jeong-guon ih, prof. - pennsylvania state university workshop_jg ih.pdfacoustics lab., novic inverse...

Acoustics Lab., NoViC

Jeong-Guon Ih, Prof.

Acoustics Lab. (http://aclab.kaist.ac.kr )

Center for Noise and Vibration Control (NoViC)

Department of Mechanical Engineering

CAV Workshop, Penn State Univ., 29 April 2014; 03:45-04:15 pm

Acoustics Lab., NoViC Acoustics Lab., NoViC

Overview Lab. Members: (as of Mar. 2014)

Ph.D. Students: 6 (1 part-time student)

MS Students: 9 (1 dual degree,1 part-time student)

Graduates: (Since Feb. 1992):

2

Thesis topics related to the Inverse Problem

Current Ph.D. Thesis Topics on Inverse Vibro-acoustics

Inverse acoustic estimation of sectional temperature distribution in a duct (TK Kim)

Inverse source identification of the fluid machine in a wide duct with flow (YH Heo)

Inverse diagnostics of wave propagation in the multi-dwelling building structures (JH Bae)

Localized vibration field suppression by utilizing array control technique (SW Hong)

Generation of localized vibration in a thin plate by using the array actuators (JH Woo)

Current MS Thesis Topics on Inverse Vibro-acoustics

Leakage detection in a long pipe by using the inverse acoustic method (E Yadollahi)

Acoustic source localization by using the multiple 3D-intensity array (F Franek)

Inverse estimation of excitation force at multiply joined structural interface (M Fayyazi)

Center for Noise and Vibration Control

Ph.D. = 26, MS = 49


Inverse problem

Problem in determining the internal physics or past state of a

system from indirect measurements

3

(finite number of vibro-acoustic data at the field)

(vibro-acoustic data in discrete form at

the source) (radiation, scattering,

reflection, diffraction in continuous domain … should

be modeled in a discrete form)

(regularization)

sensor

inversion

Vibrating source

Sound pressure

Vibro-acoustic source


4


Indirect Source Identification Using Multiple Microphones

Acoustic Intensity

NAH-spFT

NAH-iBEM

)(*)(Re2

1)( rrpr vI

• Separable coordinates only

• Hypothetical regular plane for

complex-shaped source

• Singularity prob.: regularization

• BEM modeling

fs pGv

)]],([)([1

hhs

jkx yxFyyeF fs pp

backward forward

hologramplane

sourceplane

predictedplane

SourcePlane

PredictedPlane

HologramPlane

5

1

f f s s f s sp (D D M M ) v Gv

(continued)


iFRF fs pHQ

• Singularity prob.: absorption,

regularization

• Resolution

• Same concept with TPA

Q12Q

iQ

nQ..

.

..**

**

*

*

1p

2p

jp

mp

..

...

H 1

H H H HH H

ESM

1 0

( ; ) ( ; )E J

e

f m j j m e

e j

Cp r r r

11{ } { }{ }{ } { } { }{ }{ }{ }H H

f fC U Λ W p W Λ U p

• Generalization of HELS/iFRF

• Describe complex wave fronts

• Reconstruction error

6

Indirect Source Identification (continued)

HELS • Optimal no. of expansion

terms

• Optimal selection of field pts

* *

1

N

i i

i

p c C

11 * * *

0

T T

mn mn mnC c p


40

50

60

70

(b)

Max o

nly

Window method

Rear center position

Codriver's ear position

Panel name

Bare

only

Rear

tunnel

Rear

floor

Headli

ner

Package t

ray

Heel

kic

k

Rear

doors

Fro

nt

doors

Dash

sid

es

Dash

upper

Dash

bott

om

Dash

cente

r

Fro

nt

floor

Fro

nt

tunnel

So

un

d p

ress

ure

lev

el

(dB

A)

Driver's ear position

Boundary

Microphones

Candidate

sources

mP

nQ)(H

► Calibration

( ) ( ) ( )m m n nP H Q

► Identification

Actual

sources

)H(

mP̂

nQ̂

1

H H H HH H

7

( ) ( ) ( )n n m mQ H P

+

S S fv S = H p

Passenger car interior


8

150 Hz (E3) 200 Hz (E4)

Animated

surface

velocity field

3000 RPM Surface intensity @

E/G Top

E/G Exht


12th

RPM

Fre

qu

en

cy (

Hz)

6th

3rd

2nd

RPM (625-800)

9

Measured hologram Restored velocity


100 Frequency (Hz)

So

un

d p

ressu

re le

ve

l (d

B)

300 500 700

60

900

50

40

30

20

Vibro-acoustic test with 205/65R15 tire & holography

Slick

Pattern

LF MF HF

Tread

Sidewall

Sidewall

Tread

Sidewall

Sidewall

Xoy-plane, z = 0 m

1.25 –

-1.25 –

-0.30 –

0.30 –

-1.25 1.25 -0.30 | | |

0.30

|

10

Wave propagation on tread/sidewall due to an impact


11

Acoustics Lab., ME

Fluid machines

Identification of noise generation mechanism

Difficulties

• No directly measurable reference source signal

• Propagation of high order modes, noise due to turbulence, & fast rotation

Source identification for aero-acoustic source in duct

Conventional methods bear poor spatial resolution

Source identification considering the evanescent modes and the

rotation of source?

Source

fp

farfield

60

62

64

66

68

70

Prediction of

the radiated noise

Forward propagation

Back propagation

Identification of

the spatial source distribution

Blurred Source image

12

Acoustics Lab., ME

Measurement setup

Velocity field at source plane (kR<1.84)

Compression driver

(Selenium D250-X)

Anechoic

termination

1 2

3 4

15 3045 60

n n

n n

z ,z ,z ,z

Signal analyzer and

generation system

(Pulse 3560D)

PC

z

1 800fz 2 800fz

Power Amplifier

(B&K 2716)

Aluminum

plate (10t) (All units are in mm)

-0.1

-0.05

0

0.05

0.1

-0.1 -0.05 0 0.05 0.1

68

70

72

74

76

-0.1 -0.05 0 0.05 0.1

-0.1

-0.05

0

0.05

0.1

-0.4

-0.2

0

0.2

0.4<Air blower> <Axial fan>

13


Comparison of holographic source ID & source design

Solve the problems of source design or field control by using the

equivalent concept of the acoustical holography.

Source identification Source design for a rendered field

source targetA G Hs fv G p

Known (Given)

Unknown

Known (Measured) Unknown

Noise source

sv

sv

fp

Conceptually same

principle

14


3D

boundary

z=2d z=d z=0

z

y

Control result at

the xy-plane over a

height of z=2d

3.43 m

3.43 m

Source array

3.43 m

Zone of silence

Loudspeaker array for creating

zones of silence and free wave

propagation, simultaneously

SPL r

e.

pm

ax

Re(p

)/Re(p

max) 100 Hz 300 Hz 500 Hz 700 Hz 1000 Hz

15

Zone of free wave

propagation

Uninterested zone


Op

tim

ize

d a

ctu

ato

rs

us

ing

EfI

me

tho

d

Simulation result with noise (SNR=-25 dB)

Init

ial

ac

tua

tor

se

t

16

200 Hz 300 Hz 400 Hz 500 Hz 600 Hz

- Irregular shaped room, Initial array = 36 sources, Reduced array = 18 sources

- Wall boundary condition: rigid, Room BE = 2621 nodes, 5238 triangular elements

Design: Spherical wave prop. emanating from left-lower corner


Location of Actuators: periphery of an elastic plate only

Utilization of central part for electronic components

Array actuators exciting elastic (glass) plate

Condition of target vibration field

Hot zone: a localized sector for vibration sensing

Cold zone: null field for vibration perception

Traveling wave control method

Frequency & magnitude: determined by considering the tactile

feeling

17

…

…

Elec.

comp.

Possible location

of actuators

Target

field

Hot zone

Cold zone

Hot zone Cold zone Cold zone

0

Haptic threshold

Vel.

1 1[ ] [ ] [ ]p N N pG E V †

1 1[ ] [ ] [ ]N N p pE G V

G : Transfer matrix (input signal & vel. response)

E : Matrix of input signals

V : Matrix of velocity responses

( ) : pseudo inverse

N : Number of actuators

p : Number of observation points

†

Vc,1 vh

vc,2 vc,p

… e2 e1 e3

eN


Actuators and observation points

Actuator: 20 mm spacing

• λb/8: Consideration of aliasing effect

Uniform arrangement of 34 actuators

7x11 observation points

Definition of performance index

Fulfillment of the rendered target pattern

Ratio between designed hot zone and actual controlled area (CP1) or

normally sensible area (CP2) (Also, tactile threshold & limen are considered)

Experimental results

18

CP1 = 55 %, CP2 = 35 % CP1 = 60 %, CP2 = 40 %

3x3 division 2x2 division


19

1 1

0 01 1 1

q q q

d n n n n n n

n n n

t L A ds A L ds A

†

1 1n n dq pp qA t

1 1d n np qp qt A

Sensor 1 Sensor 2

Sensor 3

Actuator

Image enhancement of reconstruction result

Stabilization of inverse solution – regularization technique

1 H

n dA VF U t

22 2

23 2 2

, ,q q

q q

F diag

Suppression of small singular value

Reference & reconstructed result Initial result Reference

1500

1550

1600

1650

1700

1750

1800

1850

1900

1950

2000

Enhanced result

1500

1550

1600

1650

1700

1750

1800

1850

1900

1950

2000

1500

1550

1600

1650

1700

1750

1800

1850

1900

1950

2000

Calculation condition=16-sensor, 64 interpolation points


Enhancement of reconstruction result II

Optimal selection of interpolation points by genetic algorithm

Reconstruction of uniform ambient temperature field

20

Initial result

0

10

20

30

40

50

60

70

80

90

100

0

10

20

30

40

50

60

70

80

90

100

Case I: Optimal interpolation points

0

10

20

30

40

50

60

70

80

90

100

Case I + regularization

: real sensor : virtual sensor

Deployment of interpolation points

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Equidistant Optimal

*: regularized reconstruction result

Reference & reconstructed result* Reference

1500

1550

1600

1650

1700

1750

1800

1850

1900

1950

2000

1500

1550

1600

1650

1700

1750

1800

1850

1900

1950

2000

Initial result

1500

1550

1600

1650

1700

1750

1800

1850

1900

1950

2000

Enhanced result

Calculation condition=16-sensor, 64 interpolation points


21

NoViC

22

0 500 1000 1500 2000 250040

60

80

100

120

Frequency(Hz)

Sound P

ressure

Level(dB

)

— ANC off — ANC on • Target

0 500 1000 1500 2000 250040

60

80

100

120

Frequency(Hz)

Sound P

ressure

Level(dB

)

— ANC off — ANC on • Target

Simulation Target freq. peaks : 70 – 2500 Hz

Simulation Experiment

ANC off ANC on Reduction ANC off ANC on Reduction

Overall level (dB) 118.0 89.5 28.5 118.0 107.7 10.3

Overall level (dBA) 113.2 81.9 31.3 113.2 105.3 7.9

Mean level at target (dB) 115.0 81.4 33.6 114.7 88.6 26.1

Experiment (Computing power limit of HW)

Target freq. peaks : Top 80 peaks only

MRI Noise

Intense noise accompanies with strong magnetic field of

MRI (Lp=130/110 dB in 3/1.5 T) for a finer image

Hearing loss potential and annoyance of patients

Proposed ANC system

Feedforward FXLMS algorithm is improper Use of

quasi-periodicity of MRI noise following SPE sequence

Actuator = Non-magnet speaker, Error mic. = ECM mic.

NoViC

23

Semi-active control of structural vibration

Truss space structure with low damping

Equivalent damping ratio represented by

normal force and modal displacement

Control the normal force by piezo actuator

Maximize damping for the 1st mode

Successful in reducing the transient vib.

Nonstationary power flow analysis

Classification of airplane engine

Analysis of vib. power with varying

engine RPM

Classification of engines by comparing

the vib. power levels

excitat ion

variable frict ion damper

xz

y

Sound visualization Sound manipulation

VISUALIZATION OF SOUND

Beam forming, Acoustic holography

Eyeglasses for the hearing impaired

Beampower of impulse

source

Sound ball System,

Spatial Equalization

Soundbar home theater system

Application to

car audio system

Dark zone (low sound

energy)

NoViC


25

All results shown in this presentation are contributed from

the former and current Acoustics Laboratory members, in

particular, by IY Jeon, A Oey, WH Cho, TK Kim, YH Heo,

and JH Woo, being conducted as their thesis and project

works.

jeong-guon ih, prof. - pennsylvania state university workshop_jg ih.pdfacoustics lab., novic inverse...

Documents