SIAM News Blog
SIAM News
Print

Computing the Noise of Turbulent Flows in Extreme Conditions

By Radouan Boukharfane, Julien Bodart, and Matteo Parsani

Computational fluid dynamics is a complex discipline that combines applied mathematics, physics, and computer science—and often high-performance computing as well—to understand and visualize the way in which a gas or liquid flows and affects nearby objects (for instance, in terms of aerodynamics, aeroacoustics, and fluid-structure interaction). Simulations provide invaluable insights into flow dynamics, which are difficult and often dangerous or even impossible for researchers to obtain with experiments alone. Furthermore, when the design of an engineering system requires rigorous and arduous design cycles or optimization approaches, experiments are deemed too cumbersome and are often limited by the availability of diagnostic instrumentation. 

Computational methods permit scientists to obtain ample amounts of data within a relatively short time frame, thus allowing them to alter and assess a plethora of key design parameters during a device’s initial development phase. However, numerical solutions can contain significant inaccuracies. The uncertainties in computational methods are associated with assumptions regarding the model and the numerical dissipation of discretization methods, the latter of which is mainly due to coarse grids.

The direct simulation of turbulence requires numerical methods that accurately reproduce turbulence’s evolution over a wide range of length and time scales, as dictated by the physics. The grid determines which scales are represented, while the numerical method establishes the accuracy of this representation. Direct numerical simulation is thus the best tool for investigating turbulent flows and computing noise sources. However, this approach is unattainable for most engineering applications. In fact, the number of spatial grid points scales approximately as \(\textrm{Re}^{13/7}\) (where \(\textrm{Re}\) is the Reynolds number). In many real applications (e.g., flows past airplanes and/or in gas turbines), \(\textrm{Re}\) can easily reach values between \(10^6\) and \(10^8\). 

One can spot the source and strength of acoustic waves by analyzing the divergence of the velocity field and iso-entropy contour.

One promising technique is standard large eddy simulation (LES)—also known as wall-resolved LES—which involves modeling the more universal small scales while explicitly computing the large-scale motions. The emergence of petascale supercomputers (i.e., computers capable of more than \(10^{15}\) floating-point operations per second) at the beginning of the last decade has led to a substantial increase in the use of LES methods for turbulent flow calculations. Researchers currently utilize LES in a wide variety of engineering applications, including combustion, acoustics, and simulations of the atmospheric boundary layer. At the dawn of exascale supercomputing (with \(10^{18}\) floating-point operations per second), LES is now becoming a powerful and more accessible computational fluid dynamics tool that is capable of enabling landmark simulations. However, even with exascale machines, LES will continue to be extremely costly in terms of computer resources for domains with large dimensions or at large Reynolds numbers. For instance, a wall-resolved LES simulation of the flow around an entire aircraft still remains out of scope at present and in the foreseeable future. 

A major component of aeroacoustics is aircraft noise, which is already an issue for many communities near major airports. It will only get worse with the continual growth of air travel and urban airborne vehicles. Therefore, in addition to air carriers’ demands for more fuel-efficient airplanes that are cheaper to operate, commercial aircraft manufacturers face increasingly stringent requirements to reduce the “community noise” that their aircrafts produce. 

Community noise stems from two main sources: engine noise and airframe noise. Airframe noise is primarily caused by airflow around the aircraft’s landing gear and high-lift devices, such as wing flaps and slats. The principal sources of engine noise are the engine fan and the jet that is downstream of the engine. In our work, we test our algorithms on the airframe noise generated by turbine blades.

In a recent study, we proposed the use of a wall-modeled LES (WMLES) strategy (not wall-resolved LES) in extreme fluid flow conditions to reduce computational costs while maintaining a high level of accuracy for the quantities of interest. The number of grid points, and hence the cost, in WMLES scales approximately as the Reynolds number. We demonstrated that WMLES is able to offer a computationally cheaper alternative that could make Reynolds numbers of practical importance both accessible and feasible. Furthermore, we showed that WMLES can effectively predict the broadband noise that aerodynamic objects generate, which is an essential requirement for fast and accurate simulations in industrial design.

The research team, which is part of the Extreme Computing Research Center at the King Abdullah University of Science and Technology (KAUST), is uniquely placed at the intersection of numerical analysis, physics, and high-performance computing to develop novel and efficient algorithms that better account for physical phenomena and efficiently utilize modern computing architectures. More information about the study is available in this press release

Radouan Boukharfane joined the Extreme Computing Research Center (ECRC) at King Abdullah University of Science and Technology (KAUST) as a postdoctoral fellow in 2019. He currently works in professor Matteo Parsani’s research group. Previously, he spent a year as a postdoctoral research fellow in the Department of Aerodynamics, Energetics and Propulsion of the National Higher French Institute of Aeronautics and Space (ISAE-SUPAERO) in Toulouse, France. Julien Bodart is an associate professor in the Department of Aerodynamics, Energetics, and Propulsion at ISAE-SUPAERO. Before joining ISAE, he was a postdoctoral fellow in the Center for Turbulence Research at Stanford University. Matteo Parsani is an assistant professor of applied mathematics and computational science at KAUST. He is also the principal investigator of the Advanced Algorithms and Simulations Lab and a member of the ECRC. Matteo holds a M.Sc. in aerospace engineering with a specialization in computational aerodynamics from Politecnico di Milano and a Ph.D. in mechanical engineering from Vrije Universiteit Brussel. After a period as a NASA Postdoctoral Fellow at NASA Langley Research Center, Matteo joined KAUST in October 2015.

blog comments powered by Disqus