Melt Pool Monitoring as a Defect Detection Method for Quality Control of AlSi10Mg Parts Produced by PBF-LB/M

Abstract

Abstract Additive Manufacturing (AM), particularly Laser Powder Bed Fusion of Metals (PBF-LB/M), has become a key technology in the aeronautical industry to produce lightweight, high-performance components with complex geometries. However, the quality and reliability of AM parts remain critical concerns due to the potential formation of process-induced defects such as lack of fusion, keyhole, and gas porosity. Ensuring structural integrity requires robust, efficient, and non-destructive quality control methods. This study investigates the use of a Melt Pool Monitoring (MPM) software as a non-destructive defect detection method for AlSi10Mg components manufactured via PBF-LB/M. By analysing thermal emission data captured during the build process, histogram-based profiles were generated and correlated with defects identified through high-resolution X-Ray computed microtomography (µCT). The correlation between both methods validates the MPM capability to detect and identify different forms of defects. Samples were manufactured under both optimal and intentionally varied process parameters to induce different types of defects. The resulting MPM histograms revealed distinct patterns associated with specific morphologies of defects, enabling rapid classification of part quality. This approach demonstrates the potential of MPM as a pragmatic and versatile solution for in situ quality assurance in AM, reducing time-consuming post-production inspections. The findings support the integration of MPM into quality control workflows for aeronautical AlSi10Mg components, contributing to improved process monitoring, defect mitigation, and certification readiness. Furthermore, this methodology lays the groundwork for future implementation of artificial intelligence (AI) models trained on MPM data. By training machine learning (ML) models on MPM derived histogram data and corresponding µCT validation, it is possible to automate the classification of parts as acceptable or defective in real time, streamlining quality control in production.