US Navy promotes ML-based quality assessment in metal 3D printing

Quality assurance for additively manufactured metal components is considered a key hurdle for productive use in a military context. Against this backdrop, the U.S. Navy has funded a project aimed at enabling the prediction of material properties directly from 3D printing process data. The contractor is the additive manufacturing–focused company Senvol, which is applying its machine learning methods to wire-based Directed Energy Deposition.

At the core of the initiative is demonstrating that mechanical properties such as strength or toughness can be reliably predicted from in-situ monitoring data. To this end, sensor data from different modalities are evaluated, including optical signals, temperature profiles, and process parameters. The underlying software analyzes these data streams, parameterizes them, and derives feature vectors that correlate with material-relevant phenomena. The goal is both to predict part performance and to identify process parameters that are highly likely to result in the required properties.

The project, titled “Additive Manufacturing Sensor Fusion Technologies for Process Monitoring and Control,” is planned for two years and runs from July 2025 to July 2027. For the Navy, the focus is less on the individual machine than on a standardized procedure for part acceptance. A data-driven approach is intended to reduce costly and time-consuming qualification campaigns while still providing sufficient evidence for operational approval. In the longer term, it is planned to integrate requirements for in-situ monitoring into guidelines issued by NAVSEA.

Senvol President Zach Simkin commented, “Quality assurance in additive manufacturing is critical. For a part to be accepted into the supply chain, there needs to be sufficient confidence regarding how the part will perform. Progress in this area continues to evolve, and we believe that developing a consistent approach to analyzing in-situ monitoring data – and developing actionable guidance from it – will enable AM users to more readily meet part acceptance thresholds.”