Fusion of Novel Sensing Methods and Machine Learning to Solve Critical Challenges in Laser Additive Manufacturing

Abstract

The metal additive manufacturing (AM) process uses high-power lasers to rapidly melt and solidify metal powder into complex 3-D shapes, but unfortunately the rapid solidification process often results in stochastic defect formation and nonequilibrium microstructures. To fully understand the AM process and ensure a high-quality, defect-free manufacturing process, novel high-speed sensing methods that can capture key physical phenomena associated with the AM process at high resolution are needed. A team at the Johns Hopkins University Applied Physics Laboratory (APL) is developing novel spectrometry techniques capable of measurement speed exceeding 50 kHz along the laser path to aid in understanding how materials are formed under different laser inputs. The team is also developing machine learning tools to interpret these signals, thus revealing features and trends that are not apparent to human analysts in the sensor data or physical postmortem inspection results of the printed components.