Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
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
Acoustics Lab., NoViC Acoustics Lab., NoViC
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
Acoustics Lab., NoViC Acoustics Lab., NoViC
4
Acoustics Lab., NoViC
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)
Acoustics Lab., NoViC
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
Acoustics Lab., NoViC Acoustics Lab., NoViC
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
Acoustics Lab., NoViC Acoustics Lab., NoViC
8
150 Hz (E3) 200 Hz (E4)
Animated
surface
velocity field
3000 RPM Surface intensity @
E/G Top
E/G Exht
Acoustics Lab., NoViC Acoustics Lab., NoViC
12th
RPM
Fre
qu
en
cy (
Hz)
6th
3rd
2nd
RPM (625-800)
9
Measured hologram Restored velocity
Acoustics Lab., NoViC Acoustics Lab., NoViC
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
Acoustics Lab., NoViC Acoustics Lab., NoViC
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
Acoustics Lab., NoViC Acoustics Lab., NoViC
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
Acoustics Lab., NoViC Acoustics Lab., NoViC
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
Acoustics Lab., NoViC Acoustics Lab., NoViC
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
Acoustics Lab., NoViC Acoustics Lab., NoViC
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
Acoustics Lab., NoViC Acoustics Lab., NoViC
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
Acoustics Lab., NoViC Acoustics Lab., NoViC
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
Acoustics Lab., NoViC Acoustics Lab., NoViC
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
Acoustics Lab., NoViC Acoustics Lab., NoViC
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
Acoustics Lab., NoViC Acoustics Lab., 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.