Predicting Defects in Additive Manufacturing

This method leverages advanced computational modeling to evaluate localized heating conditions and fusion metrics and quantify defect risks dynamically, allowing for optimized build files and process adjustments that drastically improve the final component. The model can be trained using PPF and print samples from an AM machine's prior builds to learn correlations between the machine's instructions, behavior during build, and final product quality. By integrating machine feedback metrics, model-based thermal and fusion metrics, and in-situ sensor metrics, the model learns predictive signatures that allow it to quantify localized defect probability in PBF laser beam metals. Manufacturers can employ this method to understand the reproducibility of their prints and perform defect compensation before parts are fully deployed, improving quality, reliability, and success. Particularly useful to industries that require stringent certification of safety-critical components, such as the aerospace, space, medical, and automotive sectors, this method can be flexibly deployed globally or adapted to specific AM machines. By predicting the probability of defects before and during production, 3D printing service providers and AM equipment manufacturers can save significant amounts of time and money while drastically reducing part variability. This predictive capability also allows organizations to certify parts faster and ensure consistent material properties, which is essential for meeting rigorous performance and safety standards. This is method is currently available for patent licensing (no software included). Powder-bed fusion (PBF) is an additive manufacturing (AM) method that uses a high-energy heat source, such as a laser, to selectively fuse fine metal powder particulate layer-by-layer to create complex 3D components. This is a precise process that relies on a spatially resolved dataset of process parameters known as a process point field (PPF), which maps out a laser's temperature, power, speed, and position throughout the build. A critical challenge of this process is predicting defects and inconsistencies to ensure consistent structural integrity, quality, and part reproducibility. Porosity defects resulting from spatter can be particularly harmful to successful builds, significantly impacting the mechanical properties, overall performance, and long-term reliability of printed parts. Historically, this unpredictability has held PBF technologies back from wider industry adoption and the full realization of their benefits. Overcoming these hurdles requires sophisticated methods able to accurately monitor the manufacturing process, identify anomalies, and mitigate structural defects that occur in metal during rapid fusion and cooling periods. A new ML-based software method from NASA's Langley Research Center builds upon previous work focused on detecting anomalies during AM (LAR-TOPS-368), creating an approach for determining the probability of a defect in AM PBF-laser beam metals.