Aeronautics

Modeling Aerodynamic and Acoustic Fields Around Turbofan Engines

Image from animation showing velocity traces over a rotating turbofan engine
Particles seeded from the trailing edge of each fan blade colored by streamwise velocity (red is high, blue is low), with the nacelle hidden for visualization purposes. Surfaces are colored by streamwise velocity (red is high, blue is low) at the second grid point normal to the wall. The rolling up and breakdown of each blade wake vortex sheet is apparent, as is the complex aerodynamic interaction occurring in the blade tip vortex system. Timothy Sandstrom, Luis Fernandes, NASA/Ames

NASA has set several goals for the agency’s computational fluid dynamics capabilities over the next decade, including full aircraft engine simulations. As such, the capability to accurately and reliably simulate the complex flow physics resulting from the interaction of an engine’s rotating blade cascades is highly desirable. However, time-accurate simulations of an entire airbreathing turbofan jet engine are computationally expensive, especially when researchers attempt to capture the vast amount of turbulent aerodynamic and aero-acoustic data. A team of researchers at NASA’s Ames Research Center aims to reduce the cost of these complex simulations by employing a sliding mesh technique, which facilitates the communication of flow variables between the stationary and rotating parts of the computational model. This simulation method avoids the costly connectivity task that would otherwise have to be performed at each time-step. The team at Ames has implemented this novel approach using the Launch, Ascent, and Vehicle Aerodynamics (LAVA) framework, an in-house developed structured curvilinear computational fluid dynamics solver, and the new method is currently undergoing validation using experimental data from the NASA Source Diagnostic Test campaign.

Quick Facts

To capture the main features of the noise signal characteristic of a rotating turbofan, one needs to obtain time-accurate solution data for a sufficiently long time window, usually on the order of ten or more fan revolutions. With the large size of the numerical grids needed to capture higher frequency tones, these simulations require computational resources only available in the largest HPC facilities in the world, such as the NASA Advanced Supercomputing (NAS) facility.

Luis Fernandes,
NASA Ames Research Center
  • All simulations were run on the Pleiades supercomputer at the NASA Advanced Supercomputing (NAS) facility. Each simulation ran on 20-core Intel “Skylake” nodes, using more than 480,000 core-hours and dozens of terabytes of short-term disk storage to obtain data for just ten fan revolutions.
  • One of the key results so far has been the surprising capability of Unsteady Reynolds-Averaged Navier Stokes (URANS) computations to predict the average flow-field that develops downstream of the engine’s fan blade row, especially in comparison with scale-resolving simulations performed using Hybrid RANS/Large-Eddy Simulation methods. This supports the premise that URANS might be sufficient to capture the tonal noise content generated by the turbofan assembly.
  • The results obtained in this work will be compared against experimental data obtained during the NASA Source Diagnostic Test campaign, which was performed in the Low Speed Anechoic Wind Tunnel at NASA’s Glenn Research Center, and will be used to determine how well the sliding mesh technique works within the LAVA software framework.
  • The next step of this project will be to perform a thorough comparison of the aero-acoustic field around the turbofan model predicted by the simulations and the data captured in wind tunnel experiments, with the goal to validate the LAVA solver’s capability to accurately model this complex flow environment.

Researchers

  • Luis S. Fernandes, NASA Ames Research Center
  • Jeffrey A. Housman, NASA Ames Research Center

More Information