Can We Trust the Simulation?

STAR’s CDA-funded research is testing how digital twins, physical data, and blockchain-based version control can bring rocket models closer to reality.

Minutes away from firing their rocket engine in the Mojave Desert, Aidan Rickert (EECS ’27) and his team discovered that their valves had frozen. They had three options: wait hours for the valves to thaw, vent the pressurized tank by putting a bullet through it, or fire anyway and hope for the best.

“We went with option three, which was to fire,” Rickert says. “We sort of had an inkling that it wouldn’t go well. It blew up.”

Engineering Under Pressure

Every other weekend, Rickert, the president of Space Technologies and Rocketry (STAR), Berkeley’s high-power rocketry competition team, makes the six-hour drive to the Mojave Desert to test engines. “The idea is collecting data to see if it works as you expect,” Rickert tells me. “We go for three successful times [before launch] because then you can be fairly confident it’s not going to explode and ruin your rocket.”

Valves, Rickert tells me, are one of the most complex parts of rocket-building. The rockets Rickert’s team launches utilize high pressure fluids to pressurize their tanks, making it difficult to predict exactly how those valves will behave. “It’s relatively hard to get valves that can handle that kind of pressure,” Rickert says. “And so part of that is modeling it.”

Until now, that modeling has relied on manufacturer specifications and rough physical testing. “You make a CAD model, you simulate it in ANSYS, and you’re like, ‘Okay, I think this is how it behaves,’” Rickert says. “Then you put it in the real world and it seems to work. But how do you know the simulations were correct to begin with?”

Building Smarter Simulations

Rickert’s research, which is co-led by STAR propulsion lead Carlos Bautista (Aerospace Engineering ’28), combines digital twins, repeated physical testing, and blockchain-based version control to answer this question. At its core, the project aims to build better software models by continuously comparing them against real-world data. “Over the past year, we’ve spent a lot of time trying to have better computer models for everything that we’re doing,” Rickert tells me. “Our project is…getting a better model of our whole system and adding to it.”

Rickert’s team is building digital twins of two key components in the rocket’s propulsion system. By repeatedly testing the hardware under different operating conditions and comparing those results against their simulations, the team hopes to build software models that become more accurate over time. “The digital twin side is making reliable software models of physical things and using that to better understand the physical world,” Rickert explains. “Blockchain allows us to have better version control and management.”

After Liftoff

Rickert believes that the project’s significance expands beyond rocketry, into fields like electrical engineering and medicine. “What we’re doing applies to anywhere there is a physical thing you’re trying to understand,” Rickert says. “You want a physical thing to be characterized in software so you can use the software to understand the physical thing.” Whether that physical thing is a rocket engine, an electrical circuit, or a medical device, Rickert believes that the same cycle applies: build a model, compare it against reality, and improve it over time.

Projects like Rickert’s don’t happen in traditional classrooms. Through the Center for Digital Assets, Rickert and his team have been able to explore how digital twins can be used to improve real engineering workflows. “Student teams are the best way to impact students at Berkeley,” Rickert tells me. “There’s really good support right now from the [university] towards what we’re doing.”

For Rickert and his team, the goal isn’t to build a perfect simulation, but one that moves closer to reality each time it’s tested. “You have software, then you make a physical thing, and you go back to the software afterwards,” Rickert says. “Because otherwise, you’re not really learning anything.”

In engineering, Rickert has learned, the model is never really finished.