| Abstract Scope |
Calibrating and validating simulation models invariably require executing controlled experiments. When interpreting the results of such tests, it is critical to understand which factors are actually in-control versus which uncontrolled variables may be influencing the results. Unfortunately, in L-PBF, several influencing variables are often ignored – skewing both analyses (e.g., porosity populations and microstructure) and the resultant models. These factors include sample-to-sample adjacency, local feature geometry, sample location in the build volume relative to the laser, tool paths, gas flow, and recoating direction, changes to the machine state (e.g., control errors, filter clogging, optical degradation), and inter-pass time. In this presentation, we discuss how in-situ monitoring of the process provides modelers with invaluable context when interpreting their experimental results and planning future tests. Specifically, the Peregrine software tracks key metadata, registers X-ray computed tomography and other characterization data to multi-modal layer-wise images, and uses machine learning to detect multiple anomalies. |