Tutorial: Computational Methods for Aeroacoustics

Sheryl M. Grace Dept. of Aerospace and Mechanical Engineering Boston University [email protected]

Opening comments
I refer to ANY computational method focussing on the computation of the sound associated with a fluid flow as computational aeroacoustics (CAA). The CAA methods are strongly linked to CFD CAA methods use specific techniques to resolve wave behavior well which makes this different than general computational fluid dynamics (CFD).

Kinds of applications

Rotors Helicopter noise, wind turbine noise Propulsion systems Stator/rotor, jet noise, combustion noise, propellers (underwater) Airframes Cavity noise, high lift wings Automobiles Cavities, mirrors HVAC and piping systems Fans, duct acoustics

Aeroacoustic domains

Linear equations with variable coefficients flow

Solid body

Full nonlinear equations

Computational boundary

Linear equations with constant coefficients

Main approaches being applied

Direct numerical, large eddy, and detached eddy simulationsDNS/LES/DES Useful for problems where the sound is from turbulence Euler and Linearized Euler Equations (LEE) LEE very popular when viscous effects can be considered 2nd order as a source of sound Splitting methods Based on LEE, applied to specific unsteady fluid-stucture interaction problems. Integral approaches -- need input from something (all above + CFD) Near field computation using some method above, acoustic field computed using an appropriate form of Ffowcs-Williams and Hawkings or the Kirchhoff method Other acoustic propagation methods Solve a wave equation or associated Euler equation numerically

Outline for today

Governing equations for the different approaches Application references (at end of talk) Implementation of methods for CAA applied to LEE and other CFD like methods Dispersion relation preserving method, Pad methods Boundary conditions Integral approaches Ffowcs-Williams and Hawkings, Kirrchoff method Other propagators

DNS
Two - dimensional governing equations in conservative form Cartesian co-ordinate system for a perfect gas

- density p - pressure u - velocity vector eT - total energy - viscous stress tensor T - temperature R - gas constant - Prandtl number - ratio of specific heats q - thermal conduction

Governing equations

DNS LES DES CFD Euler Linearized Euler Splitting method (at end)

Euler Equation
Two - dimensional governing equations in conservative form Cartesian co-ordinate system for a perfect gas

Moving to the conservative form of the energy equation, and noticing that the terms in the energy equation that involve can be replaced by u p, one derives a form of the equation that is useful when defining the linearized Euler equations:

LEE
Two - dimensional governing equations in conservative form Cartesian co-ordinate system
S is comprised of mass, momentum, and heat sources.

If the mean flow is uniform, H = 0.

Solution methods for aeroacoustics

DRP (spatial) Pade (spatial) Time marching Boundary conditions

Why the need for special schemes?

A numerical representation of a PDE gives rise to an approximate solution. A consistent, stable, and convergent high order scheme does not guarantee good numerical wave solutions. Vortical and entropic waves are nondispersive, nondissipative, and directional Acoustic waves are nondispersive, nondissipative, and propagate with s.o.s. All CFD techniques have some dissipation (many add artificial dissipation) to stabilize the schemes and the schemes are dispersive. (dissipation: amplitude of error) (dispersion: phase of error)
Comments and more detail found in Tam&Web, J. of Comp. Phys 107:262-281,1993.

Dispersion-Relation Preserving Schemes (DRP)

The number of wave modes and their wave propagation characteristics is found through the dispersion relation. Dispersion relation: functional relation between the and k (angular frequency and wave number) The dispersion relation is found by considering the space and time Fourier transforms of the governing equations. The LEE can be transformed into a simple matrix system:

The matrix A has 3 distinct eigenvalues. The repeated eigenvalue is associated with an eigenvector describing the entropy wave field, and an eigenvector describing the vortical wave field. The other 2 eigenvalues are associated with the acoustic wave field.

DRP scheme

(Tam,Web, J. Comp. Phys. 107:262-281,1993)

If one wants to match the dispersion relation in the numerical simulation, then one must match the Fourier transform. The finite difference representation of a derivative takes the form:

Defining the Fourier transform as The transform of the finite difference expression becomes

So we see the approximation to the wave number is DRP schemes formed by minimizing the wave number error, i.e. minimizing

DRP scheme cont.

~ For symmetric stencils (central difference type), k is a real number.
(For nonsymmetric stencils, one gets complex values and often spatially growing wave solutions.) Combination of the Taylor series finite difference approximation for N=M=3, and the minimization function, gives a fourth order accurate discretization

Scheme DRP Central Central Central

j=-3 -0.02651995 0 0 -1/60

j=-2 0.18941314 0 1/12 3/20

j=-1 -0.79926643 -1/2 -2/3 -3/4

j=0 0 0 0 0

j=1 0.79926643 1/2 2/3 3/4

j=2 -0.18941314 0 -1/12 -3/20

j=3 0.02651995 0 0 1/60

order 4th 2nd 4th 6th

Comparison of wave number accuracy

1.45

Waves with wavelengths longer than 4.33 grid spacings will be adequately approximated

DRP 6th order Central 4th order Central 2nd order Central

Pad/compact methods

(Lele, J. Comp. Phys. 103:16-42,1992)

Match Taylors series coefficients of various orders given:

Fourier transform of the space variable gives

So the approximate wave number is given by Tridiagonal schemes, 3 pt. stencil l.h.s., 5 pt. stencil r.h.s
Scheme Pad Pad a1 0.25 0.33333333 a2 0 0 b1 0.6666666 4.6666666 b2 0 0.11111111 b3 0 0 order 4th 6th

Comparison of wave number accuracy

Comment by Wolfgang Schroder : VKI lecture 04 2D: DRP 3 faster than Pad 3D: DRP an order faster

Pad (6th order) (*) Pad (4th order) (dashed)

DRP (4th order) 6th order Central 4th order Central 2nd order Central

Time discretization
One can use a time discretization scheme constructed using the DRP methodology. (Tam and Web) The method requires the addition of some artificial damping because of the existence of spurious short waves. Tam and Dong: J. of Comp. Acoustics, 1:1-30, 1993; Add artificial damping such that the damping is confined to the high wavenumber range. Then the short waves are damped leaving the long waves basically unaffected. Zhuang and Chen: AIAA J. 40(3):443-449, 2002; Use high-order optimized upwind schemes that damp out the spurious short waves automatically. Hu et al, J, of Comp. Phys, 124:177-191, 1996; Use low-dissipation low-dispersion, low-storage Runge-Kutta schemes.

LDDRK (Hu)
Consider the time discretization

applied to the one-dimensional wave equation where from the spatial transform discussed earlier so that

Define the amplification factor for the scheme

LDDRK cont.

Dissipation error

Dispersion error

Classical Rung-Kutta matches the expansion for e-i So that 1 = 1, 2 = 1/2!, 3 = 1/3!, 4 = 1/4! would give a fourth order approximation
Specifies range of

Here m chosen to minimize and to satisfy the stability limit |r| <= 1

LDDRK (cont.)

For all schemes: 1 = 1, 2 = 1/2!

Stages

3
0.1629978 0.166558 1/3!

4
0.0407574 0.0395041 1/4!

5
0.00781071 0.00781005

6
0.00132141

Order

4 5 6

0.085 1.35 1.75

2.85 3.54 1.75

2nd 2nd 4th

accuracy limit

stability limit

Hu et al. discusses the equivalence of this method to minimizing the error in the dispersion error. Also, the implementation of boundary conditions is discussed. Bogey, Bailly, J. of Comp. Phys, 194:194-219, 2004 is another good example of applying this method to define a LDDRK.

Time discretization comparison

dissipation dispersion

1-|r|

0.6

0.75

t 4th order 6 stage LDDRK (*) 3rd order DRP

2nd order 5 stage LDDRK 2nd order 4 stage LDDRK 2nd order 4stage classical RK Exact (dashed)

Boundary conditions
One must set computational boundaries that draw a line between a flow region of interest and other regions that are to be neglected. These boundaries must not produce unrealistic reflections or spurious solutions BC type Characteristicbased Asymptotic ** Pro Straightforward and robust Con Inaccurate for wave angles not perpendicular to the boundary BC surface must be in far-field, not always applicable Large zone, may produce some reflections at interface Stability has been an issue, seems to be better now. Set up for linearized eqs.

Accurate when applicable

Buffer zone

Quite familiar to CFD community Absorbs well with smaller zone, no reflections

Perfectly Matched Layer **

BCs cont.
Consider the asymptotic boundary conditions (and BCs that will apply be applied in the linearized region of the flow field) For inviscid, nonheat-conducting, calorically perfect gas, one can decouple the equations into equations that govern the vorticity fluctuation, the entropy fluctuation, and the pressure fluctuations.
Radiation boundary condition radiation boundary condition outflow boundary condition
acoustic wave source

inflow

vorticity and entropy waves

radiation boundary condition

Good reviews: Colonius, Annu. Rev. Fluid Mech. 36:315-45, 2004 Givoli, J. Comp. Phys, 94(1):1-29, 1991 Hagstrom , Acta. Numerica, 8:47-106, 1999

Asymptotic BCs
Tam and Web : LEE, uniform flow in x-direction only, simple starting point.

Transformed system

Already mentioned that A has 3 distinct eigenvalues giving rise to 4 eigenvectors, X1, X2, X3, X4 where X1 can be associated with the entropy wave, X2 can be associateed with the vorticy wave, and X3 & X4 can be associated with the two modes of the acoustic waves.

Radiation BC

(Tam/Web)

can be solved in terms of the eigenvectors: Focus on the two acoustic modes, in particular, the outgoing mode at a boundary. Transform back to space and time using

denotes acoustic part

You get a solution in the form of

V velocity of propagation in direction

In general this states that the acoustic disturbance satisfies:

Radiation (asymptotic) BC

Other conditions that are similar Bayliss, Turkel, J. Comp. Phys. 48:182-199, 1982 Hagstrom, Hariharan, SIAM J. Sci. Comput., 25(3):1088:1101, 2003 (high order!)

Outflow (asymptotic) BC

(Tam/Web, 1-D uniform mean flow)

At the outflow, entropy, vorticity, and acoustic waves must traverse the boundary. Use the same process of evaluating the behavior of the entropic and vortical parts of the solution. The density perturbation is associated with the entropic mode and the vortical mode is associated with the perturbation velocity vector.

The form of U at the outflow is then

Outflow boundary equations

BCs cont.
Tam and Dong: J. of Comp. Acoustics. 4(2):175-201, 1996; Extended the work of Tam and Web to multidimensional nonuniform mean flow. The boundary condition equations have the same form excep that

Perfectly matched layer

Following Hu, J. of Comp. Phys. 173:455-480, 2001.

PML

Computational domain

Add absorbing layer at end of computational domain where the absorption quantities are based on the plane wave solutions of the linearized Euler equations. Create the absorption quantities such that all three wave types are absorbed in the appropriate area. PML differs from the regular buffer zone technique in that the equations used in the added region will not cause ANY reflection when entering the region at any frequency and angle of incidence. Hu uses the nondimensional form of LEE, but follows the same process of forming the plane wave solutions (based on the dispersion relations) discussed previously. One forms but in nondimensional quantities, is 1.

and then find the eigenvalues of A which when set to zero give the dispersion relations.

PML cont.
Combined acoustic eigenvalues give y
wave angle

Eigenvalue for the entropy and vortical modes gives

entropy

vortical

A single Fourier/Laplace component of U is still formed from where X_j are the eigenvectors

PML cont.
Absorption coefficients are then introduced using a splitting method

absorption coefficients

Recombining the two equations, and considering a single frequency gives

Clear denominator of and reformulate to space-time

Introduction of scaled spatial parameters, allows one to write the acoustic mode for a given Fourier/Laplace component in such a way that the damping is clear.

There are similar expressions for the entropy and vortical modes

PML cont.

It was shown that there is an instability arising due to convective acoustic waves that have a positive group velocity but a negative phase velocity in the x-direction.

Another spatial transformation is used to overcome this instability, and the final equation that one uses in the PML becomes

0 in vertical layer

0 in horizontal layer
vertical layer x=0

q is only introduced in the PML domain.

PML cont.
The absorption coefficients can be varied gradually, for example (given in Hus paper): m x = 2 (x is the grid size) =2
location where PML domain starts

Term to allow absorption rates to be the same in the x and y directions

width of PML domain

Outer edge of the PML can use characteristic, asymptotic, or even very simple reflective type boundary conditions.
yl

xl

Solution methods using the integral approach

Acoustic analogy Ffowcs-Williams and Hawkings Kirchhoff

Flow field quantities are known in a region near the source, use the integral approaches to find the propagation of the acoustics to the farfield.

Lighthills Eq.
creation of sound generation of vorticity refraction, convection, attenuation, known a priori Lighthill stress tensor mean speed of sound mean density

excess momentum transfer

attenuation of sound wave amplitude nonlinearity mean density variations

Solution to Lighthills Eq.

Quadrupole like source! Direct application of Greens function

Far-field expansion, integration with respect to retarded time

Some example applications of Lighthills analogy

Uzun et al. AIAA Paper No. 2004-0517. Applied to jet flow, coupled to an LES. Source term for acoustics propagating in a specific direction (figure)

Colonius, Freund, AIAA J. 38(2):368-370. Applied to jet flow, coupled to a DNS. Oberai et al. (AIAA J., 40(11):2206-2216, 2002) - airfoil selfnoise, coupled to FEM LES

FWH
Differential form of the FWH Eq.

Turbulence is moving Two distinct regions of fluid flow Solid boundaries in the flow

unsteady surface pressure Reynolds stress i i viscous stresses

rate of mass transfer across the surface

Only nonzero on the surface

FWH cont.
Integral form of the FWH Eq. Quadrupole

Dipole Monopole Square brackets indicate evaluation at the retarded time

If S shrinks to the body dipole = fluctuating surface forces monopole = aspiration through the surface

Comments on FWH method

There is a formulation for moving surfaces (some discussion included as

an appendix of this presentation)

There is a formulation for permeable (or porous) surfaces developed for use with CFD where the surface has to be placed
quite close to the body, but not on the body Paper to appear in AIAA J (in near future) - A. Morgans et al. CFD + permeable surface FWH for transonic helicopter noise

Brentner, Farrassat, Progress in Aerospace Sciences 39:83-120, 2003 Great review/overview of use of FWH in rotor noise studies Gloerfelt et al., JSV 266:119-146, 2003, FWH and porous FWH for 2D
cavity problem coupled to DNS Acoustic part computed in the frequency domain (no need for retarded time variable this way.)

Kim at colleagues at FLUENT, AIAA Paper No. 2003-3202, couples

FWH to their LES solver.

Comment on Kirchhoff method

Solve the homogeneous wave equation using the free-space Greens

function approach

All sources of sound and nonuniform flow regions must be inside the
surface of integration. Integration surface must be placed in the linear region of the flow.

FWH is same if the surface is chosen as it is for the Kirchhoff method FWH superior Based on the governing equation of motion (not wave equation) Valid in the nonlinear region Both methods when used in the linear regime, may capture a lower
maximum frequency (CFD method may use grid-stretching).

Kirchhoff and FWH cont.

Brentner, Farrassat AIAA J. 36(8):1379-1386, 1998 Compare FWH and Kirchhoff FWH does separate contributions to the noise (if surface is placed
closed to or on body)

Patrick(Grace) et al.ASME FED 147:41-46, 1993, used the Kirchhoff

method coupled to the splitting method for fluid/airfoil interaction noise

Gloerfelt et al. also compares the two methods (for the cavity problem) Lyrintzis gives a great review of coupling CFD to FWH and Kirchhoff
in Int. J. of Aeroacoustics 2(2):95-128, 2003

Uzun et al., AIAA Paper No. 2004-0517, shows LES coupled to open
FWH and Kirchhoff surfaces for jet (meaning jet outflow not enclosed by the surface - see figure above)

Rahier et al. Aero. Sci. & Tech. 8:453-467,2004, open vs. closed
surfaces for jet analysis. Open surface makes more sense.

Alternative couplings
Solve the linearized Euler equations (with an additional term) using near-field DNS or LES as boundary information. Additional term drives the density towards the Navier-Stokes value. Applied to jet M = 0.9, JFM 438:277-305,2001)

Freund, J. of Comp. Phys. 157:796-800, 2000.

Grace, Curtis. ASME NCAD, 1999. Low M applications, compute

solution to wave equation in appropriate region using CFD input

Some more CAA applications in the literature

DNS examples

Gloerfelt, Bailly, Juve (JSV 266:119-146, 2003) - subsonic cavity Use DRP to discretize equations Use non-reflecting boundary conditions + absorbing layer Couple to an integral approach

Colonius, Freund, Lele (AIAA J, 38:2023, 2000) - supersonic jet Use Pade methods for discretization Use non-reflecting boundary conditions

LES examples

Bogey, Bailly, Juve (Theor. Comp. Fluid Dyn, 16:273-297,2003) - jet Use DRP to discretize equations Use non-reflecting boundary conditions Uzun, Lynrinztis, Blaisdell (AIAA Paper No. 2004-0517) - jet Use DRP to discretize equations Use non-reflecting boundary conditions Couple to an integral approach Sheen, Meecham (ASME Fluids Div, Sum. Mtg, 2:651-657, 2003) - jet Coupled to an integral approach Oberai et.al. (AIAA J., 40(11):2206-2216, 2002) - incompressible airfoil Use finite element incompressible LES Use non-reflecting boundary conditions Coupled to an integral approach

Euler, examples
Main applications found in literature -- rotor type simulations where flow disturbance is periodic and dominant in the creation of sound.

Lee, et al (JSV 207(4):453-464, 1997) - rotor noise

Coupled to integral approach

Lockard, Morris (AIAA J 36(6):907-914, 1998) - airfoil/gust

allowed for viscosity in some calcs

Hixon (AIAA Paper No. 2003-3205), Golubev - cascade,airfoil/gust

Different approach, no time marching, space time mapping

LEE, examples

Florea, Hall (AIAA J, 39(6):1047-1056,2001) - cascade/gust Low-order discretization, finite volume Bailly, Juve (AIAA J, 38(1):22-29,2000) - apps. DRP scheme Longatte, et al (AIAA J, 38(3):389-394,2000) - sheared ducted flow Lim, et al (JSV, 268(2):385-401,2003) - diffraction from impedance barriers High order discretization Ozyorok, et al (JSV, 270(4-5):933-950,2004) - turbofan noise Chen, et al (JSV, 270(3):573-586,2004) - sound from unflanged duct Mankbadi et al (AIAA J, 36(2):140-147, 1998) - jet

Splitting (LEE-based)

(Atassi, Grzedzinski, JFM 209:385-403)

This new unsteady velocity splitting appears as

The vortical part of the velocity still satisfies exactly as it did in the original splitting, but the boundary conditions are now defined so that there is no singularity

The potential function is now governed by

the boundary condition along the surface is the jump of the potential velocity in a wake must be 0 far upstream:

Splitting, examples
9 The vortical part is first solved analytically or numerically, and then the potential part is found numerically. 9 Most often these problems are computed in the frequency domain. 9 As shown on last slide, valid for subsonic, nonswirling flows

Scott, Patrick/Grace, Atassi, (JCP 119(1):75-93, 1995, AIAA J. 31(1):12-19, 1993) - airfoil/gust Low order-finite difference Coupled to an integral approach Fang, Atassi, (JFE 115:573-579, 1993) - cascade/gust Low order-finite difference Novel non-reflecting boundary conditions Verdon, Hall (AIAA J. 29(9):1463-1471, 1991)- cascade/gust Peake et al. - (JFM 463(25):25-52, 2002) cascade/gust Golubev, Atassi (AIAA J 38(7):1142-1158,2000) - cascades/swirling flow Vortical velocity is no longer the solution to a homogeneous equation

Summary

Weve consider the main features of computational aeroacoustic methods Governing equations - hierarchy of approximations Discretization schemes

DRP, Pad--spatial, LDDRK--time, methods for damping not described

Boundary conditions Acoustic propagation methods for coupling near/far fields

Many choices are problem dependent -- makes it difficult to incorporate good acoustic calculations in general CFD type codes If one is implementing these methods, it is good to use the CAA benchmark problems as preliminary method checks.

Thanks: Atassi, Tam, Bogey/Bailly, AME Dept., NCAD, Sondak

The End

More questions?

LES
Two - dimensional governing equations in conservative form Cartesian co-ordinate system for a perfect gas Spatially filtered (overbar), Favre (or density weighted) average (tilde) Smagorinksy turbulence model

- density p - pressure u - velocity vector eT - total energy - viscous stress tensor T - temperature R - gas constant - ratio of specific heats q - thermal conduction t - turbulent viscosity T - subgrid scale stress tensor Cs - Smagorinsky constant

DES

Bisseseur et. al. (Aerospace Science Mtg. Proc. pg. 1673-1685, 2004) Use high-order compact difference scheme Couple to an integral approach

CFD, examples
Turbulence is modelled, usually high level of dispersion in discretization

Kim, et al (AIAA Pap. 2003-3202) - general apps FLUENT for near-field Coupled to an integral approach Hendriana, et al (AIAA Pap. 2003-01-1361) - sideview mirror FLUENT for near-field Coupled to an integral approach Grace, Curtis (ASME, IMECE, NCAD 26:103-108, 1999) - cavity FLUENT (URANS) for near-field Coupled to solution of wave equation

Splitting Technique

(Goldstein, JFM 89(3):433-468)

Ideally (uniform mean flow) the unsteady velocity can be split into solenoidal (vortical/entropic) and irrotational (potential) parts with separate governing equations. The components are linked through boundary conditions. For realistic flows (no shocks or swirl), the velocity components cannot be split as such. Potential governed by a single inhomogeneous, non-constant coefficients, convective wave equation forced by the solenoidal component Vortical part governed by homogeneous, non-constant coeffeicient, convective wave equation Entropic part governed by energy equation.

Used when the disturbance is vortical or entropic not acoustic. Upstream where flow is uniform:

So entropy and velocity upstream are the boundary conditions imposed on the flow.

Splitting cont.
Equations for the split variables are derived from the nonconservative form of the governing equations with the energy equation expressed in terms of entropy.

One can show that the solution to this set of equations can be written as: where
The components of the argument are Lagrangian coordinates of the steady flow fluid particles. The components of X are defined as independent integrals of

Lighthill drift function, time for particles to travel along a streamline

Splitting cont.
Solenoidal part of flow is defined and then the single equation:

must be solved, subject to the boundary condition on the surface: and far upstream:

This splitting above leads to singular behavior of the solenoidal part along the solid boundary.

Time discretization
Consider the time discretization:
3 bs chosen so that the Taylors series are satisfied to 3rd order. Leaving one free parameter b0 to minimize dispersion error.

Transform the discretized equation

b0 chosen to minimize E1. allows one to adjust the degree of emphasis placed on wave propagation (real part) or damping characteristics (imaginary part). Tam uses a value of 0.36

Time cont.
~ is not unique based on . Spurious numerical solutions exist because
Optimization range has been selected based on the behavior of the ~t is well behaved for values approximated frequency. In particular, less than 0.6, and the optimization ranges from -0.5 to 0.5. For stability, the entire computation must be restricted to the range of t from -0.6 to 0.6. Time step must be chosen to ensure numerical stability:

M is the mean flow Mach number, c0 is the mean flow speed of sound, 0.4 is the value under which all of the roots of are damped. Numerical damping due to the small imaginary part of the approximate frequency. This gives a more stringent requirement on the numerator above. (Details in Tam)

What we can learn from the far-field form

For low Mach number, M << 1 (Crow) If the source is oscillating at a given frequency The far-field approx to the source in Lighthills equation can be written as

Therefore the solution becomes Scalings: velocity -- U, length -- L, f of disturbance -- U/L

Acoustic wavelength/source length >> 1

Acoustic field pressure fourth power of velocity Acoustic Power eighth power of velocity

FWH Eq.
Need a more general solution when:
Turbulence is moving Two distinct regions of fluid flow Solid boundaries in the flow Define a surface S by f = 0 that encloses sources and boundaries (or separates regions of interest) Surface moves with velocity V Heavy side function of f : H(f) Rule:

Curle Eq.
When the surface is stationary the equation reduces to Curles Equation

Comments on analogy
Analogy is based on the fact that one never knows the fluctuating fluid flow very accurately Get the equivalent sources that give the same effect Insensitivity of the ear as a detector of sound obviates the need for highly accurate predictions Just use good flow estimates Alternative wave operators that include some of the refraction etc. effects that can occur due to flow nonuniformity near the source have been derived: Phillips eq. , Lilleys eq. Using these is getting close to the direct calculation of sound.

FWH cont.
Multiply continuity and Navier Stokes equation by H Rearrange terms, add and subtract appropriate quantities Take the time derivative of the continuity and combine it with the divergence of the NS equation
Differential form of the FWH Eq. Dipole type term

i Quadrupole

Monopole type term

Example of application of FWH/Curle

We have a BEM for calculating the near field (surface forces)
(Wood, Grace)

3D BVI (rotor-type problem)

BEM computes unsteady pressure on the wing surface

Curles Eq.

Ex. cont.

far-field expansion integration of pressure is lift interchange space and time derivatives

Acoustic pressure in non-dimensional form

Ex. cont.
our CL vs. t Curle acoustic calc using our CL

our acoustic calc using our CL

Curle acoustic calc using our CL Purely analytic acoustic calc (based on analytic CL

Additional info FWH moving frame

FWH moving frame

Introduce new Lagrangian coordinate Inside integral, the function depends on now and

Where the additional factor that appears in the denominator is =

because

angle between flow direction and R = unit vector in the direction of R

FWH moving cont.

Volume element may change as moves through space

density at = 0 Volume element affected by Jacobian of the transformation

When control surface moves with the coordinate system becomes ratio of the area elements of the surface S in the two spaces

Retarded time is calculated from

FWH moving cont.

When f is rigid: uj = Vj When body moves at speed of fluid: Vj = vj

Square brackets indicate evaluation at the retarded time e > 1 for approaching subsonic source < 1 for receding subsonic source

Doppler shift

accounts for frequency shift heard when vehicles pass

Turbulent noise sources

Stationary turbulence (low M)

Moving turbulence (high M)

Stationary turbulence (Low M)

Far-field form

From before. Pressure goes as fourth power of velocity and power as eighth power of velocity

Moving turbulence

Pressure goes as scaled fourth power of velocity Power goes as eighth power scaled by

Example
Method based on : Unsteady CFD -> forced wave equation solved numerically Goal : Make use of existing CFD through a hybrid method for computational aeroacoustics (i.e. no integral formulation for the acoustics) viscous acoustic

Results for U = 33.1m/s, L/D = 8.0, Re=8100

Unsteady non-dimensional source term

Acoustic pressure field

t=0.5168 s

t=0.5247 s

t=0.5326 s

t=0.5405 s