NASA Researcher Wins CFD Flow Visualization Award
Earlier this summer, a researcher in the NASA Advanced Supercomputing (NAS) Division received the award for Most Quantitatively Descriptive Flow Visualization Animation at the CFD Flow Visualization Showcase held during the 2017 AIAA Aviation Forum and Exhibition in Denver, CO. Patrick Blonigan, who studies complex computational fluid dynamics (CFD) methods with the University Space Research Association (USRA) at the NAS facility, presented his work at the event to a committee of government, industry, and academic researchers in aeronautics scientific visualization.
The video, titled “The Adjoint of a Near-Wall Turbulent Flow” (which you can watch below), shows a basic simulation of turbulent airflow through two parallel plates, known as a channel. Turbulent airflow is characterized by chaotic changes in pressure and velocity, and can be observed on nearly all air- and spacecraft during flight, such as the airflow inside a jet engine or around a capsule during atmospheric reentry from space. Turbulent aerodynamics are difficult to simulate accurately and efficiently, even with the computational power of agency supercomputers like Pleiades, operated at NAS, so researchers like Blonigan are looking into methods to improve NASA's CFD tools to better and more efficiently model turbulent airflow around aircraft of all types.
In particular, researchers have encountered a problem with the application of adjoint sensitivity analysis to simulations of three-dimensional, high-fidelity turbulent flows using “eddy,” a specialized research software that was developed at NASA's Ames Research Center. Adjoint sensitivity analysis is a mathematical method of computing that allows engineers to estimate errors in CFD simulations used to evaluate aerodynamic performance, and shows how to improve aerodynamic designs, such as where and how to modify the shape of a wing to improve its aerodynamic performance. Traditional adjoint methods don't work well with simulations that exhibit chaotic dynamics because the butterfly effect causes the adjoint calculation to grow exponentially. These results aren't very helpful to aeronautics researchers, so CFD experts like Blonigan are looking for novel ways to counter this growth in order to simulate a useful adjoint field.
Blonigan's visualization shows the first successful application of a new approach to computing the adjoint called "shadowing," which he helped develop at the Massachusetts Institute of Technology (MIT), to the eddy framework. These proof of concept simulations show that the shadowing adjoint approach reveals additional information previously unavailable in the simulation of turbulent airflow using eddy, including when, where, and how a flow can be disturbed to increase or decrease some quantity an engineer might be curious about, such as kinetic energy, as shown in Blonigan's test case. These features could be used to estimate the error in kinetic energy computed by the simulation or to develop a controller that researchers could use to amplify or reduce the intensity of turbulence within the airflow.
Now that Blonigan and his colleagues have shown that the shadowing approach works for small turbulent simulations like basic channel flow, they're looking to extend it to larger simulations, such as a jet nozzle, for a further proof of the success of the new method. They also plan to demonstrate that it can also be used to estimate simulation errors and help identify their sources. This, in turn, would help engineers improve the accuracy of high-fidelity simulations run on NASA supercomputing resources.
Watch for yourself:
This work was sponsored by NASA’s Transformational Tools and Technologies (TTT) Project of the Transformative Aeronautics Concepts Program.
