Han Hu is driven by order. What excites Hu, an assistant professor of mechanical engineering at the U of A, is finding a formula that explains the world around us.

"I want to see all the dots connected, either by mathematics, physics or some logical connection," Hu says. "That's the dream of a researcher like me."

In Hu's field of thermal fluids, which studies how fluids transport heat, mathematical models exist, but they are not complete. Adding another piece towards solving that puzzle motivates Hu's research.

Hu admits, though, that his decision to focus on two-phase cooling systems was largely chance. It was the subject of his first research project in graduate school, and he stuck with it.

"Maybe if my first project was different, then my research would be a little bit different," Hu says.

With the assistance of U of A's Technology Ventures, he has filed a pair of patent applications related to two-phase cooling systems and is working on a third patent.

Two-phase systems are nothing new. Turning liquid — one phase — into steam — a second phase — is the process behind distilling and steam engines. Today, though, two-phase systems have begun to be used to cool increasingly powerful, and tightly packed, computers in data centers necessary for growing technologies like AI. The electronic components are submerged in a fluid with a low boiling point. As the computer components heat up, the liquid turns to steam, drawing away heat that is then transferred to a condenser.

"Liquid cooling has been there for decades. But for data centers in the '90s, air cooling was efficient," Hu says.

The same process can also cool lasers, nuclear power plants or equipment on spaceships.

A two-phase cooling system can fail. When boiling liquid reaches critical heat flux, excessive bubbles become insulation that traps heat in the electrical components instead of carrying it away to the surface. Right now, no effective, non-intrusive technology exists to detect that failure.

The inventions in Hu's patent applications solve that problem. His inventions will detect these failures by listening to acoustic signals and analyzing the sounds with deep learning. Like an experienced cook who can judge the temperature of oil in a pan by its sound, Hu's invention will hear when the cooling system's fluid has overheated. The next technology he is developing will use an AI regression model to predict a critical heat flux, so expensive computer equipment can be shut down well before it overheats.

"If we keep the hardware similar to what we have been doing, can we have a more sophisticated AI model so that we can give a quantitative prediction?" Hu wonders.

Hu makes outreach to students a priority. He has invited junior high students into his Nano Energy and Data-Driven Discovery Laboratory, part of the U of A Power Group. He has been a faculty adviser for high school students in the Arkansas Data Analytics Teacher Alliance program, and he has taught engineering workshops to seventh and eighth grade students at U of A's Explore Engineering camp.

Most of the young students Hu works with will not become engineers, he realizes. But those with interest in the field need to experience it first before they know it's right for them. And all students, Han believes, can benefit from learning to solve problems like an engineer.

"Engineering is not trial and error. It's a hypothesis validation process. What we do is debug a lot," he says. "And I feel like people should be equipped with some level of debugging capability even in real life."