Researchers at Washington State University and the University of Minnesota have developed an AI-guided experimental design method that identified feasible Directed Energy Deposition (DED) parameters for GRCop-42 at laser powers as low as 500 W.
Called Bayesian Experimental design for Additive Manufacturing (BEAM), the method uses previous print results to select promising parameter combinations for testing. Applied to NASA-developed GRCop-42, it identified successful configurations from 950 W down to 500 W within ten trials at each power level.
Using AI to search for feasible DED conditions
Metal AM process development can involve millions of parameter combinations, making trial-and-error testing costly and slow. BEAM narrows this search using an adaptive loop that learns from previous successful and failed experiments, selects promising configurations, and updates the model with each new result.
Unlike conventional Bayesian optimization, which typically searches for an optimum, BEAM aims to identify multiple feasible configurations while using domain knowledge to exclude unrealistic combinations. For the GRCop-42 study, the model was initialized with 37 previous failed experiments and used a probabilistic five-nearest-neighbor approach.
Finding a lower-power process window for GRCop-42
GRCop-42 is a copper-chromium-niobium alloy developed by NASA for high-temperature aerospace applications. Its thermal conductivity and elevated-temperature properties make it suitable for components exposed to high heat flux, but also complicate laser processing.
Copper absorbs relatively little infrared radiation from conventional fiber lasers and rapidly dissipates heat away from the processing region. The researchers therefore describe establishing a DED process window below 900 W as particularly challenging.
The team used a five-axis powder-fed DED system with a 1000 W fiber laser to deposit GRCop-42 on Inconel 718. The search covered powder feed rate, gas flow, Inconel thickness, scan speed and layer height, while other variables were held constant.
BEAM tested fixed laser powers of 950 W, 700 W, 600 W and 500 W, with an initial budget of ten experiments per level. It found at least one feasible configuration at each power, including three at 600 W, compared with 37 earlier failed experiments conducted over several months.
High scan speeds emerge from successful experiments
Successful configurations generally paired high scan speeds with low layer heights, which the researchers identified as a key process insight. At 700 W, one viable setup used a 0.2 RPM feed rate, 7 L/min gas flow, 1600 mm/min scan speed and 0.11 mm layer height.
The lowest-power result was obtained at 500 W, using a 0.075 RPM feed rate, 7 L/min gas flow, 250 mm/min scan speed and 0.3 mm layer height.
Failed samples developed thin pillar-like structures or large beads during deposition, while successful settings were first demonstrated using test blocks before being used to produce larger cylindrical specimens.
After machining a cylinder printed using the successful 700 W conditions, the researchers reported a solid interior with minimal defects. A cross-section also showed an even transition between the Inconel 718 and GRCop-42 regions.
The three-month campaign established feasible GRCop-42 processing from 500 W to 950 W, potentially widening access to lower-power DED systems. The study does not, however, qualify the parts for aerospace service. Future work will extend the approach to additional alloys and compositions.
Expanding the GRCop-42 process window
Previous work has shown that GRCop-42 can be processed using conventional infrared laser systems, although copper’s high reflectivity and thermal conductivity continue to constrain the usable process window. In 2025, Auburn University’s NCAME reported that GRCop-42 had been successfully printed on EOS M290 systems using standard infrared fiber lasers, while Nikon SLM Solutions developed LPBF parameters for the alloy that achieved 99.97% density.
Process conditions also affect how GRCop-42 behaves alongside other alloys. Earlier this year, University of Nottingham researchers studying multi-material LPBF of GRCop-42 and Inconel 718 found that depositing IN718 onto GRCop-42 caused lack-of-fusion defects as the copper alloy rapidly dissipated heat. Reversing the sequence avoided these defects, although alloy mixing occurred at the interface.
The BEAM study extends the process-development problem to DED, using AI-guided experimentation to identify feasible GRCop-42 conditions down to 500 W. Published for the Fortieth AAAI Conference on Artificial Intelligence, it was conducted by researchers from Washington State University and the University of Minnesota, including Azza Fadhel, Nathaniel W. Zuckschwerdt, Aryan Deshwal, Susmita Bose, Amit Bandyopadhyay and Jana Doppa. It focuses on reducing the time and resources needed to establish workable metal AM process windows.
3D Printing Industry is inviting speakers for its 2026 Additive Manufacturing Applications (AMA) series, covering Energy, Healthcare, Automotive and Mobility, Aerospace, Space and Defense, and Software. Each online event focuses on real production deployments, qualification, and supply chain integration. Practitioners interested in contributing can complete the call for speakers form here.
To stay up to date with the latest 3D printing news, don’t forget to subscribe to the 3D Printing Industry newsletter or follow us on LinkedIn.
Explore the full Future of 3D Printing and Executive Survey series from 3D Printing Industry, featuring perspectives from CEOs, engineers, and industry leaders on the industrialization of additive manufacturing, 3D printing industry trends 2026, qualification, supply chains, and additive manufacturing industry analysis.
Feature image shows printed successful structures from 700W laser power using DED-based metal AM. Image via Fadhel et al.
