Non-destructive Evaluation Methods for Detecting Major ...€¦ · Internal Composite Structures...
Transcript of Non-destructive Evaluation Methods for Detecting Major ...€¦ · Internal Composite Structures...
Non-destructive Evaluation Methods
for Detecting Major Damage in
Internal Composite Structures
New Project Title: Impact Damage
Tolerance Guidelines for Stiffened
Composite Panels
Francesco Lanza di Scalea
Professor, Dept. of Structural Engineering
University of California San Diego
Charlotte Convention Center, NC
2019 JAMS Technical Review
May 22-23, 2019
2
• Principal Investigators & Researchers
– PI: Prof. Hyonny Kim
– Co-PI: Prof. Francesco Lanza di Scalea
– Graduate Students
PhD: Eric Hyungsuk Kim, Margherita Capriotti, Ranting Cui
MS: none
• FAA Technical Monitors
– Lynn Pham, Ahmet Oztekin
• Other FAA Personnel Involved
– Larry Ilcewicz
• Industry Participation
– Boeing, Bombardier, UAL, Delta, DuPont, JC Halpin
Participants
3
Motivation• High energy Blunt Impact Damage (BID) of main
interest:
• involves large contact area, multiple structural
elements
• internal damage (cracked shear tie, frame, stringer
heel crack) can exist with little/no exterior
visibility
• External-only NDE needed
Heel
GSE Impact/Contact
Overall Objectives:
• Quantify detectable and
non-detectable damage
characteristics
• Relate Ultrasonic Guided
Wave NDE measurements
to damage state and
residual strength
Ultrasonic Guided Waves: structure is a natural
“waveguide”
4
Ultrasonic excitation
@ 1.666E-04 sec
Guided-Wave Transfer Function: Single-Input-Dual-
Output Scheme (SIDO)
1 1 1( ) ( ) ( ) ( ) ( ) ( ) ( )SAM f S f T f TS f H f SR f R f Receiver 1:
Receiver 2: 2 2 2( ) ( ) ( ) ( ) ( ) ( ) ( ) ( )SA ABM f S f T f TS f H f H f SR f R f
2 2 2
1 1 1
( ) ( ) ( )( )
( ) ( ) ( )AB
M f SR f R fDeconv H f
M f SR f R f ( ) ( ) eAB ABH t H f df
i 2 f t
(time domain)
SIDO Transfer Function Scanning Systems
“AIR-COUPLED” SCANNERHYBRID “IMPACT/AIR-COUPLED”
SCANNER
skin-stringer assembly
Piezocomposite transducers
“high” and “narrow” frequency
band (110 – 210 kHz) Mini-impactor + micro-machined
capacitive transducers
“low” and “broad” frequency
band (40 – 270 kHz)
Test Panels [45/-45/0/45/90/-45/0/90]s CFRP stiffened panels
with hat-shaped, co-cured stringers
Panel “A”Panel “B”
A (stringer flange)
B (stringer cap)
Impact Locations
Results: stringer heel slit and stringer cap slit
Air-coupled scanner
(110 – 210 kHz)
Hybrid impact/air-coupled
scanner
(40 – 270 kHz)
Results: stringer flange impact
Air-coupled scanner
(110 – 210 kHz)
Hybrid impact/air-coupled
scanner
(40 – 270 kHz)
Hybrid impact/air-coupled
scanner
(low vs. high frequencies)
(80J)
Results: stringer cap impacts
Air-coupled scanner
(110 – 210 kHz)
Hybrid impact/air-coupled
scanner
(40 – 270 kHz)
11
Residual Compression Strength Tests of Impacted Panels
Cap impactsFlange impacts
12
Residual Strength Estimation from UGW Scattering
Finite Element Analysis
(isotropic plate. Holes from 0.05 mm
to 50 mm dia, 150 kHz freq)
30 cm
10 cm
HoleTransmitter Receiver
UGW Experiments
(Hexcel [0]10 plain weave 282/SC780.
Holes from 2.5 mm to 25 mm dia,
various frequencies)
13
UGW Transmission Strength
Hole Radius (mm)
Norm
. G
uid
ed W
ave A
mplit
ude
Frequency
(experimental)
FEA
30 cm
10 cm
HoleTransmitter Receiver
Elastic Constants Identification from UGW Testing
• Use Guided-Wave Phase Velocity Dispersion
Inversion and Simulated Annealing
Optimization.
• Use SAFE method to solve forward problem.
• Utilize three fundamental guided-wave modes
(S0, A0 and SH0) propagating along a single
direction (x).
Transversely isotropic lamina: five unknowns
E11, E22, n12, G12 and n23
Engineering laminate properties from CLT:
seven unknowns
Ex, Ey, nxy, Gxy, Kx, Ky, Kxy
X
Y
Z
2
3X
Dispersion curves
“in-plane” “out-of-plane”
Impact Damage causes change in UGW transmission.
Why? Presence of damage directly relates to change in
Elastic Constants inverse problem.
Initial Parameters
Initial increments
SAFE Analysis for dispersion curves Objective function for ,
Set as accepted transition
Set as accepted incremental transition
Set as
Define trial test with last accepted transition
SAFE Analysis
Updating Increments
(Simulated Annealing)
Yes
Metropolis Criterion
r
No
1000 iterations
YesNo
Identified Lamina Constants
3 times Simulated Annealing
Identified Laminate Engineering Constants
ph ph ( )c c 2
ph,pred ph,true
1 ph,true
( ) ( )1
( )
N i i
i i
c cd
N c
Constants Identification
Flowchart
Cui, R. and Lanza di Scalea, F., “On the identification of the
elastic properties of composites by ultrasonic guided waves and
optimization algorithms,” Composite Structures, in press, 2019.
Objective function to minimize:
mismatch of phase velocity dispersion curves
2
ph,pred ph,true
1 ph,true
( ) ( )1
( )
N i i
i i
c cd
N c
for S0, A0 and/or SH0
Constants Identification Results: anisotropic
laminate
1
2
3
X
Y
Z
X
T
hic
kn
ess(
mm
)T
hic
kn
ess(
mm
)T
hic
kn
ess(
mm
)
Normalized Strain
Normalized Strain
X
Z
X
Z
X
Z
sxx
sxx
txy
1
2
3
X
Y
Z
X
Thi
ckne
ssT
hick
ness
Thi
ckne
ss
Normalized Displacement
Normalized Displacement Normalized Strain
Normalized Strain
Normalized Stress
Normalized Stress
txy
sxx
txy
txz
szz
“normal - shear” coupling and
“longitudinal - transverse” coupling !!SAFE analysis
S0
SH0
A0
Constants Identification Results: anisotropic
laminate – lamina properties (1D inversion)
(a) (b) (c)
(d) (e)
Anisotropic Laminate
Lamina constants identification(1D Inversion)
X
Y
Z2
3 XS0
S0
A0SH0
A0 A0
SH0
SH0
d
Constants Identification Results: anisotropic
laminate engineering properties (in-plane)
Anisotropic Laminate
Engineeringconstants
identification(in-plane)
X
Y
Z
23 X
Constants Identification Results: anisotropic
laminate engineering properties (out-of-plane)
d
Anisotropic Laminate
Engineeringconstants
identification(out-of-plane)
X
Y
Z
23 X
Summary• Methods of UGW testing investigated for inspection of impact
damage in composite stiffened panels
• Two scanning systems based on UGW dual-output scheme:
• non-contact “air-coupled” system (damage in skin and
stringer flange)
• hybrid “impact/air-coupled” system (damage in stringer cap)
• UGW studies in plates with holes show relation between wave
scattering and residual compression strength
• Inverse procedure based on matching phase velocity UGW
dispersion curves for identifying elastic properties of composite
panels (lamina constants and laminate engineering constants)
Ongoing/Future Work
21
• Package mini-impactor into scanning system for automatic scan
• Expand elastic constants identification to impact damage for residual
strength estimation
• Conduct additional analyses of wave scattering through various
damage types/severity for residual strength estimation
GLOBAL
(SAFE)
GLOBAL
(SAFE)
LOCAL
(FEM)
Incoming
(m1, m2)
Reflected
(m1, m2,
m3, m4)
Transmitted
(m1, m2,
m3, m4)
GLOBAL-LOCAL
MODELS
EXTRA SLIDES
22
Guided-Wave Transfer Function: Semi-
Analytical-Finite-Element (SAFE) method
𝑢(𝑒) 𝑥, 𝑦, 𝑧, 𝑡 =
𝑁𝑗 (𝑦, 𝑧)𝑈𝑥𝑗
𝑛
𝑗 =1
𝑁𝑗 (𝑦, 𝑧)𝑈𝑦𝑗
𝑛
𝑗 =1
𝑁𝑗 (𝑦, 𝑧)𝑈𝑧𝑗
𝑛
𝑗 =1
(𝑒)
𝑒𝑖(𝑘𝑥 𝜔𝑡 )
11 2C R C R
2
0A B QM
k
Displacement field
Lamina stiffness matrix
Eigenvalue problem
Eigenvalues (, k): dispersion curves
Eigenvectors U: cross-sectional mode shapes
Transfer function (band-limited) – freq. domain
𝐩 U
A B1
exp[ ( )] withp
U
LM mRup
m m S mm m
ik x xB
Transfer Function Comparison: Experimental
vs. Numerical
CFRP laminate, 16 plies, [45/-45/0/45/90/-45/0/90]s.
(representative of B787 fuselage “skin”)
SIDO scheme, Mistras PICO PZT transducers,
170 kHz excitation Exp. transfer function (freq. domain)
Transfer Function Comparison
(time domain)
25
Non-Contact NDE Scanning Prototype
• Line scan approach with non-contact sensors on moving carriage
• Air-coupled piezocomposite transducers (170 kHz)
Excitation Detection
26
Mini-Impactor (probes interior + portable)
• Frequency range up to 500 kHz and peak
intensity at 42 kHz
Aluminum Tip – 0.56 mm thick
Uni-directional Carbon/Epoxy
[0]8 Layup; 0.56 mm thick
6.35 mm
////
////
Courtesy of Eric Kim and Dr. Hyonny Kim
Thermography for Independent Damage Survey
Thermography (TSR): ground truth of damage for quantitative damage survey
Dent 1:
Energy Level = 30 J
Dent 2:
Energy Level = 50 J
Dent 1: 56.84mm Dent 2: 69.7mm
FLANGE
CAP CAP
SKIN SKIN
FLANGE
S0 (1)
skin+stringer
A0 (3) skin
A0 (4) stringer
S0 (1)
skin+stringer
A0 (3) skin
A0 (4) stringer
Statistical Analysis
Outlier Analysis:
Feature
Super-Vector
𝑥
𝑀𝑎𝑥 𝐴𝑚𝑝𝑙𝑀𝑎𝑥 𝐼𝑛𝑑𝑉𝑎𝑟𝑖𝑎𝑛𝑐𝑒 𝑢𝑟𝑡𝑜𝑠𝑖𝑠
… 𝑚
𝑀𝑎𝑥 𝐴𝑚𝑝𝑙𝑀𝑎𝑥 𝐼𝑛𝑑𝑉𝑎𝑟𝑖𝑎𝑛𝑐𝑒 𝑢𝑟𝑡𝑜𝑠𝑖𝑠
… 𝑚
𝑀𝑎𝑥 𝐴𝑚𝑝𝑙𝑀𝑎𝑥 𝐼𝑛𝑑𝑉𝑎𝑟𝑖𝑎𝑛𝑐𝑒 𝑢𝑟𝑡𝑜𝑠𝑖𝑠
… 𝑚𝑁
Test Vector
𝑥
Test Signal
(six possible time gates)
𝒕𝒉 𝒆𝒔𝒉 𝒍𝒅
Baseline Signal
(six possible time gates)
𝑰𝒇 𝑫𝑰 > 𝒕𝒉 𝒆𝒔𝒉 𝒍𝒅 ⟹ DEFECT
• Multivariate
• Multi-modeSuper-Vector for mode compounding
Known Undamaged
Region:
Any Location
d
Constants Identification Results:
anisotropic laminate – lamina constants
(5D inversion)
Anisotropic Laminate
Lamina constants identification(5D Inversion)
X
Y
Z
23 X
Property Identification Results: quasi-
isotropic laminateSAFE analysis
X
Y
Z 1
2
3X
Normalized Strain
Normalized Strain
Th
ick
nes
s(m
m)
Th
ick
nes
s(m
m)
Th
ick
nes
s(m
m)
X
Z
X
Z
X
Z
d
𝟓
𝟔
Property Identification Results: quasi-
isotropic laminate
Quasi-Isotropic Laminate
Lamina constants identification(5D Inversion)
X
Y
Z
23 X
d
Property Identification Results: quasi-
isotropic laminate
Quasi-Isotropic Laminate
Engineeringconstants
identification(4D Inversion)
X
Y
Z
23 X
33
Non-Contact NDE Scanning Prototype
Statistical Analysis Results:
Cracked
skinDisbonded
stringer
Detached/
cracked
stringer
Noise floor
(Skin modes only)
ROC curves for performance assessment
Every point is
different
threshold level –
typically, lower
threshold yields
higher detection
but more false
alarm
Cracked Skin
Disbonded Stringer
Excellent detection :
90% POD with 0% PFA
Excellent detection :
93% POD with 0% PFA
Ok detection :
90% POD with 20% PFA
Detached/Cracked Stringer
34
Non-Contact NDE Scanning Prototype
Outlier Analysis Results:
Cracked
skinDisbonded
stringer
Detached/
cracked
stringer
Noise floor
(Skin + Stringer modes)
ROC curves for performance assessment
Cracked Skin
Disbonded Stringer
Perfect detection
Perfect detection
Detached/Cracked Stringer
Perfect detection
35
Mini Impactor on Built-up Panel• Excitation and measurement (R15 contact transducer) on exterior skin-side
• S0 waves through skin path move faster (~150 kHz);
• A0 waves through C-frame path move slower (~50 kHz);
• Specimen with C-frame removed has only skin modes content
Skin Path
Skin+Stringer
Path
Shear Tie +
C-Frame Path
Panel Exterior View
Panel Underside View
With C-Frame
W/out C-Frame
Skin Modes
Frame Modes
36
• Internal shear tie damage detection using mini-impactor excitation
Mini Impactor on Built-up Panel
Undamaged shear-tie
Damaged shear-tie
37
Residual Strength Estimation: Validation
• Three new stringer panels fabricated
– T800/3900-2 uni-directional tape plies. Skin thickness = 3.175mm
– Panel dimensions: 1m x 1.3m
– Five stringers with 0.26m spacing
– Various impact energy levels
A (stringer flange)
B (stringer cap)
Impact Locations
38
Residual Strength Estimation Plans: Flat Stringer Panel
Stringer Cap
Impact
Skin&Stringer Flange
Impact
Flat Stringer Panel Impact Plan
– Stringer cap impacted portion
will be trimmed into 0.3m
specimens for compression w/o
buckling
– Stringer flange impacted
portion will be trimmed into
0.48m specimens for
compression w/ buckling
39
Residual Strength Estimation: Wave Scattering
Caprino, Giancarlo. "On the prediction of residual strength for notched laminates." Journal of Materials Science 18.8 (1983): 2269-2273.
Empirically determine the exponential value e, and relate values to estimate residual
strength
Wave_Amplitude= (Dam_Size)-e σcrack /σpristine = (L0/Dam_Size)m [Caprino]
Looking forward
Correlate the features with damage location and type: preliminary results of defect characterization (extension,
severity, type) by UGW.
40
Skin Damage Disbond Damage Undamaged
Looking forward•
Further investigation on internal structural wave penetration. (Global Local modeling)
41
GLOBAL
(SAFE)
GLOBAL
(SAFE)
LOCAL
(FEM)
Incoming
(m1, m2)
Reflected
(m1, m2,
m3, m4)
Transmitted
(m1, m2,
m3, m4)
Looking forward
Further investigation on internal structural wave penetration. (Global Local modeling)
42
SKIN + STRINGER: Heel
crackIncoming
(m1, m2)
Reflected
(m1, m2,
m3, m4)
Transmitted
(m1, m2,
m3, m4)