AM4Sight: Additive Manufacturing, Model-based, Multi-resolution, Machine Learning defect risk visualization tool

While AM systems, especially metal AM, bring revolutionary capabilities and have the potential to reduce supply chain issues and enable new designs through unique layer-by-layer fabrication capabilities, AM technologies currently suffer from defects that exist within the components. Defects such as porosity, inclusions, large-scale voids, and chemical inconsistencies can inhibit the functional performance of a part and reduce confidence in designing parts for AM. While NDE methods exist to identify defects, such as X-ray CT, they are very costly and time consuming. FTL's previous work, Volumetric AM Metadata Engine (VAME), is Air Force-funded analysis software that provides a framework for AM knowledge capture that is adaptable to different metallic AM processes and design pipelines. Building on that code base, the proposed AM4Sight (AM4 refers to AM-targeted Model-based, Multi-resolution, Machine Learning) tool adds novel 3D build-time data aggregation, Machine Learning (ML) defect detection, and probabilistic defect risk mapping to guide the CT operator and test designer, improving the efficiency, cost-effectiveness, and successfulness of AM NDE/I. AM4Sight uses FTL’s voxel visualization engine to identify the probability of a defect at every volume sample of the resulting AM part, as well as the severity of the defect in terms of associated failure modes of the part while in service. This provides “foresight” of defect type and location to the NDE/I technician to guide decisions on resolution, integration time, and test setup. This is a significant improvement to current commercial efforts to quantify the effects of defects on additively manufactured components focus on “brute force” testing, with an emphasis on expensive destructive testing to qualify a printed component.