1
Structural Health Monitoring anStructural Health Monitoring an Großanlagen unter Anwendung von
Multisensorsystemen undMultisensorsystemen und fraktalbasierten Auswerteverfahren
Jürgen Schreiber
h f i fFraunhofer Institut fürZerstörungsfreie Prüfverfahren
Institutsteil Dresden
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
2Introduction
Challenge:
Save and economic service of
Characterisation of the material damage state without baseline
Save and economic service of safety relevant industrial components
Maximal usage of the whole Reliable evaluation of the fatigue gservice life time, definition of actions
gdamage , forecasting of the residual life time
Search for suitable NDT-methodseither for
regular inspection
Structural Health Monitoring
or
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
3Introduction
Fatigue Monitoring in nuclear power plants
Up to 150
AREVAErlangenp
thermocouples in 20-50 measurement sectionsplus signals of existing p g goperational instrumentationfor residual life timeestimationestimation
Materials damage Safty
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
4Introduction
Components of brown coal open cast mining
Vattenfall EuropeC ttb
Cyclic load
CottbusSystem of strain gauges are displaced to control load and
h
Aging
to assess the service strength
Materials damage Residual service life time ?Materials damage Residual service life time ?
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
5Introduction
Materials damage Actions?
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
6Introduction
Integrity of Railway Structures Wiener Linien
strain
gauge
Tests with mobile eddy current sensor system for crack inspection
Rolling Contact Fatigue Repairing in time before cracking
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
7Introduction
Available NDT-methods
• Micro-thermography: Crack indicator, but too week signals from deformation structure
• US-backscattering: Multiple scattering and long propagation path strongly mask the effect of deformation meso-structures, while cracks are seen well.
• Acoustic emission: Low effect of plastic deformation, signals only due to force changes
P t ti l W k iti it f d f ti l k• Potential sensor: Weak sensitivity for deformation, only crack indication
• Optical methods: Dirty surface prevent simple application • Eddy current: Good for crack indication but deformation?• Eddy current: Good for crack indication, but deformation?• Barkhausen noise: Testing locally, simple and continuous noise
excitation, magnetic parameters sensitive to stress and microstructureand microstructure.
What can we learn from standard Barkhausen noise measurements?
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
8State of the art in NDT
l b ( )
Barkhausen noise parameter versus fatigue
Material: 15NiCuMoNb5 (WB 36)
250
300
a] m/m
]
100
125
250
ess
σ xx
[MP
stic
str
ain
[µ
50
75
resi
dual
str
ey
[a.u
.], p
las
stress controlled (R=0; f=3 Hz, stop at necking)
0
25
load
, B
Hx,
BHstop at necking)
a) N=0-1000, σa=240MPa
b) N=1000-2000, σa=260MPa0 1000 2000 3000 4000 5000 6000 7000
number of cyclic load Nc) N=2000-6000, σa=260MPa
c) N=6000-6229, σa=300MPa
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
9State of the art in NDT
0.20.3
150
Barkhausennoise-Signal
Barkhausen noise parameter versus fatiguem
eter
-0 4-0.3-0.2-0.10.00.1
Integrated sumRMS 50
100
0.20.3
BH
N-P
aram
2500 3000 3500 4000 4500 5000 5500 6000 6500
-0.5-0.4 RMS
0
50
Magnetic feld
0 4-0.3-0.2-0.10.00.1
Width of the peakA t f th k
0 500 1000 1500 200050
Time [µs]
2500 3000 3500 4000 4500 5000 5500 6000 6500
-0.5-0.4 Asymmetry of the peak
Number of cyclic load N
This method is not able to estimate the actual stage of fatigue!
Resume:
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
10New developments in NDT for SHM
Thermocouples
Mounting StrapMounting StrapCover shell
Multi-sensor systems to detect ti h t i ti f
Sm - magnetizing coilsDS - detecting sensors for
Barkhausen noise
time characteristics of materials response to
thermo-mechanical load
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
11New developments in NDT for SHM
Concept of 16-element eddy current line array developed
at IZFP-Dresden.
Schematic of 64-element two-dimensional eddy current array planned
as foil sensor systemy
Array sensor system to detect St t l i dd tStructural varying eddy current
(“Grain noise”)
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
12New developments in NDT for SHM
Scheme of fibre Bragg grating sensor (see IPHT, Bartelt)
Meander type fiber Bragg grating sensor system for
space time characteristics of loadspace-time characteristics of load as well as
materials damage
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
13
Mesoscopic deformation structures
It is possible now to measure very much, however, what is the key information for the assessment of materials damage?
Micro M MMicro Meso Macro
Panin-Idea:
CT-Image of fatigue crack
Mesomechanics
TEM-Image
Eine wirklich gute Idee erkennt man daran, dass ihre Verwirklichung von vorneherein
ausgeschlossen erscheint.g
Albert Einstein (1879 - 1955)
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
14
Mesoscopic deformation structures
Austenitic sample Ferritic sample
a) b)
c)
a) N = 2000, b) N = 7000, c) N = 13000
)
What does characterise these structures?
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
15
Mesoscopic deformation structures
Theoretical und experimental basis
Mesomechanics (Panin since 82) Depinning Models ( self-organising criticality“ 90ies)Plastic deformation: Gliding locally due to stress concentrators accompanied by
i f i hb i i
(avalanches, sand pile, domain motion)Distribution of jump heights, duration and energyas well as the power spectrum for Barkhausen noise in
hrotation of neighbouring grains formation of closed structures
(new structure elements) hierarchical structure
amorphous magnets:
D(s) ~ sα , P(k) ~ k β scaling behaviour
Symmetry arguments point at 3 universal classes:hierarchical structure•Dislocation pattern (Micro)• Slip band formation (Meso I)• Folded multi-bands (Meso II) S i
Symmetry arguments point at 3 universal classes:
• α 1 ≈ 1.70 ⇒ 3D domains , dipole-dipole-interaction• α 2 ≈ 1.44 ⇒ 2D domains in local fields• α 3 ≈ 1 30 ⇒ 2D multi-domains in local fields
Suggestion:Scaling behaviour of micro-meso-structure may characterise fatigue, fractal dimension DF as a
α 3 1.30 ⇒ 2D multi domains in local fields
However, continuous transformation between the structures are obtained experimentally. There are similarities between deformation and fatigue, fractal dimension DF as a
measure of damage? magnetic structures!
Scaling behaviour
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
16
Mesoscopic deformation structures
Scaling behaviour?
150
.]
50
100
usen
noi
se si
gnal
[a.u
.
0 500 1000 1500 200050
0
time [µs]
Bar
khau
Topography Z(Ri) or time signal S(t) Z(λRi) ~ λH Z(Ri) i i
S(λti) ~ λH S(ti)Fractal dimension: DF
(r) = 3- H(R) or D(t)F = 2 – H(t) ,
Area length: F(s) s H-1Area, length: F(s) ~ s H 1, Correlation function: C(τ) = <|S(t+ τ)-S(t)|2>t ~ τ 2H
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
17
Mesoscopic deformation structures
Load direction
FOnline SEM-Images
X6 CrNiTi 18 10Panin, Kuznetsov, Schreiber (1998)
σa ~ 240, 250, 260 MPa
ΔD = 0 04ΔD = 0 07 B
ΔDF32 = 0.04ΔDF21 = 0.07
What is a suitable NDT-method?
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
18
Mesoscopic deformation structuresTopography and micro-magnetic structures
N =622AFM-Topography N = 622, DF = 2.53Fatigue experiment
0 7 % d /d 3 %/ i
15 NiCuMoNb 5
N=0εa = 0.7 %, dε /dt = 3 %/min
M t tMesostructures
MFM-Image DF = 2.50
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
19Fractal analysis of Barkhausen noise
150
]
Magnetic Barkhausen noise
50
100
ise
sign
al [a
.u.
680 685 690 695 700
0
50
Bar
khau
sen
nof ld
0 500 1000 1500 200050
time [µs]
BMagnetfeld
Integral parameters1. Integrated Amplitude 2. RMS-value3 ( )
Time series1. Distribution of chumps2. Powerspectrum
3. Asymmetry (Moments) 3. Correlation function
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
20Fractal analysis of Barkhausen noise
30
40
50Simulation of real fractals:
Finite time series and random noise
Snowflake + noise
10
20
Cq r( )N r−
Si r+ Si−( )q
N r−∑:=S
0 100 200 300 40010
01i =
15 1 5
Time [µs]Analysis for the integrated time series
101
1.5
Cq(r)q = 3
5
0.5
Snowflake - exactDF = ln4/ln3 = 1.262Fit ~ r a
DF = 2-a/q
0.1 1 10 100 1 .103 0 10 20 30 40 500
qTime difference r [µs]
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
21Fractal analysis of Barkhausen noise
Online fatigue experiments(variable strain amplitude and strain rate) 15 NiCuMoNb 5
440
472
503
534
566
Last
max [M
Pa]
and strain rate) 15 NiCuMoNb 5
1,40
1,45
1,50
Probe 342 ε
α = 1.2%, dε/dt = 12%/min
Probe 354n
377
409
440
log(N/NA)
1 20 15 NiCuMoNb 5 (WB36), dε /dt = 12 %/min, R = -1)
1,15
1,20
1,25
1,30
1,35 Probe 354ε
α = 1.0%, dε/dt = 12%/min
Probe 365N
B = 348, ε
α = 1.2%, dε/dt = 6%/min
rakt
ale
Dim
ensi
on
1.15
1.20 15 NiCuMoNb 5 (WB36), dεa/dt 12 %/min, R 1) εa = 0.4 % bis N = 2000 (NB ~ 45000) εa = 1 % bis N = 2315 (NB ~ 2500)
X6 CrNiTi 18 10 (online, REM) f 2H 250MP R 0)m
ensi
on D
F
-3,5 -3,0 -2,5 -2,0 -1,5 -1,0 -0,5 0,0 0,51,00
1,05
1,10Fr
log(N/NB)
1.05
1.10f=2Hz, σa ~ 250MPa, R
σ=0)
Frak
tale
Dim
gB
ΔDF(1 2) = 0.082 (Topografie: 0.07) ΔDF(2 3) = 0.045 (Topografie: 0.04) -5.0 -4.5 -4.0 -3.5 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5
1.00
log(N/NB)
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
22Fractal analysis of Barkhausen noise
1.20Effect of cracks on DF
Riss
1.10
1.15
men
sion
120
1.00
1.05
S üb Ri
Frak
tale
Di
100
80
1.0 1.5 2.0 2.5 3.0 3.5
0.95Scan über Riss
Scan 90° versetzt
Position [mm]
80
60
40
Computer-Tomography
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
23Applications
22 NiMoCr 37Y
RMS R id l
CT-samplelike a component
RMS ~ Residual stress Fractal dimension
crack121.4
1.5Fractal d
X
0
1
6
8
10
4
RMS
X
0
1
1.1
1.2
1.3
1.4
4
dimension D
F
X
2
3
4
-4
-2
0
2
4
Y-Ach
se [c
m]
X-Achse [cm]
1
2
3
4-2
0
2
4
Y-Ach
se [c
m]
X-Achse [cm]
45
55
-4
2
3
4-4
-3-2
-10
12
34
Y-Ach
seX-Achse Asymmetry parameter
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
24
Applications
Brown coal open cast mining
Support of the 600 m long bridge F60 pp g g
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
25
RissR120 RissR80 RissR120 RissR80
Along the crack Perpendicular direction
Applications
2
1
sitio
n
1.139
1.166
1.194
1.221
1248
Riss R120 Riss R80DF
2
11.139
1.166
1.194
1.221
1.248
Riss R120 Riss R80DF
DF
4 6 8 10 12 14 16 18 205
4
3
y-P
os
1.248
1.275
1.303
1.330
1.357
4 6 8 10 12 14 16 18 205
4
31.275
1.303
1.330
1.357
Riss R120 RissR80 Riss R120 Riss R80
F
3
2
1
ositi
on
2.080
3.105
4.130
5.155
6.180
Riss R80RMS [r.E.]
3
2
1
RMS [r.E.]
2.080
3.105
4.130
5.195
6.108
RMS
4 6 8 10 12 14 16 18 205
4
3
y-P
o
P i i
7.205
8.230
9.255
10.28
4 6 8 10 12 14 16 18 205
4
3
y
7.205
8.230
9.255
10.28
Riss R120 Riss R80RMS[rE]
Riss R120 Riss R80
Beginning cracks?
3
2
1
ositi
on
191.0
234.5
278.0
321.5
365.0
RMS [r.E.]
3
2
1
Asym[r.E.]
191.0
234.5
278.0
321.5
365.0AsymFatigue
versus
4 6 8 10 12 14 16 18 205
4
3
y-P
o
x-Position
408.5
452.0
495.5
539.0
4 6 8 10 12 14 16 18 205
4
3
x-Position
408.5
452.0
495.5
539.0
versus power crack?
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
26Applications
First results for “Wiener Linien” rails
Y
X
Y ~ roling direction
Io : 70 120
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
27Applications
1.5
1.6
1.7in service
lina Aline B
PerpendicPerpendicular directionRMS ~ Residual stress
0.7
0.8
0.9
1.0
1.1
1.2
1.3
1.4 line B
new line A line B
MS
~ R
esid
ual s
tres
s
A1 A2 A3 A4 A5 A60.2
0.3
0.4
0.5
0.6
Measuring position
RM
Roling d
2.2
2.4 in servicelina A
Roling direction
1 0
1.2
1.4
1.6
1.8
2.0
lina A line B
new line A line B
MS
~ R
esid
ual s
tres
s
A1 A2 A3 A4 A5 A60.2
0.4
0.6
0.8
1.0
RM
Measuring position
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
28Applications
1.40
Perpendicular direction1.40
DF ~ Fatigue damage
1 20
1.25
1.30
1.35
tale
Dim
ensi
on D
F
line A line B
1.20
1.25
1.30
1.35 line A line B
tale
Dim
ensi
on D
F
ewervi
ce
1 2 3 4 5 61.10
1.15
1.20
Frac
tMeasuring Position
1 2 3 4 5 A61.10
1.15
Frac
t
Measuring Position
ne
in s
e
Roling direction
1 35
1.40
line A
1.40
line A
1.20
1.25
1.30
1.35 line A line B
acta
le D
imen
sion
DF
1.20
1.25
1.30
1.35 line A line B
ract
ale
Dim
ensi
on D
F
1 2 3 4 5 6
1.10
1.15Fra
Measuring Position1 2 3 4 5 6
1.10
1.15
Fr
Measuring Position
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
29Further applications
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
30Further applications
Fractal dimension of Barkhausen noise
1 35
1.40 Neu Gebraucht 9.0
Sample A1.20
1.25
1.30
1.35al
dim
ensi
on
7.0
7.5
8.0
8.5
RMS RMS
MS
1.05
1.10
1.15Frac
ta
5.5
6.0
6.5
RM
1.0 1.5 2.0 2.5 3.0 3.5 4.01.00
Messpunkt
1 2 3 4 55.0
Messpunkt
1.30
1.35
1.40
n DfNeu 5.4
5.6
5.8
6.0
DfNeu DfGebrau
1.15
1.20
1.25
30
Frac
tal d
imen
sio DfGebraucht
4.2
4.4
4.6
4.8
5.0
5.2
RM
S1.0 1.5 2.0 2.5 3.0 3.5 4.0
1.00
1.05
1.10
1.0 1.5 2.0 2.5 3.0 3.5 4.0
3.4
3.6
3.8
4.0
Sample B
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
31Outlook
τ = -log(N/NB) Estimation of residual life
1.20
1.25
DF(N
/NB)
Dm
DF(N/NB)-function independent on NB and R universal curve! 1.10
1.15
ract
al d
imen
sion
D DmF
NB and R universal curve!
-3.5 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.51.05
1.10
Life time parameter −τ
τm
Fr
DMS Load collectives +Wöhler-curve NB, α(Δti) = ΔN(Δti)/Δti
mM(m)B
m mi 1 i
NT 1/M(m) (1 exp( ))( t )=
= − −τα Δ∑
Residual service life time
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
32Outlook
Ti l d S kl Ph
Nonmagnetic materials
Time resolved Speckle-Photometry
He-Ne laser
CCD camera
He-Ne laser
CCD camera
CO2 laser heatingImaging Optics
S kl
CO2 laser heatingImaging Optics
S klSpecklesSpeckles
1.6
1.7
Al-alloyF
Thermal waveThermal wave
1.1
1.2
1.3
1.4
1.5
al d
imen
sion
DF
0.01 0.1 10.8
0.9
1.0
1.1
Frac
ta
log(N/NB)
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
33Outlook
Fractal dimension DF as a function of running time
y 3 000 MPa, T = 60 °C
Direction x-axesalong roling
trace
y-axesx
100 h 1.059 1.18
200 h 1.129 1.083
300 h 1.37 1.265
400 h 1.126 1.349
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
34Outlook
<j(t)> = <j(to> – (<j(to>- <j(t ∞)) (1 – R(t-to))R(t t ) Δj(t) Δj(t ) / Δj2 d Δj(t) j(t) j(t )
Eddy current relaxation
R(t-to) = <Δj(t) Δj(to)>/< Δj2> and Δj(t) = j(t) - j(t ∞)
Pos 2
Pos 3
Pos 4Pos 2
Pos 3
Pos 4
Pos 1Pos 4.1
Pos 4.2Pos 1Pos 4.1
Pos 4.2
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
35Conclusion
SHM – characterization of materials damage before cracking and without baseline!Fractal analysis of deformation structure new working concept based on DF!
For steel DF increases with fatigue damage, however, the appearance of cracks and the stress release can locally reduce DF.
as a function of cyclic load number was found.Symmetry arguments point at the existence of three universality classes.
The fractal analysis of the offers the possibility to introduce the fractal concept into the inspection practise. Besides of DF the RMS-value of Barkhausen noise has to be measured in order for crack indication
Search is ongoing for similar approach to nonmagnetic materials (eddy current noise, Speckle-Photometry), it looks promising!
IZFP is now developing a special device “FrakDim” with optimized working regime including fast soft ware. Specialized sensors will be manufactured.
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008
36
Than you very much !
Sensorsysteme 2008 Lichtenwalde, 16-18. Oktober 2008