WSU Researchers Use AI to Cut NASA Alloy 3D-Print Tests

What Happened

Researchers at Washington State University used an artificial intelligence system to identify viable 3D-printing parameters for a NASA rocket alloy, reducing more than 100 million possible experimental combinations to just 40 targeted tests. Of those 40 experiments, six produced successful prints, including the first-ever successful print of the alloy at 500 watts, according to reporting by the Times of India citing the research.

Background

The alloy in question is used in NASA rocket applications and presents significant manufacturing challenges. Additive manufacturing, commonly known as 3D printing, requires precise control over variables including laser power, print speed, layer thickness, and material feed rates. For complex aerospace-grade alloys, the number of possible parameter combinations can reach into the hundreds of millions, making conventional trial-and-error testing prohibitively expensive and time-consuming.

Washington State University has maintained an active aerospace and materials engineering research program. The work intersects two areas of growing federal and industry interest: AI-assisted materials discovery and advanced manufacturing for space applications. NASA has funded and collaborated on multiple university-based research initiatives aimed at reducing the cost and lead time of producing components for rockets and spacecraft.

How the AI System Worked

The research team applied an AI model to survey the full parameter space associated with 3D printing the NASA alloy. Rather than testing combinations at random or relying on established engineering intuition alone, the system evaluated the landscape of more than 100 million options and selected 40 experiments calculated to yield the most useful results. The approach is consistent with a class of AI techniques known as active learning or Bayesian optimization, in which a model iteratively selects experiments most likely to improve its predictive accuracy, though the specific methodology used by the WSU team was not detailed in the available wire reports.

Six of the 40 AI-selected experiments produced prints meeting success criteria. One of those six represented the first documented successful 3D print of the alloy at a laser power level of 500 watts, a threshold noted as significant in the reporting.

What It Means in Practice

The reduction from more than 100 million possible tests to 40 represents a substantial decrease in laboratory time and material costs. In aerospace manufacturing contexts, physical test iterations for high-performance alloys require specialized equipment, skilled technicians, and significant quantities of costly raw material. Each failed or uninformative experiment consumes resources without advancing the manufacturing process.

For NASA and its suppliers, the ability to establish reliable 3D-printing parameters for rocket alloys more quickly could shorten development timelines for new components. Additive manufacturing is already in active use in rocket engine production, with companies including Rocket Lab, Relativity Space, and others having printed structural and propulsion components. The challenge of qualifying new alloys for those processes has remained a bottleneck.

The 500-watt successful print is also noted as a distinct technical milestone. Lower power thresholds in laser powder bed fusion and directed energy deposition processes can affect the microstructure and mechanical properties of the finished part. Establishing a successful process window at that power level extends the range of equipment configurations on which the alloy could be produced.

Scale and Efficiency

The ratio of AI-selected experiments to the full parameter space, 40 tests drawn from a field exceeding 100 million, illustrates the scale of efficiency gains that proponents of AI-assisted materials research have cited in recent years. Similar approaches have been applied to battery electrolyte discovery, pharmaceutical compound screening, and structural materials for civil engineering, though direct comparisons across domains depend heavily on the complexity of each parameter space.

Washington State University has not publicly released a timeline for follow-on research or indicated whether the findings have been submitted for peer review publication, though further validation and broader experimental testing would typically follow an initial discovery of this kind.