Methodology
++ This page describes the methodology being used for sample fabrication and in-situ IR + pyrometry being used by the AM-D-Model project. Data contributions to the project + should follow a similar methodology to this, to ensure data consistency in the + resulting benchmark. As part of our commitment to openness and transparency, we + encourage collaborators and interested parties to review our methodology and engage + with our team. We welcome any and all suggestions for potential improvements and + discussions on how we can ensure that this project produces a valuable benchmark + dataset that can drive the future development of metal AM. +
+Component Geometry and Specification
++ The geometry of the test AM components (samples) shown in Figure 1 + consist of a 7mm × 7mm × 7.2mm (L x B x H) cuboid, complemented with a + tapered support structure of height, 2.85 mm. Additionally, sample + numbers with a height of 300 µm were incorporated into the top of the + cuboid geometry, resulting in a total sample dimension of 7mm × 7mm × + 10.35 mm. This geometry, along with the support structure design, was + specifically chosen to enable straightforward sample removal using a wet + abrasive cutter equipped with 2.8 mm thick blades, while also ensuring + easy handling during characterization. +
+
+ AM Pilot Line and Process Parameters
++ The samples were fabricated using the AM pilot line at Dublin City University + (DCU). This pilot line is the AconityMINI® 3D powder bed fusion-laser system. + It is equipped with a 200 W, 1068 nm Yb- fiber laser. This system supports laser + scanning speeds of up to 2000 mm/s and features a minimum laser beam diameter of + 32 µm. High-purity argon gas (99.99999% purity) was used as shielding gas, and + the "Skywriting" feature was enabled during the fabrication process. A + comprehensive PBF-LB process parameter window was defined through a + full-factorial experimental design. This approach systematically explores the + effects of key process parameters by varying four factors at four levels each (4 + factors × 4 levels = 44), resulting in a total of 256 unique + parameter combinations. The factors and their corresponding levels are as + follows: +
+-
+ {#each laser_params as param}
+
- {@html param.name}: +
- [ {@html param.values.join(", ")} ] {@html param.unit} +
+ The range of energy densities calculated from these parameters spans the + following: +
+-
+ {#each ved_params as param}
+
- {@html param.name}: +
- + [ {@html param.min} to {@html param.max} ] {@html param.unit} + +
+ The metal powder feedstock used was the PowderRange® 316L stainless steel + powder supplied by Carpenter Additive®. The particle size of the powder + feedstock ranged between 15 and 45 µm. For this reason, the layer thickness (L) + parameter was maintained at a constant value of 50 µm. At this layer thickness, + each sample with dimensions of 7 mm × 7 mm × 10.3 mm required 206 layers to + complete fabrication. A bidirectional scanning strategy was applied, starting + with a layer angle of 45 degrees, followed by a 90° rotation for each subsequent + layer. +
++ Two sets of the 256 unique parameter combinations were fabricated, + yielding a total of 512 samples. Figure 2 shows representative pictures + of the first 100 samples produced. During the final print session, 56 + remaining samples from the first set were combined with 56 remaining + samples from the second set, creating a final print batch with 112 + samples. This method was implemented to maintain consistent interlayer + printing times and achieve thermal distribution uniformity across all + samples, comparable to the other print batches of 100 samples. +
+
+ + Volumetric Energy Density (VED) is a derived thermodynamic metric widely used in + AM industry and reported in metal AM research. It is typically expressed as + either the spot size variant (VEDf) or the hatch spacing + variant (VEDH), calculated using the below equations, + respectively: +
++ However, these two VED variants are not directly comparable, as neither + exclusively captures the full complexity of melt pool physics. This includes + melt pool formation and propagation, solidification geometry (width and depth), + and the intricate mass and heat transfer between the melt pool and the + surrounding material. Given these limitations, both VEDf and + VEDH are included as PBF-LB process-based features. The inclusion + of these two distinct measures of volumetric energy density, further enriches the + experimental space, enabling a nuanced exploration of the relevance of energy distribution + to the ML model performance. +
+Infrared (IR) Pyrometry
++ IR pyrometry measures thermal radiation from the melt pools during the PBF-LB + process, enabling real-time recording of temperature distributions. This helps + identify issues such as overheating, insufficient melting, or thermal gradients + that may cause defects or residual stresses. +
+
+ + In-situ temporal IR monitoring was performed using two + Kleiber® KG 740-LO pyrometers + fitted to our AconityMINI machine (Figure 3a). The optical path of both pyrometers + was configured to be in line with the laser beam path (see Figure 3b). This + enabled the capturing of the meltpool positions (x, y) with the corresponding + relative temperature of the meltpool (T) of all samples being fabricated. + The pyrometers were calibrated to a black-body standard to the same temperature + range, and features: +
+-
+ {#each pyrometry_params as param}
+
- {@html param.name}: +
- {@html param.desc} +
Relative Density
++ The target material property is the observed relative density (RD) of the test + samples, expressed as a percentage of the theoretical density of a perfectly + dense specimen. The RD of each sample was determined using Archimedes’ + principle, following the ASTM B962-23 standard. Ethanol served as the immersion + fluid for the measurements, and the assumed theoretical density of 316L -SS was + to be 8000 kg/m3 for these calculations. +
+Sensor Data
++ The collected data from the sensors form a comprehensive dataset designed to + capture various aspects of the PBF-LB process. This dataset integrates + measurements from multiple sources, ensuring a detailed representation of the + process parameters and their influence on the target material property + prediction. +
+Infrared Pyrometry (IR) Data
++ IR data is stored as layer-by-layer .pcd files, where the filename represents + the layer thickness (e.g., 5.24.pcd corresponds to IR data at a 5.24 mm layer + thickness). Each file contains meltpool coordinates (x, y) and relative meltpool + temperature (in degrees Celsius) from the two pyrometers, as illustrated in + Table 1. +
+| {header} | + {/each} +
|---|
| {row[header]} | + {/each} +