Statistical process monitoring and control methods for in-situ detection and localization of defects in laser powder bed fusion

Key industrial sectors for the implementation of metal additive manufacturing (AM) systems, like aerospace and bio-medical industries, involve stringent quality and certification requirements that are difficult to meet at the current technological maturity level. One major barrier is represented by the limited stability and repeatability of the AM processes. This motivates the development of in-situ monitoring and control solutions for a zero-defect oriented production. Most efforts in the literature and in industry have been focused on gathering in-situ sensor data so far, but what is still lacking is the availability of data analytics tools able to make sense of big amounts of acquired signals and yield automated defect detection and localization capabilities. Automated alarm rules represent a first necessary step to design novel closed-loop control strategies for defect mitigation or even defect repair methods that are still not available in commercial systems. In this framework, we present statistical image-based methods for in-situ monitoring of various kinds of “process signatures” aimed at characterizing the melting state and detecting local defects during the layer-wise production of the part. The proposed methodologies are applied to real case studies in Selective Laser Melting (SLM).