength (ILSS) of carbon, glass and Kevlar, by 33%, 22% and 12% respectively, compared to those materials printed under atmospheric pressure [53]. Another method of both decreasing the porosity of the FFF printed parts, as well as enhancing interlayer adhesion, is the use of atmospheric plasma surface activation treatments. This was demonstrated through the use of an in-line air atmospheric plasma jet treatment for the activation of sized basalt fibres, immediately prior to the application of polypropylene by extrusion coating, to form the polymer-coated filaments [25]. The flexural modulus

and the maximum shear stress values of the resulting FFF composites were found to increase by 12% and 13%, respectively, compared with those obtained for composites fabricated using untreated fibres. It was concluded that this increased mechanical performance is likely due to the enhanced interfacial bond strength between the fibres and the polypropylene polymer, with an associated reduction in the level of air incorporation around the basalt filaments.

#### *3.1. 3D Printed Composites—Mechanical Performance Comparison*

Agarwal et al. [54] demonstrated how the 3D printed composites can outperform those produced using conventional approaches, particularly when the printing is carried out using optimised fibre orientations. Figure 7 provides an overview of the tensile properties of continuous reinforcement composites versus short, particle and unreinforced polymers, based on values reported by a number of authors in the literature. The wide range of tensile properties obtained for a given composite type is likely to be influenced by parameters such as fibre content, as well as composite processing technique used.

**Figure 7.** Literature values for tensile strength and modulus for short and continuous fibre-reinforced composites, as well as unreinforced polymers for comparison. A comparison between similar additive manufactured (AM) and compression moulded (CM) woven PA66/CF composites is highlighted, with tensile performance being comparable. Key: Author, matrix, reinforcement, fibre % [18,27,29,34,43,46,51,55–58].

#### *3.2. Continuous Fibre Printing—Pathing Behaviour*

#### 3.2.1. Open Source Programme

Printing parameters such as temperature, speed, nozzle height, cornering radii and filament overlap can have an impact on the printed composite's mechanical performance. As the majority of composite printing systems under development are based upon a 3-Axis CNC (Computer numerical control) platform, they utilise Gcode to control movements. Toolpath logic and printing behaviour must accommodate the unique behaviour of a filament, containing a continuous unbroken reinforcement. As the fibre reinforcement remains in a solid state throughout the printing process, its mechanical properties should be unaffected by the printing process, however this assumes that no destructive printing behaviour has occurred during deposition. "Destructive printing behavior" refers to any movement or printing parameter that results in a reduction of the final properties of the composite, compared with those of the pre-deposition material (such as porosity or fibre discontinuity). FFF slicing software (GrabCAD, Ultimakers CURA, Slic3r, simplify3D etc.), has developed significantly over the last 30 years, since its invention in 1989 by Scott Crump, founder of Stratasys Ltd. [59]. The majority of studies performed on custom-built continuous composite systems use a simple raster deposition pattern, which is a back and forth movement separated by raster gaps [42,44–46], as shown in Figure 8a. As raster patterns can be easily programmed using existing FFF software, or with milling CNC software, they allow for the creation of rectangular specimens in a short period of time. This is likely the reason for many studies focusing on tensile and flexural testing, as samples can be manufactured without complex pathing software. The tight corners of 180◦ introduce major fibre damage and dislocation during printing, however the attachment of tabs during testing usually obscures this fact [52]. Another popular printing method is to use FFF software perimeter-following logic to form rings of material in a spiral like motion. This mode can mitigate most of the problems of a raster pattern by taking corners at reduced angles [47]. It can be used to make simple geometries, such as cylinders, wing cross-sections or dog bones, or any solid shape that contains no internal features (i.e., no hollow spaces) (see Figure 8b,c). The primary reason behind the use of these patterns is that the toolpath generated is continuous, as most custom-built systems do not include a cutting apparatus, which is necessary for the printing of more geometrically complex shapes.

**Figure 8.** (**a**) Example of a raster pattern generated for printing tensile testing samples [41]. (**b**) Objects printed using "spiral" generated by an FFF slicer software [47]. (**c**) Printer producing a bowl-shaped component from PLA/CF [45].

#### 3.2.2. Proprietary/Commercial Programs

In order to facilitate the commercialisation of the 3D printing technology for composite production, several companies have simplified the tool pathing software and aligned the method of operation

more closely with that of FFF systems. Between 2014 and 2019, only one such software was available commercially; this was the Markforged slicer software "Eiger". This software allows for automated fibre placement based upon existing slicer logic, such as perimeter placement, but can only be utilised with Markforged Mark series composite printers [60]. As of 2019, additional companies have released open sourced equivalents of slicing software, with a specific focus on turning FFF printers into composite printers. For example, 9T labs of Zurich, Switzerland and Anisoprint of Moscow, Russia have both released open sourced slicer software for use on a wide range of 3-axis printing systems [61,62]. 9T labs is in early beta testing of its "CarbonKit" and accompanying slicing software; this system serves as a drop-in kit for existing FFF printers, expanding their capabilities into composite printing. The accompanying software "Fibrify" appears similar in function to Markforged "Eiger", with a greater emphasis on fibre optimisation. Anisoprint has followed a similar route to that of Markforged with its "Aura" slicer, which enables fibre printing with their desktop "Composer" series printers [61,62]. As with "Eiger" and "Fibrify", this slicer is fundamentally an FFF slicer, with the added functionality of fibre inclusion. These commercial systems cater to a model of "improving printed part performance" rather than "improving composite part performance". Figure 9 compares the Markforged "Eiger" and Anisoprint "Aura" slicer software for fibre composite printing. As demonstrated in this literature review, the inclusion of fibres will ultimately lead to an increase in the strength and stiffness of FFF parts, however the performance of the resulting composite part can be significantly enhanced by optimising the fibre placement.

**Figure 9.** Comparison of (**a**) Markforged "Eiger" and (**b**) Anisoprint "Aura" slicer software for fibre composite printing. The Eiger software facilitates fibre placement in tighter spaces, with blue lines indicating fibre paths.

#### *3.3. Geometrically Complex Composite Fabrication through 3D Printing*

The majority of the studies to date on composites have focused on improving their mechanical properties. 3D printing systems, however, also facilitate greater design freedom than automated fibre placement (AFP), automated tape laying (ATL) and moulding techniques, particularly regarding fibre placement, localising volume fractions and part geometry.

Several studies have produced novel continuous fibre pathing programs to print unique structures. Hou et al., for example, printed a corrugated structure for use as a core material in composite sandwich paneling, by printing the panel sideways in the Z-direction of the printer [63]. These were reported to exhibit superior compression strengths and lower core densities to those of aluminum corrugated structures. Similarly, Sugiyama et al. printed sandwich structures from carbon fibre;

however, these were printed in the XY plane utilising fibre tension to print over the gaps in the core structure [64]. Utilising the continuous fibre filament under tension to form a bridge allows for printing with very little support material. A multitude of core patterns and shapes were tested in 3-point bending, with supporting abutment material being removed using a saw prior to testing. The printing was performed in a single continuous movement with no cutting apparatus present. These studies utilised a layer-by-layer approach for deposition, however a number of studies have taken this a step further to produce 3D curvilinear composites [65,66]. Liu et al. [65] produced a core structure for sandwich paneling, utilising a novel method of depositing small sections of composite in mid-air, without the need for a mould. In this study, the latticed core was printed and then adhesively bonded to epoxy composite face sheets. Compressive strengths were low, however the repeatability and filament placement error were significantly improved during the course of the study. To make a similar curvilinear component, Tse et al. [66] utilised a spring-loaded mechanism to reciprocate a heated nozzle over a 3D printed dissolvable mould. A number of contoured composite parts were produced, and in this case a 2D coordinate system was used to guide the head, with the spring mechanism following the mould in the Z-direction [66]. The exact method of achieving these printing paths is not disclosed in these studies, presumably to prevent replication, however it is likely that each study would have required bespoke programs/scripts to be written.

The use of 3D printing to minimise stress concentrators such as holes and notches has been explored in a number of modelling studies [67,68]. Stress concentrators are a major cause of the early/catastrophic failure of fibre composites. These can be associated, for example, with the machining of holes and notches in conventional composites. 3D printing has the potential to create holes within a part by reorienting fibres around the opening of a hole, rather than breaking the fibres through machining. Yamanaka et al. [67] produced a preliminary model of such a structure, indicating that by 3D printing the composite structure and preventing fibre breakage a significantly increased tensile strength could be achieved, compared with that obtained by the cutting of unidirectional reinforcements (Figure 10c) [67]. This study did not consider the width of a 3D printed fibre filament, and filaments are still cut at the point they reach the perimeter of the hole, meaning that fibre damage was also not accounted for at the point of cutting. Zhang et al. [68] performed a similar study; however, material properties were taken from the 3D printed composite literature, rather than the conventional composite values used by Yamanaka. Laminates were tested in single ply and cross ply configurations. In contrast to the procedure of Yamanaka, the fibres were not cut when the hole perimeter was reached, and instead formed a densified region on either side of the hole perimeter. This densified region prevented a major strain riser formation during testing, increasing the laminates strength and stiffness. Figure 10a,b represent the optimised and cut samples respectively.

Both studies are based upon the simulation of non-woven materials, with precise fibre placement, which are needed in order to reduce the effects of the induced off-axis forces generated by the stress concentration around the hole. Woven materials, however, typically exhibit superior through-thickness and off-axis properties [69]. It is therefore reasonable to conclude that woven laminates would be potentially favourable for resisting the forces developed around a notch.

**Figure 10.** Modelled fibre layouts (left) and resulting stress heat maps (right). (**a**) represents an ideal fibre placement scenario with reduced strain concentration [68]. (**b**) represents a drilled/cut sample with discontinuous fibres resulting in large strain concentrators. (**c**) represents a cut sample containing fibre vorteces to reduce the strain concentration at the centre hole [67].

#### *3.4. Six-Axis Robotics in the 3D Printing of Composites*

In order to facilitate the wider adoption of 3D composite printing as a manufacturing process, automation, including robotics, is vital. The use of robotic techniques has been utilised for the layup process of composites for a number of years, and most are utilised in combination with ATL systems which, for example, are widely used in the aerospace industry [40]. A multi-axis articulating robot can be used to hold the ATL head, which is in turn connected to a gantry or rail. This can allow 5 to 10 degrees of freedom. This type of system is efficient for the depositing of large non-complex (flat or single curvature) parts; however, for non-flat structures with higher curvature, a second forming process is required after placement. This second forming step has a tendency to cause the warpage and buckling of the laminates, leading to fibre damage [70]. The tape width also precludes tight cornering, and results in machining being required for small feature inclusion.

Many companies have incorporated robotics as part of their composite printing activities, with a wider range of designs made possible by multi-axis printing. Arevo Labs, for example, reported on the printing of short fibre composites as early as 2016, and has since unveiled a new robotic printing system for the printing of PEEK/CF composites [71]. Arevo have also produced a demonstrator of their technology in the form of a 3D printed CF bicycle frame [71]; this frame is made of layered CF filaments that are deposited using a custom-built, laser-heated deposition head, the details of which are not disclosed (Figure 11). 9T labs have demonstrated a robotic system for the placing of CF/PA12 composite onto curved surfaces, however the details of this design have also not been disclosed publicly [62]. Atropos is a robotic technology demonstrator that originated at the Politechnico Milano +Lab, and features a unique thermosetting resin print head [72]. This system is reported to utilise carbon, glass, or basalt fibres, and can use a number of UV curable resins to produce complex geometric parts (Figure 11). The processing efficiencies obtained through the further automation of FFF composite printing are clearly key for the wider adoption of this 3D printing

technology for the manufacture of individualised consumer products, such as sports bicycles, tennis rackets and golf clubs, as well as medical devices such as prostheses.

**Figure 11.** The Arevo labs robotic AM system printing a portion of a bicycle frame (**left**) [71]. The Atropos Robot system printing a glass/epoxy turbine blade without a mould for support (**right**) [72].

#### **4. Summary and Conclusions**

This paper provided an overview of the use of the fused filament fabrication (FFF) technique for the manufacture of both short and continuous fibre-reinforced polymeric materials, along with details of the mechanical properties of the resulting composites. The most widely reported short fibres in the literature are those of carbon and glass fibres, which is primarily due to the application focus within aerospace and automotive. Several other reinforcement fibres, including basalt, aramid and jute (and other natural fibres), are also reported, and have also been shown to improve the mechanical properties of polymer composites. The addition of short fibres into neat thermoplastics polymers can significantly improve their stiffness and strength. However, the maximum achievable properties of these composites are limited due to the presence of porosity in the printed parts. Despite the identified mechanical limitations, the FFF printing of short fibre-reinforced thermoplastic composites shows potential in 3D printing to produce some "end-use" components, such as moulds for tooling. Whilst short fibre composites have excellent utility, and can be processed through a standard FFF process, continuous fibre composites offer orders of magnitude higher strength and stiffness, versus neat polymers or short fibre composites. These materials contain continuous unbroken strands of reinforcing fibre, which allows for greater load capacity, but also requires specialist hardware for processing, such as cutting devices and multi-input printheads for proper polymer–fibre infusion. Two primary methods of continuous composite printing have been highlighted. "In-situ fusion" shows great promise for the rapid manufacturing of composite parts, with potential to produce variable volume fraction parts with a single manufacturing process. This technique, however, typically produces inferior quality parts due to entrained air contents and poor polymer permeation. "Ex-situ prepreg" provides superior quality parts with lower air contents and excellent polymer infusion, at the expense of a more complex multi-stage manufacturing process. A difficulty with the resulting 3D printed composites, however, is the presence of porosity, which can significantly impact on mechanical performance. Amongst the methods of addressing this are the application of pressure during printing, fibre bundle overlapping, the use of low-pressure printing conditions, as well as the use of atmospheric plasma pre-treatments.

The wider adoption of 3D printing for consumer products is facilitated through the use of robotic printing techniques. Developments in this area were reviewed and, when combined with the superior material properties of the high strength and low weight of the 3D printed composites, have the potential to produce a wide range of individualised parts, particularly for sectors such as sports goods and medical devices.

**Funding:** This work was funded by Science Foundation Ireland, through the I-Form Advanced Manufacturing Research Centre, grant number [16/RC/3872]. And the APC was funded by I-Form.

**Acknowledgments:** This work is partially supported by the Irish Centre for Composite Research (IComp) and the I-Form Advanced Manufacturing Research Centre (Grant Number 16/RC/3872).

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study, in the collection or interpretation of data, in the writing of the manuscript, or in the decision to publish.

#### **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* **Novel Method for the Manufacture of Complex CFRP Parts Using FDM-Based Molds**

#### **Paul Bere 1,\*, Calin Neamtu 2,\* and Razvan Udroiu 3,\***


Received: 8 September 2020; Accepted: 24 September 2020; Published: 27 September 2020

**Abstract:** Fibre-reinforced polymers (FRP) have attracted much interest within many industrial fields where the use of 3D printed molds can provide significant cost and time savings in the production of composite tooling. Within this paper, a novel method for the manufacture of complex-shaped FRP parts has been proposed. This paper features a new design of bike saddle, which was manufactured through the use of molds created by fused deposition modeling (FDM), of which two 3D printable materials were selected, polylactic acid (PLA) and acrylonitrile butadiene styrene (ABS), and these molds were then chemically and thermally treated. The novel bike saddles were fabricated using carbon fiber-reinforced polymer (CFRP), by vacuum bag technology and oven curing, utilizing additive manufactured (AM) molds. Following manufacture the molded parts were subjected to a quality inspection, using non-contact three-dimensional (3D) scanning techniques, where the results were then statistically analyzed. The statistically analyzed results state that the main deviations between the CAD model and the manufactured CFRP parts were within the range of ±1 mm. Additionally, the weight of the upper part of the saddles was found to be 42 grams. The novel method is primarily intended to be used for customized products using CFRPs.

**Keywords:** CFRP; PLA mold; additive manufacturing; fused deposition modeling; vacuum bag technology; 3D scanning; bike saddle

#### **1. Introduction**

Carbon fiber-reinforced polymers (CFRPs) often offer greater advantages than most other commonly used materials and are frequently used in many fields such as the aerospace, automotive, railway, naval, sports industry, medical, and civil construction industries [1,2]. Within the sports industry (CFRP bicycles, CFRP tennis rackets etc.) due to their excellent physical and mechanical properties, such as being lightweight, high strength, etc., CFRP products offer a good opportunity to improve the performance of the participants.

The uses of polymers are very popular within additive manufacturing (AM) processes having applications across many domains [3]. Fused deposition modeling (FDM) or fused filament fabrication (FFF) is defined as a material extrusion process within ISO/ASTM 52900:2015 standard [4]. Recently, due to the low production costs and the ability to feature a high degree of automation, the technology surrounding FDM/FFF has developed in a relatively rapid fashion [5–8]. This technology allows for the manufacture of complex parts, at a relatively low production cost, from various thermoplastic filaments including polylactic acid (PLA), acrylonitrile butadiene styrene (ABS), polycarbonates, nylons, etc. However, there are some limitations to the FDM process, for example the thermoplastic parts have poor mechanical properties when compared to traditional manufacturing methods, such as injection molding [9]. PLA is often treated with more importance in FDM, over other comparable materials, due to its relatively low production cost, good three-dimensional (3D) printability, and biodegradability; in addition to this, PLA has a low thermal tension, which reduces warping during the FDM process [10]. ABS material contains three monomers: acrylonitrile (A) in a proportion of 15%, butadiene, (B) 35% and styrene, (S) 60%. It should also be noted that ABS has good thermal and mechanical properties, resulting in ABS being used to produce jigs, fixtures, and tooling. The styrene from this material contributes to the achievement of a shiny surface. ABS is also compatible with soluble support materials for easy support removal, allowing more complex prints to occur.

Post-processes methods and variation of different printing parameters can enhance mechanical properties and surface finish of the part made by FDM. Porter et al. have investigated the optimal infill percentage of the beams from PLA that can maximised specific flexural rigidity, and obtained an infill percentage in the range of 10% to 20% [11].The post chemical treatment using acetone (99%) and dichloroethane (98%) solutions dramatically improves the surface finish and dimensional accuracy, and reduce the tensile strength of FDM specimens made of ABS [12]. Several researchers have analysed the effect of different annealing times and temperatures on thermoplastic materials which was 3D printed by FDM/FFF techniques [13–15]. The results showed that annealed PLA shape memory polymer at 75 ◦C increased the tensile and compressive strength (6% increases in ultimate compressive strength) compared to as-printed samples [13]. Wach et al., have printed PLA specimens by the FDM process and annealed they over its glass transition temperature determined an increase of 11–17% in the flexural stress of the parts [15].

There is a great diversity of available manufacturing methods and processes for the production of composite material (CM) parts. Parts with a complex geometry often require the use of molds, which can be manufactured using various methods and technologies (Figure 1). Traditionally, molds are made by computer numerical control (CNC) milling a block of raw material or of various epoxy resins, of which, this technological process is time and manpower intensive, usually with high associated costs. An additional method of making molds is by using CM, where it is necessary to manufacture a master model by a separate method, tailored to the composite material the mold is made from. This process involves the deposition of FRP layers on to the part model, followed by the polymerization of the manufactured part then extraction of the mold from the master model. This process has a relatively high level of accuracy, however, it is time intensive.

**Figure 1.** Methods of mold manufacturing for composite materials.

Rapid tooling (RT) [8,10,16,17] is a technology designed for short manufacturing runs that adopt rapid prototyping (RP) 3D printing techniques and apply the techniques to making soft or hard tooling, determined by the material used. RT can produce tooling and molds indirectly, through use of a master 3D printed model, or directly though AM. Thus, the use of AM allows for a direct and rapid way to produce hard tooling, jigs, and molds from different strong materials of the designer's choice (metals, resins). A mold was manufactured by Udroiu, using 3D inkjet printing techniques, fabricated out of the following materials: ZP 131 composite powder, ZB60 binder, and Z-max epoxy resin [16]. The mold was then used for the manufacturing of a part for an aircraft joystick, comprised of rubber. A research study of acrylic thermoplastic molds based on polymer jetting technology concluded that

3D printing is a rapid manufacturing method which can be used in a small series production of parts by using reaction injection molding (RIM) and casting [17]. Rodriguez et al. investigated the use of FDM technology to create molds using various types of material, for the thermoforming process [18], and found that the properties of the mold can transfer the surface defects of FDM technology to the thermoformed material sheets. Compression molds were 3D-printed using PLA and ABS filaments and then polished to produce personalized dental fillers [19]; this study demonstrated that 3D printed molds are able to be used in dental applications. Hay et al. used a FDM-printed mold of PLA to cast a customized PMMA (polymethyl methacrylate) implant for an external surface of a cranium [20]. Sieminsky et al. analyzed FDM/FFF technology used to manufacture mold inserts from PLA [21], to increase the life of the inserts and reduce the production time cycle, the inserts should be cooled using compressed air, after the removal of the molded part. Lusic et al. has analyzed, through the use of FEM (finite element method) simulations, the potential of CFRP-laminating molds to be used as rapid tooling through use of FDM [22]. It was concluded that thermoplastic molds comprising ULTEM 1010 material, manufactured by FDM, could be used for the manufacturing of CFRP components in autoclaves; however, they did not validate their FEM results. Additionally, Sudbury et al. investigated an industrial-grade approach to additive manufacturing, similar to extant FDM process used to produce molds used for hand lay-up of the fiberglass part [23]. The 3D-printed molds were machined by a CNC mill to improve the quality of the finished surface, and this paper concluded that 3D printed molds could be an effective approach for limited production runs of composite parts, although it must be noted that the molds used in this paper were not exposed to temperature or pressure.

The reinforcement of fibers on the thermoplastic filaments used within the 3D printing process can be created by combining short fibers and continuous fibers. The rapid manufacturing process of CFRP using FDM technology is treated in [24]. The thermoplastic matrix and the CFRP powder is used in this case and the process parameters and mechanical properties are investigated. Zhang et al. proposed a method for the 3D printing of continuous carbon fiber-reinforced plastics with a pressure roller, and found that there was an improvement in the strength of parts printed as compared to traditional FDM parts; however, this printer is considered in an experimental stage and thus cannot be used for 3D printing of molds at present [25]. Nanoparticles and short or long fibers in the thermoplastic filament can increase the specific desired mechanical properties of PLA, ABS, nylon etc. [26–28], although the 3D printing of continuous fiber-reinforced composites manufactured through an FDM process is a relatively new technique still in its infancy; thus there is a lack of experimental data on the mechanical performance of structures manufactured using this process [26]. The internal defects [29–31] resulting from the FDM process decrease the desired mechanical properties of the specified thermoplastic-reinforced fibers, and they cannot be compared with traditional FRP. Thus, traditional fiber-reinforced plastic materials with light weight and high strength are often the best option to manufacture parts to improve the performances of the athletes.

At present, the measurement of tools and molds through optical 3D scanning has gained much interest in the quality control process for different processes, e.g., forging [32,33], injection molding [34], and CFRP parts [35].

The use of advanced materials within performance cycling has led to outstanding results, and specialist bicycle components are now becoming lighter whilst retaining desired special mechanical properties. Liu et al. proposed to manufacture a bike frame comprising CFRP, using finite element analysis (FEA) the stacking sequence of the layers was investigated, however, they did not validate the results of the FEA [36]. The use of high-performance MC, especially CFRP, through the preferential arrangement of the layers can lead to both low total end mass and desired mechanical properties. Depending on the individual, some cyclists require more rigid parts, so that as much energy as possible is transmitted directly to the pedals. Alternatively, some solutions are centered around more elastic elements, in order to dampen vibrations encountered through use. It is accepted that elements made of CFRP can satisfy both of the aforementioned main requirements, through distinct approaches and the correct orientation of the material filaments for preferential reinforcement can lead

to spectacular results. The position of the body on the bike during seated sprint cycling is studied in [36–39]. The optimization of lower limb joint cinematics of the participants under maximal intensity cycling was investigated. Also, a bicycle saddle 3D printed in colors can shown simulated pressure distribution of a rider [40]. A bicycle saddle with a special shape was tested by Piazza et al., where the effects on perineal compression, blood perfusion, genital sensation and sexual function of the cyclist were investigated [41]. A manufacturing method for a bicycle saddle was presented in [42]; this was created based upon a master 3D printed model of the bike saddle manufactured from ABS material. Using the ABS master model, the authors have applied a traditional technology by hand layup to manufacture the mold from glass fiber and epoxy material. The accuracy of the parts obtained and this traditional mold was not studied.

Traditional molds (metal or epoxy resin block, composite molds), used for composite materials are often very robust; however, they are only cost-effective when used in conjunction with mass production. For production of small batches or customized parts, there is a growing need for molds which can be produced quickly and at a relatively low cost. 3D printing through the use of FDM/FFF processes, utilizing low-price filaments, such as PLA or ABS, could be an approach that will significantly reduce production time and costs to produce molds for CM.

In this paper, the authors propose a novel method to obtain a mold comprising a thermoplastic material, used for laminating the CFRP parts, using an additive manufacturing technique. Two materials were selected for evaluation, PLA and ABS, which were used to create the mold(s) using FDM. The design of the novel mold and CFRP bike saddle were then presented within this paper, where the newly manufactured CFRP parts and mold(s) were then scanned using non-contact 3D scanning methods in order to perform dimensional evaluation of the results; these results were then presented and compared against the original CAD model. It was found that the manufacturing technology used to create the mold and CFRP parts allowed for good results to be achieved.

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

#### *2.1. Design of a Novel Bike Saddle and its Mold*

The proposed method consists of several steps as shown in Figure 2. The 3D CAD model of the mold is converted to a STL (standard triangulation language) file format. Within the second phase of the method, the mold is 3D printed by FDM using two different materials on two different 3D printers. The molds are prepared for vacuum bag forming methodology and oven curing, within the third step. Following the manufacture of CFRP, specimens are subjected to a quality inspection, using non-contact 3D scanning techniques, in order to perform the dimensional evaluation of the results. Three 3D scans were performed as follows for the 3D printed mold, the final mold, and the CFRP parts.

**Figure 2.** Novel method for manufacturing complex carbon fiber-reinforced polymer (CFRP) parts.

Saddles' dimensions used worldwide for performance bicycles are in the range of 110–170 mm in width and 240–280 mm in length, respectively, depending on the type of bike, kind of competition, and individual customization. Chen investigated the measurement of the external ischial tuberosity width and the anthropometric data using a group of men and women to determine the bicycle-seat sizes [43]. The proposed predictive model assists riders to determine their seats size for bicycling.

A saddle with a width of 145 mm and a length of 270 mm was taken into consideration in this research. The surface area of the bike saddle is 3300 mm2. The design constraints of the saddle shape were established by analysis of the lateral and longitudinal position of the cyclist on the bicycle and by studying the anatomical particularities of the ischial bones (pelvic bones) and the perineum area of a group of cyclists (Figure 3). The authors cannot provide exact details on the anatomical dimensions of the analyzed individuals for reasons of protecting individual data. The geometric shape of the saddle (Figure 4) was determined taking into account the best possible contact of the pelvic bones with the saddle surface. A longitudinal channel designed on the central part of the saddle reduces the pressure applied in the perineum area on the prostate of male cyclists. Two CFRP bars of 8 mm in diameter are glued on the lower surface of the saddle, allowing the assembly to link with the bar inserted into the frame of the bike (Figure 5).

The 3D model of the saddle was designed based on non-uniform rational B-spline (NURBS) curves, and the multi-section surface technique within the CATIA V5 software (Dassault Systems, USA), as is shown in Figure 4a–c. A CAD model of the saddle mold with 4 mm wall thickness was obtained based on the 3D model of the saddle (Figure 6).

**Figure 3.** Determination of the design curves based on the human bones position on the bike saddle: (**a**) front view; (**b**) side view.

**Figure 4.** The CAD model of the novel saddle: (**a**) isometric view; (**b**) side view; (**c**) front view.

**Figure 5.** The 3D model of the assembled saddle: (**a**) upper side; (**b**) lower side.

**Figure 6.** The 3D model of the mold saddle.

#### *2.2. Manufacturing Method Using 3D Printing by Fused Deposition Modeling (FDM) Technology of the Mold Bike Saddle*

Initially, a comparative study between two different mold manufacturing techniques was performed. Firstly, CNC milling using two different material blocks, aluminum and epoxy was considered. The processing time, the cost, and the masses of the mold were estimated by simulation. Two technological operations, roughing with a 6 mm end mill and finishing with a 3 mm ball-nosed tool were considered for manufacturing simulation within CATIA software. The mold manufacturing simulation by CNC milling is presented in Figure 7.

**Figure 7.** Mold manufacturing simulation by CNC milling.

The new method of manufacturing molds involves making them from plastic materials by 3D printing. Two thermoplastic molds were 3D printed using two different printers. Thus, the first mold marked with 'A' was 3D printed from SMARTFIL PLA filament (Smart Materials 3D, Alcalá la Real, Spain) on the Leapfrog Creatr XL printer (Leapfrog 3D Printers, Alphen aan den Rijn, The Netherlands) and the second one marked with 'B' from ABS-M30 (Stratasys, Eden Prairie, MN, USA) on a STRATASYS Fortus 380 mc dual extruder printer (Stratasys, Eden Prairie, MN, USA). The diameter of each type of filament was 1.75 ± 0.05 mm. The properties of printed materials provided by the suppliers are presented in Table 1 [44,45] and the processing parameters of the 3D printing are shown in Table 2. The mechanical properties of FDM parts from PLA and ABS were investigated by many authors in different printing conditions [11,14,15,24,27,28,46,47].


**Table 1.** Properties of the material used in 3D printing [44,45].

**Table 2.** The 3D printing process parameters.


The STL file of the mold, saved from CATIA V5 software, was imported to every specific software of the two printers. This was used to generate G-code and 3D print settings. In the first study, Ultimaker Cura software [48] was used to generate a G-code file for mold A. The mold A was positioned with the functional surface upward on the build plate (Figure 8a). The mold B was preprocessed using the Insight 3D printing software (Stratasys, Eden Prairie, MN, USA). The build orientation of the part was optimized in diagonal position with the functional surface downward (Figure 8b). The support structure of PLA-based mold was mechanically removed in the post-processing stage. The SR30 soluble support material used for 3D printing of the ABS-based mold was removed by dissolving it in a very mild NaOH solution.

The surface roughness of the mold has been measured using a Namicon TR-220 (Namicon Testing S.R.L., Otopeni, Romania) roughness tester, as per the ISO 4288 standard [49]. The roughness was measured perpendicular to the deposition direction, following a measurement procedure similar as in [50]. As is concluded in [51] the surface obtained by FDM technology is the surface composed of high peaks. The surface roughness of the FDM-based mold is not appropriate for laminating and polymerizing CFRP. The active surface of the mold requires mechanical processing and deposition of filler materials for its uniformity. Thus, layers of polyester gel coat were applied and then polished to obtain a glassy surface.

**Figure 8.** The 3D printing simulation of the molds: (**a**) mold 'A' made of PLA on Leapfrog Creatr XL; (**b**) mold 'B' made of ABS on STRATASYS Fortus 380 mc.

The PLA material does not make a very good chemical connection with the polyester gel coat. A chemical surface treatment or a deposition of an intermediate layer of another material would be suitable. The surface of the PLA-based mold was initially treated additionally with a layer of epoxy gel coat. Thus, OH 38 hardener and CH-3 epoxy gel coat (Ebalta Kunststoff GmbH, Rothenburg o.d.t. Germany) that contains aluminium (Al) powder was used, where the mixing ratio is 100: 35 parts by weight. The treated PLA mold was polymerized for 24 h at 20 ◦C, and then T35 polyester gel coat (Havel Composites CZ s.r.o., Czech Republic) layer was applied by spraying on both sides; 2% methyl ethyl ketone peroxide (MEKP) catalyst was used to polymerize it. The epoxy layer allows a better adhesion of the polyester layer to the surface of the mold, and creates a good interface between the T35 gel coat layer and the PLA. Also, Al powder creates high stiffness properties of the mold. The covered layers were applied on both sides of the molds.

The ABS mold was sprayed with T35 gel coat layers. Also, the styrene from ABS allows a good chemical adhesion with the styrene component from the polyester gel coat. A heat treatment in the oven at a temperature of 85 ± 2 ◦C for 24 h was applied for both molds in order to prepare them for the lamination of CFRP layers (Figure 9). Finally, the PLA mold was mechanically processed with abrasive paper up to 1000 grids and polished.

**Figure 9.** Heat treatment of the molds in the oven.

#### *2.3. Manufacturing Method of Carbon Fiber-Reinforced Polymer (CFRP) Bike Saddle*

The CFRP bike saddle specimens were manufactured by vacuum bag technology and oven curing. The used prepreg materials are codified as GG245T-DT806W-42 (Delta Tech S.p.A., Rifoglieto Italy) and consist of carbon fiber Twill prepreg fabric type 2 × 2 by 245 g/sq, (245T), 3K threads, and epoxy resin type DT806W. This prepreg material contains 42 wfr. epoxy resin with a curing range from 65 ◦C to 140 ◦C.

The mold was prepared by applying mold sealer layers, and mold release layers that prevent the adhesion of the CFRP material at the mold surface. A layer of mold sealer type S31 (Jost Chemicals, Laudenbach, Germany) was applied in order to close the material pores and create a sleek surface. Then, the surface was treated with five layers of liquid mold release type Frekote 770NC (Loctite, Düsseldorf, Germany). Each layer was applied at 20 min interval to allow evaporation and drying of the solvent.

Three CFRP prepreg layers were applied on the mold surface following a stacking sequence distributions of [0/90/±45/0/90] degrees (Figure 10a). The edges of the CFRPs prepreg were cut according to the PLA mold shape (Figure 10b), and the whole assembly was covered with release foil and the breather fabric (Figure 10c). The mold was inserted into a vacuum bag which was sealed on the separation plane using an electric welding machine. A uniform pressure of −0.09 MPa was applied for 30 min on the composite surface. Then, the vacuum pump was stopped in order to check the bag tightness. The composite material was cured in an oven at 80 ± 20 ◦C for 5 h, where the vacuum pressure was applied.

**Figure 10.** Application of CFRP prepreg layers on the PLA mold: (**a**) applying of a piece of CFRP prepreg; (**b**) cutting the edges of the CFRPs prepreg according to the PLA mold shape; (**c**) mold and composite material in the bag under vacuum pressure.

The mold was cooled down for 30 min. The auxiliary materials have been removed (vacuum bag, the breather fabric, and the release foil) and the CFRP bike saddle was released from the mold. The CFRP part was then sanded on the borders by glass paper and the central longitudinal channel of the part was cut.

Five specimens denoted by A1-A5 from CFRP and two specimens denoted by C1-C2 carbon-Kevlar fiber-reinforced polymer (CKFRP) were manufactured using the PLA mold. The mass of all samples was determined using a precision scale.

#### *2.4. Quality Control by 3D Scanning Technique of Molds and CFRP Parts*

The molds and parts were subjected to a quality inspection, using non-contact 3D scanning techniques. The Creaform GO!Scan 20 (Creaform, Québec, Canada) 3D optical scanning system used in this study integrated with VXelements software (Creaform, Québec, Canada) provides a fast acquisition speed at 550,000 points/second, and extensive capture of complex surfaces at 0.1 mm accuracy.

The parts sprayed with white powder were positioned on a rotary table on which positioning targets were placed (Figure 11). The scan was performed in one go resulting in a 3D mesh of points. VXelements software was used to acquire and optimize the 3D scanning data. In this case study, the 3D mesh of points was saved as an STL file and then transferred to the CATIAV5-6 2019 software

to dimensionally compare the scanned 3D data and CAD model. The evaluation of the dimensional deviations was performed with the help of the deviation analysis tool within the CATIA software, where a quality control report was generated. Color comparison charts of 3D scanned model to CAD surface model which show the dimensional deviations were obtained. The dimensional deviation values of the 3D-printed molds, final molds, and CFRP specimens were determined. The sources of possible errors in the evaluation process can result from the scanning process, mesh processing, or the alignment procedure of the models being compared. The results of the dimensional deviations were statistically analyzed.

**Figure 11.** Positioning of the bike saddle on the rotary table of the scanning system

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

#### *3.1. FDM-Based Molds Evaluation*

Following the FDM process, two rigid molds of PLA and ABS were obtained as shown in Figure 12. The average roughness Ra of the upper mold surface was 25 ± 5 μm in both cases, as in the results obtained by Alsoufi et al. [52].

The surface aspect of treated molds was analyzed. The upper and lower surfaces aspect of the treated molds is shown in Figure 13. Some cracks were detected within the gel coat layer on the surface of the ABS mold after the heat treatment, as shown in Figure 13b. Thus, the results show that the treated ABS mold cannot be used for processing laminated composite materials. The unexpected behavior of the treated ABS mold can be attributed to the heat treatment and the expansion coefficient of ABS being different than that of the gel coat, although the deformations of 3D printed ABS parts are expected to appear at over 90 ◦C. The PLA mold had a different behavior than the ABS mold, the gel coat layer was adherent, and no cracks or mold deformation were detected.

The surface roughness (Ra) of the PLA mold was 0.03 μm after mechanical processing and polishing. The resulted surface of the PLA mold was smooth and continuous.

The results of the comparative study between different mold manufacturing techniques showed that the PLA mold weighed 180 g, the mold from epoxy 1200 g, and the mold of aluminum 2800 g, respectively. Thus, the molds obtained by FDM have a noticeably lower total mass than the molds

conventionally used. The estimated cost of the molds and detail about the estimated manufacturing time of the molds can be found in Table 3.

**Figure 12.** PLA mold (top) and ABS mold (down): (**a**) active surface of the molds (cavity); (**b**) back side of the molds.

**Figure 13.** The molds after gel coat application and heat treatment: (**a**) PLA mold; (**b**) ABS mold.


**Table 3.** Estimate of mold cost, lead time, and the mold weight.

The estimated cost of the PLA mold was 150 euros. The PLA mold had a lower cost than the molds conventionally used, -10×, and -8.6× less in cost than CNC milled molds from the Al blocks and epoxy blocks, respectively.

The PLA molds can be heated up very easily at the same time as the laminated composite material. Due to the chosen solution, shell-type mold heating occurs rapidly. The mold surface has the same temperature as the CFRP material in the curing period time. This causes the first layer of the resin to fluidize and be pressed at the same time as the other layers. This removes the pores from the CFRP surface parts. This cannot be said of metal molds or epoxy blocks that are much harder to heat, where the outside layer is heated firstly and becomes fluid. Thus, the outside layer passes to the gel phase, seals the top layers, and the pores within the first deposited layer on the mold cannot be removed. They remain on the surface of the mold and CFRP part. It is undesirable that the surface of the CFRP specimen contains pores after the polymerization process. That generates important manual labor to cover pores and supplementary costs. From this point of view, the shell-type molds made of polymers have very good behavior.

It should also be noted that from the mechanical point of view they are quite fragile. For this reason, when the mold and CFRP are inserted into the vacuum bag and during vacuuming, the bag must be very well arranged. The bag must not strain the mold during the vacuum and curing process. It is important to leave as many as possible crimps of the bag on the surface of the mold. These can compensate for stresses generated by the vacuum pressure on both sides of the mold.

The mechanical properties of 3D printed materials by FDM (PLA and ABS) which have been post-processed by heat treatment (annealing) were studied by many authors [13–15]. Also, the mean ultimate strength at compression for annealed samples which were 3D printed with 100% infill was determined as 69.85 MPa [13]. The main stress of the mold during the vacuum bag-forming process and oven polymerization of CFRP is compression. The applied pressure during the CFRP part polymerization process is 0.09 MPa. In the most disadvantaged cases, this value is much lower than the compressive strength of the mold material (PLA in our case) which was 3D printed by FDM and thermal treatment.

#### *3.2. Manufacturing of Bike Saddle Specimens of CFRP*

The results have shown that the FDM-based mold made of PLA successfully withstands the manufacture of CFRP components. It has good behavior for the laminating of the composite materials. Heated and simultaneously exposed to a vacuum pressure of –0.9 bar, the mold remains rigid and can be used for the curing of CFRP prepreg in the oven.

Five samples from CFRP and two from CKFRP were manufactured as shown in Figure 14. A very well pressed and compact composite material of the saddle was the result. After manufacturing by oven vacuum bag curing technology of seven FRP specimens, no cracks of the gel coat layer were noticed, the mold successfully supporting the manufacture of the bike saddles.

**Figure 14.** The CFRP and carbon-Kevlar fiber-reinforced polymer (CKFRP) bike saddle specimens.

The mass of the CFRP bike saddle specimens was 42 ± 2 g. The bike saddle assembly weight was 85 ± 2 g, where two fixing bars and the structural adhesive were included supplementary.

The proposed method that uses FDM-molds is suitable for obtaining, in a rapid way, CFRP parts in limited series. Customized CFRP parts can be manufactured at a low cost compared with traditional manufacturing. However, it should be mentioned that it must be undertaken with great care, especially when the edges of the CFRP part are cut at the same dimension as the borders of the mold. The borders of the mold can deteriorate easily or the release agent layer can be damaged.

The saddle assembled on a road bike (Figure 15) has resulted in a very comfortable biker position. The saddle position can be adjusted by changing the saddle tilt angle or by moving the saddle forward or backward.

**Figure 15.** The CFRP saddle prototype assembled on a road bike: (**a**) view about the bicycle; (**b**) detail of the saddle.

#### *3.3. Results of the Quality Control by 3D Scanning Technique of Molds and CFRP Parts*

The molds obtained using the FDM process from ABS and PLA were dimensionally evaluated. The results indicate a mean deviation of −0.94 mm and 86.14% of the scanned points are within the ±1 mm range, in the case of the ABS mold (Figure 16a). In the case of the PLA mold, the scanning results indicate a mean deviation of −0.219 mm, and 74% of the scanned points are within the ±1 mm range (Figure 16b). These negative values indicate a contraction of the materials for these two molds. In the case of the PLA mold, the deviations of the measurements indicated that the lower limit of deformations is within acceptable limits for such parts. The heat treatment can affect the dimensional tolerances of FDM-printed parts (PLA or ABS), as was mentioned by [14]. Therefore, it is necessary to take the shrinkage or expansion of the material into consideration when designing molds to be printed by FDM and subjected to heat treatment afterward. Thus, the shrinkage of the material can be reduced by scaling the 3D model of the mold in the design process; however, in this case study the model was not scaled to eliminate this issue. A scaling factor can be determined by dimensionally comparing the CAD model and the final part after post-treatment, as is investigated in [53].

Although the process has very good accuracy, after the deposition of the material we measured a small contraction of the real parts. It should be noted that this contraction can be amplified/decreased during the scanning process, as it is known that each measuring process has an uncertainty associated with the equipment and method. The appearance of the obtained pieces is glossy and black, due to which during the scanning process a series of errors can appear, and the points taken by mistake in the air must be removed. In some scans, there were areas where the scanned points are obviously erroneous due to the geometry (the part is very thin) and the appearance of the part. In Figure 16 it can be seen that a small percentage of points is marked with red, which are positioned on the side of the saddle that theoretically cannot be deformed only in that area.

**Figure 16.** The comparison regarding the dimensional deviations from the CAD model and the FDM-based molds: (**a**) ABS mold; (**b**) PLA mold.

The dimensional deviations between the final PLA mold and the CAD model are shown in Figure 17. The mean deviation is −0.33 mm. The new contraction is due to the layers of gel coat applied on the mold. In this case, 87.45% of the scanned points are between ±1 mm. The value decreases from −0.21 mm to −0.33 mm regarding the mean deviation.

The deviations analysis between the CFRP saddle and the mold from PLA for A2 specimen is shown in Figure 18. It is found that 85.67% of the scanned points on the surface of the composite parts are between ±1 mm from the surface of the mold. In this case, the mean deviation is −0.233 mm. Similar analyzes were performed for all the specimens and the results are summarized in Table 4. The values of the mean deviation are in the range of −0.356 mm and −0.0296 mm.

A comparison between the designed CAD surface and the CFRP bike saddle is shown in Figure 19. It is found that 61% of the scanned points on the bike saddle's surface are in the range of ±1 mm from the surface of the CAD model. Some scanning deficiencies can be noticed here as well. Part of the scanned points is also in the area of the scanned surface and another part is on the edges of the workpiece. The mean deviation, in this case, is −0.102 from the CAD model.

**Figure 17.** Dimensional deviation between the CAD model and the gel-coated PLA mold.

**Figure 18.** Dimensional deviation between a CFRP bike saddle specimen (A2) and the final PLA mold.


**Table 4.** Statistical parameters of comparison between the final PLA mold and CFRP parts.

**Figure 19.** Comparison between the CAD model and specimen A3 obtained through CFRP.

Detailed results of the comparison between the CFRP bike saddle and the designed CAD model for all the specimens are shown in Table 5. The values of the mean deviation are in the range of −0.349 and −0.102 mm. The results of the statistical analysis for all the five specimens are shown in Table 5.


**Table 5.** Statistical parameters of comparison between mold CAD model and CFRP parts.

Also, the interval plots of the main deviations in the range ±1 mm denoted as Dev[−1,1] of the active surface for all the samples is shown in Figure 20. Individual standard deviations were used to calculate the interval plot.

**Figure 20.** Interval plot of Dev[−1,1] CAD-CFRP and Dev[−1,1] PLA-CFRP; bars are one standard error from the mean.

The results show that the coefficient of variation is lower than 10% that assures the data heterogeneity and expresses the repeatability of the experiments, as shown in Table 6.

**Table 6.** Statistical parameters of percentage deviations Dev[−1,1] for PLA-CFRP and CAD-CFRP for five samples.


The coefficient of variation (CV) is a measure of spread that describes the variation in the data relative to the mean.

#### **4. Conclusions**

This paper presented a novel method for the manufacturing of molds used to laminate complex geometry parts using CFRP prepreg. Specifically, the application this paper focused on was the production of a bike saddle for performance bicycles, where the design of a prototype bike saddle fabricated from CFRP was presented. The new mold and the CFRP prototype were created utilizing different technologies, where the mold was manufactured using additive manufacturing FDM processes and the CFRP fabricated part utilized vacuum bag-forming technologies and polymerization. This paper investigated the use of two materials (PLA and ABS) to create the molds through FDM.

The following conclusions can be drawn:


mold material, mold coating material and its thickness, constant thickness of the mold walls, uniform distribution of the vacuuming bag, vacuum pressure applied, oven baking temperature, cooling and demolding of the part.

• A future direction of research for these types of mold(s) would be to use them in the curing autoclave process of FRP. The same temperature can be used for the mold(s); however, various additional pressures that may rise up in the range of 4–5 bars will be needed. Another important study would be to investigate the design of these types of mold(s), so that they do not deform under the external pressures applied, whilst also retaining conditions resembling a low wall thickness. Additionally, studies would also need to investigate the use of different types of thermoset polymer on the mold surfaces in order to improve the stiffness and the thermal expansion qualities of the mold.

**Author Contributions:** Conceptualization, P.B., C.N. and R.U.; methodology, P.B., R.U. and C.N.; software, C.N.; formal analysis, R.U.; investigation, P.B. and C.N.; data curation, P.B., C.N. and R.U.; writing—original draft preparation, P.B., R.U. and C.N.; writing—review and editing, R.U. and P.B.; visualization, P.B., R.U. and C.N.; supervision P.B. and R.U.; project administration, P.B.; All authors have read and agreed to the published version of the manuscript.

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

**Acknowledgments:** This article has benefited from the support of the project 345PED/2020 "Smart hospital bed - HoPE" funded by the Executive Agency for Higher Education, Research, Development and Innovation Funding (UEFISCDI), Romania. Also, the authors acknowledge of Technical University of Cluj-Napoca and Transilvania University of Bra¸sov, for providing the infrastructure used in this work.

**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/).
