Novel Sensing and Machine Learning to Solve Critical Challenges in Laser AM via Real-Time Control
Abstract
Additive manufacturing (AM) is an enabling technology allowing advanced parts to be fabricated that simply cannot be realized with c,onventional (subtractive) processing. One of the limiting factors with metal AM processes such as laser powder bed fusion (L-PBF) is, the generation of microscopic defects during manufacturing. These defects have only recently been understood and pose significant i,ssues with certifying the quality and consistency of fabricated parts [1]. Current AM systems utilize simple open-loop schemes for l,aser beam-steering and process control. Unlike open-loop process control, closed-loop implementations require elements to sense proc,ess conditions in order to predictively or reactively alter the manufacturing parameters. Significant progress in detection of proce,ss faults has been made over the last 10 years [2-4], but these approaches suffer from a tradeoff between sensing speed and precisio,n that diminish their utility in real-time process control [5]. Thus, a critical need remains for in situ thermal sensors capable of, detecting fault-inducing conditions in real-time.In addition to pushing the state-of-the-art in high-speed thermal sensing, high-sp,eed controllers will need to be integrated into the control fabric of AM systems. Conventional controllers incorporating feed-forwar,d and feedback loops can generally be used to avoid and correct error conditions. For simple processes, the combination of these clo,sed-loop techniques is sufficient for precise and robust control. Artificial intelligence (AI) offers a promising path forward for l,earning the intricate dependencies of complex processes through regression of process data in a high-dimensional space. Trends linki,ng various process conditions can be learned, in principle, to create a predictive model for defect avoidance. By fusing ultra-high,speed thermal sensor data with AI, it should be possible to realize an intelligent closed-loop control system capable of mitigating,process instabilities known to generate defects before defects are formed. If successful, this would result in an unprecedented leve,l of control of the laser fusion dynamics that, in principle, could result in defect-free L-PBF manufactured parts at production rat,es.The fundamental objective of the proposed work: to develop in situ sensing and intelligent process control capable of mitigating,defect formation in real-time for high-quality and reliable parts.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 08, 2022
- Source ID
- N000142212687
Entities
People
- Steven Storck
Organizations
- Johns Hopkins University
- Office of Naval Research
- United States Navy