Jonathan Poggie

Hammer and anvil were used to forge early metal objects, the anvil an iron block on which the metal was hammered into shape. The modern world is not so different in intent but vastly different when it comes to technology, and Purdue’s Anvil supercomputer is a new “iron block” on which solutions to complex problems can be “hammered” into shape via powerful computational simulations.

Researchers at the School of Aeronautics and Astronautics will be taking advantage of the new resource.

“Supercomputers allow us to explore problems that are difficult and expensive to address in laboratory and wind tunnel experiments,” said Jonathan Poggie, professor of Aeronautics and Astronautics. “We can ask ‘What if?’ questions to try to solve difficult challenges.”

Anvil, built in partnership with Dell and Advanced Micro Devices and funded by a $22.5 million grant from the National Science Foundation, is sited alongside and complements Purdue’s existing campus computing infrastructure. While that already provides powerful clusters of computing power, the largest practical job that can be run is about 1,000 cores, Poggie said.

“That itself is tremendous capacity, but it is not enough to run cutting-edge simulations of flow turbulence," he said.

Haryl Ngoh

In computing, each core refers to an individual processor within the central processing unit.

“Anvil provides a total of 128,000 cores, so a job of 20,000 cores would be routine on that machine,” Poggie said. “Working at that scale, my students can explore the influence of flow turbulence on the thermal and mechanical loads experienced by an airplane.”

For example, Poggie’s PhD student Haryl Ngoh is exploring methods to control the aerodynamics that occur when the airflow separates, or detaches, from the aircraft surface. When airflow no longer follows the surface of the aircraft it recirculates, generating vortices. The resultant flow can pound on the aircraft “like a hot hammer, leading to problems with structural fatigue and flight control,” Poggie said.

Ngoh has already obtained promising results on Purdue’s existing computational platforms that may lead to novel methods to reduce drag and aerothermal loading on aircraft structures.

“Shockwaves form around the vehicle body when flight vehicles travel at supersonic/hypersonic speeds,” he explained. “Interactions between these shockwaves and the boundary-layer flow over the vehicle surface can result in intense, fluctuating, thermal and mechanical loads.”  

His research focuses on the shockwave/boundary layer interactions induced by a blunt fin protruding from a plane surface, which if left unchecked can result in structural failure.

“The computing power on Anvil will allow me to use more resource-intensive and accurate numerical methods,” Ngoh said. “This will let me ‘unveil’ more of the underlying physics, and explore concepts to control these interactions.”  

The Anvil rollout is in the early stages, in “shakedown” mode, and experienced users like Poggie — who has also accessed supercomputer time through the Department of Defense and Department of Energy — are helping to test it. 

“The first step is to make sure our programs compile and run correctly,” Poggie said. Compiling involves translating programs written in a higher-level programming language into the binary-speak understood by the computer. “Our software makes full use of standard parallel programming libraries, and problems often arise on new machines,” he explained.

In the test phase, Poggie will time how long it takes for a program to run a typical size problem on different numbers of cores and measure the speed differential.

“We can often make improvements at this stage with small changes in the settings of the compiler or the programming script,” he said. “We include plots of these speedup measurements when we submit our requests for access hours on Anvil.”

Once the supercomputer is in production mode, Poggie will tap it for several of his ongoing research programs in fluid dynamics and flow control.

“The topics that I plan to explore using Anvil include separation unsteadiness and plasma-based flow control,” he said. “One solution could be an array of sensors and phased actuators that help to stabilize the separation motion. A promising technology for implementing this idea is a plasma actuator, which introduces a perturbation in the flow through an electrical discharge and can operate with very rapid response times, even nanoseconds.”

The new supercomputer holds great promise for addressing aerospace challenges going forward, including continued pressure to reduce emissions and increase fuel efficiency.

“Commercial aircraft are already quite efficient,” Poggie said. “To shave off more drag will require creativity and methodical exploration of small changes. Supercomputer simulations will help us find and test these new ideas.”

Another crucial initiative in the coming years is the development of efficient hypersonic flight for military applications and commercial space access.

“Flight tests can be prohibitively expensive, and involve a certain amount of risk,” Ngoh said. “Powerful computational resources like Anvil enable designs and concepts to be evaluated numerically, complementing ground and flight test efforts in order to better manage the risks and costs associated with research and development.”

In hypersonic flight particularly, said Poggie, “There are opportunities for huge improvements in efficiency. For example, introducing a hypersonic airbreathing stage can greatly increase the payload fraction of a launch system.”

A rocket carries both its fuel and an oxidizer; for example, liquid hydrogen and liquid oxygen. An airbreathing engine takes its oxidizer (air) from the atmosphere. That saves considerable weight, which can be used for payload.

“With less fuel and more payload, the cost of sending people and material to orbit could be greatly reduced,” Poggie said.