AI-ASSISTED OPTIMIZATION OF ADDITIVE MANUFACTURING: PROCESS CONTROL, DEFECT PREDICTION, AND SUSTAINABLE PRODUCTION

Abstract

Although complex and customised parts can be manufactured by AM, the process can be easily affected by various process parameters, thermal behaviours, material properties and machine conditions. Monitoring and prediction of defects as well as parameter optimization and adaptive control are now being increasingly improved by the use of artificial intelligence (AI) and machine learning (ML). The current review highlights the use of AI in process optimisation in AM, focusing on 'In-situ Monitoring', 'Intelligent Process Control', 'Defect Prediction', and 'Sustainable Production'. The literature shows that supervised learning is employed for quality prediction and defect classification tasks, and reinforcement learning and hybrid physics-data-driven approaches have been proven to have the potential for dynamic process optimization tasks. Other benefits of using AI-assisted systems include fewer failed builds, less material waste, no rework, and lower energy use. Data imbalance, sensor variations, domain change, cross-machine generalisation, model understanding, computational delay, and certification, however, are the factors that limit industrial application. In the future, the research emphasis will be on the standardisation of datasets and physics-informed models, digital twins, explainable AI, the quantification of uncertainty, and robust closed-loop control systems. Overall, AI shows great promise in the continuous improvement of the reliability, adaptability and sustainability of AM.