Deposition strategies for generating cuboid volumes using extreme high-speed directed energy deposition Open Access

With laser-based additive manufacturing methods maturing and their range of applications steadily growing but also through tightening regulations to implement sustainability targets and constant pricing pressure in manufacturing, the demand for highly efficient, environmentally friendly, and high-performance, high-quality processes is increasing constantly, and significant growth potential for this demand is projected.^1,2 Against this backdrop, extreme high-speed directed energy deposition (EHLA) on rotationally symmetric components has become well-established in industrial applications such as coatings of hydraulic cylinders, especially to increase wear and corrosion resistance in the oil and gas industry, and the application of functional layers, such as coatings for brake disks.^3,4 With this background, the goal is to use the advantages of EHLA not only for single-layer coatings but also for multilayer additive buildup. To achieve this goal, challenges must be solved not only in the process development domain but also in machine development, since the high speeds characterizing EHLA need to go hand in hand with high acceleration and precision.⁵ 

Arguably, if implemented successfully on suitable hardware, such a high-speed process variant, has the potential to combine the advantages of high build rates and fine buildable structures, as illustrated in Fig. 1.

In a previous experimental study, it was first demonstrated that EHLA 3D can be used for the additive manufacturing of 1.4404 cuboid volumes. It is possible to build up virtually pore-free volumes at speeds up to 50 m/min. The mechanical properties that can be achieved are on the level of conventionally manufactured probes.⁵ Also, this preliminary study served as a foundation for the conceptualization, development, building, setup, and initial operation of a high-speed tripod system in cooperation between ponticon GmbH and Fraunhofer ILT. With the system now available, this study marks the start of a more specific experimental work on the novel process variant of EHLA 3D.

Near net shape buildup is a key factor to the economic efficiency of additive manufacturing processes, as postprocessing is often necessary before putting a component to use. The closer the additively manufactured part is to the final component, the fewer postprocessing steps and materials are required and the costs per part decrease. When building large volumes, the achieved accuracy of the individual layer is a key factor contributing to overall build success due to the error propagation of defects and height differences from layer to layer.

For directed energy deposition (DED-LB), previous studies have shown that by varying contour-hatch strategies in the build process and by generally varying buildup strategies, it is possible to produce parts that are closer to the final contour. In one approach, the contour is run twice per layer and a compensation layer is applied after a certain number of layers.⁶ In a different study, the highest final contour accuracy was achieved using a spiral strategy from the inside to the outside.⁷ 

Heat input during volume buildup is of similar importance as the geometric buildup strategy. Not only does the thermal history influence heat-induced part distortion, but it also affects the microstructure and, hence, the mechanical properties of the built part. When the process window is narrow, tailoring heat input, e.g., by trying to achieve a distribution as uniform as possible, can determine whether the part can be successfully built at all, as well as influence the occurrence of residual stress and defect formation.

With both geometric and heat input strategies being important for building functional parts, simulation is an attractive option to avoid entirely experimental approaches. However, due to the DED-LB process involving solid, liquid, and gaseous phases as well as a complex interaction of feed gas, shielding gas, and powder particles in the formation of the powder gas jet, simulation efforts up to now cannot fully replace the experimental approach. An existing simulation model for DED-LB,^8,9 which has also been tested for EHLA in a pilot experimental comparison study,¹⁰ relies on simulating a multitude of individual particles. The particle density is modeled based on measurements of the powder gas jet with respective shielding and feeding gas settings using a patented setup¹¹ that relies on a coaxial camera and a lateral illumination line laser.¹² The laser intensity distribution can be measured with state-of-the-art monitoring equipment or alternatively be modeled e.g., as a super-Gauss distribution. With these prerequisites, particle trajectories are evaluated with an integral over laser intensity yielding total energy input, hence, temperature for each particle. The energy transmitted through the powder gas jet to the surface is calculated as well, again yielding a local temperature. In conjunction, the temperature of the surface and the particle is used to predict a particle's contribution to track formation.

Even more than DED-LB, the EHLA process, which is characterized by more transient deposition mechanisms and more complex phase transformations, is difficult to implement for robust predictions of real-world experimental results. The physical process complexity in EHLA is manifested in the solidification and material buildup process on the substrate surface but also in the interaction zone between the powder gas jet and the laser beam. This interaction zone is decisive for the outcome of the EHLA process and is characterized, just like the buildup zone on the substrate, by an interplay of solid, liquid, and gaseous phases.

Solid powder particles are molten, beginning on the top side from which the laser radiation hits them,^8,10 that is before taking into consideration the possible spin of particles in the powder gas jet. Crucially, particles may be solid on their lower side, partly molten by risen temperature through laser beam absorption, and may partly evaporate if they absorb enough energy. This effect of partial evaporation leads to particles being accelerated by repulsion.^13–15 This mechanism affects the interaction time of the concerned particles in the laser beam and possibly through collisions, with the surrounding particles. High-speed camera footage of the powder gas jet indicates that the propulsion effect causes a significant number of particles to escape the powder cone, hence not contributing to track formation.¹⁰ Despite its importance for the correct modeling of DED-LB and EHLA processes and likely due to its complexity, the propulsion effect has not been incorporated into state-of-the-art simulation models. If implemented in future models, it is expected to significantly influence simulation results and help to better match experimental results.

Few other simulation approaches for DED-LB besides the one mentioned above also exist and manage to make qualitative predictions by using multiphysics simulation approaches, however, do not come close to a process prediction that can be universally used to solve real application tasks.^16,17 In face of these current difficulties of process simulation, experimental studies remain indispensable. By compiling a catalog of base parameter sets or process parameter ranges (process windows) within which desirable results are to be expected for a certain case and geometry, EHLA 3D as a novel AM process can be made more accessible to new users seeking easy setup and transferability, and for industrial applications. At the same time, beginning with simpler geometries and materials sets the stepping stone for more complexity further downstream.

Hence, as a first and most simple geometric shape, this research focuses on the buildup of cubes and cuboid volumes.

In line with the preceding study,⁵ powder material 316L-A by Oerlikon Metco and hot-rolled 1.4301 plates as substrates were used. The substrates were sandblasted prior to the experiments.

In Table I, the composition of the austenitic steel 1.4404 (316L) in wt. % according to DIN EN 10088-3 is given.

According to the manufacturer's specification, the particles are spherical and their size is between 15 and 45 μm.¹⁹ The liquidus temperature of 1.4404 is 1448 °C; specific heat capacity at 20 °C is 500 kJ/(kg  K); and the hardness is around 225 HV at 20 °C for solution-annealed block material. Oerlikon Metco specifies a hardness in the range of 193 to 239 HV after PBF-LB (laser powder bed fusion) processing,¹⁹ with the possibility to increase it to 950 to 1600 HV through hardening processes.²⁰ 

All experiments were conducted on the high-speed 3-axes handling system pE3D by ponticon GmbH. The maximum speed of the pE3D system is 200 m/min and the maximum acceleration (for typical applications within the working envelope, a cylinder of around 700 mm diameter and 800 mm height) is 50 m/s². The powder feeder model used was a Twin-150 by Oerlikon Metco. A Trumpf BEO D70 welding optics with motorized collimation was used, which is connected to a Trumpf TruDisk4002 disk laser with a maximum laser power of 4000 W, a wavelength of λ_L = 1030 nm, and a beam parameter product of 8 mm mrad, using a fiber with 200 μm core diameter. A coaxial powder nozzle of the type ILT-Aachen-EHLA-Coax D40 was used.

In the preceding study on a different machine and laser setup, process parameter sets for different speeds were selected.⁵ In this work, these process parameters were varied and adapted to the given system technology to examine process stability at different speeds. The buildup volumes were analyzed with regard to density, process powder deposition efficiency, visual process appearance, and perceived stability. The main process parameters were selected in an iterative process with adaptions based on these criteria. The parameters used throughout this study are given in Tables II and III.

Cubes with an edge length of about 15 mm (nominal) were built using different geometric buildup strategies. The following buildup strategies were examined:

• hatch strategy with rotation by 90° after every layer (Hatch 0°–90°),

• hatch strategy without rotation (Hatch 0°),

• contour-hatch strategy with rotation by 90° after every layer (Contour-Hatch 0°–90°),

• contour-hatch strategy without rotation (Contour-Hatch 0°),

• contour-hatch with rotation after every layer and with different numbers of contour runs per layer,

• spiral from outside to inside, and

• spiral from inside to outside.

The buildup strategies are illustrated in Fig. 2. After initial evaluation of all strategies, cubes were built using all strategies except the spiral variants. The speed of the contour varied between 20 and 60 m/min.

Due to the high speeds and the requirement to deposit at constant velocities to obtain weld beads of constant size and a near-stationary deposition, the turn pattern in Fig. 3 was used in every corner to ensure the right speed and accuracy for the deposition welding process while the laser was on.

The geometric shape of the deposited cubes was evaluated by taking measurements of the cubes with a VR 5200 laser profilometer by Keyence. The measurements were compared to the original CAD data to obtain the deviation.

After the cubes were built, a thermographic image was taken about 3 s after completion of the build process. The position of the measurement is shown in Fig. 4. The camera used was of the type ImageIR 8300 by Infratec.

Key experimental results of the geometric buildup tests and the thermographic evaluation are presented in the following section. A special emphasis is laid on the comparison to conventional DED-LB processes to facilitate comparability of the performance against established AM processes.

As can be seen in Fig. 5, the melting tracks of the different buildup strategies were analyzed. The spiral-like strategies were not suitable for further experiments using EHLA 3D with the available system technology, since gaps appeared in the corners, where apparently, the machine setup including the laser beam source and the communication interface between machine and laser did not allow fast and precise enough on/off switching. Similarly, an apparent limit to the minimum track length of welding could be observed, which was too long for the use of spiral-like strategies. The hatch and contour-hatch strategies were examined more deeply in the following experiments.

The aim of contour-hatch strategy is to get smooth, even and upright surfaces on the outside and to build the inside with high productivity. With the separation of these two parts, different parameters can be used to achieve both targets, e.g., thicker tracks for filling the inside of a volume and fine tracks to get an outside surface that needs little or no postmachining.

The following geometrical parameters were used:

Buildup strategies without rotation resulted in asymmetrical cubes. When building the hatch first, material accumulation was noticeable in the corners, which did not happen when building the contour first.

To approach near net shape deposition, the number of contour runs (on top of one another) per layer was varied. Figure 6 shows that the best result was achieved with 1.5 contour runs per layer, realized by alternating layer-wise between a single contour run in one layer and two runs on top of one another in every other layer.

To validate this outcome, cuboids with double length and double height were built using 1.5 contours per layer. Examples for the two cases and the corresponding micrographs can be seen in Fig. 7. The cuboids had relatively even surfaces on the sides and the upper face. Especially, the cuboid with double height showed that the flatness is constant over the entire height with 1.5 contours per layer, further proving this strategy to be beneficial.

On the vertical surfaces of the cubes, diagonally inclined, porous, loosely attached material buildup can be observed in the micrographs. This effect seems typical for EHLA 3D, as it has been observed in several experiments, also for different materials. Often, this excess material can be removed easily, e.g., using a wire brush, to obtain a smoother surface.

In a final experiment series with one contour run per layer, the deposition speed was varied for the contour run, resulting in differently sized weld beads for the contour. The goal of this experiment series was to understand the influence of these differently sized weld beads on the ability to deposit near net shape cubes, in particular, vertical, flat wall surfaces. At 20 m/min, the upper edges of the deposited cube turned out too high for a flat surface (see Fig. 8, top left). At speeds of 40 m/min and faster, the edges turned out more and more rounded and the walls' evenness and verticality deteriorated (Fig. 8, bottom left and right). A speed of 30 m/min resulted in a flat top surface. By again adjusting the number of contours per layer, a flat surface could be built with all speeds from 20 m/min to 60 m/min. Hence, adjusting the number of contours per layer is again proven to be an effective approach to compensate for a process parameter setup (i.e., resulting in weld bead geometry) that would otherwise lead to imperfect wall buildup.

To compare the results of the 1.5 contour-hatch strategy to the results from an established AM process, additional cubes were built with DED-LB. Of all variants tested in this study, the cubes with 1.5 contours per layer showed the best flatness on the top surface. Based on the visual inspection of etched-sample microscope images, it was found that the microstructure of EHLA 3D resembles more to PBF-LB than to DED-LB. This effect might be explained by the fact that the cooling rates in EHLA processes are at ca. 10⁶ K/s comparable to those of PBF-LB rather than DED-LB, where cooling rates are 1–2 orders of magnitude lower.²¹ In the etched micrographs, two of which are shown exemplarily in Fig. 9, a different grain structure can be observed between EHLA 3D and DED-LB. While for EHLA 3D, epitaxial grain growth can be identified with grains growing over several layers, in the DED-LB sample, grain growth occurs mainly within single weld tracks.

With the exemplary samples presented in Fig. 9, the sharper edges of the EHLA 3D cuboids become apparent. Crucially, in this case, the EHLA 3D samples have a build time of 11:20 versus 7:40 min for DED-LB. This is due to the higher number of layers (80 versus 14), leading to many acceleration phases, i.e., dead times where no powder is deposited; and is aggravated by the fact that the powder feed rate in the EHLA 3D sample was almost twice as high (29.6 versus 15 g/min). Powder deposition efficiency was not measured for these two samples because it was not the optimization target. Generally, process-related deposition rates (not considering machine-related dead times depending on geometry and build strategy) above 95% have been achieved in the course of this work for EHLA 3D.

With higher numbers of contour runs, the surface temperature of the cubes—measured about 3 s after completion of deposition—decreased. Without contour runs, the final temperature was around 480 °C. With two contour runs per layer, the temperature was around 410 °C (see Fig. 10).

These results suggest that the final temperature is mainly dependent on the travel time without laser. The turn pattern, which is needed for the contour runs, has a large ratio of laser-off to laser on time.

Effectively, the cubes were too small and the process too long to see bigger local temperature differences in the x- and y-direction.

In this study, most samples were built only once, so the results have a low confidence level. Also, the experiments were limited to a single powder material. Nevertheless, trends should be independent of sample size and material, and to some extent, transferrable and universally applicable for EHLA 3D.

The individual results are hard to directly transfer onto different DED systems, since the system technology components like powder nozzle, laser beam source, handling system, or powder feeder have a significant influence on the process outcome. However, as shown through the comparison with DED-LB, general findings are transferable.

Because of the high deposition speeds, the switching time of the laser is critical for small features, which led to the conclusion that deposition using spiral-like strategies is not feasible with the used system. With better understanding and technological changes to improve switching times, further experiments with the spiral strategies could be reasonable and lead to new findings.

The absolute temperatures of the thermographic images can deviate from the actual values because the surface condition, the angle of exposure, and the material, influence the measurement. Such a measurement deviation could also be the reason why the process with the buildup strategy Contour-hatch 0°–90° and one contour does not follow the otherwise recognizable trend. The thermographic image for this strategy was taken at a slightly smaller angle than the others due to a slightly different camera orientation in a different measurement campaign. The thermographic images are suitable for comparing different buildup strategies and process variants with each other, as the absolute temperatures are irrelevant if all images for the relative comparison were taken under identical conditions. In addition, comparative measurements with a digital contact thermometer proved that the temperature values obtained from the thermographic system were sufficiently accurate (max. ±10 K deviation).

An analysis of the microstructure of the built-up cubes suggests that the temperature differences in EHLA 3D have no noticeable influence on grain growth. On the other hand, there is a big difference between the samples built using EHLA 3D and those built using DED-LB. Since both process variants differ greatly regarding thermal input and cooling rates, an influence on the resulting microstructure is immanent. The grains in EHLA 3D are narrow and grow over many layers in the build direction. With DED-LB, the grains grow toward the center of the individual weld tracks and only grow beyond a few layers. They are larger than in EHLA 3D and less elongated. Recent research also suggests that the EHLA microstructure resembles more PBF-LB than DED-LB due to the high cooling rates.^21–23 

The exemplary comparison between an EHLA 3D sample and the same geometry built with DED-LB reveals that the “high-speed” process may, in fact, be slower regarding total build time. This is due to the thinner layers and higher lateral speeds that require intermittent acceleration. For small geometries, the high number of acceleration times and the requirement to accelerate to and from high-end speeds can mean that the biggest part of the build time is spent in those dead times where no powder is deposited. The counterintuitive contrast presented helps to understand how EHLA 3D processes must be planned with the geometric situation in mind, when small structures are greatly influenced by dead times. In the case of the given 15 mm cubes, a way to prevent this inefficiency for productive use is to build multiple cubes next to one another so that only in small gaps between the cuboids, the laser is switched off, and only before and after each row, an acceleration or deceleration phase is needed. On the contrary, for long and thin structures that can be built alongside their spine, this problem is not relevant.

In industrial use and for productive application, a combination of DED-LB with EHLA 3D for filigree areas of the part to be built may be advisable. As an example of such a combination, consider a near net shape part hull (contour) and a high-power DED-LB inner volume filled with larger weld beads and thicker layers (hatch).

Several buildup strategies that are common in DED-LB were compared regarding their applicability for precise, near net shape deposition. Spiral buildup strategies turned out not to be transferable to EHLA 3D due to process interruptions in the corner points, which were inevitable with the high feed rates and restrictions to the minimum length of the weld tracks. The well-established contour-hatch strategy, however, was successfully transferred to EHLA 3D. By varying the number of contour runs per layer and the orientation of the hatch tracks, the geometric accuracy of the deposited volume was significantly improved.

The best result concerning geometric accuracy and flatness and verticality of the cube walls was achieved with 1.5 contour runs per layer and a layer-wise alternating 0°/90° hatching, with the contour runs executed prior to the hatching of a layer. Other buildup strategies resulted in rounding or protruding of the upper edges of the cube. Without the 0°/90° orientation change, the cubes turned out asymmetrical. By adjusting the number of contour runs, near net shape cubes can also be built with contour speeds that deviate from the hatch speed. Still outstanding, subject to future improvements regarding laser switching times and precision, are spiral-like strategies. Together with turn strategies that induce lower jerk and, hence, vibration to the mechanical system, buildup both in the corners and with short weld tracks should be improved. As a consecutive step to this research, the results of the cubes are to be transferred to free geometries to validate the knowledge gained for generalized shapes.

When comparing the final temperatures depending on the used buildup strategies, differences of 30 K were observed: The highest final temperature was measured at approx. 440 °C with the buildup strategy Contour-hatch 0°–90°. The number of contour runs per layer has a greater influence on the final temperature. The more contour runs per layer, the lower the temperature at the end of the process. The temperature difference between the setup without contour run and the setup with two contour runs per layer was 70 K. This correlation was also observed in the comparison with DED-LB samples. The temperature difference between buildup with 1.5 contour runs per layer and without the contour run was 110 K. The greatest influence on the component temperature was found to be caused by the fly-in distances (for acceleration of the handling system to the deposition speed) and the runs with the laser switched off, e.g., indirect turning in the corners. The longer these intermediate laser-off times are, the lower the final temperature.

When analyzing the microstructure, no dependence on the temperature could be determined between the examined buildup strategies. However, depending on the process variant used, different microstructures can be observed: With EHLA 3D, narrow grains that grow vertically over many layers form similar to the microstructure with PBF-LB. In the case of DED-LB, larger grains are formed, which grow toward the surface of the respective melt pool and only over a few layers. Within the scope of this work, no dependence of the microstructure on the final temperature was found within the two process variants. An influence of the buildup strategy on the component temperature was noted, but to a small extent of approx. 70 K in relation to an annealing temperature of over 1000 °C of the powder material. In further experiments, the influence of the final temperature on the material properties can be investigated. For this purpose, a more detailed analysis of the microstructure could be carried out. Such a study can help to better understand the differences and similarities regarding the propensity to defect formation between the EHLA 3D process and DED-LB, where a bigger amount of data is already available. In general, the results need to be validated for further or more complex geometry elements and to distinctly account for and quantify laser on or off times.

The findings of this study can serve as the foundation for the automated application of the EHLA 3D process. If research in this direction is pursued further, arbitrary given volumes can be deconstructed into several base geometry features, for each of which a functioning buildup strategy can be taken from a database. Joined together, those geometry features yield the final near net shape part.

The experimental work in this study was part of the master's thesis of Co-Author David Hausch. His work at Fraunhofer ILT was kindly supported by “Stiftung Industrieforschung” through a thesis scholarship. The authors would like to thank the foundation, which is aimed at fostering research that is relevant for small and medium enterprises, for this valuable support.

The authors have no conflicts to disclose.

Jonathan Schaible: Conceptualization (lead); Formal analysis (equal); Funding acquisition (supporting); Investigation (equal); Methodology (lead); Supervision (lead); Writing – original draft (lead). David Hausch: Conceptualization (supporting); Formal analysis (equal); Investigation (equal); Methodology (supporting); Resources (supporting); Visualization (lead); Writing – original draft (supporting); Writing – review & editing (supporting). Thomas Schopphoven: Formal analysis (supporting); Funding acquisition (lead); Resources (supporting); Supervision (supporting); Writing – review & editing (supporting). Constantin H㥮er: Funding acquisition (supporting); Investigation (supporting); Resources (supporting).