Slicing Solutions for Wire Arc Additive Manufacturing ^†

Abstract

Both commercial and research applications of wire arc additive manufacturing (WAAM) have seen considerable growth in the additive manufacturing of metallic components. However, there remains a clear lack of a unified paradigm for toolpath generation when slicing parts for WAAM deposition. Existing toolpath generation options typically lack the appropriate features to account for all complexities of the WAAM process. This manuscript explores the key slicing challenges specific to toolpaths for WAAM geometry and pairs each consideration with multiple solutions to mitigate most negative effects on completed components. These challenges must be addressed to minimize voids, prevent bead collapse, and ensure deposited components accurately approximate the desired geometry. Slicing considerations are grouped into four general categories: geometric, process, thermal, and productivity. Geometric considerations are addressed with overhang compensation, corner-sharpening, and toolpath-smoothing features. Process considerations are addressed with start point configuration and controls for the bead lengths and end points. Thermal and productivity considerations are addressed with island optimization, multi-material printing, and connected insets. Finally, tools for the post-processing of generated G-code are explored. Overall, these solutions represent a critical set of slicing features used to improve generated toolpaths and the quality of the components deposited with those toolpaths.

1. Introduction

The study and adoption of wire arc directed energy deposition (Arc-DED), more commonly referred to as wire arc additive manufacturing (WAAM), has grown significantly in recent years [1,2]. This process combines the well-studied properties of the arc welding process with advanced robotics and control to deposit large-scale structures out of molten metals. The rapid expansion of research into the WAAM process relates to its ability to fill a key niche in the space of additive manufacturing processes. Among the metal additive processes, it is uniquely positioned towards the rapid manufacturing of large components as it can deposit fully dense parts without an inert enclosure at a rate that exceeds most comparable metal deposition processes. The WAAM process has also been demonstrated as inherently scaleable to massive geometries through the collaborative use of multiple robots [3].

However, this WAAM technology is limited without a suitable strategy for the generation of toolpaths that ensures completed components are accurately deposited and fully dense. Many WAAM researchers are still heavily dependent on commercial CNC software, manually coded toolpaths, or modifications of slicing software intended for fused deposition modeling (FDM) to generate WAAM geometries [4,5,6,7,8]. Manually coded toolpaths can incorporate WAAM considerations, but they are cumbersome to generate and only suited to simple geometries. Conversely, existing CNC toolpathing or FDM slicing packages handle complex geometries, but they lack the settings to account for the arc welding process. There are a multitude of considerations for WAAM to achieve a high-quality deposition which do not exist in polymer or other metallic manufacturing methods. As an example, most thermoplastics have low melting temperatures, around 105 °C, while mild steel requires temperatures of ≈1350 °C or more to flow. Especially for complex geometries, the existing slicing tools often produce poor toolpaths for WAAM systems and fail to account for the intrinsic features of the welding process.

Therefore, WAAM slicing requires tailored solutions which allow tuning of key toolpath generation parameters that affect the deposited beads. The Oak Ridge National Laboratory (ORNL) has developed the ORNL Slicer which continuously incorporates end-user feedback to implement relevant tuning parameters into the WAAM slicing process. Since each WAAM component may require a different combination of slicing solutions, all features have been developed to be fully cross-compatible with one another. Additionally, all solutions presented in this manuscript have been successfully validated in large scale geometries constructed using WAAM.

This manuscript attempts to bridge the gap between the unique WAAM considerations required to maximize part quality and existing slicing solutions by highlighting tailored features which were developed for use in functional components. It begins with an overview of the WAAM process and the general slicing paradigm utilized to generate WAAM toolpaths throughout. This slicing paradigm employs three fundamental bead types to generate G-code from a solid model: insets, skeletons, and infills. After this slicing overview, the fundamental challenges are outlined which influence the generation of WAAM toolpaths. These considerations are then paired with tailored slicing solutions. One consideration that is not directly addressed within this work is the control of material properties through toolpath selection. Although the impact of toolpathing on material properties has been quantified for different materials [9,10,11,12], it is difficult to account for these properties when slicing without knowledge of the corresponding welding parameters used during deposition. Instead, this manuscript focuses primarily on toolpathing solutions for deposition challenges which reduce the part quality independently of the material properties.

Solutions for geometric considerations include an overhang compensation method as well as corner-sharpening and smoothing algorithms. For process considerations, this work introduces start point configuration options, control of bead segment end points, and control of bead lengths. Afterwards, a method of island optimization is introduced to address thermal concerns as well as solutions for multi-material printing and connected insets to increase deposition productivity. Finally, the manuscript explores options for post-processing of the generated G-code with splicing and UI-based tools to fine-tune generated toolpaths.

2. Additive Process and Slicing Fundamentals

2.1. The WAAM Process

The canonical wire arc additive manufacturing process builds on the well-established principles of manual welding and shares process fundamentals with widely deployed robotic welding equipment. However, unlike the former processes, WAAM is able to generate complex metallic geometries in multiple dimensions composed of many individual weld beads [13]. The typical components found in a WAAM production cell are illustrated in Figure 1. The WAAM process consists of the deposition of independently welded beads deposited in sequence according to the generated toolpaths. The welding process generates an electric arc between an electrode and the substrate, which generates high heat and ionizes the inert shielding gas to generate plasma. The generated heat creates a pool of molten metal in the substrate, and wire is fed into the melt pool to generate a weld bead. Before the deposition of each bead, inert welding gas is purged for 1–2 s before an arc is struck to ensure complete coverage of the weld. A similar purge occurs after the welding arc is extinguished. The mixture of gas utilized for WAAM is a key factor in achieving a strong weld and quality surface finish [14]. After a welding bead has been completed, the welding torch may undergo multiple types of routine maintenance, including anti-spatter spraying, torch reaming, wire cuts, and contact tip replacement. The anti-spatter spray and reaming prevent welding spatter from building up on the nozzle, while wire cuts between deposited beads improve the surface finish of the part. For deposition with gas metal arc welding (GMAW), the welding contact tip also wears out after a certain weld duration and must be replaced before the process can continue. While plasma arc welding (PAW) and gas tungsten arc welding (GTAW) are also used for WAAM deposition, the GMAW process is most widely adopted and was the primary process used to develop the solutions within this work. However, many of the outlined considerations and solutions are still relevant to WAAM deposition with other welding processes.

2.2. CAD to Part Process

Similar to most additive manufacturing processes, the generation of WAAM parts follows a CAD-to-part process, as illustrated in Figure 2. The WAAM slicing process starts with a faceted model, typically a stereolithographic (STL) file, which acts as a universal abstraction of a computer-aided design (CAD) model that approximates all surfaces in the CAD model with triangles. The primary focus of this manuscript is the middle segment of the process which takes the faceted representation of the geometry and slices it into layer-based toolpaths to prepare for deposition on WAAM equipment. Once the toolpaths have been generated, the G-code is typically loaded onto the robot controller of a WAAM system, which coordinates the layer-by-layer construction of the part. At the the completion of the process, a near-net-shape geometry is created, which can either be immediately deployed or post-processed according to application requirements. The shape of WAAM beads typically results in high surface roughness, and the functional surfaces of most parts will be machined before installation. Some efforts have been made to improve the WAAM surface finish in situ, but these techniques are not yet a substitute for finish machining [15].

2.3. Slicing and Toolpath Generation

Before detailing the tailored solutions the WAAM process necessitates, it is important to outline the fundamentals of toolpath generation. First, an STL file is loaded into the slicing software, where the adjustment of specific machine settings and/or process tailoring is performed. Slicing begins on user command, and the software executes a sequence of steps to generate the G-code. The first step in slicing is the intersection of the STL file with a sequence of planes (one for each layer) that are evenly spaced along the z-axis in a typical 2.5-dimensional print. A point is created at each location where these planes intersect the edge of a triangle representing the outer surface of the object. These points are connected together in sequence to form the outer bounding polygon(s) of that layer [16].

After the bounding polygon(s) have been created for each layer, the next step is the generation of layer toolpaths. Toolpaths are usually generated starting with the outside perimeter and working towards the center to ensure dimensional accuracy, but this may vary according to the geometrical requirements of a particular part. In this method, a perimeter bead which traces the bounding polygon is the first deposited bead in a layer. The toolpath for this bead is determined by offsetting the bounding polygon inward by half of a bead width to locate the center line [17]. The bead width is the nominal width of a single bead of deposited material. Parts can contain multiple bead widths corresponding to different process parameters. If any regions are generated where the offset polygon for the perimeter bead intersects with the bounding polygon to create a negative area between contours, these regions are deemed too small to generate a toolpath and eliminated. From here, the slicing software must decide how to fill the remaining area within the bounds of the perimeter bead. Within the slicing paradigm presented in this work, there are three fundamental bead types (as shown in Figure 3), which the slicing process employs to fill the remaining area: insets, skeletons, and infills. Each of these beads serves a particular function in ensuring that layers in the completed part are appropriately flat and without voids.

2.3.1. Insets

Inset beads are the primary building blocks of most WAAM parts. These beads are always printed in closed loops to maximize part strength and are constructed as shown in Figure 4. The perimeter toolpath(s) can also be referred to as perimeter inset(s) as the remaining insets are constructed as offsets from these initial beads. These perimeter inset(s) form the visible exterior surface of the completed part. The perimeter inset(s) are generated by offsetting the part outline by 1/2 the nominal bead width. From here, the remaining inset beads are generated by marching inwards from the perimeter beads with interior insets being offset from an adjacent inset by the nominal bead width.

The user is able to define the total number of insets that are offset from the perimeter beads using slicer functionality. Depending on the amount of material expected to be machined away, additional insets can be added to ensure that the finished part retains a uniform surface finish. Once the prescribed insets are generated, any unfilled area within a layer can be completed with either infill and/or skeleton beads depending on the contour shapes and selected parameters. In some cases, the geometry of a part may promote the sole usage of insets to fill the layer bounds without the introduction of skeletons or infill, as shown in Figure 4.

2.3.2. Skeletons

The skeletonization process is used to fill any small gaps between insets. These skeleton beads are open-loop paths that cover the remaining area where geometric restrictions will not allow for the generation of a closed-loop inset [18]. Skeletons follow the part curvature and, in most cases, will not be generated adjacent to another skeleton. Skeletons are critical to ensuring that all sections within a layer are deposited uniformly to form a fully dense geometry. They fill small internal areas within a layer, which would create voids if additional beads were not added. This is critical to the success of future layers as the existing WAAM process is unable to bridge large gaps while maintaining the arc.

Once a skeleton region is identified where a closed loop inset does not fit, the straight skeleton algorithm generates a path line. First, defined for simple shapes [19] and later defined for arbitrary two-dimensional inputs [20], the straight skeleton, also known as an angular bisector network, is a method for representing a topological skeleton. The straight skeleton is defined through the continuous shrinking process illustrated in Figure 5 in which each of the edges of the skeleton region are moved inward at a constant speed and remain parallel to the corresponding edge of the region. Each vertex of the polygonal region moves along the angle bisector of its incident neighbors. The edges continue to shrink until either their length is reduced to nothing and their vertices merge into one or if an edge intersects with another vertex which splits the shrinking edge [18]. The resulting line segments and vertices located at the center of the skeleton region denote the straight skeleton. The straight skeleton is then transformed into a toolpath for a skeleton bead by extending each end point along its corresponding line segment until it intersects the edge of the skeleton region. This ensures that the generated skeleton bead traverses the entirety of the empty contour.

2.3.3. Infill

The role of infill is to rapidly fill in any area of the part not covered by insets or skeletons to create a fully dense layer. Note that the order of deposition for skeletons and infill is not particularly important as they are generated within different areas of the layer. Skeletons cover small gaps due to geometrical limitations on insets, while infill typically covers larger areas where insets were not added. The primary infill pattern in WAAM is the raster fill, which oscillates between the boundaries of the given area to rapidly deposit material. Infill is also suitable for filling irregularly shaped contours that are not well suited to a closed-loop inset or would require multiple skeleton beads. A key feature of infill beads printed with the raster pattern is the retention of additional heat during deposition, which causes material to flow more within the bounds of the layer. This phenomenon is a result of each infill bead segment being deposited adjacent to the preceding bead segments, which increases heat input within a reduced area. This enables infill to be deposited at a higher rate to rapidly solidify the large internal geometries within a layer. However, this additional heat input generated by these toolpaths may be undesirable depending on the target material properties. In some cases, it may be preferable to use insets to fill the remaining geometry as they are distributed in a way that enables the toolpath to transverse the full perimeter of an enclosed polygon before being deposited adjacent to the previous bead, which reduces thermal input. Depending on the part geometry, the layer can be populated with solely concentric insets or a specified number of inset beads with the remaining area deposited using an infill pattern as shown in Figure 6.

2.4. Slicing Hierarchy

Although insets, skeletons, and infills form the foundation for the toolpaths generated within the sliced part, there are many parameters which control the exact geometry and bead types utilized to fill the bounds of each layer. To maximize flexibility in toolpath generation, there must be appropriate control over the slicing and bead parameters used in each layer of the part. To this end, the slicing process implements a hierarchical model, as shown in Figure 7, which contains three nested levels of control over toolpath generation. Bead settings are the lowest level of the hierarchy and contain the geometrical settings for a single bead profile, such as the nominal bead spacing, layer height, and resultant bead width. This setting also allows for each bead to be tagged with a corresponding weld mode and a torch designation to enable multi-material printing. The next level of the hierarchy encompasses material profiles, which contain settings controlling geometry for the inset, infill, and skeleton beads. Within a material profile, a different bead profile can be assigned to each of the fundamental bead types to account for geometrical variations in deposition. This structure also enables the usage of different materials for different bead types. The top level of the hierarchy includes layer profiles, which determine what material profile is used for a specified range of layers. These layer profiles allow the user to fully customize the slicing parameters as the cross-section of the part varies along the z-axis. Using the three levels of the slicing hierarchy, the user is empowered to tune toolpath generation according to their engineering judgment.

2.5. G-Code Generation

Once all the toolpaths within the part have been generated, the final step in the slicing process is G-code generation. This process utilizes the toolpaths to create machine-specific instructions for the deposition of the 3D object. Prior to this point, the generated toolpaths are typically agnostic to the system used for deposition, but the generation of a G-code file requires that the outputted code be formatted according the expectations of the WAAM controller. This G-code may include position, speed, tool, and deposition parameters that the machine controller interprets in order to accurately build a part. With the initial slicing and generation of G-code being complete, editing tools allow for post-processing, which may include combining G-code or the removal of slicing inconsistencies. Once the finalized G-code is complete, the code can be exported from the slicing software and loaded into the controller for WAAM deposition.

3. Slicing Considerations and Solutions

WAAM is distinguished from other metal additive processes by its use of arc welding apparatus and wire feedstock. This operational modality allows one to easily draw comparisons to the well-studied fused deposition modeling (FDM) process, which is a polymer additive method. However, there are many challenges unique to WAAM introduced by the high temperatures required for the deposition of metals and the complexities of the arc welding process. This section outlines some of the fundamental difficulties associated with the WAAM process and introduces corresponding slicing solutions, which help to address those challenges. Although most WAAM parts will not need to incorporate all of these solutions, most geometries will require one or more of the tools introduced below.

3.1. Geometric Considerations

Many of the most important factors to consider when designing for WAAM relate to geometric variations resulting from the unique shape of an arc welding bead. The slicing process must properly account for this bead geometry to ensure that completed parts are fully dense. Comparing the FDM process with WAAM, there is a key distinction in the shape of the deposited beads, which is illustrated in Figure 8. Unlike FDM beads, which are typically small in diameter and modeled as cylinders, beads in WAAM have a distinctive parabolic shape [21]. This parabolic shape requires the nominal bead spacing as well as the layer heights to be tuned for each combination of material and corresponding welding parameters to ensure that layers are evenly deposited throughout the vertical cross-section of the part.

Another key distinction between polymer deposition and WAAM arises from their differing abilities when printing bridges and overhang structures. WAAM parts are strongly limited in the achievable overhang angle between layers by the tendency of the melt pool to drip when it is not fully supported along the gravitational axis. On FDM machines, bridges can be generated across small unsupported gaps with appropriate deposition settings and cooling. These bridges are currently less feasible on a WAAM machine as the welding arc must make contact with existing material and therefore cannot usually deposit beads across an air gap. Additionally, many FDM machines are able to deposit support structures by utilizing secondary materials [22]. Similar support structures for the WAAM process are an ongoing challenge due to the need to deposit on a conductive material that can easily be removed after a part is completed.

3.1.1. Overhang Compensation

Overhang geometry occurs on perimeter insets where the centerline of the bead being deposited does not vertically align with the corresponding bead from the preceding layer. The angle between a line segment connecting the centerlines of these beads and the vertical axis determines the nominal overhang angle of an overhang bead segment. Similar to FDM additive manufacturing, the difficulty of depositing overhang bead segments increases noticeably with increasing overhang angles. When deposited with a gravity-aligned torch without compensation, overhang geometry can be subject to dripping at angles exceeding 15° due to gravitational effects on the unsupported melt pool [23]. However, compensation methods have been developed to enable the stable deposition of overhang geometry exceeding 90° via an adjustment of the welding torch during overhang deposition [24,25,26,27]. Overhang geometry may also be improved through the use of a multi-axis positioning table. However, these positioners are limited in their maximum payload and build volume. With torch compensation, shifting the toolpath of the outer perimeter towards the part center allows the melt pool to flow into the intended geometry and compensates for the gravitational forces exerted on the bead. Additionally, rotating the torch from its gravity-aligned position to match the angle of the overhang bead aligns the arc force of the welding process to assist in counteracting the gravitational forces.

Within the slicing environment, these methodologies are implemented as an overhang compensation feature, which allows for the addition of toolpath shifts, torch reorientation, or a combination of the two methods. The key terminology relating to this compensation feature is illustrated in Figure 9. This feature tags bead segments that exceed the specified overhang angle and applies the desired compensation. The bead shifting parameter allows detected overhang toolpaths to be shifted inwards by the specified amount towards the adjacent inset bead to avoid degradation of the overhang surface and eliminate dripping. The appropriate amount of toolpath shift varies based on welding parameters and the material being deposited, but it typically varies between 10% and 25% of the nominal bead width with higher overhang angles requiring larger shifts. Another feature reorientates the torch for detected overhang geometry. The rotation angle of the torch from a vertical orientation typically matches the overhang angle, but a lesser angle may be used to prevent the torch from colliding with existing geometry. A transition distance in which the torch rotates between overhang and vertically aligned deposition is incorporated to prevent excessive robot acceleration.

The bead shifting and torch reorientation features can also be combined. Using these compensation features, the part shown in Figure 10 was printed with overhangs up to 35° without collapse of the overhanging surface. However, this is simply one example of a functional geometry constructed with overhang compensation, and the methodology can be effectively extended to overhangs at higher angles [27].

3.1.2. Corner Sharpening

As corners within a sliced part become sharper, such as the 15° corner shown in Figure 11, triangle-shaped voids tend to appear due to the fundamental geometrical limitations of the inset generation method [28,29]. One can see that these gaps could also be filled with the skeletonization algorithm, but the short length of the skeleton beads and the triangular shape of the contour would result in the underfilling and overfilling of different segments of the void, thereby decreasing the geometrical accuracy of a sharp corner.

To mitigate this issue, a feature was implemented which enables the detection and compensation of these sharp corners. The corner-sharpening feature detects corners with angles below a specified threshold and extends the point where two beads intersect towards the adjacent bead, as shown in Figure 12. This feature allows for the extended beads to violate the overlap threshold specified for the inset generation to ensure that corners are accurately filled. Bead segments connecting the legs of the corner which are shorter than the defined close points threshold are eliminated. Then, the merge point is calculated by determining where the axes of the existing corner beads intersect. This point is extended along the axis bisecting the corner by the defined extension length to generate the intersection point for the sharpened geometry. Finally, the sharpened bead geometry is generated by connecting the intersection point for the sharpened geometry to the points on the original geometry determined by the defined sharpening leg length. The portion of the original beads along the sharpening leg length is deleted, which leaves a continuous toolpath containing the sharpened geometry.

Figure 13 demonstrates the practical application of the corner-sharpening method with the slicing and deposition of a star-shaped object with multiple sharp corners. The sharpening feature was applied to the star-shaped object with a corner threshold of 33°, corner extension of 8 mm, leg length of 6 mm, and close points threshold of 7 mm. The physical validation was performed by depositing 410 SS wire with a nominal bead spacing of 4.5 mm, wire feed rate of 170 mm/s, and travel speed of 17 mm/s. Without the corner-sharpening feature enabled, there are clearly voids in the layer where the corners do not properly join. With the feature enabled, the layer is deposited without voids and without overfilling of that layer. The corner-sharpening feature is particularly applicable to ensuring the fully dense deposition of turbine blades, such as those similar to the component shown in Figure 13e, which was completed using this feature.

3.1.3. Smoothing

Another geometric concern with WAAM is the total number of intermediate path points generated by the slicing process and the smoothness of the resulting toolpaths. For larger builds, the total number of points in the resulting G-code can number in the millions, which introduces computational concerns regarding the efficiency of the slicing process. Additionally, there may be geometries which result in jagged bead paths that induce high accelerations during welding deposition. These problems can be alleviated through the use of an appropriate smoothing algorithm, which reduces the number of intermediate points within the G-code to create longer path lines for each bead without a noticeable loss of the dimensional accuracy. Because the WAAM process typically generates beads on a larger scale compared to other processes, some path points within the G-code can be eliminated without a significant impact on the deposited part. Within the slicing environment, smoothing thresholds can be independently set for inset, infill, and skeleton beads. For example, perimeter beads which trace the visible surface of a part may require a greater positional accuracy than infill beads, which mostly fill the internal voids within a layer.

One versatile algorithm for smoothing is the Douglas–Peucker algorithm, which starts by generating a line segment between the first and last points of a bead [30]. Next, it computes the distance of each of the intermediate points within the bead to the line with the furthest point (with a distance larger than a specified threshold value) denoted as a key. This process occurs recursively between keys until each point on the original path line is within the threshold tolerance. The performance of this algorithm is illustrated in Figure 14, with the smoothed slice on the right containing fewer path points, especially within curved segments. The outcome of the smoothing operation is clearly visible within the green perimeter circles where the number of bead segments was reduced from 36 to 20 segments, which is a 47% reduction in path points, while the circular shape was maintained.

3.2. Process Considerations

Individual beads deposited with FDM are generally stable throughout the deposition with minimal variation in the bead shape. Conversely, the welding process cannot be turned on and off instantaneously. There will always be a transition period at the start and end of each bead segment due to the inherent instability of the welding process [31,32]. As a consequence of these transition periods, there is a need for a fine adjustment of the bead geometry and welding parameters towards the start/end of each bead.

Looking at material strength, FDM generates anisotropic parts that are weaker along the print axis due to limits on layer adhesion [33]. This is not typically the case with WAAM. The heat input from a nominal arc welding process will remelt a portion of the material deposited in the previous layer, ensuring that the beads are fully fused. This results in parts being constructed via WAAM displaying quasi-isotropic properties with a minimum strength equivalent to the as-welded standards [34]. However, due to the nature of WAAM beads, the weakest areas in a layer are typically where two ends of a bead join together. In order to avoid weak points in the finished part, the slicing process should maximize the individual bead length and ensure that beads do not join together at the same location in each layer [35]. Finally, polymer additive manufacturing is a clean process where consecutive beads are deposited with minimal maintenance. On the other hand, arc welding generates high amounts of spatter, which necessitates periodic torch maintenance and limits the maximum bead length.

3.2.1. Start Point Configuration

Since the initial and final regions of beads deposited via WAAM experience transient instabilities, placement of the bead ends is critical within each layer. The instability of these end points increases the chance for bulging or dripping of the weld pool, especially on the perimeter insets [31]. Certain bead segments, such as those contained within overhanging geometry, are especially sensitive to weld pool stability. Eliminating start points within these areas potentially reduces the risk of part failure. Additionally, some excess material may be deposited at the end points of a given bead. If successive layers contain vertically aligned starting locations, extra material may build up in this region of the part and possibly result in failure of the deposition. Conversely, there may be hidden or non-functional surfaces of the part where it is suitable to stack start points on top of each other. The impact of start points on geometry is also material-dependent, so different start point selection methods may be preferable for different materials. Controlling the locations of the start points alleviates their impact on part geometry and maximizes accuracy.

To this end, multiple solutions for controlling start point locations have been studied and tested utilizing various geometries. The first feature consists of methodologies for arranging start points in successive layers and includes three options, constant, rotating and random, as shown in Figure 15. Constant is the simplest method with start locations remaining in approximately the same relative location on successive layers. This feature is typically selected if a particular location on the external surface of the part is hidden or otherwise unimportant to the final geometry or strength of the part. The rotating method allows for the start point of the current layer to be set at some specified distance from the previous start point. In this method, the start point will travel a fixed distance around the boundary of the part with each successive layer. Finally, the random setting selects an arbitrary position for start points on every layer. This setting ensures some degree of distribution and is well suited to complex geometries where the length of the perimeter bead varies along the build axis. However, the random distribution can also lead to start points sometimes being grouped together in successive layers.

Another feature that fine-tunes the generation of start points allows for the definition of a bounding box within the x/y plane, which restricts the start point generation to that rectangular area for every layer. Start points will vary within the bounded area according to the fixed, rotating or random selection, but the slicing process avoids generating points outside this area. This setting allows for start points to be consistently placed according to the functionality of the finished geometry. For example, it can prevent start points from being placed near the functional surface of a part or near complex geometry, such as beads printed at a large overhang angle. An example of a hydro impeller deposited with rotating start points that are constrained to a small section of the center hub is shown in Figure 16.

3.2.2. Control of Bead Segment End Points

Since the welding process is known to deposit inconsistently at the starts and ends of beads, an adjustment of the toolpath may be required to ensure that the correct amount of material is deposited. Depending on the material being deposited and its thermal properties, it is necessary to tune where the welding process starts and ends relative to the idealized bead geometry employed by the initial slicing process. Tuning of the end points is also a factor in overall part strength as the lack of fusion at the bead ends leads these areas to be weaker within each layer of the component.

To compensate process inconsistencies, slicing parameters were developed which enable the fine-tuning of bead start and stop points for the material being deposited. The pre-start feature moves bead start points along the axis of the bead away from the nominal start location defined by component geometry. This setting helps account for any discrepancies between the time when the robot motion begins and when a stable arc is established by the welding torch. Correspondingly, bead end points are adjusted with the tip wipe feature which allows welding to continue a short distance beyond the nominal ending location of the current bead, as illustrated in Figure 17. The tip wipes ensure that closed-loop insets are fully bonded and that there are no voids at the ends of deposited beads.

To demonstrate the effectiveness of tip wipes in improving the geometrical accuracy of a part, two circular layers composed of 14 insets were printed with the same process settings, one with tip wipes disabled and one with them enabled. Note that it is difficult to compare start points 1-to-1 in the experimental layers since the generated start points are not aligned between the compared layers, so averaged deviations across the layer were utilized instead. Deviation maps of the two slicing configurations are shown in Figure 17 with all 14 of the start point locations denoted. It is clear that, while some geometric variation is still visible after the addition of tip wipes, there is a clear reduction in the amount of deviation from the nominal layer surface. As recorded in

Table 1. Comparison of the deviations at start point locations and the layer surface roughness with and without tip wipes.

	Without Tip Wipes	With Tip Wipes	Reduction in Deviation
Average Maximum Deviation at Start Points:	−0.98 mm	−0.32 mm	67%
Std. Deviation from Nominal Surface:	0.96 mm	0.86 mm	10%

, when tip wipes are enabled, there is a considerable decrease in the average maximum deviation for the start points and a significant improvement in the surface roughness as measured by the standard deviation of the exposed surface layer.

3.2.3. Control of Bead Lengths

In addition to controlling the deposition at bead end points, it is important to control both the minimum and maximum length of beads and the placement of beads to fill gaps between toolpaths. In the case of WAAM printing, it is generally undesirable to place a bead within a gap that is less than half of the nominal bead spacing. Assuming that bead parameters have been properly tuned, this gap will be fully encapsulated by the succeeding layers, while attempts to deposit additional material in a narrow gap would likely result in excess deposition. Additionally, a large number of extraneous small beads considerably increases downtime during deposition while also increasing the computational load and memory requirements of the sliced part.

With this in mind, a key aspect for slicing is the ability to set minimum and maximum threshold lengths as well as gap thresholds to control bead generation. Independent parameters can be set for insets, infills, and skeletons to allow for finer control of which bead type is generated within small contours. The minimum length threshold allows for the filtering of extraneous short bead segments, and it is illustrated for skeletons in Figure 18a. For infill beads, there is a secondary parameter which controls the minimum length for each bead segment within the raster pattern. Additionally, skeleton beads also employ a minimum and maximum gap threshold, which is illustrated in Figure 18b. Gaps larger than the maximum threshold will be filled with inset beads, gaps smaller than the minimum threshold will generate no bead, and only gaps in between these user-specified values will generate a skeleton bead. Finally, there exists a maximum bead length threshold to break long beads into smaller segments to address the maintenance needs of the welding process. The maximum bead length threshold ensures that all beads are shorter than the specified threshold and allows for the system to travel for torch maintenance before finishing the remaining area within the layer. An exaggerated example with a maximum bead length of 300 mm used to split a long infill bead into multiple shorter beads is shown in Figure 19. In a practical scenario, the maximum bead length is configured to some value less than the toolpath length corresponding to the wear time for a single contact tip (typically ≈30 min for GMAW, depending on process parameters).

3.3. Thermal Considerations

Managing thermal input is key to deposition in all metal additive processes. WAAM, in particular, is very sensitive to thermal conditions because of its high deposition rates and power input compared to other processes [36,37]. Most thermoplastics have low melting temperatures, i.e., ABS melts at ≈105 °C, while steel has a minimum melting point of ≈1350 °C. Unlike FDM deposition where the heat input is low and the bead cools to a steady state temperature, the heat input from WAAM deposition conducts through all of the prior layers of the part. Without downtime between layers, the increasing thermal mass means that beads take an increasingly long time to cool. This phenomenon can result in beads in the upper layers overheating, which may cause the melt pool to flow or drip before it has sufficiently solidified, as exemplified in Figure 20. The thermal sensitivity is also alloy-dependent, with most low-carbon steels having poor high-temperature stability and stainless steels exhibiting improved high-temperature stability [38]. In some cases, wait times between layers or temperature thresholds are sufficient to prevent overheating at the cost of increased print duration [39]. Additionally, some alloys are sensitive to thermal gradients within deposited material, and heating the part unevenly may induce thermal cracking. Therefore, WAAM slicing should account for inconsistent thermal inputs to balance the heat distributions in deposited components.

Island Optimization

Islands are isolated segments of a layer that are not connected to other parts of the layer, as illustrated in Figure 21. These islands exist where part geometry results in a fully connected layer diverging into multiple disconnected segments or where disconnected segments may join together in one of the succeeding layers. The order in which these islands are printed is particularly important to managing thermal input in WAAM as it affects the heat distribution within the previously deposited material. Typically, the smaller an individual island is, the more susceptible it is to overheating and the greater the need for thermal management.

One solution to control the thermal input for island geometry is the incorporation of three methods of optimization (closest, furthest, even) for selecting the order in which islands are printed in successive layers. These selections represent a trade-off between minimizing travel times when printing a part and allowing the heat to dissipate from the most recently printed island. Figure 22 illustrates the order is which each optimization method would print four islands over two successive layers, assuming that the robot starts at an origin near Island 1. In the closest method, the first island printed on a new layer is the shortest distance from the last island of the previous layer. This method minimizes the amount of robot travel between layers, but it does not allow time for the previous layer to reach a thermal equilibrium. Typically, the closest method would be selected for large geometries where minimizing print times is of critical importance. The closest method may be an appropriate choice when each of the island contours cover a sufficient area. However, this selection is somewhat problematic when printing small islands in materials with poor high-temperature stability as the first bead in the succeeding layer may be deposited over the bead that was most recently printed. In that case, depositing new material on a recently printed island may inhibit proper solidification of the newly deposited bead and cause the melt pool to drip, eventually leading to layer failure. As an additional concern, depositing the next layer after different amounts of cooling within each island can vary the microstructure within that section of the geometry. It may also result in uneven layer heights as the deposited material will flow more on the hotter areas of the part.

Another method of optimization for islands is the furthest method, which selects the island that is the greatest distance from the last island completed. The furthest method is a good choice for reducing overheating on complex parts where the number of islands changes often between layers. However, the furthest method may still select to deposit material on an island that was recently printed, depending on where the welding torch finishes the previous layer. This method also results in the highest travel time for the robot between islands.

The final method of optimization, even, prints the islands in the same order on every layer. This method strikes a balance between the furthest and closest methods by ensuring that each island will always having a cooling time equal to the time that it takes to complete each layer. The even method is the best selection for balancing thermal input without noticeably increasing printing times, but it may be less effective in geometries where the number and location of islands changes across layers. Since the even method results in consistent heating of the part across each layer, it is typically the preferred solution to minimize hot cracking and overheating.

3.4. Productivity Considerations

Productivity in WAAM is a representation of the overall efficiency at which parts are constructed and impacts the total time required to complete a given geometry. One common metric for productivity is the mass of metallic material deposited per unit of time. Another productivity metric is the percentage of time spent depositing material compared against downtime. Productivity plays a bigger role in the WAAM process than with FDM, as net productivity for the polymer process is primarily a function of nozzle diameter since there is minimal downtime. One factor that has a large impact on WAAM productivity is the interval between services for wire cuts and torch cleaning. This maintenance interval has an outsize impact on productivity because no welding can occur while the torch is being serviced. Also, travel moves, which occur between beads and during service, add to print times and decrease productivity. Therefore, minimizing the frequency of maintenance and travel moves can significantly improve productivity.

Productivity can also be a function of the bead profiles utilized for generating toolpaths within a given layer. The closed-loop structure of inset beads typically defines an upper limit on how much of the layer can be filled by a single inset. Alternatively, the open-loop raster pattern of the infill bead can fill a much larger area without stopping and starting the welding process. Maximizing the length of each bead and the time spent continuously welding will result in a corresponding increase in productivity. Finally, the bead width in WAAM is a variable function of the welding parameters which allows for different bead widths within the same layer with some trade-off existing between the deposition rate and resulting part quality. In areas where geometrical accuracy is less critical to a part, such as large internal regions, increased productivity can be achieved by depositing thicker beads.

3.4.1. Multi-Material Printing

A key advantage of WAAM over traditional methods of manufacturing metals is its ability to leverage multiple materials deposited within a solid geometry [40,41,42,43,44]. One application of multi-material deposition is to reduce costs by introducing less expensive materials in non-critical areas of the part. However, using multiple materials also enables enhanced productivity when one material can be more rapidly deposited than another one. An example of this flexibility is shown in Figure 23, where the structural inset beads are composed of 410 SS, with the remaining internal geometry being filled by ER70S-6 infill beads.

One challenge that arises from this form of multi-material printing is that the different deposition parameters utilized for each material result in varying layer heights. If the same nominal layer thickness was utilized for slicing each of the materials, eventually the height of the two areas of the part would diverge. Here, the proper slicing solution is to slice each material according to its nominal layer height and then merge the layers accordingly as shown in Figure 24. When the vertical position of a layer for Material 1 matches the position of a layer for Material 2, those layers are combined into a single layer. Otherwise, the layers for Material 1 and Material 2 are ordered into a combined sequence of layers according to their relative positions along the vertical axis. This process is similarly extended to any n number of materials being deposited.

3.4.2. Increasing Productivity with Connected Insets

As previously defined, insets are closed-loop beads which define the boundaries of a part and can sometimes compose all toolpaths within a layer. However, the closed-loop nature creates a geometrical limitation on insets as the maximum bead length is fixed by the length of the inset contour. In many cases, maintenance must be performed on the welding torch, and the robot must perform travel moves before depositing consecutive beads, which affects overall productivity. This means a layer composed of concentric insets will take considerably longer to print and undergo more thermal cycling compared to an equivalent layer filled with an infill pattern. Additionally, since the welding process is not continuous, each inset bead will experience the transient effects that occur when stopping and starting the weld. These transient effects result in less consistent bead geometry throughout a layer composed mostly of insets.

The solution to these problems is a slicing feature referred to as connected insets. Connected insets allow the toolpath to cross between concentric insets from the end of the current inset to the start of adjacent insets without an interruption to the welding process. As the welding torch reaches the end point of the current inset, it rapidly crosses over to an adjacent inset bead at a specified angle (typically 45°) and continues welding. This process continues until the welding torch has reached the final concentric inset within that layer. The angled crossover between insets allows for the ends of the previous inset to be fully joined while also ensuring that no excess material is deposited at the junction of the two inset paths. The connect inset feature greatly improves the functionality of inset geometry and allows insets to be deposited with an efficiency approaching that of an infill pattern. Simultaneously, connected insets allow the part to maintain a more even heat distribution compared to infill as the toolpath traverses the entire closed loop path before depositing material next to a recently welded bead. This contrasts the infill raster pattern where a sequence of linear bead segments are deposited adjacent to one another. Finally, surface uniformity is improved due to the minimization of starts and stops during the welding process.

A comparison of print times for a circular geometry with and without connected insets was conducted with the sliced geometry and resulting deposition shown in Figure 25. The time to deposit a single circular layer with connected insets being enabled was compared against the filling of the layer with disconnected insets similar to the toolpath shown in Figure 17d. Times were compared for two operational modalities: one with a wire cut service operation between each bead and one without. Looking at the layer times recorded in

Table 2. Comparison of the total time required to print a layer with and without connected insets for the geometry shown in Figure 25. Times were compared with and without a wire cut service step between individual beads.

	Without Connected Insets	Connected Insets	Reduction in Time
With Wire Cuts:	487 s	234 s	52%
Without Wire Cuts:	364 s	218 s	40%

, geometries with many concentric insets can see reductions in layer times of over 50%. It should be noted that even a small reduction in layer times can have a dramatic impact on the total print times when propagated across the many layers composing large parts.

3.5. Solutions for Post-Processing

Typically, post-processing operations within a slicing environment are defined as those that occur after initial G-code generation has been completed by the slicing process according to the defined parameters. However, the slicing process is rarely perfect, and the resulting toolpaths may necessitate minor refinements, simplifications, or corrections to the generated G-code using post-processing tools. A G-code splicing tool can automatically combine multiple G-code files into a single file while accounting for varying layer heights. In addition, UI-based tools allow the user to fine-tune the bead geometry and eliminate small errors that might exist after the generation of the initial slice. Used in combination, these tools ensure that the final slice written to the outputted G-code is fully robust before being transferred to the WAAM system.

3.5.1. G-Code Splicing Tools

While the intent of the slicing process is to generate the complete G-code for a given part or parts without additional operations, there may be some cases where the multiple G-code files must be joined to print multiple parts simultaneously or to combine different portions of a complex geometry. Printing multiple parts simultaneously increases productivity by reducing setup times while also allowing additional cooling time for parts that are sensitive to thermal input. To fulfill the need to combine G-code, a splicing tool can be incorporated into the slicing environment, which automatically combines any number of previously generated G-code files into a single file. Layers with matching vertical positions are merged to allow for simultaneous printing, and the remaining layers with differing vertical positions are sequenced in an order corresponding to their heights, as highlighted in Figure 26. While multiple parts can be sliced together within the slicing environment, splicing G-code is critical when the user may not have access to the original CAD files utilized in generating the separate G-code files. In addition to printing multiple parts simultaneously, this feature enables the user to merge multiple slices corresponding to different segments of a single geometry. This capability allows the user to divide a geometry within CAD software when finer control of the slicing is required and then combine the sliced segments into a single G-code file. In particular, this need arises with parts where multiple materials are unevenly distributed within the layers, limiting the applicability of the aforementioned multi-material bead assignment.

3.5.2. UI Tools for User Tuning of G-Code

While the majority of slicing and toolpath generation should be automated according to the selected parameters, any remaining compensation can be manually implemented as a last resort through a suitable UI interface within the slicing environment using tools such as the ones shown in Figure 27. The move and rotate bead features allow for the relocation or reorientation of beads to ensure that beads are located in the optimal location for deposition. If a few segments of a bead are shifted, the slicing engine will attempt to adjust the connected bead segments to ensure that the toolpath remains continuous. The delete bead tool allows for the deletion of any beads or bead segments which are extraneous and automatically turns insets with deleted segments into skeleton beads. The insert bead tool attempts to insert a skeleton bead at the user-specified location. The break beads feature allows the user to break a bead segment into two shorter bead segments, and the slicing engine adjusts the G-code accordingly. Used in combination with the delete bead segment feature, the bead-breaking feature allows the user to precisely trim away any portion of a larger bead.

While many UI tools deal with modifying the bead geometry, other tools allow for adjustment of the toolpath to account for process considerations. The start point tool allows for the manual adjustment of any start point locations that the user has determined as being in an undesirable location. When selecting an existing bead, the slicing engine will relocate the start point from its current location to the selected location and also regenerate the existing G-code so that the rest of the bead is ordered according to the adjusted start point. The connect inset tool allows for the selection of bead segments from two separate insets and generates a connected inset feature between them. Finally, the reverse bead tool allows for the reversal of the travel direction of any toolpath to allow for fine-tuning of the thermal input or geometrical variations resulting from that bead.

4. Conclusions

Without prior knowledge of the WAAM process fundamentals, one might assume that toolpath generation is nearly identical to the slicing methodologies applied to other additive methods, such as FDM. However, this is generally not the case as there are many unique process considerations specific to WAAM that are not properly addressed in general-purpose slicing solutions. To maximize the quality of parts deposited using the WAAM process, a tailored slicing solution must account for the unique process considerations associated with this type of additive manufacturing. One must consider the full spectrum of geometric, process, thermal, and productivity factors that impact WAAM deposition.

The slicing solutions that were outlined within this manuscript inform the development of a slicing process, which maximizes the accuracy and success rate of geometries produced with WAAM. Solutions for specialized geometries, such as overhangs and sharp corners, ensure that the slice is properly tuned to improve accuracy within these features. Control of the start point locations, bead end points, and bead lengths helps to account for the inherent inconsistencies of the arc welding process. Island optimization allows for the control of thermal input to manage thermal gradients and prevent overheating. Connected insets and multi-material printing maximize the productivity of the WAAM process during deposition. Finally, post-processing tools enable fine control over the tuning of toolpaths after the generation of the initial G-code. Taken together, these slicing solutions enable the creation of complex parts far beyond the capabilities of a naive process-agnostic approach to toolpath generation.