AI monitoring for WAAM: Virginia Tech demonstrates real-time quality control in the printing process;

A team led by Prahalada Rao at Virginia Tech describes in Materials & Design a method that couples Wire-Arc Additive Manufacturing (WAAM) with AI-based process monitoring. The goal is to inspect large metal components for defects during the build and initiate corrections immediately. The background is the need for rapidly available spare parts in safety-critical areas such as maritime and aerospace, where custom components often have long lead times and defects are noticed only after fabrication is complete.

Technically, the group relies on sensor-acquired features of the melt pool and models the relationship between process signals and subsequent part defects. Cameras and data acquisition provide thermal and geometric indicators that a machine-learning model evaluates in real time.

“We’ve always relied on conventional machining, but it takes months to produce even a single part,” Rao said. “Additive manufacturing gives us the ability to make those parts much faster and with less waste, which opens up a new way of thinking about how we build.”

“Wire-arc additive manufacturing is basically welding in 3D,” Rao said. “If laser powder bed fusion additive manufacturing produces a pint of material a day, wire-arc is keg-sized. You can deposit 40 or 50 kilograms of material in just one day. The challenge is making sure that much metal goes down without a single flaw.”

The high deposition rate—sometimes 40 to 50 kilograms per day—makes robust control necessary, as pores, cracks, or phase misalignment can quickly propagate.

“When the melt pool looked good, the part turned out how we wanted. When it looked bad, we knew what would happen,” Rao said. “So we built a machine learning algorithm that was able to predict with about 90 percent certainty when things were going wrong.”

In addition to WAAM, the team is investigating laser-wire processes to compare parameter windows and correction strategies. The crucial factor is the combination of monitoring and actuation: process deviations are not only detected but compensated for by adjusting current, wire feed, or toolpath planning.

“Faster, better, cheaper,” Rao said. “I want to make it better through quality control, faster by not wasting time redoing parts, and cheaper by reducing defects. That’s something we do very well in process control.”

“Giving students access to the same machines they’ll see in industry is critical,” Rao said. “That’s why the laser powder bed fusion and laser wire system in the Learning Factory is so valuable. Additive is something you can train quickly, and it prepares our students to step right into the future of manufacturing.”

The work is embedded in the Virginia Tech Made: Center for Advanced Manufacturing and uses university facilities such as the Learning Factory. For operators of complex fleets, the study suggests that large-format wire processes with embedded AI control can be a viable option for quality-assured, conformal preforms—from repair to pilot production.