At California’s Lawrence Livermore National Laboratory, the world’s most powerful computers are working on some of our most fundamental questions about the universe. The Sierra supercomputer, for example, is delving into the Big Bang and trying to figure out why elementary particles have mass.

tumblr_inline_moving blue on black image

But Sierra is also solving problems that are closer to home. This supercomputer and more recently the world’s second most powerful computer called Titan at Oak Ridge National Laboratory in Tennessee have been helping GE engineers to build a better jet engine.


This image shows a snapshot from a numerical simulation of a generic aircraft engine injector. Top Image: This animation shows a numerical simulation of a jet fuel spray performed on Sierra in collaboration with Cornell. Researchers used between 500,000 to 1 million CPU hours of simulation time. (One CPU hour is equal to one hour used by one computer processor for simulation.)

Jet engines started out as complicated creatures ever since GE built the first one in the U.S. in 1941, and their design has gotten exponentially more intricate since.

Madhu Pai, an engineer in the Computational Combustion Lab at GE Global Research, is working on an elaborate part in the jet engine combustor called the fuel injector. “It delivers the lifeblood of a jet engine combustor,” he says.

Injectors atomize liquid jet fuel and spray it into the combustion chamber where it burns and generates energy for propulsion. “They are one of the most challenging parts to design and very expensive to produce,” Pai says. (The next-generation LEAP jet engine is the world’s first engine with 3D-printed injectors.)

tumblr_inline_jet nozzel image

This fuel nozzle for the LEAP jet engine was 3D-printed from a special alloy.

Pai has teamed up with researchers from Arizona State and Cornell universities to use Titan and Sierra to study what exactly happens inside a fuel injector. The time and processing power the engineers have at their disposal is equal to running 10,000 computer processors simultaneously for over 9 months. “The supercomputer gives us a microscopic view of the inside of the injector,” Pai says. “We can study the processes occurring in regions hidden behind the metal or where the fuel spray is too dense. This allows us to better understand the physics behind the design.”

This is physics with practical implications. Pai says that small changes to fuel nozzle geometry could lead to significant changes in engine performance. “These high-fidelity computer simulations help us understand how air and fuel mix and burn, and eventually reduce the number of trials,” Pai says. “Ultimately, we want to build more powerful engines that consume less fuel and have lower emissions.”

Pai’s simulations could also yield new insights beyond jet engines and improve injectors used in locomotives, land-based gas turbines, and potentially find applications in healthcare. “This is just the beginning,” he says.

tumblr_inline_water bubbles on black image

A still from a supercomputer simulation of a jet fuel spray.

Henry Sapiecha

Take a look at other GE research involving supercomputers here.

Leave a Reply