Laser Powder Bed Fusion Optimization of Inconel 939 for Improved Part Production Quality

Sammanfattning

This thesis presents an experimental and simulation-based investigation into the effect of contour scan velocity on the surface quality, relative density, and dimensional accuracy of Inconel 939 (IN939) components fabricated via Laser Powder Bed Fusion (LPBF). IN939 is a γ′-strengthened nickel-based superalloy with outstanding high-temperature capability. Yet, it exhibits a characteristically narrow processing window and pronounced susceptibility to solidification cracking. These characteristics make contour parameter selection especially critical for producing dense, well-finished parts on application-relevant geometries. Experiments were conducted on an EOS M270 Direct Metal Laser Sintering (DMLS) system equipped with a 200 W ytterbium fibre laser operating under a nitrogen atmosphere. Core hatch parameters were held constant throughout (laser power P = 195 W, hatch spacing h = 100 μm, layer thickness t = 40 μm, beam offset = 50 μm, hatch scan speed = 600 mm/s), thereby isolating the influence of the contour scan velocity, which was varied across 18 discrete levels from 300 to 1150 mm/s. A prismatic test geometry incorporating a 45° chamfer enabled simultaneous evaluation of three representative surface orientations: down-skin, vertical, and up-skin. Relative density was quantified using the Archimedes method in isopropanol (ρ = 0.78134 g/cm³), and surface roughness Sa was measured using a Zygo NewView 7300 White Light Interferometer. Surface roughness exhibited strong orientation dependence. Vertical surfaces were least sensitive to velocity variation and achieved the lowest areal roughness (Sa ≈ 3.29 μm) at 400 mm/s. Up-skin surfaces also performed best at moderate velocities (Sa ≈ 9.03 μm at 400 mm/s). Down-skin surfaces were most challenging due to limited thermal dissipation and powder adhesion effects; the minimum profile roughness (Ra ≈ 0.972 μm) occurred at 350 mm/s, while the minimum areal roughness (Sa ≈ 13.6 μm) was recorded at 800 mm/s, a divergence that underscores the limitations of Ra as a sole optimisation criterion for complex surface textures. A multi-objective equilibrium was identified in the contour velocity range of 400–500 mm/s, which simultaneously maintained relative density above 98% and produced satisfactory surface roughness across all three orientations. Dimensional validation on a benchmark part confirmed accuracy within ±3% for features exceeding 1 mm, systematic undersizing of 5–8% for features in the 0.4–0.6 mm range, and sub-resolution conditions for features below 0.3 mm. The use of contour scanning consistently improved the definition and repeatability of fine geometric features relative to hatch-only builds. Complementary ANSYS Additive simulations, employing the inherent strain method, were performed on cantilever geometries at build angles of 0°, 15°, 60°, 75°, and 90°. Simulations demonstrated that the build plate effectively suppresses distortion during fabrication, but significant deformation occurs upon part removal. Low-angle orientations (0° and 15°) exhibited the greatest post-cut-off displacement, consistent with experimental measurements obtained via image analysis. The findings establish a contour scan velocity of 400–500 mm/s as the recommended process window for IN939 fabricated on the EOS M270, offering a robust balance of density, surface quality, and dimensional accuracy. These results provide a practical foundation for industrial parameter qualification and motivate future integration of orientation-specific scan strategies, support structure optimisation, and residual stress mitigation measures.