Fluid Dynamics / Aeroacoustics

Turbulente Strömung um einen Zylinder, der in einem LES-Solver simuliert wird. Die turbulente kinetische Energie k wird durch Iso-Oberflächen in der Färbung von blau nach rot dargestellt. Hohe k (rot angezeigt) an den Strömungsablösungspunkten auf beiden Seiten des Zylinders weisen auf eine Turbulenz hin, die zur Schallquelle führt. Der Hintergrund des Bildes stellt den Druck farbig dar. Er impliziert, dass die Strömung vor dem Zylinder stagniert, und eine Reihe von Wirbeln im Nachlauf der Strömung erzeugt wird.
© Fraunhofer IBP
Druck auf der Oberseite (Druck 1) und auf der Unterseite(Druck 2) des Zylinders. Diese schwanken wechselweise, d.h die erzeugte Geräuschquelle ist ein Dipol.
© Fraunhofer IBP
Schallspektrum mit tieffrequenten Komponenten mit einem Pegel über 90 dB sowie höherfrequente, ca. 20 dB leisere Anteile.Durch Anklicken dieses Bildes ist das simulierte Geräusch zu hören.
© Fraunhofer IBP

Computational fluid dynamics (CFD) and Computational aeroacoustics (CAA) study for noise estimation

A new method for aeroacoustic noise estimation has been introduced to the Fraunhofer IBP. It is a simulation study based on computational fluid dynamics (CFD) and computational aeroacoustics (CAA). CFD is a tool to simulate a flow numerically. Aeroacoustics is a sub-field of acoustics in which sound gererated from flows s studied. In CAA, such sound is computed numerically. Both CFD and CAA require a massive computational resource. Simulations are usually done on a cluster or a workstation that has multiple central processor units working in parallel.

Aeroacoustic noise is generated by an interaction between a flow and a object. In CFD, a turbulent flow around an object is first simulated to determine locations and sprectra of sound sources. The incompressible Navier-Stokes equations are solved numerically with an assumption that the flow velocity is sufficiently smaller than the sound velocity. A model of turbulence such as Reynolds-averaged NavierStokes equations (RANS) [1] and large eddy simulation (LES) [2], [3] is usually employed. Although direct numerical simulation (DNS) without any turbulence models is attempted. It is not a realistic solution: It requires huge computational resources, which sometimes surpass the ability of the state-of-the-art super computers. Once sound sources are identified in the flow simulation, sound radiated from the sources is then calculated. In CAA, linearized acoustic wave equations are solved numerically. It is usually performed in the time domain, although it was preferred in the past to do in the frequency domain. The finite-difference an time-domain (FDTD) method is the most popular these days, but the finite-volume and time-domain (FVTD) method is here adopted so that the cases in which objects having more complex shapes are involved can be examined.

One of typical aeroacoustic noises is a hiss generated from a flow passing a cylinder. The source of this noise is considered to be turbulence occurring at a point where the flow separates from the cylinder surface. Figure 1 shows a CFD simulation result of this turbulent flow. In this example, a cylinder of 30 cm diameter is placed in a flow having uniform velocity of 50 m/s. As shown by the iso-surfaces of the turbulent kinetic energy k (the energy is larger as the color turns from blue to red), flow separation and turbulence occur on the both sides of the cylinder. Pressure drawn in color in the background of this figure indicates that the flow is stragnant in front of the cylinder and that a series of vortces are generated in the wake of the flow. Figure 2 plots pressure on the upper side (pessure 1) of the cylinder and that on the lower side (pressure 2). These oscillate alternately. This means that the generated noise source has dipole nature. Spectrum of pressure 1 implies that low-frequency components at about 40 to 50 Hz and a broadband noise are generated (Figure 3).

A CAA simulation reveals how an acoustic pressure field or sound is radiated from the cylinder. Figure 4 shows the radiated sound a one frequency. Due to thedipole nature, the radiation ist not isotropic. Larger radiation in the direction perpendicular to the uniform flow and smaller radiation in the parallel direction can be observed. The radiated sound is actually simulated based on Curle's equation [4] from pressure exerted on the cylinder surface, which has already been found in the CFD simulation. Figure 5 plots sound waveform at a point 10 meter away from the cylinder in the direction perpendicular to the uniform flow. The spectrum (Figure 6) implies that low-pitch tones at the levels larger than 90 dB are generated as well as a broadband noise 20 dB smaller than these tones. This simulated noise can be heard also by clicking on this figure.

In the field of aeroacoustic noise estimation, research has traditionally been done by experiment. But, it takes a lot of work and time to examine a problem in such an experiment. Considerable efforts are made not only in flow and acoustic measurements but also in reprocucing a particular case of the relevant problem. It is not unusual to take a few months for preparation such as constructing a scale model of an object, and another few months for measurement with occupation of a laboratory. If an examination is all done in a computer, much work and time would be saved or more optimized solution could be supplied by testing large variations of the shape of the object. This is probably the reason why CFD and CAA are rapidly developed these days. It is, however, too early that the computational methods can take over the experimental method completely. Simulation results may sometimes vary depending on various parameters such as the numerical algorithm and the computational mesh. To acept the results, verification by an experiment is always needed. In this respect, CFD and CAA still remain a matte of research. Especially, CAA has not fully been established yet as compared with CFD. Not many studies have been done in estimating sound comparable to what we can actually hear from flow simulation results. To span the gap in our knowledge, our research focuses on this point. By using  CFD and CAA together with the traditional experimental method, the Fraunhofer IBP will work on various aeroacoustic problems in machines, vehicles, aircafts and others.


[1] D.C. Wilcox: Turbulence modeling for CFD, DCW Industries (1993).
[2] J.H. Ferziger: Large eddy numerical simulation of turbulent flows(1977), AIAA J., 15(9), 1261-1267.
[3] M. Lesieur; O. Metais: New trends in large-eddy simulations of turbulence (1996), Ann. Rev. Fluid Mech., 28, 45-82.
[4] N. Carle: The influence of solid boundaries upon aerodynamic sound (1955), Proc. Roy. Soc., London, Series A, 231, 505-514.