Applying machine learning to wire arc additive manufacturing: a systematic data-driven literature review

Due to its unique benefits over standard conventional "subtractive" manufacturing, additive manufacturing is attracting growing interest in academic and industrial sectors. Here, special emphasis is given to wire arc additive manufacturing (WAAM), a directed energy deposition process that employs arc welding tools and wire to build metallic components by deposition of weld material. The WAAM process has several advantages, e.g., low cost, rapid deposition rate, and suitability for large complex metallic components. However, many WAAM challenges such as large welding deformation, undesirable porosity, and components with high residual stress remain to be overcome. Multidisciplinary cross-fusion research involving manufacturing, material science, automation control, and artificial intelligence/machine learning (ML) are deployed to overcome the above-mentioned problems. ML enables improved product quality control, process optimization, and modeling of complex multiphysics systems in the WAAM process. This research utilizes a data-driven literature review process, a defined and deliberate approach to localizing, evaluating, and analyzing published studies in the literature. The most relevant studies in the literature are analyzed using keyword co-occurrence and cluster analysis. Numerous aspects of WAAM, including design for WAAM, material analytics/characterization, defect detection/monitoring, as well as process modeling and optimization, have been examined to identify state-of-the-art research in ML for WAAM. Finally, the challenges and opportunities for using ML in the WAAM process are identified and summarized.