There are billions of ways to build an engine. Here's how they found the best one.

RTX-built software helps researchers sift through a sea of engineering options

On paper, the point of the project was to build an aircraft engine powered by hydrogen. But there was a twist.

Rather than simply showing hydrogen can do the job - Pratt & Whitney engineers already did that decades ago - the team went a different route: They wanted to see what kind of engine hydrogen could make possible.

"We were really trying to reimagine, from the ground up, what a hydrogen engine could be," said Neil Terwilliger, HySIITE principal investigator and technical fellow for Advanced Concepts at Pratt & Whitney, an RTX business.

To help answer that question, the team tapped into the broad realm of artificial intelligence, using a program called DISCOVER - built by researchers at the RTX Technology Research Center and Collins Aerospace - to quickly analyze design options and suggest promising new architectures.

"What we found," Terwilliger said, "was that we can take advantage of some of the unique properties of hydrogen to do things that can't be done with jet fuel - and fundamentally get a much more efficient engine."

How did DISCOVER help?

Steve Taylor and Joe Turney of the DISCOVER team calculated the number of possible ways they could arrange the engine's roughly 70 components.

The result: There were about a quattuorvigintillion possibilities - that's 1 with 75 zeros.

"Just to put that in perspective, there are 10 to the 80 atoms in the universe," said Larry Zeidner, a technical fellow at the research center who led the development of DISCOVER.

Naturally, the next step was to narrow the results. They gave DISCOVER lists of the engine's components and how they work together along with a set of design rules. From there, 4,202 combinations emerged as possible winners.

And that was a number they could work with.

"We put up a scatter plot that had 4,000 dots on it, and we said, 'those are all the possible architectures that you have,'" Taylor said, "and you can point to one dot and say, 'that one did the best.'"

That gave the Pratt & Whitney team a starting point to begin testing their design. Before DISCOVER, Taylor said the team would likely have been able to explore about five different architectures. DISCOVER allowed them to validate thousands in the same amount of time - and find not just what works, but why it works.

"DISCOVER doesn't tell us ideas that we couldn't have written down," Taylor said. "It sifts out the ones that didn't work and tells you which ones were good, which you had no other way of knowing."

Finding pearls in the ocean

Zeidner, who's been working on DISCOVER for 20 years, describes finding the optimal design as trying to find a pearl in the ocean. The challenge, he said, isn't recognizing pearls - it's the water that makes it difficult. Hear Zeidner describe how they reveal the best ideas.

A new design is born

The team's work resulted in the design of an engine known as HySIITE, which could improve performance by 35% while eliminating nearly all nitrogen oxide emissions. They demonstrated their design on rig tests at the RTX Technology Research Center in East Hartford, Connecticut, supported by the U.S. Department of Energy's ARPA-E program. The engine's efficiency hinges on its ability to capture water from the exhaust - one gallon every three seconds - and use it to control how the liquid hydrogen burns.

The concept relies on recuperation, which has been explored in the past, but it's never been a practical option for large commercial aircraft. Once the team started considering steam injection to recuperate heat instead of a heat exchanger, DISCOVER helped the team recognize that shrinking the engine core was the key to making the steam injection work.

"A small core helps because it makes it easier to capture water to convect, and having water that is convected makes it feasible to have a smaller core," Terwilliger said. "Understanding this virtuous cycle helps engineers continue making changes that further lean into this strategy to really bring out benefits. That's something that might not have been obvious at first sight."

Overcoming the challenge

Terwilliger said there were two main challenges to overcome. First, hydrogen burns so hot and fast that it's hard to control the flame. Second, they need steam in the burner to increase efficiency, but steam makes it difficult to sustain combustion.

As Terwilliger described it, it's like trying to make a fire while spraying it with a hose.

Hear him describe how they used those problems to their advantage and found a solution where hydrogen's high flame temperature and speed allowed them to inject more steam for better efficiency while effectively managing the combustion.

"The challenge was to burn hydrogen in that mixture, and it turns out that the steam actually helps control the combustion. It controls the flame speed. It mitigates flame temperature. So, we saw essentially an elimination of NOX production, which is a pollutant that combustion can create, which is fantastic, and the steam helped control the stability of hydrogen combustion, so we saw no problems with the flame going back onto the hardware or melting anything. It was one of those rare situations in systems engineering where two problems actually kind of cancel each other out. That is very fortunate when it happens and seems to never happen enough."

In his words

What's next?

While hydrogen-powered aircraft are still decades away from flight, the team's concept proved that the new engine architecture could triple net energy savings compared to synthetic aviation fuels. It's also informing an industry-wide conversation about what's possible - something Terwilliger said was part of the team's intent.

"One of the questions we look at in advanced concepts is what kind of fuel would we use in the future," he said. "We felt that as the engine people, what we can do is figure out how much better an engine can be with hydrogen."

DISCOVER's job is to find feasible design options through a method called design-space exploration, which relies on models and rules. Once DISCOVER finds the feasible options, it evaluates them with numerical models, which simulate the physics and other metrics including cost and reliability.

But DISCOVER doesn't just present one ideal solution. Researchers can use the mounds of data generated to evaluate what works from each solution and what doesn't.

"That cloud of 4,000 points has more information in it than what solution is the best. It has information about which features are helping more than others," Terwilliger said. "If you can understand which feature was more or less important, you can prioritize where to spend technology dollars. If you can understand why an architecture isn't performing well, maybe you can think of a way around that problem."

The team is working on a new program that will use artificial intelligence and machine learning to explore larger design spaces and streamline how they evaluate and visualize the data.

As the team trains more people across RTX to use DISCOVER, the program's benefits could go beyond designing high-performing products and into finding ways to make them easier and less costly to manufacture, said Mike Ikeda of RTX's Chief Technology Office.

"How do we better inform designers so that the architectures we're choosing are more manufacturable, less expensive or more durable in the field?" Ikeda said. "Can we actually bring some of that understanding earlier in the process?"

The program's benefit, Zeidner said, is straightforward: DISCOVER allows subject matter experts across RTX to come up with better solutions, do it faster and with less risk.

"It's tough to imagine that you could rapidly provide excellent solutions for a wide range of applications across RTX," Zeidner said, "but that's exactly what DISCOVER does."

News Across RTX