Turbomachinery Seminar - Aero-Acoustic

© 2014 ANSYS, Inc. September, 2014 1

Aero-acoustic assessment of turbomachinery using advanced turbulence modelling methods

Satish Patange

ANSYS UK Ltd


Outline

Acoustics modeling

Applications of rotating machines

Sound propagation

Fan flow

SRS


Acoustics – Key featuresMagnitude of acoustic waves is very

small compared to aerodynamic

pressure.

Acoustic radiation contains only a tiny

fraction of primary flow energy.

• Most unsteadiness in flow is ‘pseudo sound’

and does not radiate!

Acoustic problems are unsteady!

Frequency range of interest is quite

large:

� Frequency range 20Hz – 20kHz

• Temporal resolution for acoustics is orders of

magnitude larger than the interesting time

scales in the flow.

• Small eddies need to be captured, requires

spatial resolution.


Aero acoustics – Source Classification

Monopole simple source

Quadrupoletwo dipoles

Unsteady mass

injection

Acoustic ~ U 3M

Power

Unsteady external

forces

Acoustic ~ U 3M 3

Power

Unsteady turbulent

shear stresses

Acoustic ~ U 3M 5

Power

Monopole and dipole sources dominant at low Mach numbers.

Scaling valid for acoustically compact sources, λ >> L!

psurface = psurface(t) τ = τ(t))(tmm && =

Dipole two mopoles


Turbomachinery NoiseDiscrete + Broadband

Steady Rotating Forces

(Lowson/Gutin Models)

Discrete

Unsteady Rotating Forces

Discrete + Broadband

– Steady flow, discrete – Unsteady flow, discrete + broadband

– Secondary flow, discrete + broadband

– Vortex shedding, narrowband + broadband

– Turbulent BL, broadband

MonopoleBlade Thickness NoiseDiscrete

DipoleBlade Loading NoiseDiscrete + Broadband

QuadrupoleTurbulence NoiseBroadband

Aero acoustics – Approaches


Aero acoustics – Simulation Basics

Aeroacoustics modeling involves simulation of

two aspects

• Sound source

– Provides source characteristics and rankings

• Sound propagation

– Propagation of sound from the source to the receiver

• Requires input of source characteristics

• Provides

– Sound spectrum and receiver

– Sound directivity


Sound source simulation is done with detailed CFD analysis of

flow around the blades, hub, shroud, etc.

Can be done in two ways

• Steady State

• Transient

Advantages/Disadvantages

• Steady State

– Computationally cheap, fast, but not very accurate

• Transient

– Computationally expensive, slow, but more accurate

• After all, sound generation is a highly transient phenomenon



Practical usage of Steady State and Transient methods

Simple hand

calculations

Steady StateTransient and

Experimentation

Design

Possibilities

Final

Design

Design

Screening

Methods

Increasing Accuracy and Expense




Sound propagation can be calculated in different ways

CAA (Computational Aero Acoustics)

• Direct sound computation

• Uses the transient turbulence modeling capability in CFD

– LES, WMLES, DES, Detached DES and Scale Adaptive

simulation

SSPM (Segregated Source-Propagation Methods)

• Propagation is decoupled from source

– Source and propagation are treated as mutually independent

– Models many be used for computing propagation

• Lighthill-Curle Method

• Ffowcs-Williams-Hawkings Method

• FEM/BEM (Solution of Lighthill’s equation/Wave equation)


Computational Aero Acoustics (CAA)

In fact, wave equation is a special case of Navier-Stokes equations.

� CFD solves the Navier-Stokes equations.

� In theory, sound generation as well as propagation can be simulated by:

• Transient, compressible CFD simulation

With computational domain spanning from sources to receivers!

• Monitor static pressure at the receiver locations as function of time

SRS (or URANS if tonal noise)

� No further models involved!

10 100 1000Frequency [Hz]

0

10

20

30

40

50

60

70

80

90

100

110

120

130

140

SPL [dB]

Freestream Velocity = 140 km/h

Experimental data

SAS model

Sensor 121

SAS of a side view mirror

(ReD = 520 000)


Acoustic analogy

Acoustic analogy assumes acoustics can be decoupled from flow dynamics.

On the basis of Lighthill’s analogy:

• Noise Sources are assumed in a uniform fluid at rest

• Acoustic field at observer is described by wave equation

• Resolution of acoustic and dynamic flow field are decoupled

Based on two steps:

• Simulate transient flow field accurately using CFD to get the acoustic sources

location and intensity

• Propagate noise from sources to receiver by solving wave equation

analytically

CFD domain

Wave EquationAcoustic sources

Acoustic

receiver


Acoustic Analogy – Integral FW-H

Williams, J. E. F. and Hawkings, D. L. (1969): Sound generation by turbulence and surfaces in

arbitrary motion. Philosophical Transactions of the Royal Society, Vol. A264, pp. 321-342.

The solution contains surface integrals over source surfaces and a

volume integral

� Less sensitive to proper placement of permeable source surfaces

than other integral methods (e.g. Kirchhoff)

Volume integral not directly solved – too time consuming

� Collect all sources inside permeable surface

Noise generated in the fluid volume

(Quadrupole)

Loading noise

(Dipole)

Thickness noise

(Monopole)

where


FW-H – Example: Canon Loudspeaker

Bass-reflex loudspeaker to increase

efficiency of the system at low

frequencies

� Low frequencies sound radiated

through the port and added in phase

with the driver front wave

3 million cells, Δt = 8*10-6 s

deforming zone

moving zone

Courtesy of Canon

M.Younsi, G.Kergourlay, V. Morgenthaler, Near Field and

Far Field Prediction of Noise in and around a

Loudspeaker: A Numerical and Experimental

Investigation EURONOISE 2012, 10–13 June, Prague


Lowson/Gutin ModelLowson

• Noise level @ specific location

• Unsteady load replaced by steady load

multiplied by exponential decay function

(semi-empirical):

− Wave propagation from rotational

symmetric geometries (Bessel

function)

Gutin

• Steady loading noise of blades

• Solver rotates signal, not geometry

• Considers thickness noise (monopoles)

and loading noise (dipoles)

( )( )

h

0y

0x

y

x

F

F

F

F−λ

=

λ

λDirectivity plot of

1st harmonic

Steady-state vs. transient for 2-bladed fan


Aero acoustics – Approaches

Features &

LimitationsCAA

Acoustic

analogy

(FWH)

Lowson/

Gutin

Modal-

Analysis

Broadband

noise

modeling

Computation

costMost Fair Moderate Moderate Least

Account for

reflectionYes No No No No

Account for

effect of

sound on

flow

Yes No No No No

Solution

schemeTransient Transient Steady State Steady State Steady State

AccuracyVery

GoodGood Limited Limited Limited

3rd party

coupling

(1-way)

Fair

Yes

No

Transient

Good

Decreasing computational effort

Increasing accuracy


Applications in rotating machines


Fan Noise Macro in CFD-Post

Specific to ANSYS CFX

Based on the Lowson Noise Model

• Low speed machines : Tip Mach Number < 0.35

• High Mach number : Less accurate

• Forces acting as punctual force on gravity center

• Small blade span

• Usage of semi-empirical coefficient to define loading decay

Acoustics Pressure at mth Harmonic

( ) ( ) ( ) ( )ϕ

λλ−−λϕ−

πω

= λ−

+∞=λ

−∞=λ

λ−∑ sinmzMJM

F.

mz

mzFcosi

r.c..2

imzp mz

y

X

mz

1o

2

m

Where, λ = Harmonic Mode; M =Mach Number; z = Blade Number; ω = Rotational Speed (rad/s)

h = Loading Coefficient (2.0 ~ 2.5)

( )( )

h

0y

0x

y

x

F

F

F

F−λ

=

λ

λUnsteady Force Components Steady State Force Components : Fx0 & Fy0


Fan / Turbo Noise Macro [1]

Typical Fan Noise Output Results

Fan / Turbo Noise Calculation Examples


Broadband Noise Models

Two kinds of models are available in ANSYS Fluent

• Broadband models based on averaged quantities

– Proudman’s formula for turbulence noise

– Turbulent boundary layer noise model

– Jet noise model (2D axisymmetric only)

• Broadband models based on reconstruction of flow field fluctuations

– Source terms in Linearized Euler equations (LEE)

– Source terms in Lilley’s equation

ANSYS CFX

• Estimate of noise source strength

– Monopole sources

– Dipole or rotating dipole sources

– Quadrupole sources


ANSYS CFX-Pre setupANSYS CFD Post

Monopole Terms

Sound Source Strength Prediction [1]



Dipole Terms




Quadrupole Terms



Ori

gin

al

De

sig

n

Radial

Forward

Op

tim

ize

d

De

sig

n

Forward

radial

Design point

Measurements

Forw

ard

: Lo

w N

ois

e a

t D

esi

gn

Po

int

Design Comparison



Case Study #1

• Aeroacoustics Modeling of a Centrifugal Fan Using ANSYS CFX

– N = 3000 rpm Z = 39 Blades

– Near Field & Far Field Noise Prediction

– Steady Flow Simulation using SST

Turbulence Model

– Unsteady Flow Simulation using Scale

Adaptive Simulation (SAS) Turbulence

Model

– Node Count = 2.177 Million

– 6 Near Field Microphone (Two used to

Capture Noise Spectra)

– 1 Far Field Microphone

– Far Field Noise Modeling using ANSYS

CFX Turbo Noise Macro Based on

Lowson Model


Near Field Microphones

Far Field Microphones

Case Study #1



Near Field Noise Prediction

Microphone #1 Microphone #4

Case Study #1



0 2000 4000 6000 8000

Frequency, Hz

0

10

20

30

40

50

60

70

Sound Pressure levels, dB

Experimental data

TurboNoise macro

At BPF SPL [dB]

TurboNoise 56.8

Experiments 55.9

Near Field Noise Prediction

ANSYS CFX Turbo Noise (Based

on Lowson Model)

Case Study #1



Case Study #2

• Aeroacoustics Modeling of an Automotive Electric Cooling Fan

Using ANSYS Fluent

– Free-standing fan (open to

atmosphere on all sides)

– Nine, evenly spaced blades Fan Speed

= 2000 rpm

– Single Reference Frame

– Single Blade Modeling

– Cell Count ~ 10 Million

– LES Turbulence Model

– FW-H Model For Far-Field Sound

Propagation


Low values

occur at

higher radius

due to higher

flow velocity

• Grid & Temporal Resolution Verification:

– Height of First Cell Center on the blade

is roughly equal to the Taylor length

scale (λλλλ)

– Grid is fine enough to capture eddies in

the inertial sub-range

– Therefore the grid is good for

conducting a “true” LES computation

– Time step required for LES ≈ λλλλ/U

– Steady state results indicate that the

timestep for transient LES solution

should be roughly 1E-6 second.

Taylor Length Scale λ:λ:λ:λ: Blade Pressure Side


Using ANSYS Fluent

Case Study #2



Using ANSYS Fluent

Source Pressure Spectra

0

20

40

60

80

100

120

140

160

0 500 1000 1500 2000 2500

Frequency (Hz)

SPL (dB)

pt01 pt02

pt03 pt04

pt05 pt06

Near Field Sound Spectra

Case Study #2



Using ANSYS Fluent

0

10

20

30

40

50

60

0 200 400 600 800 1000 1200 1400 1600 1800 2000

SPL (dBA)

Frequency (Hz)

2000 RPM, 0 degrees, 1 meter

Far Field Sound Spectra

Flow Structure

Instantaneous Iso-

Surface of 2nd Invariant

of Velocity Gradient

Rotation

Rotation

Vortices in Near Wall

Region

Case Study #2


Conclusion

• Unsteady simulations are the future for many CFD applications.

• A wide spectrum of Scale Resolving models are available in ANSYS CFD :

o LES,

o WMLES,

o (D-)DES,

o EMBEDDED LES,

o SAS.

• Such models can be combined with different acoustic approaches,

particularly:

o Direct CAA,

o Acoustic analogy (FW-H).

• Question is: Which approach is best suited for which type of flows?

� Best ratio of cost vs. performance.

� Safest environment for user (limited sensitivity to mesh, time step, …).

• User feed-back is always welcome and appreciated!

Documents

Turbomachinery Seminar - Aero-Acoustic