High-speed Infrared Imaging and Multiphysics Modeling for Prediction of Three-Dimensional Thermal Characteristics during Selective Laser Melting

The microstructures and hence mechanical properties of laser based additively manufactured metals are known to be sensitive to the thermal gradients and solidification velocity at the melt pool solid-liquid interface. However, experimentally-validated transient three-dimensional (3D) melt pool thermal characteristics during the solidification process remain elusive. Our unique approach couples high-speed infrared (IR) imaging at ~15000 fps, of the melt pool surface, with 3D multiphysics simulation to predict the sub surface transient temperature distribution and solidification conditions during selective laser melting of MAR-M247. Briefly, first, we capture the melt pool surface radiation using high-speed IR imaging and convert it to melt pool surface temperature using a Planck’s law-based custom calibration. The experimentally obtained spatio-temporally varying melt pool surface temperature is then imported as a boundary condition into a COMSOL model. Based on the experimentally obtained top surface temperature, the coupled COMSOL model solves the 3D sub surface temperature distribution in and around the melt pool. Importantly, the coupled COMSOL model concurrently solves the complex melt pool multiphysics which involves conduction, phase change (solid-liquid), and thermal Marangoni induced flow inside the melt pool. We validate the predicted melt pool cross-sectional dimensions, including depth and width, using ex-situ scanning electron microscopy (SEM). Additionally, we quantify the critical local 3D solidification conditions such as thermal gradient, solidification velocity and solidification direction at the melt pool solid-liquid interface. Through ex-situ SEM analysis, we confirm that during solidification, the grain growth direction closely follows the 3D thermal gradient direction at the solid-liquid interface, as predicted by the coupled model. Further, based on these local solidification conditions, crucial to microstructure growth, the coupled model suggests a reduction in cell width of the solidified microstructure as the laser scan speed is increased from 500 mm/s to 1000 mm/s at a constant laser power of 200 W. The microstructure cell width predictions from the coupled model are validated through a comparative cell width analysis of ex-situ SEM for 200 W – 500 mm/s and 200 W – 1000 mm/s laser scanning parameters. We believe the multi-dimensional and data rich outcomes of our methodology can readily be integrated with machine learning and present a platform that can viably be translated to commercial applications to improve SLM processes.