Metal Additive Manufacturing Conference - MAMC 2023

The actual state of the art in SLM is linked to the utilization of certain parameters according to the metal powder. Nevertheless, the shape and geometrical features of the parts to print as well as unexpected defects during the fusing process may lead to inconsistencies in the final part properties. In addition, there is a lack of monitoring tools that allow to analyze the process in-situ and stop it or apply correction strategies. The main objective of this work is to develop a combination of digital tools, including parts segmentation, process monitoring and defect detection, acting from the build planification to print validation. The proposed methodology includes process data and images capture with InfiniAM© as well as the application of computer vision techniques, machine learning (ML), 3D data representation and artificial intelligence (AI) for the extraction high-level information of process issues and their mitigation. An interactive interface was developed to be able to segment CAD geometries according to relevant geometrical features (down-skin, up-skin, thicknesses, etc.) and thus, to be able to assign the adequate process parameters to them. The acquired data during a build was represented with a high-detail 3D view with tunable resolution (up to 15 μm/voxel) and employed to detect layer defects, gas footprint and the dimensional deviation of printed parts within each layer. Also, 2D regression models were built aiming to simulate and predict the detector measurements. To conclude, the developed tools can successfully be used to validate SLM jobs and ensure the lack of macro-defects.