structural health monitoring: basic approaches and ...€¦ · prof. dr.-ing. c.-p. fritzen...
TRANSCRIPT
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Structural Health Monitoring: Basic Approaches and Applications
Claus-Peter Fritzen
Uppsala, Sweden, 10.11.2010
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Introduction
Basic principles of Structural Health Monitoring
Model-based damage localization
Damage detection with Null Space-Based Fault Detection
Method (NSFD)
Load identification
Ultrasonic Waves: Spectral Element Method (SEM)
Damage localization in high frequency domain
Conclusions
Outline
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Members of the working group
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Researchers:MSc. Maksim KlinkovDipl.-Ing. Philipp KösterDipl.-Ing. Peter KraemerDipl.-Ing. Martin KübbelerDipl.-Ing. Jochen MollDipl.-Ing. Rolf T. SchulteMSc. Kejia XingMSc. Cheng YangDipl.-Ing. Erion Zenuni MSc. Yan NiuMSc. Miguel Torres
Graduate Students:Inka BütheIbraim DzhaferovTorsten FischerPhillipp HilgendorffBernd SchenkelDaniel GinsbergDavid GossenRannam Chaaban
Technical Staff:Gerhard DietrichDipl.-Ing. Wolfgang Richter
Secretary:Gisela Thomas
Director:Prof. Dr.-Ing. C.-P. Fritzen
External PhD Students:MSc. Dongsheng Li, Univ. of Dalian, ChinaDipl.-Ing. Benjamin Eckstein, EADS/AirbusMSc. Carsten Ebert, Fa. Wölfel
Members of the working group
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
SHM system components
Diagnosis
Sensor signals
Environ./Operat. cond.
Excitation(active/environ.)
Prognosis
Load identificationExtreme events
Material modelsDamage models
Decision,Action
PredictionRUL
Damage type and extent
Damage detection,
localization
Sensor self-diagnosis
Different measurement equipments and systems
System coordination
Choice of sensors• Types• Number• Positions
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Mechanical:Static measurement (strain, displacements,...)Vibration measurements using either modal data,
frequency or time responsesStress wave measurements:
Ultrasonic guided waves (Lamb waves)Acoustic emission
Electric, Electro-mechanical, Electro-Magnetic:Impedance based methods with PZTEddy current methodsElectric resistance-based methodsElectro-magnetic
Basic Principles of SHM methods
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Global and Local Methods
• Monitoring of structural parts • Dense sensor network• Sensors close to damage• critical location required (hot spots)
• High frequency• Sensitive to small damage
Local:Global:
• Monitoring of whole structure • Rough sensor network• Sensors not necessarily close to damage
• No knowledge about critical location required
• Low frequency• Less sensitive to small damage
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Operational/Ambient Excitation
Test Excitation
Environmental & Operational Conditions
Structure(System) Response
Feature Extraction
ReferenceFeatures
Differences, Projections, ...
Residuals(Symptoms)
Residuals evaluation:• Damage Indicators• Statistics,Thresholds
Inverse Modeling:• Localization• Extent of Damage
Decision
Damage identification scheme Measurements Pattern Recognition
Decision
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Structural Model
Dynamics of the mechanical system(nonlinear eqn. of motion)
),,(),,,,(),( ededed tt θθfθθxxgxθθM =+ &&& ),,,,( tedd xxθθΓθ && =
),,,,( edt θθxxhy &=
Dynamics of thedamage growth
Measurement Equation
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Model-based damage localization
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Eigenvalue Problem
( ) 0MK =− ii ϕω 2
Analytical modal data set is obtained from solutionof Eigenvalue Problem using FE-matrices M and K
Assume: Damage causes a local change of stiffness:
∑+=+= jjΔΔ aKKKKK 00
Correction factor approach:
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Eigenvalue-/eigenvector residuals and sensitivities
Eigenvalues Eigenvectors
⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
∂∂
∂∂
∂∂
∂∂
=
k
i
S
a
a
aa
1
22
1
1
1
λ
λ
λλ
λ
M
O
K
⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
∂
∂∂∂
∂∂
∂∂
=
k
j,i
,
,,
S
a
a
aa
1
212
11
1
11
ϕ
ϕ
ϕϕ
ϕ
M
O
K
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
−
−−
=
i,undami,dam
,undam,dam
,undam,dam
r
λλ
λλλλ
λ M22
11
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
−
−−
=
i,undami,dam
,undam,dam
,undam,dam
r
ϕϕ
ϕϕϕϕ
ϕ M22
11
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Inverse Problem: Solution
( ) ( ) rWSaISWS TT =Δ+ γγ 2
Ill-posed problem → regularization (2. term)
Tykhonov-Phillips-Regularization
εraS +=Δ
MinJ aTT →ΔΔ+=Δ aWaWa εε)(
( ) rWSaWSWS Ta
T =Δ+
Use subset selection to find most significant components of Δa
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Z24-Bridge: Use of Modal Data from Ambient Vibrations
Three span concrete box girder bridge over the A1 Bern-Zürich, CH60m length, two lanes, one sidewalk
Photos: KU Leuven
EU COST F3 Action: Benchmark example for testing damage diagnosis algorithms
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Results of Reference Measurements
Modes 3 and 4Mode 1 Mode 2 Mode 5
Output-Only Modal Data: Stochastic Subspace Identification with MACEC
Provided by: B. Peeters & G. DeRoeck, KU Leuven, Belgium
Meshgrid and Setup
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Results of Reference Model Updating
Mode 1 Mode 2 Mode 3 Mode 4 Mode 5Meas. [Hz] 3.9 5.0 9.8 10.3 12.7Model [Hz] 3.9 5.3 9.8 10.4 12.0
Comparison of eigenfrequencies:
ModelMode 1 Mode2 Mode 3 Mode 4 Mode 5 Mode 6 Mode 7 Mode 8
Meas. Mode 1 99.6 0.0 0.0 0.0 0.2 0.0 0.0 0.0Mode 2 0.1 97.0 1.6 1.0 0.0 0.0 0.6 0.2Mode 3 0.2 3.0 96.2 2.9 0.2 0.3 0.5 0.0Mode 4 0.0 2.3 4.9 94.8 0.2 1.1 0.7 0.4Mode 5 0.0 0.2 0.1 0.3 91.4 0.0 0.6 0.2
MAC-values [% ]Table of MAC-values:
Updating using SQP-algorithm (Matlab-function „fmincon“)
( )∑∑==
−+⎟⎟⎠
⎞⎜⎜⎝
⎛ −=
eses n
iiiiMAC
n
i measi
imeasii MACwwJ
modmod
1
2,,
1
2
2,
2mod,
2,
, )(1)(
)( pp
pωωω
λ
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Damage Scenario: Pier Settlement
Settlement of the „Koppigen-pier“ with hydraulic jackshere: settlement of about 95mm
Simulation of the dangerous case of undercutting of the pier,this type of damage can cause a collapse of the structure
Damage: Cracks in the concrete box near the connection to the pier
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Damage Scenario: Pier Settlement
Settlement of the „Koppigen-pier“ with hydraulic jacks (95mm)Simulation of the dangerous case of undercutting of the pier
Damage: Multiple cracks in the concrete box near the connection to the pier
Mode 1 Mode2 Mode3 Mode4 Mode5Reference [Hz] 3,9 5,0 9,8 10,3 12,7Pier settlement [Hz] 3,7 4,9 9,2 9,7 12,0
Change of Eigenfrequencies
MAC [%] Mode 1 Mode 2 Mode 3 Mode 4 Mode 5Mode 1 99,8 0,2 0,2 0,1 0,8Mode 2 0,0 98,8 1,8 1,6 0,0Mode 3 0,3 5,0 86,4 19,1 0,0Mode 4 0,0 3,8 1,0 88,7 1,8Mode 5 0,1 0,1 2,3 5,0 90,1
MAC Meas. undamaged/damaged
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Identification Results
Result:Identification of 7elementswith reduced stiffness
Data: Eigenfrequency and eigenvector residuals and sensitivities (5 modes)
Solution of inverse problem with subset selection (regression analysis)
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Software: Model-based damage localization
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Steelquake-Structure (ELSA-JRC, Ispra, I)
Sensor positions
Cooperation during the EU Cost Action F3 „Structural Dynamics“
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Damage position Results of localization
Steelquake-Structure: Results
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
measured input yi
Residual Generator
Damage Indicatorζ D
y2
y1
yi
… ( )HKvec T=ζ
Correlation function:
Nullspace-based Fault Detection (NSFD)
Ref.: Basseville et.al. 2000 Fritzen/Mengelkamp 2001
( ) ( )⎟⎟⎠
⎞⎜⎜⎝
⎛+
−−= ∑
−
=
qN
k
Tq kyqky
qNR̂
111
ζζ 1ˆ −Σ= TD
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Null-Space based Fault Detection (NSFD)
Important remark: • Constant environmental and operational conditions (EOC)• One Reference• However: changes in EOC can reveal changes of features in same order of magnitude
like damage
⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢
⎣
⎡
=
−+
+
1
132
21
βαα
β
β
βα
RR
RRRRRR
H
ˆˆ
ˆˆˆˆˆˆ
,
LL
MOMM
L
L
Hankel matrix (R = auto and cross correlations)
( )i,,i vec βαHKζ ⋅= 0
Residuals (K0 = left kernel space of the reference H0)
iTiiˆD ζζ 1−∑=
Damage indicator (Σ = Covariance matrix)with for intact structure0≅iD
Damage detection
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Weighted Hankelmatrix:
∑=
=c
jji
EOCi
jcw1
)(00 )(μ
HH
])[( 0 iEOCii vec HKζ =
Classification with fuzzy-k-means
Training data
Calculate DI with one Reference
Classifikation of DI w.r.t EOC: fuzzy-k-means
New calculation of DIwith more references and threshold setting
Weights: jiw
Calculation of and DIfrom test data
jiw
Damage detection under consideration of changing EOC
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
SHM of a Complex Composite Structure
ARTEMIS Satellite
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
ARTEMIS Satellite Antenna
Cooperation with the Univ. of Madrid (Profs. Güemes, Lopez-Diez, Cuerno)Active PZT sensors and actuators for testing
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Crack
Simulated Damage Scenarios
Small hole
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Experimental Results: NSFD
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Simulated data: 168h (no fault) + last 34h (progressive damage).Input: node 24 (force from measured wind speed, here supposed to be unknown).EOC: changing wind speed, wind direction, position of the nacelle, rotational speed of blades. Output: s1-s8 (response included 8 bending modes); Measurement time for one data set: 10 Min.;
Sample rate: 50Hz.
Damage: Young‘s-modulus reduction (1%, 5%, 10%and 25% at element 59)
Example of damage detection
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Damage detection with one reference
Training data and results of classification
Damage detection with more referencesand fuzzy-k-means
classification
Example of damage detection
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
External Forces Identification - An Inverse ProblemImpulse response-matrix (calculated or measured)
U H Y( t ) ( t ) ( t )∗ +Δ = Δ Δ
• Solely ill-posed (Dimension of H)
• Ill-posed and rank deficient (Dimension of H and sensors locations)
• Unstable (Noisy measurements)
0
y H ut
t( t ) ( t ) ( )dτ τ τ= −∫
F
Existing approaches:Modal based – SWAT(Batemen,1992), Input force identification in time domain (Genaro,1998), Optimization dynamic programming (Doyle 2001), Inverse structural filter - ISF (Steltzner & Kammer,2001), Time delay method for input estimation (Nördstrom, 2004) ,Observer based methods MPIO (Söffker & Idriz, 2003, Ha & Trinh, 2004), Neuronal Networks (Ljang 2001, Uhl, 2006), Impact identification of stiffened composite panels (Seydel & Chang 2000), …
Source: Areva-Multibrid GmbH
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
External Forces Identification - An Inverse ProblemImpulse response-matrix (calculated or measured)
0
y H ut
t( t ) ( t ) ( )dτ τ τ= −∫
F
Simultaneous State and Input EstimatorGiven System
Observer systemDesign of asymptoticallystable Observer withN, L, T and Q matrices
( )ˆ ,ˆN Ly Tf y
Qy
ω ω ξ
ξ ω
= + +
= +
&
( ),,
x E M f yu y H
ξ ξ ξξ
ξ⎡ ⎤ = +
= ⎢ ⎥ =⎣ ⎦
&
0 0T Tˆe , V e Pe, P P , Vξ ξ= − = = > <& & &&
(( ) ), ,x Ax Bu f x u yy Cx Du= + += +
&
Source: Areva-Multibrid GmbH
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Example: Laboratory structure 1- Wind Load
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Example: Laboratory structure 1- Wind Load
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Example: Laboratory structure 1- Wind Load
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Example: Laboratory structure 1- Impact Load
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Measured & Reconstructed Impulse Excitation
Example: Laboratory structure 1- Impact Load
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Example: Tripod Laboratory Structure
Wind load
Force sensor
Accelerometer
Strain gauge
Observer
FE-Model
Reconstructed forces
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Example: Tripod Laboratory Structure
Measured and Reconstructed Force
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
M5000-2 Wind Load Reconstruction
Rated power 5.000 kW
Cut-in wind speed 4 m/s
Rated wind speed 12 m/s
Cut-out wind speed 25 m/s
Nacelle height ~97m
Rotor diameter 116 m
Blade weight 16.500 kg
Hub weight 60.100 kg
Nacelle weight 199.300 kg
Design life time 20 years
Source: Areva-Multibrid GmbH
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
M5000-2 Wind Load Reconstruction
Observer
F
F̂
__ Acceleration norm. Nacelle__ Acceleration orth. Nacelle
( )
xu
E M f yy H
ξ
ξ ξ ξξ
⎡ ⎤= ⎢ ⎥⎣ ⎦= +=
& ,
Source: Areva-Multibrid GmbH
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
M5000-2 Measured operational dataNacelle directionWind direction
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
M5000-2 Estimated Wind Force
0 500 1000 15000
2
4
6
8
Time [s]
Ang
le[d
eg]
Pitch
212Betz windF c v Aρ=
200 400 600 800 1000 1200 1400
8
10
12
14
Time [S]
Vin
d V
eloc
ity[m
/s]
Wind Velocity
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
∑ ⋅=i
i tt )()(),( iurrF ϕ
Distributed force with unknown force spatial distribution
: Orthogonal force basisfunctions. (e.g. mode shapes)
)(xiϕ
)(tiu : Time history coefficient.
Distributed Forces: Basic idea
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Application: Guangzhou TV Tower
Ambient vibration measurement data (24 hours) is available.
Fig. 2 Guangzhou New TV Tower (610 m). Fig. 3 3D model in ANSYS. Fig. 4 Reduced 3D beam model.
Source: Hongkong Polytechnic University
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
SHM Using Ultrasonic Waves
Goal:
• improved system development process
• virtual SHM system design
• efficient simulation tools
• many SHM approaches based on propagation of elastic waves
• data-based, model-free• setup: time consuming, costly,
many pre-tests, “optimization” by trial and error
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
• finite elements with high degree of interpolation polynomial
• carefully chosen nodal base and numerical integration rule
• combines advantages of both methods
global pseudospectral-method
„classical“ FEM
numerically efficient, high accuracy
geometric flexibility
spectral element method, SEM
Ultrasonic Waves: Spectral Element Method
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
),,,(),,(),,(),,,(~ 0 tzyxwxz
tyxztyxutzyxu y ⋅∂∂
+⋅+= θ
),,,(),,(),,(),,,(~ 0 tzyxwyz
tyxztyxvtzyxv x ⋅∂∂
+⋅−= θ
),,,(),,,(~ tzyxwtzyxw =
Kinematics:
Nodal Base:
)()1( 12 ξξ −⋅− NLo
)()1( 12 ηη −⋅− NLo
Lobatto polynomials of order (N-1)
A Spectral Element for Flat Shells
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Approximation of displacements:
⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢
⎣
⎡
⋅=Ψ=
⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢
⎣
⎡
∑∑∑∑+
=
+
=
+
=
+
=
),(ˆ),(ˆ),(ˆ),(ˆ),(ˆ
)()(),(
),(),(),(),(),(
1
1
1
1
1
1
1
1
)(
ji
ji
jiy
jix
ji
j
N
i
N
ji
N
i
N
j
eijijy
x
vu
w
vu
w
ηξηξηξθηξθηξ
ηψξψηξ
ηξηξηξθηξθηξ
q
Selected shape functions:
)(ηψ)(ξψ1D Lagrange inter-polation polynomials
and :
Important property:ijji δξψ =)(
A Spectral Element for Flat Shells
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Element stiffness matrix:
( ) [ ] ( ) .det),(),(det),()],([1
1
1
1
)( ∑∑∫∫+
=
+
=Ω
≈Ω=N
i
N
jijij
Tijijjie
Te yxyxwwdyxyx JBDBJBDBK
Element mass matrix:
( ) [ ] ( )∑∑∫∫+
=
+
=Ω
≈Ω=1
1
1
1
)( det),(),(det),()],([N
i
N
jijij
Tijijji
Te yxyxwwdyxyxe JΨHΨJΨHΨM
for symmetrical material lay-up: diagonal mass matrix
A Spectral Element for Flat Shells
( ) [ ] ( )∑∑∫∫+
=
+
=Ω
≈Ω=1
1
1
1
)( det),(),(det),()],([N
i
N
jijijmat
Tijijjimat
Te yxyxwwdyxyxe JΨCΨJΨCΨC
Damping matrix:
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
UD-GFRP plate:
Comparison of Numerical and Experimental Data
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
UD-GFRP plate:Excitation:
5 cycle burst, 100kHz hann-windowed,actuator: P5
Comparison of Numerical and Experimental Data
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Comparison of Numerical and Experimental DataCFRP plate with delamination: low-velocity impact
energy: 15J
Material data:
Stacking: [0, 90, -45, 45, 0, 0, 90, -45, 45]sE1 [GPa] E2 [GPa] G12 [GPa] G13 [GPa] G23 [GPa] ν12=ν13=ν23 ρ [kg/m3]
155,0 8,5 4,0 4,0 4,0 0,3 1600
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Separated upper and lower elements:
− conforming mesh
− nodes remain in midplane
− offset to midplane
− additional coupling
− mode conversion
− with small gap
Modelling of Delamination
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
CFRP plate: undamaged stateExcitation:
5 cycle burst, 60kHz hann-windowed,actuator: P5
SEM model:approx. 160,000 dofs
Comparison of Numerical and Experimental Data
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
CFRP plate: damaged state Sensor P2; excitation of P5 at 60kHz
Sensor P2; excitation of P5 at 120kHz
vid
Comparison of Numerical and Experimental Data
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
3 damage scenarios in a stiffened panelStiffened panel under investigation:
• D1: base-plate delamination• D2: stringer delamination• D3: stringer debonding
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
3 damage scenarios in a stiffened panelDamage case D1, base-plate delamination:
• excitation: 4 cycle burst, 70kHz
• visibility of damage is isolated by stiffeners
• quantitative results
z-displacement after 100 μs
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
3 damage scenarios in a stiffened panelDamage case D2, stringer delamination:
x-displacement after 210 μs
• stringer sensors: higher sensitivity
• factor 10 in amplitude
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
3 damage scenarios in a stiffened panelDamage case D3, stringer debonding :
z-displacement after 150 μs
• debonding: energy-based scheme
vid1
vid2
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Damage Localization in Anisotropic Plates
Sum-Travel Time:
Geometric Solution (upper triangle):
1 2TOF TOF TOF= +
Group Velocity:
1,2(f d, , OC, E )Grc f θ τ=
1 2
2 1
1 2
sin( )sin( ) sin( )
( , , ) ( , , )Gr Gr
TOFL L
c fd EOC c fd EOC
π− θ − θ =θ θ
+θ + τ π − θ + τ
L
Properties of anisotropic approach: Damage Location found through intersection of non-elliptic curvesAccurate estimation of time of flight (based on Hinkley-Criterion)Accurate velocity model (Experiments, Simulation, DISPERSE etc.)
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Damage Localization in Anisotropic PlatesS0-Mode has been selected for localization (faster than A0-Mode)
Velocity Profile for S0-Mode at 100kHzSolution for two actuator-sensor pairs
Curves are piecewise linear
Mirror damage position
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Damage Localization in Anisotropic PlatesAnalysis of all actuator-sensor combinations
Distribution of Curve-IntersectionsStatistical Evaluation by means of probability density function (pdf)
Mirror damage position artifacts
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Lamb Waves in Isotropic PlatesWave propagation analysis using Laser-Doppler-Vibrometry
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Lamb Waves in Isotropic Plates
Excitation Signal: Frequency: fc=90kHzNumber of cycles: nS=5
Defect Properties:Two Cuts: 40mm x 1mm
Video of wave propagation:
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Lamb Waves in Isotropic PlatesInteractive GPU-based visualization of wave field:
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Conclusions• A large variety of approaches for monitoring of structuresexists
• Improve safety concepts and reducecosts by continuous inspection and earlydamage detection, change of design concepts
• SHM not yet as far developed as ConditionMonitoring methods for rotating machines
• Highly multidisciplinary task
• SHM has great potential for various applications: civil, aeronautical, mechanical eng., wind energy plants
University of Siegen
Institute of Mechanics and Control Engineering – MechatronicsProf. Dr.-Ing. C.-P. Fritzen
Thank you for your attention!