Watching what happens inside the metal while it is being built — testing AI process control on a tungsten-tantalum alloy

1Ending the build-then-inspect order

Metal materials development has long run as a cycle: design, build, break, examine, design again. Each turn takes time, and the reason for a failure emerges only after the part is finished. This award tries to change that order. Additive manufacturing builds material layer by layer, so in principle the interior can be observed while it forms. If it can be seen, it should be correctable on the spot.

2The seeing and the correcting

1. 1SeeBackscattered electron and x-ray detectors integrated into an open-architecture platform monitor material properties and microstructural evolution during the build
2. 2LearnAI models learn compact representations of complex microstructures and predict evolution from manufacturing and sensor data
3. 3MeasurePrediction uncertainty is quantified, with Bayesian neural networks as the initial implementation
4. 4CorrectProcessing strategies for desired properties are recommended, and sequence-learning models identify adaptive scanning strategies

Worth noting is that quantifying the uncertainty of a prediction sits alongside the prediction itself. For a decision made mid-build, what matters is not only whether the model is right but how confident it is. Changing the scan pattern on a prediction the model is unsure of risks damaging what it meant to fix. Choosing an uncertainty-aware Bayesian method for the initial implementation reads as fitted to that use.

3How the material was chosen

The demonstration uses a tungsten-tantalum alloy, a material with potential for extreme environments that is difficult to produce with consistent internal structures. A hard case was chosen deliberately: succeeding on an easy material would carry limited weight. The framework is to be evaluated across multiple alloy compositions and component geometries and integrated with the manufacturing platform for real-time optimization.

4The HBCU-EiR program

The program is HBCU-EiR (HBCU Excellence in Research), aimed at raising research capacity at historically Black colleges and universities. This award has Prairie View A&M University strengthening capacity in collaboration with Texas A&M University, and includes undergraduate and graduate research participation, curriculum development, seminars and hands-on outreach to high school students.

By state, Texas accounts for 13 of the 120 NSF awards this site holds as of 2026-09-02, second to California at 24. Amounts are the obligated amount as of the check date and may change.