Deep learning-based applications in metal additive manufacturing processes: Challenges and opportunities–A review
Tuğrul Özel
Abstract
In metal additive manufacturing (AM), parts often exhibit quality variations, defects, intricate surface topography , and anisotropic properties influenced by factors such as process parameters, energy and fusion interactions, and material physics . These complexities make metal-AM processes challenging to manage consistently, leading to unacceptable levels of inconsistency. To address these issues and predict quality, in-situ process sensing and monitoring as well as post-process measurements are commonly employed, aiming to enhance process understanding, control, and reliability. This review paper surveys literature on deep learning (DL) methods used in AM processes , discussing current research challenges and future directions. The ultimate objective is to develop intelligent AM systems capable of using real-time process data for automated control decisions and interventions, advancing towards more reliable defect-free manufacturing outcomes.