But today’s gains have not come without problems. Jet engines rely on large amounts of fossil fuels, creating greenhouse gases and noise pollution. And local noise restrictions are hampering the expansion of green wind-turbine technology.From the earliest sailing ships to the latest wind-turbine generators, humanity has long harnessed the power of wind. In aviation, we generate our own wind – powerful enough in commercial jet engines, for example, to produce 100,000 or more pounds of thrust.
Improved control of flow and turbulence generated by these technologies promises to lead to increased fuel efficiency, reduced greenhouse gases and less noise.
A government-industry partnership is helping solve these problems, with help from a Department of Energy (DOE) supercomputer housed at Argonne National Laboratory. Nicknamed Intrepid, the IBM Blue Gene/P is among the world’s fastest for this kind of research.
Aerodynamics experts once relied solely on wind tunnels, mathematical modeling, empirical observations and expensive, hit-or-miss physical engineering experiments. But now they’re able to fine-tune the complex world of air turbulence as never before. Expanding supercomputing capabilities have enabled a new field called large eddy simulation (LES) – incredibly detailed, direct computation of turbulence, noise sources and heat transfer that is paving the way for innovative, environmentally sensitive product designs.
‘For the first time, we are able to compute noise sources accurately.’
Last year, DOE’s Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program awarded GE Global Research’s Aerodynamics and Aeroacoustics Laboratory 20 million computer processor hours on Intrepid. INCITE grants universities, national laboratories, government agencies and private companies time and expertise on supercomputers at Argonne and Oak Ridge national laboratories.
A single jet flow or wind-turbine airfoil calculation would have taken months to complete on computers available just a few years ago. Now, with Intrepid, the GE team could make the same calculation in a day or two, “a huge improvement in turnaround time,” says Umesh Paliath, a GE mechanical engineer interested in high-fidelity computational techniques for jet noise prediction. He worked with team leader Anurag Gupta, who has more than a decade of experience studying jet noise.
Pared to the bones, jet engines work like this: Fans ingest large amounts of air and compress it at the front. That air is combined with fuel for combustion, then passed though a turbine to extract thrust before it’s expelled through a large nozzle at the back of the engine. This thrust also powers on-board electrical and hydraulic flight-control systems. But the fast-moving exhaust flow generates noise and pollution.
Designing a simulation is no easy task. It requires long hours validating mathematical methods on simpler cases. New designs for wind-turbine airfoils and jet nozzles are then simulated using the validated codes.
To run their simulations, GE researchers accessed Intrepid remotely. The output was further processed to predict the amount of noise generated. The end product allowed the GE team to visualize the flow to get at what occurred at scales ranging from a few millimeters to meters.
As Paliath knows from work during his graduate school days, the advent of modern computing has greatly improved turbulence research. But the fidelity of models that were possible even just a few years ago was fuzzy compared to what can be achieved now through LES work on machines like Intrepid.
“The earlier computer studies were limited in what you could resolve and how fast you could resolve it,” he says. “You had to use simplified models. In noise research, for example, these models weren’t accurate enough to get you the source diagnostics, and sometimes not even accurate enough to give you crude designs.”
But with the combination of LES and enhanced computer power, the scale and level of resolution has increased dramatically. “For the first time, we are able to compute noise sources accurately as well as air flow parameters from jet and wind turbines,” Gupta says. “Before, others were able to do one of them accurately but not the other.”
Once simulations for nozzle flows and airfoils start on Intrepid, the researchers can monitor how the turbulence flow evolves in real time. To capture the flow field and noise from just one jet nozzle, for instance, requires about 750,000 CPU hours, Paliath explains.
Wind-turbine noise research uses even more computing time – on the order of 2 million to 3 million CPU hours. In this area, the team looks at how much lift a particular design generates and how noisy it is. For jet nozzles, they examine how much heat and thrust are generated and the noise levels involved.
“We can actually figure out exactly what’s happening at various scales and due to what,” Gupta says. “We can closely visualize phenomena in much less than millimeters or at many meters, enabling the redesign of these technologies to make them more efficient and less noisy.”
Although a wind-turbine airfoil, or blade, may be 50 meters long and the airfoil itself 1 to 2 meters wide, the researchers are interested in noise generation at scales much less than a millimeter. This work will enable design of quieter, more efficient airfoils for wind turbines as well as nozzles for jet engines.
The team’s also interested in film cooling. After fuel is burned in modern jet engines, temperatures are so hot – more than 1,000 centigrade – they are beyond the normal melting point of materials in the engine.
“The only way the engine survives is by cooling it,” Gupta explains. To do this, engineers need to control air movement over the engine parts, creating a film of cool air. “Traditional methods have failed miserably to predict this phenomenon accurately, but with LES we can do so.”
LES enables researchers to examine problems in scalable dimensions ranging from a several-millimeter jet nozzle, for example, to a full-scale jet nozzle. “The computing power we have through the INCITE grant enables us to examine larger problems in a more realistic time frame,” Gupta says.
“The simulation size is typically expressed in terms of how many grid points you use to resolve the problem,” he continues. “For example, jet nozzle problems use 200 million discrete grid points in the region of interest, where you are trying to resolve the flow field. This could be as much as a 1-meter by 1-meter by 3-meter region for a 50-millimeter-diameter nozzle.”
Paliath and Gupta believe that LES research is just beginning.
“We’ve shown that this works,” Gupta says. “If we can reduce the noise of a single turbine by even one decibel, for example, we can translate this into 2 or 3 percent more yield per turbine. And being able to fine-tune thermal mixing in these systems will lead to greater fuel efficiency and reduced emissions. These will be big improvements for mankind.”
The human brain contains a vast expanse of unmapped territory. An adult brain measures only… Read More
Computer scientists are democratizing artificial intelligence, devising a way to enable virtually anyone to train… Read More
Electrifying transportation and storing renewable energy require improving batteries, including their electrolytes. These fluids help… Read More
During the pandemic turmoil, Margaret Cheung reconsidered her career. At the University of Houston, she… Read More
Growing up in the remote countryside of China’s Hunan province, Z.J. Wang didn’t see trains… Read More
Princeton Plasma Physics Laboratory (PPPL) scientists are creating simulations to advance magnetic mirror technology, a… Read More