A passenger jet that’s propelled by electricity? Artificial intelligence that can learn to reason as children do? A robotic “worm” that can dig a tunnel longer than five football fields? Those were just some of the ambitious concepts — with important real-world applications — that the scientists and engineers at GE Research set their minds to in 2020. In between, they used supercomputers to figure out how to make engines run hotter and wind turbines operate more efficiently. Here are just a few of the things that some of the world’s smartest people were cooking up this year at GE’s research campus in Niskayuna, New York.
If you saw a 6-foot earthworm popping up out of a tunnel it had just dug, you might express some surprise — but don’t worry, this worm is here to help. At GE Research, Deepak Trivedi and his colleagues have been working on a soft, bio-inspired robot as part of a $2.5 million, 15-month project funded by the Defense Advanced Research Projects Agency’s Underminer program. The idea is to create a burrowing machine that could one day save lives — say, by tunneling in to resupply troops.
Earthworms are small but mighty, able to leave a tunnel twice their diameter in the dirt they’ve dug through. Trivedi, an expert in soft robotics, has drawn inspiration from the worms’ internal dynamics among other physical attributes. In the prototype he and his colleagues are working on, fluid flows inside seven internal chambers that act as muscles, flexing as water is pumped in and out. The machine’s nose contains a piloting tool to stir and soften the earth as it moves forward. Once it’s fully operational, the robot’s engineers expect the Underminer to be able to dig a tunnel longer than five football fields in just 90 minutes. “There’s something beautiful in the mathematics and the biological examples used in understanding these soft structures,” Trivedi said. “In addition to being able to make something really cool, really fast.”
A native of Ukraine who came to the U.S. in the 1990s to pursue studies in analytical chemistry, Radislav Potyrailo is known to his colleagues at GE Research as a scientist who looks for answers wherever he can find them — and isn’t afraid to challenge assumptions. Potyrailo is fond of hitting the library, and during one project to develop a film that would detect toxic chemicals, he and his colleagues even found inspiration in the structure of the wings of an iridescent butterfly. Another recent achievement also relates to a sensor to detect dangerous chemicals, which could someday be used in wearable applications to help keep workplaces and people safe. It’s making waves: In May, Potyrailo’s discovery was featured on the cover of the journal Nature Electronics.
Conventional sensors can help users avert fire or other accidents, but their performance is affected by temperature and humidity, and they typically have only one output. Looking past conventional technologies, Potyrailo and his team used a principle called dielectric excitation, which is driven by an alternating current to activate the sensors. To explain how it works, Potyrailo uses the analogy of looking at a famous painting: Leonardo da Vinci’s “Mona Lisa.” Examine this painting turned on its edge, he says, and you see a single line. That represents the single output of a traditional sensor. But turn this painting perpendicular to show its surface, and all the crucial details suddenly appear. “If you have the right variable when you turn, you see multiple colors in the painting,” he says. Or, in the case of the gas sensor, multiple outputs that tell users much more about the atmospheric conditions they are trying to measure.
How much potential power is blowing in the wind? That’s a question facing GE engineers as they seek to design wind turbines that can try to get the most energy from offshore breezes without becoming overloaded. In 2020 they enlisted a new ally: Engineers at GE Research were granted access in August by the U.S. Department of Energy to use Summit — the powerful supercomputer housed at Tennessee’s Oak Ridge National Laboratory — to help them with their calculations. Over the next 12 months, GE Research lead aerodynamics researcher Jing Li and her team will mine reams of existing wind data, run simulations on Summit to model wind strength and speed, then use those outputs to model airflow through a theoretical wind farm — ultimately calculating optimal load capacities and power production.
Li and her colleagues are specifically seeking to understand coastal low-level jets, currents that are of interest to the wind power industry — they’re thought to relate to the way that wind flows from land out to sea — but aren’t well understood. Summit will make the job easier, allowing researchers to home in on wind dynamics both in finer detail and at a massive scale. “With a supercomputer,” Li said, “you are no longer restricted to looking at how the wind flows through one blade. You can get that information for dozens, if not hundreds, of turbines in a big wind farm.”
How hot is a jet engine? So hot that scientists have enlisted the world’s most powerful computers to figure out how to cool them without sacrificing the energy the heat can deliver.“ Just like biologists use microscopes or astronomers use telescopes, high-fidelity simulations empower researchers to see what they otherwise could not,” says Rick Arthur, an engineer at GE Research. In 2020 that fidelity got even higher when Arthur and his colleagues were granted the use of Summit, the Oak Ridge National Laboratory’s superfast supercomputer.
“It opens up a whole new area of predictions we never would have been able to do,” says Michal Osusky, a lead thermosciences engineer at GE Research. Osusky and Arthur say the supercomputer simulations will help point them to breakthroughs in jet engine and gas turbine design, as well as create a “gold standard data set” that will allow scientists to analyze previously hidden problems and illuminate potential solutions.
When he was a kid growing up in Toronto, Peter Tu liked to run around the neighborhood with his dog or oversee the reptile pit he built in his backyard. He particularly liked to teach tricks to his animal friends. Now 53 years old, Tu is still in the business of teaching new tricks — only now, his trainees aren’t dogs but computers. Tu is the chief scientist for artificial intelligence at GE Research and lately, he’s been looking at new ways to get artificial intelligence to learn, including by the kind of basic reasoning that dogs — and kids — excel at. “As children, we do a large number of inductions,” Tu explained. “A child with limited examples is very good at drawing out general rules.” Can AI become good at that too?
AI is great at processing massive amounts of data and spotting subtle patterns. It’s less prepared for situations where it faces a “poverty of stimulus,” as Tu describes it — simply not enough information to go on. That’s where it could take a cue from kids as they’re learning their way through the world. Tu has been experimenting with providing young AI systems with the equivalent of parents — experts to offer gentle corrections, including exceptions to seemingly general rules, like those that govern language. If AI-enabled machines can master nuances in communicating with humans, and eventually with one another, Tu thinks, it’ll be a step toward machines that comprehend all sorts of problems. It’s all about working together: “I’ve always been of the opinion that we need more companions to walk down this path with us,” Tu said.
In terms of fuel efficiency, commercial aviation has come a long way: The amount of fuel used per passenger has dropped 80% since 1960. But those savings have been offset by the surging growth of passenger aviation in the same period, leading aircraft and engine designers to search for new ways to reduce aviation’s impact on the environment. “We need something fundamentally different to take the next leap,” said John Yagielski, senior principal engineer at GE’s Research Center in Niskayuna. Yagielski and his colleagues are at work on something fundamentally different indeed: an electrically driven propulsion system powerful and light enough to keep aloft a 175,000-pound commercial airliner and its 175 passengers.
That goal is being backed up by $4.8 million in new research grants from the U.S. Department of Energy’s Advanced Research Projects Agency-Energy (ARPA-E) — and it will be no small feat. The challenge is figuring out how to convert a cleaner-burning biofuel into megawatts of electricity, and then how to turn that electrical energy into enough thrust to fly a Boeing 737-class jet. But that challenge is also an invitation for the GE engineers to reimagine what an aircraft engine looks like, drawing up new designs that might be more efficient for flight than the traditional model of engines beneath each wing. “It’s about proving the feasibility of a number of these technologies and convincing ARPA-E to invest in building a complete prototype and testing it,” Yagielski said. “This is for aircraft in the 2050s.”