����ֱ��

A Major Leap Forward

This breakthrough builds on years of work at APL to advance additive manufacturing. When Steve Storck, the chief scientist for manufacturing technologies in APL’s Research and Exploratory Development Department, arrived at the Laboratory in 2015, he recognized the practice had its limits.

“Back then, one of the biggest barriers to using additive manufacturing across the Department of Defense was materials availability — each design required a specific material, but robust processing conditions didn’t exist for most of them,” Storck recalled. “Titanium was one of the few that met DoD needs and had been optimized to match or exceed traditional manufacturing performance. We knew we had to expand the range of materials and refine processing parameters to fully unlock additive manufacturing’s potential.”

APL spent years refining additive manufacturing, focusing on defect control and material performance. Storck's team , an effort that led to a patent filed in 2020. In 2021, the APL team published a study in the ����ֱ�� APL Technical Digest examining how defects impact mechanical properties. 

This framework — designed to significantly accelerate the optimization of processing conditions — provided a strong foundation for the latest study. Building on that groundwork, the team leveraged machine learning to explore an unprecedented range of processing parameters, something that would have been impractical with traditional trial-and-error methods.

The approach revealed a high-density processing regime previously dismissed due to concerns about material instability. With targeted adjustments, the team unlocked new ways to process Ti-6Al-4V, long optimized for laser powder bed fusion.

“We’re not just making incremental improvements,” Storck said. “We’re finding entirely new ways to process these materials, unlocking capabilities that weren’t previously considered. In a short amount of time, we discovered processing conditions that pushed performance beyond what was thought possible.”

AI Finds the Hidden Patterns

Titanium’s properties, like those of all materials, can be affected by the way the material is processed. Laser power, scan speed and spacing between laser tracks determine how the material solidifies — whether it becomes strong and flexible or brittle and flawed. Traditionally, finding the right combination required slow trial-and-error testing.

Instead of manually adjusting settings and waiting for results, the team trained AI models using Bayesian optimization, a machine learning technique that predicts the most promising next experiment based on prior data. By analyzing early test results and refining its predictions with each iteration, AI rapidly homed in on the best processing conditions — allowing researchers to explore thousands of configurations virtually before testing a handful of them in the lab.

This approach allowed the team to quickly identify previously unused settings — some of which had been dismissed in traditional manufacturing — that could produce stronger, denser titanium. The results overturned long-held assumptions about which laser parameters yield the best material properties.

“This isn’t just about manufacturing parts more quickly,” Croom said. “It’s about striking the right balance among strength, flexibility and efficiency. AI is helping us explore processing regions we wouldn’t have considered on our own.”

Storck emphasized that the approach goes beyond improving titanium printing — it customizes materials for specific needs. “Manufacturers often look for one-size-fits-all settings, but our sponsors need precision,” he said. “Whether it’s for a submarine in the Arctic or a flight component under extreme conditions, this technique lets us optimize for those unique challenges while maintaining the highest performance.”

Croom added that expanding the machine learning model to predict even more complex material behaviors is another key goal. The team’s early work looked at density, strength and ductility, and Croom said it has eyes on modeling other important factors, like fatigue resistance or corrosion.

“This work has been a clear demonstration of the power of AI, high-throughput testing and data-driven manufacturing,” he said. “It used to take years of experimentation to understand how a new material would respond in our sponsor’s relevant environments, but what if we could instead learn all of that in weeks and use that insight to rapidly manufacture enhanced alloys?”

New Possibilities

The success of this research opens the door to even broader applications. The recently published paper focused on titanium, but the same AI-driven approach has been applied to other metals and manufacturing techniques, including alloys specifically developed to take advantage of additive manufacturing, Storck said.

One area of future exploration is so-called in situ monitoring — the ability to track and adjust the manufacturing process in real time. Storck described a vision where state-of-the-art metal additive manufacturing could be as seamless as 3D printing at home: “We envision a paradigm shift where future additive manufacturing systems can adjust as they print, ensuring perfect quality without the need for extensive post-processing and that parts can be born qualified.”