From gantry-based machine to robot-based fused deposition modelling: A state-of-the-art

Over the last decade, significant literature has emerged that advocates the potential of different Additive manufacturing (AM) technologies and printable polymeric materials. Nevertheless, a large-scale printing and complex geometric shapes, with curvatures and non-planar layer deposition, are challenging for the traditional gantry-based machine. The 3 degrees of freedom cartesian configuration restricted their capability to planar layered printing and restricted part dimensions. To date, many researchers have used industrial robots to overcomes this limitation. This review gives the reader a good overview of the FDM technique due to its scalability, cost efficiency and a wide range of material printability. A strong emphasis is laid on the PLA and PLA-based composites as promising materials for the FDM process applications. The second part of this paper links the successful use of these materials in the traditional printing process to large-scale printing using the robot-based FDM process. This survey presents representative setups for robot-based AM and works that have been used these setups for non-planar material deposition. Finally, we conclude this paper by identifying opportunities for realizing new functional capabilities by exploiting robot-based AM, and we also present the future trends in this area.


Introduction
Additive manufacturing (AM), or more commonly known as 3D printing, differs from subtractive manufacturing processes (Abdulhameed, Al-Ahmari, Ameen, & Mian, 2019). The final products are usually built by materials deposition in a layer-by-layer process without any conventional tool (Carlier et al., 2019). AM makes it possible to build complex objects in one single process step, eliminating production steps and accelerating time to market with marginally increasing production costs. Whether it consists of Stereolithography (SLA) using photopolymer liquids ( Fused Deposition Modelling (FDM) is the most commonly used 3D printing technique (Ning, Cong, Wei, Wang, & Zhang, 2015). FDM consists of melting the thermoplastic polymer's filaments through a hot nozzle followed by cooling and solidification to construct the final structure in a layer-by-layer process ( Additive manufacturing has successfully left prototyping steps to be used in various applications (Urhal, Weightman, Diver, & Bartolo, 2019). However, the limited construction platform dimensions and the three-axis Cartesian coordinate-based machines limit applications to small build volume that the process permits (Easter, Turman, Sheffler, Balazs, & Rotner, 2013;Huang, 2013). Consequently, the limited building volume mitigates the primary benefit of single-step part printing. It might force portioning the design into smaller parts, which adds further assembly steps to the manufacturing process, particularly in large part applications like those for aerospace ( Several attempts have been made to overcome these limitations that hampered the broader adoption of FDM printing. Multi-axis robots (MARs), usually used for welding and pick-and-place tasks, are emerging as an alternative to extending the dimensional boundaries of the three-axis cartesian coordinate robots (Urhal et al., 2019). Coupling a six-degree-of-freedom (6-DOF) robot with an extruder head for FDM printing improves the final part dimensions and the geometric complexity for extensive multi-scale applications.
The paper is comprised of three sections. Section 1 presents the fused deposition modelling (FDM) process, as the most frequently used in additive manufacturing, and discusses the most commonly used polymers in the FDM-printed parts. Section 2 summarizes the literature review of the development status and the application of PLA polymers and PLA-based composites in FDM. In the third section, the robot-based FDM is emphasized and the corresponding research work using these materials. Finally, the paper ends with a discussion and conclusion on the robot-based AM and its future development trend.

FDM using three-axis Cartesian coordinate machine
Fused Deposition Modelling (FDM) is one of the most frequently used printing methods in 3D printing of polymers and composites (Yao, Deng, Zhang, & Li, 2019). As a common concept of all AM methods (Figure 1), the FDM printed part is in the first step prepared as a 3D model using CAD modelling software, such as Auto Cad Solidworks, Catia, Creo etc. Most of the FDM printing setups use the 3D model in .STL file format where a slicing software converts it into the appropriate file type and parameters (G-code), by which the tool path plan is generated to control the machine hardware  The printed part volume results from the up-or-down movements of a nozzle of an extruder head or a printing plate ( Figure 2). Depending on the printing setup, the part's volume is mainly built by a movement of the nozzle and/or the printing plate in XYZ directions. Besides, the printing quality, such as surface finish and precision, depends on the printing parameters. The standard settings of different printing machines are the layer thickness, infill density, raster orientation, build orientation, printing temperature, and speed. These parameters should be tuned based on the printed polymer's flow properties (Kaveh, Badrossamay, Foroozmehr, & Etefagh, 2015). The printed part increases in height by continuous feeding of quasi-liquid state thermoplastic polymers through at least one liquefier nozzles. The melted polymer cools down during the layer by layer deposition due to the lower surrounding temperature, hardens, and consolidates with the neighbouring layer. Support materials, deposited onto the printing plate to form a foundation for the part, can also be loaded to a secondary nozzle. The support materials are dissolved in a water-based cleaning solution after the printing is completed ( Figure 3) (Dikshit, Goh, Nagalingam, Goh, & Yeong, 2020).

Thermoplastics polymers for FDM
For FDM printing, thermoplastic polymers are generally extruded and spooled in a filament of two standard sizes, 1.75 and 3.0 mm in diameter (Rahim, Abdullah, & Md Akil, 2019). The most commonly used thermoplastic polymers are shown in Table 1. The polymers' melting temperature is a crucial parameter to define the process parameters that lead to improved part quality/properties. One of the most critical parameters related to the polymer melting temperature is the temperature of the nozzle. Low melting will lead to inconsistent layer-to-adjacent and layer-to-layer bonding. However, overheating beyond proper melting temperature leads to reduced viscosity and a lengthier time to solidify. Additionally, printing velocity and nozzle temperature should be set considering the melting temperature, where higher speed leads to reduced residence time in the nozzle and a shorter time for heating (Liu et al., 2019). In the cases of amorphous polymers such as ABS, there is no crystallization. Thus, the shrinkage due to cooling the polymer from the melting to the ambient temperature is typically significant. Moreover, due to the higher melting and printing temperatures, amorphous polymers printed parts may suffer dimensional precision and an eventual warpage event (Liu et al., 2019).
Depending on their molecular structure and the applied cooling rate, some semi-crystalline polymers have a low degree of crystallinity, typically between 15 and 80% (Ehrenstein, 2012;Sender et al., 2007). This behaviour allows using some polyamides polymers in FDM printing due to their lower degree of crystallinity, between 20 and 32% (Athreya, Kalaitzidou, & Das, 2011). Polyamide molecules have a high tendency to be cooled in a relatively short time, which is crucial to achieving a successful print with successive additions of thin layers. However, many semi-crystalline polymers, such as polyurethane (TPU), have a high degree of crystallinity. This behaviour makes them less attractive in FDM because of the high degree of shrinkage (Carneiro, Silva, & Gomes, 2015). During the printing process, a high degree of crystallinity will lead to a fast cooling of the circumferential volume than the rest of the part. Thus, the increase of shrinkage will cause the piece to pull up from the platform.
Also, a high degree of crystallinity can provoke the printed part's warping, mainly in the sharp corners and due to high shrinkage. Even though warping is located at corners and the printed part is still attached to the platform, the printing's uneven surface will prevent flowing at an equal thickness, which causes clogging of the polymers in the nozzle. Besides, polymer clogging will cause the melt flow's local failure, increase the porosity, and reduce the part's mechanical properties (Rahim et al., 2019).

Polylactic acid (PLA) polymer
The growing interest in 3D printing for industrial-level use, the increasing awareness of the environmental impact of polymers and fibre-reinforced polymers and the limited range of biomaterials increase the attractiveness of PLA for its use in . The low glass transition temperature and melting temperature, lower coefficient of thermal expansion, and non-adherence to the printing surface make it a promising thermoplastic for printing purposes (Cuiffo et al., 2017). However, low thermal stability, high degradation rate during processing, low toughness, and moisture sensitivity limit its application (Jo, Kwon, & Moon, 2018). PLA can be found in semi-crystalline or amorphous grades. Pure poly (l-lactic acid) (PLLA) or poly(D-lactic acid) is semi-crystalline, whereas PLA with 50-93% L-lactic acid is amorphous. Amorphous PLA exhibits better processability but poor mechanical properties as compared to crystalline.
Several studies have already paid attention to PLA and PLA-based composites because they are critical for industrial and general use. However, the high complexity of setting optimal printing parameters, which highly depend on the polymers properties and their interaction with the printer setup and environment, is one of the major factors that hinder further growth of the PLA usage in FDM. PLA polymers are affected by processing methods. Their influence on mechanical, thermal and rheological properties results in poor mechanical properties of the FDM printed parts compared to those processed by injection moulding (Garlotta, 2001; Rahim et al., 2019).
As mentioned above, the success of polymers in FDM printing depends on the printed part's quality and strength. Polymers extrusion temperature, printing orientation, layer thickness, combined with other printing settings such as print speed and building platform temperature, drastically escalate the number of experiments needed to produce guidelines for the optimal mechanical performance of FDM printed parts.
Valerga et al. (Valerga, Batista, Salguero, & Girot, 2018) investigated the relationship between the printing parameters and PLA polymer conditions. The manufactured parts result in dimensional terms, surface quality, and mechanical strength. The obtained results clearly showed that the relative humidity was the most relevant variable, where the stored PLA in an atmosphere with low relative humidity has resisted higher tensile forces. Also, polymer melting temperatures have caused water boiling, which is transformed into bubbles, leading to cracks in the mechanical tests. Jo et al. (Jo et al., 2018) showed that the mechanical properties of PLA printed parts are highly correlated to the layer thickness. However, externally applied heat and pressure have increased the PLA parts performance due to the bond between raster to raster and layer to layer.
Regarding the effect of nozzle diameter, liquefier temperature, extrusion velocity, filling velocity, and layer thickness of PLA printed parts, Yang et al. (Yang, Li, Li, Yang, & Yuan, 2019) concluded that the nozzle diameter and layer thickness are the most influencing factors on tensile strength, surface roughness, and build time of printed parts. Other authors such as Alafaghani et al. (Qattawi, Alrawi, & Guzman, 2017), which studied the effect of process parameters (such as building direction, printing speed, extrusion temperature, layer height) on the mechanical properties of FDM printed PLA parts, concluded that building direction, extrusion temperature, and layer height were more influencing parameters. Despite the several efforts of the research community in controlling and optimizing FDM key process parameters, the intrinsic mechanical behaviour of pure PLA still the bottlenecks that restrict the further development of 3D printing technology. Several attempts have already been made the PLA modification, which involves improving mechanical properties by 3D extrusion of its composite materials.

PLA-based composites
Although not well-established concepts, FDM printing of reinforced PLA, by addition of particle, filler, fibre or nanomaterial, can well-increase the strength of 3D parts.
The latter compounding PLA with graphene and carbon nanotubes has enhanced the mechanical behaviour of FDM printed parts.  Figure 4. A preprocessing of carbon fibres was realized using a methylene dichloride solution containing 8% of partially dissolved PLA particles to improve the interfacial strength between fibre and polymer. The results indicated that modified carbon fibre's tensile strength and flexural strengths were 13.8% and 164% higher than the original carbon fibre reinforced PLA. The modified carbon fibre reinforced samples' storage modulus was also higher than the PLA and original fibre reinforced samples for about 166% and 351%, respectively.
Using   The obtained results from the experimental investigation, using tensile properties, microstructure, thermogravimetric analysis (TGA), differential scanning calorimetry (DSC) and X-ray diffraction (XRD), have clearly shown an enhanced deformation resistance of the composite after adding WF. Also, the starting thermal degradation temperature of the composites decreased slightly. However, the final thermal decomposition residual ratio of the composites increased with no effects on the melting temperature of the PLA 5 wt%.

Inspired from conventional 3-axes FDM printers, the multi-axis Additive Robot
Manufacturing System (ARMS) has been used to achieve dimensional extension, giving designers more design freedom and an extension of the printed part dimensions to meet the industrial application requirements. Also, the additional degrees of freedom (DOF) over a conventional gantry-based machine (3-axis CNC machines) can be used for multi-axis printing to accommodate parts larger than the arm itself (Keating & Oxman, 2013). Additional DOF is useful to reduce production time, as fewer setup changes are required and improved the printed parts' quality (L. Li, Haghighi, & Yang, 2018). The development of an ARMS is challenging as it requires applying reverse kinematics to program the mechanism's kinematic configurations. However, employing an industrial robot is advantageous, where the link and joint motions are already programmed. There are several commercial multi-axis Computer-Aided Manufacturing (CAM) packages for conventional machining, but there is a lack of such for FDM printing (Isa & Lazoglu, 2019). However, sequentially FDM printed parts in discrete build orientations can be re-alized using conventional slicers. Hence, the 3D model can be partitioned algorithmically and fabricated at different build orientations using a multi-axis arm robot (Lee & Jee, 2015).
Researchers have focused on different ways of multi-axial additive manufacturing. Some researchers use multi-axis CNC machines to get more than 3 DOF, while others focus on robot-based additive manufacturing for greater flexibility. Both approaches have their respective advantages. The higher flexibility and scalability of the robotic systems attenuate the effect of their challenging higher initial investment compared to a multi-axis CNC system (Prahar M. Bhatt, Malhan, Shembekar, Yoon, & Gupta, 2020). Table 2 summarizes the most recent research work on robot-based additive manufacturing of PLA parts.
Ishak et al.  integrated an existing FDM extruder with a 6 DOF industrial robot arm to create a 3D printer with a multi-plane layering capability. They have successfully shown that inner and outer toolpaths can be printed in the horizontal plane and vertical planes. Two years later, the same research team developed a three-dimensional lattice structure generator to generate a toolpath for multi-plane FDM printing applications (Ishak IB, 2018). As shown in Figure 6, the geometric input model was printed using a 6 DOF arm platform.  Figure 7. The first arm is equipped with a large diameter nozzle to print interior regions of the part at high build rates. The second arm is equipped with a small diameter nozzle to print exterior regions of the part with a smooth surface finish. The combination of two arms reduces the build time of large parts without sacrificing surface finish. Also, the high flexibility in manipulating the print head enables to deposit of materials on a non-planar surface. Similarly, Wu et al. used a 6 DOF UR3 robotic arm to provide FDM printed parts from PLA filament, with minimal or no support structures. A nozzle with a 1 mm hole was used for quick fabrication, and another one of 2 mm hole was used for enhanced surface quality and more geometric details. A computational tool was developed to decompose the 3D part model into multiple support-free smaller parts using a coarse-to-fine decomposition algorithm, which first segments a model into multiple parts according to the skeleton-based shape analysis. The hardware system allows printing sub-parts, incorporating the collision-free constraint, before applying a fine-level partition to refine the sequence of printing. Zhang et al. (Zhang et al., 2016) used a robotic simulation and offline programming software RobotStudio, an ABB platform for robotic additive manufacturing process simulation and virtual part representation, to simulate FDM with different printing parameters ( Figure 8). They showed that the part building path and bead size could be tuned based on the AM path, FDM equipment, robot model, extrusion head types, configuration, and building temperature and material properties. In another work, Fry et al. (Fry NR, 2020) used two Denso VS-068 6-axis robot arms to investigate: the effect of PLA parts printing at different orientations with respect to gravity, the effect of dynamically changing build orientation concerning the build plate when printing overhanging features, and the effect of printing curved parts using curved conformal layers. The first arm robot was designated as Extruder Robot (RE), allowing the nozzle to be positioned relative to the build plate. The second arm robot was designated as Build Plate Robot (RP), moving the printing plate, as shown in Figure 9. They have proved that the printed part's surface roughness is independent of the print orientation concerning gravity, where the orientation does not affect print quality. Also, dynamically changing build orientation allows overhangs up to 90° to be cleanly printed without support structures. The use of two arms has allowed concluding that curved layers improve an arch's strength, which is steeply curved and printed with the nozzle remaining normal to the curvature.

Conclusion: Challenges and Opportunities
Improving the flexibility, productivity and agility of AM techniques are the keys to the competitive manufacturing industry. Gantry-based printing machines have great potential to reduce time to market, increasing product customization and broadening the design options. However, limitations (i.e. limited product size, built rates and the need for a support structure for regions with overhang) drive researchers to develop enhanced additive manufacturing strategies. Non-planar printing processes using 6-DOF articulated robot arms can significantly expand the capabilities of the additive manufacturing processes. As described in this paper, adding an extra DOF to the additive manufacturing process allows changing the direction of material accumulated during the printing process, material deposition on complex non-planar layers and overhang features without printing support structures.
Unlike conventional gantry systems, robotic arms are ideal for creating large parts. They can be placed anywhere, allowing part production in a controlled environment or a combination of multiple techniques (e.g. robotic systems printing different materials and robotic systems performing inspection tasks) (Bandari, Williams, Ding, & Martina, 2015). Since a single robot can handle various extruding heads or multiple robots can work in the same station, robot-based printing can build multi-material and multi-resolution parts. However, this will require complex motion planning and control where collision-free trajectories need to be generated for the robots under constraints (Prahar M Bhatt, Rajendran, McKay, & Gupta, 2019). Also, robot-based AM setup must ensure high accuracy where they usually use control systems that can potentially need sophisticated compensation algorithms.
High DOF robot setups can be developed using multi-robot systems, which can work synchronously and asynchronously to print the parts faster using conformal and multi-resolution printing. Although multi-robots can successfully coordinate with each other, printing one or multi-part sections will require solving a complex motion planning and control problem. Collision-free trajectories need to be generated for the robots under constraints while the optimum trajectory and process parameters for supportless AM to move the robots efficiently (Prahar M. Bhatt et al., 2020;Urhal et al., 2019). The rapid development of the motion planning algorithm will allow robots to adapt their functions in real-time due to changes in the surrounding environment, where multiple robot-assisted additive manufacturing systems could be used both off-site and on-site. This will also facilitate the development of smart additive manufacturing systems by optimizing the production processes and selecting the most suitable fabrication strategy to produce a part using artificial intelligence (AI).