GE Gas Power is committed to creating world class, lower-carbon solutions for the energy transition. Additive manufacturing (AM) enables the delivery of affordable, reliable, fl exible, and sustainable products for power providers around the world (1) . GE Gas Power has over thirty Direct Metal Laser Melting (DMLM) printers for development and production at our campus in Greenville, South Carolina (USA) and is in volume production of over fi fteen combustion and hot gas path components to reduce emissions, lower fuel consumption, and increase power output (2) . The value of AM is achieved through increased component performance, reduced cost through parts combination and process elimination, and faster speed-to-market product introduction. For example, GE recently introduced the AM produced 9HA.02 DLN2.6e advanced pre-mixer combustion system that expands fuel flexibility on both rich and lean fuels, accommodates 50 % hydrogen capability, and extends turndown, the load range possible while in low emissions mode, to a park mode below 15 % load (3) . On the 7HA.02, GE introduced an additively produced, high performing stage 2 turbine shroud that helped achieve world class 64 % combined cycle plant output and effi ciency (4) . These are high impact, high customer value driven applications that will positively impact gas turbine power production for the next twenty years.
To strengthen the value proposition for future products, GE Gas Power partnered with the Mechanical Engineering department at Clemson University to innovate new and creative AM build plate techniques. The ability to improve support structures to reduce residual stress deformation drives increased yield and decreases product cost. Traditionally, support structures were minimized to save material and print time. However, applying support structures with an understanding of geometric feature behavior allows them to decrease build-induced part deformation, thereby reducing cost and development time while increasing the overall quality of the component.
In a short initial trial, two ad hoc support strategies were applied to a small, concave component with thick-to-thin transitions and the resulting effect on displacement was measured (5) . The resulting reductions in distortion served as the basis for the current study. Four common part features were identified through employee interviews and support structure guidelines were generated to support the part features regardless of their position in components. Rather than basing design decisions on purely process-driven limitations, the design of the novel supports instead took into account the mechanical requirements of the part and the need to prevent distortion arising from repeated thermal expansion and compression that is known to occur as layers are printed during the build process (6) . The design guidelines were generated to help standardize the application of supports to reduce development time and costs while delivering consistent and successful prints.
While AM enables many complex geometric opportunities, four common geometries were initially selected for study of displacement behavior and for design of support solutions. ー 31 ー
Power identified the most challenging part features to print.
This pool of interviewees provided a holistic understanding of DMLM challenges from both the design and manufacturing points of view. Analysis of the interviews allowed for the selection of four features based on number of times identifi ed and for the naming scheme based on the language used by the interviewees.
The selected features were modeled into testable geometries that allowed for lightweight simulation and effective test prints while remaining representative of feature behavior in end-use parts. The first feature, the bottom surface (Fig. 2a), was created as a 72 x 12 x 9 mm part based on an existing geometry commonly used to calibrate part simulation software. The roof feature (Fig. 2b) was defined as having an opening surrounded by three vertical walls.
Its dimensions were 45 x 43 x 27.5 mm with a 25 mm deep recess surrounded by 1.5 mm thick side walls and a 2.5 mm rear wall. The overhang feature (Fig. 2c) was built with 5 mm thick horizontal surfaces but with only one 1.5 mm thick vertical wall. The overall size was 40 x 25 x 30 mm. The build of the hole feature (Fig. 2d) produced a 15 mm hole with 3.5 mm of material at the horizontal edges and 7.5 mm of thickness at the top, for a total envelope of 26.25 x 22.5 x 15 mm. Support structures were leveraged to reduce part displacement. First, a process-limit defined support strategy was used as a baseline. This was based on the traditional approach to supporting parts for print, addressing DMLM process restrictions while using the minimum amount of volume to reduce print time and costs (7) . This baseline was constant across all four geometries and is shown in orange in Fig. 2. All supports used in this strategy were thin plates with toothed attachments to the part. The dimensions of the plates were based on machine capabilities, and the toothed ends maximized ease of removability of the supports.
Commercial dual-laser, high-volume DMLM machines with customizations were used to print the test geometries. Print parameters defined by GE Gas Power were held constant during the prints, and all parts were printed on the same build plate. The operating requirements of power generation parts demand unique properties, so a high γ', nickelbased superalloy was used. The use of DMLM allows for components to be made with the design freedoms associated with AM while also creating very low porosity parts that, after heat treatment, rival the density of wrought parts (8) .
In order to quantitatively compare the printed parts to the nominal CAD models, a blue-light scanning method was employed. Rather than making contact with the part like in many traditional tribology approaches, this process uses light reflected back to the camera from the part to create a 3-dimensional surface representative of the printed parts (9) .
The measured point-cloud was then overlaid onto the original CAD file used in the print to evaluate distortion at specific points on the printed part. This direct comparison between support approaches determined their effectiveness in reducing undesirable shape change.
Because of the expensive nature of a print-and-check approach to DMLM, simulation software has become an increasingly popular solution (10) . With the ability to tune the software to match both machine parameters and material properties, the use of these software solutions can output a variety of results including stress experienced during the build and the resulting distortion amounts. For this work, the stress output was analyzed to design supports capable of addressing specific stress concentrations, and the distortion output was used to gain a better understanding of the effects the supports had on the resulting part. The displacement output is shown in Fig. 3 for the baseline hole feature in comparison with the printed part. 3. Results
The design of support structures for each geometry needed to balance process limitations and mechanical strength. The traditional approach to applying support structures to parts was driven solely by process limitations and production objectives, including minimizing material and build time and maximizing removability (11), (12). The spacing of the supports at the maximum bridging distance allowed by the machine complemented these goals. However, this study introduced a novel approach to supporting, which was based on a need for mechanical reinforcement. Leveraging the results of simulations and an increased understanding of feature behavior, supports could be placed strategically to directly address movement by the part during build (e.g., from thermal growth). Fig. 4 shows supports placed in a roof geometry to balance the mechanical and the process needs according to anticipated deformation, shown by the dotted white lines.
The two different approaches were explored for each of the four geometries in order to create support design guidelines that would minimize the amount of costly printand-check steps needed and that would eliminate subjective understanding of part displacement based on engineer experience. Instead, the guidelines would standardize the practice of support design, regardless of engineer's experience level.
The bottom surface featured a geometry often used to stress-test materials and machines in order to calibrate simulation software. Its long horizontal surface was susceptible to the uneven cooling of thickthin transitions leading to tension in the top layer creating a bowing of the part. For guideline design, the bottom surface was described as having two vertical walls connected by a raised horizontal surface. For support design, the entirety of the part was susceptible to curling, making mechanically driven supports necessary along the entire length. The supports span from the part to the build plate (Fig. 5a).
Based on these needs, two example supports were created.
They alternated checkered columns and plates along the length of the bottom surface in order to have one of the shapes in tension and one in compression. Initial simulations showed that only the plate-majority supports improved part displacement, so the column-majority support was not further utilized. to address the tensile mechanical need at the free end of the overhang and a process limit need closest to the vertical wall. In attempting to satisfy these needs, two examples were developed (Fig. 5c). In the process limit half of the support design space, thin plates similar to the baseline were used. At the free end, one approach used a checkerboard pattern of circular cross-section columns (named "cylinder") and the other used a square cross-section (named "box"). Both were equal in volume and spaced according to process limits.
Finally, the hole geometry was simplified as Kirsch's solution (13) . The hole geometry was defined as a circular cavity extending completely through the body. This shape resulted in an alternating tension-compression pattern around the 90° marks of the circle. The shrinking of the top layer of the geometry as it cooled confl icted with the bottom of the hole constrained by the build plate. This created tension at the horizontal 0-and 180-degree areas and compression at the vertical 90-and 270-degree areas. It was determined that mechanical supports were needed at the vertical and horizontal edges of the hole. In the central portion, the large volume could be filled with process-limit based supports instead. As an example, a box and a cross approach were taken (Fig. 5d). The box connected the vertical and horizontal portions of the circle through vertical and horizontal members, with plates in the middle. The cross instead connected them through diagonal members joining in the center, with empty space allowed by process limits at the diagonal portions of the hole.
The parts were all printed and measured to assess the impact of the support structures on the distortion. For the bottom surface geometries, the parts were partially cut from the build plate using wire electrical discharge machining (EDM). The resulting vertical displacement was measured from the build plate with a vertical gauge. The support strategy using a plate-majority approach with columns in compression in the middle reduced the vertical curl compared to the baseline by 0.95 %.
Subsequently, the remaining parts were all removed from the build plate using wire EDM, and the supports were removed by hand. Blue light scanning was used to quantitatively measure the distortion of each printed part compared to the baseline and to yield maximum and average values. Each part was printed twice, and the average of the maximum distortion for each pair was compared to the baseline. These results are summarized in Table 1.
The average distortion was lower than the baseline for all advanced support strategies. The maximum distortion was reduced in all cases except one -the boxed overhang supports.
The bottom surface showed a 6.06 % reduction in maximum distortion with an 11.10% increase in support volume. In was also compared, and the results are summarized in Table 1. Supports without an entry for a change in distortion were not able to be removed, emphasizing the importance of manufacturability in support design. (14), (15) .
Design practices reduce the build plate design time for early career AM engineers and provide structure and consistency for new talent to enter the industry and make an impact as soon as possible (16) . An example guideline generated in this research is shown in Fig. 7. DLN2.6e advanced combustor and the 7HA.02 stage 2 turbine shroud. AM is also opening new pathways in research and design for engineers to create novel components capable of burning very high levels of hydrogen while managing the risk of burning such highly reactive fuels. The combination of highly efficient and low carbon burning gas turbine components enables GE Gas Power to create a new generation of lower carbon power generation products and allows power producers to adapt to a future decarbonized world.
Download service for the GTSJ member of ID san114, via 165.225.111.36, 2022/04/25.