Brincker_1

  • Upload
    almarfe

  • View
    218

  • Download
    0

Embed Size (px)

Citation preview

  • 8/12/2019 Brincker_1

    1/31

    Recent advances of Operational ModalRecent advances of Operational ModalAnalysis and applications in StructuralAnalysis and applications in Structural

    Health MonitoringHealth Monitoring

    Part one (OMA)Part one (OMA)Rune Brincker

  • 8/12/2019 Brincker_1

    2/31

    MotivationsMotivations

    2

    The mechanical engineer The civil engineer

    SISO

    MIMO

    ODS

    SISO

    MIMO

    ODS

    OMAOMA

  • 8/12/2019 Brincker_1

    3/31

    Classical Modal AnalysisClassical Modal Analysis

    3

    Input

    Output Signal

    Time(Excitation) - Input

    Working : Input : Input : FFT Analyzer

    0 40m 80m 120m 160m 200m 240m

    -200

    -100

    0

    100

    200

    [s]

    [N]

    Time(Excitation) - Input

    Working : Input : Input : FFT Analyzer

    0 40m 80m 120m 160m 200m 240m

    -200

    -100

    0

    100

    200

    [s]

    [N]

    Time(Response) - Input

    Working : Input : Input : FFT Analyzer

    0 40m 80m 120m 160m 200m 240m

    -80

    -40

    0

    40

    80

    [s]

    [m/s]

    Time(Response) - Input

    Working : Input : Input : FFT Analyzer

    0 40m 80m 120m 160m 200m 240m

    -80

    -40

    0

    40

    80

    [s]

    [m/s]

    FFT

    FFTInput Signal

    Frequency Response H1(Response,Excitation) - Input (Magnitude)

    Working : Input : Input : FFT Analyzer

    0 400 800 1,2k 1,6k 2k 2,4k 2,8k 3,2k

    10m

    1

    100

    [Hz]

    [(m/s)/N]

    Frequency Response H1(Response,Excitation) - Input (Magnitude)

    Working : Input : Input : FFT Analyzer

    0 400 800 1,2k 1,6k 2k 2,4k 2,8k 3,2k

    10m

    1

    100

    [Hz]

    [(m/s)/N]

    Output Spectrum

    Small structures

    - easily tested

    Artificial input provided no problemOutput

  • 8/12/2019 Brincker_1

    4/31

    Thinking of Civil EngineersThinking of Civil Engineers

    4

    Larger structures More slender structures Bigger loads

    longer service New materials

    Codes are getting bigger and bigger

    Buildings are falling down

    Even civil enginners NEEDS to know reallity

    Cheaper and more accurate equipment

    September 11, 1916.

    Quebec Bridge (Canada)

    *Images from http://www.engineeringcivil.com/theory/civil-engineering-disasters

    December 15, 1967.

    Silver Bridge (USA)

    March 17, 1945.

    Ludendor ff Bridge (Remagen, Germany)November 7, 2005.

    (Almunecar, Spain)

  • 8/12/2019 Brincker_1

    5/31

    Idea of operatinal modal analysisIdea of operatinal modal analysis

    5

    MeasuredResponses

    Stationary

    Zero MeanGaussian

    White Noise

    Excitation Filter

    (linear,

    time-variant)

    Structural System

    (linear, time-

    invariant)

    Unknown excitation

    forces

    Combined System

    Loading modes

    Time-variant

    Broad banded

    Structural modes

    Time-invariant

    Narrow banded

  • 8/12/2019 Brincker_1

    6/31

    History on OMAHistory on OMA

    6

    Bendat & many others: Basic Frequency Domain

    Andreas Felber PhD thesis

    =

    NNN

    N

    GG

    GGGGG

    L

    OM

    K

    1

    2221

    11211

    G

  • 8/12/2019 Brincker_1

    7/31

    Time domain techniqusTime domain techniqus

    7

    Sam Ibrahim and the Time Domain

    Random Decrement Technique for Identification of Structures, J.

    Spacecraft and Rockets, Vol. 14, No. 11, 1977

    Sam Ibrahim and the Time Domain

    Random Decrement Technique for Identification of Structures, J.

    Spacecraft and Rockets, Vol. 14, No. 11, 1977

    Henry Cole was looking for...

    A simple and direct method for translating the time histori into a form

    meaningful to the observer (1971)

    Henry Cole was looking for...

    A simple and direct method for translating the time histori into a form

    meaningful to the observer (1971)

    Vandiver, Brincker and Assmussen and the Random

    Decrement (1982-1990)

    Vandiver, Brincker and Assmussen and the Random

    Decrement (1982-1990)

  • 8/12/2019 Brincker_1

    8/31

    0 2 4 6 8 10 12-5

    0

    5

    10

    15

    20

    25

    30

    SystemO

    rder

    Frequency(Hz)

    20 40 60 80 100 120 140 160 180 2000

    0.5

    1

    1.5

    2

    2.5

    3

    3.5Singular values of the system Hankel Matrix

    index number of the singular values = 2*Model Order

    IdentificationIdentification

    8

    PRCEH. Vold et al et al,1982

    PRCEH. Vold et al et al,1982

    ERAJuang and Pappa, 1985

    ERAJuang and Pappa, 1985

    MIMOMIMO

    Could be used for OMA

    Use RDD or correlation functions

    Could be used for OMA

    Use RDD or correlation functions

    Developed on the basis of traditional

    modal testing

    Developed on the basis of traditionalmodal testing

    But it did not (really) happenBut it did not (really) happen

    Reasons: mode selection problem, no software...

    Civil engineers were still sleeping...

    Reasons: mode selection problem, no software...

    Civil engineers were still sleeping...

  • 8/12/2019 Brincker_1

    9/31

    Begining of aplicationsBegining of aplications

    9

    First real test reported in

    1993

    Felber uses

    the frequency domainsuccessfully

    Felber uses

    the frequency domainsuccessfully

  • 8/12/2019 Brincker_1

    10/31

    ApplicationsApplications

    10

    A.J.Felber & R.CantieniIntroduction of a new Ambient Vibration Testing System, EMPA,

    1993-1996.

    A.J.Felber & R.CantieniIntroduction of a new Ambient Vibration Testing System, EMPA,

    1993-1996.

    C.E.Ventura, A,J, Felber, S.F.StiemerDetermination of the dynamic characteristics of the colquiz Bridgebe full-scale testing, 1992.

    C.E.Ventura, A,J, Felber, S.F.StiemerDetermination of the dynamic characteristics of the colquiz Bridge

    be full-scale testing, 1992.

    Tom Carne 1986Called it NExT= Correlation

    functions+ Polyrefernce and

    ERA

    Tom Carne 1986Called it NExT= Correlation

    functions+ Polyrefernce and

    ERA

  • 8/12/2019 Brincker_1

    11/31

    ApplicationsApplications

    11

    Basic freq domain

    SSI/ERA?

    Updating

    M.Hoeklie, O.E. HansteenMeasured and Predicted Dynamic Behavior

    of the Gulfaks A Platform, 1988.

    Basic frequency domain

    Random decrement + Polyreference

    J.C.Asmussen, R.Brincker, A.RytterAmbient modal testing of the vestvej bridge using random

    decrement, 1998.

  • 8/12/2019 Brincker_1

    12/31

    Stochastic Subspace Identification: revival ofStochastic Subspace Identification: revival of

    the time domainthe time domain

    12

    BUT: It sounds like black magic; Kalman gain matrix??? Projection of the Hankel matrix???

    If you just need modal parameters forget about the Kalman gain matrix !

    Bart de Moor and Overshees book in 1996 Data driven SSIBart de Moor and Overshees book in 1996 Data driven SSI

    Very much like Random Decrement and time domain solved by SVD???

  • 8/12/2019 Brincker_1

    13/31

    Frequency domain decomposition: revival of theFrequency domain decomposition: revival of the

    frequency domainfrequency domain

    13

    =

    = UU

    GGG

    GGG

    GGG

    nnnnn

    n

    n

    0

    00

    00

    2

    1

    21

    22221

    11211

    OMM

    L

    L

    L

    MOMM

    L

    L

    G

    ( ) { } ( ) ( ) ( ) =

    =

    ++ =

    =

    N

    k

    N

    ik

    jiji

    kN

    n

    j

    kn

    i

    nkjiji ekRGyykNyyEkR 0

    2

    ,,

    1

    0

    *

    ,

    1

    ,

    PSD Mag. Example SVD of PSD Matrix Decoupled Modes

    Modal coordinate

    auto spectral density( )1

  • 8/12/2019 Brincker_1

    14/31

  • 8/12/2019 Brincker_1

    15/31

    OverviewOverview

    15

    Reliable frequency domain techniques

    Reliable time domain techniques

    Using several data sets mode shapes

    Many applications

    Still problems: mode shape scaling... software, harmonics....

    Reliable frequency domain techniques

    Reliable time domain techniques

    Using several data sets mode shapes

    Many applications

    Still problems: mode shape scaling... software, harmonics....

    What do we have (around year 2000)?What do we have (around year 2000)?

  • 8/12/2019 Brincker_1

    16/31

    Scaling mode shapes (Mass change method)Scaling mode shapes (Mass change method)

    16

    Un-scaled mode shapes 1=T

    =

    1=MTScaled mode shapes

    Scaling factor

    Parloo et. al

    MT

    2=

    Nomenclature :

    Method:

    Classical equation

    KM =21

    KM)(M =+

    2

    2

    22

    22

    21 )( M 2M =

    22221

    MT22

    =

    Make a mass change

    Subtraction

    Solve forApproximation

    Solve for

    M

    T2

    2

    22

    21

    =

    Brincker et. al

    1 2

  • 8/12/2019 Brincker_1

    17/31

    Approximation errorsApproximation errors

    17

    Simulation study

    =

    =

    21

    1..

    .21

    12

    1

    ..

    10

    .01

    km KM

    20 DOF chainlike system

    Mass change 0-20 % of DOF mass

    Mathematical Model

    22

    2

    2

    1

  • 8/12/2019 Brincker_1

    18/31

    Approximation errorsApproximation errors

    18

    Errors should vanish for mass changes in increasing DOFsErrors should vanish for mass changes in increasing DOFs

    Improved equationImproved equationM

    T22

    22

    21

    =Parloo equationParloo equation

    MT

    2=

  • 8/12/2019 Brincker_1

    19/31

    Error reductionError reduction

    19

    MM =

    How to do it!

    Do good ID and make large shift

    Use improved equation

    Do good ID and use many DOFs

    Distribute mass change

    How to do it!

    Do good ID and make large shift

    Use improved equation

    Do good ID and use many DOFs

    Distribute mass change

    Error types

    Random Error on frequency shift

    Linearization error

    Random Error on mode shape

    Mode shape change error

    Error types

    Random Error on frequency shift

    Linearization error

    Random Error on mode shape

    Mode shape change error

    How can it be done (in an easy way) ?

    Make a mass change and estimate the frequency shift

    Easy: Shift masses around while using several data sets

    How can it be done (in an easy way) ?

    Make a mass change and estimate the frequency shift

    Easy: Shift masses around while using several data sets

    What is the accuracy ?

    In a lausy test the typical uncertainty will be

    around 10-20 %

    In a well prepared test the typical uncertainty will

    be around 2-5 %

    What is the accuracy ?

    In a lausy test the typical uncertainty will be

    around 10-20 %

    In a well prepared test the typical uncertainty will

    be around 2-5 %

  • 8/12/2019 Brincker_1

    20/31

    Harmonic removalHarmonic removal

    20

    WithHarmoni

    c

    W

    ithoutHarmonic

    Harmonic Removal

    Algori thm

  • 8/12/2019 Brincker_1

    21/31

    Harmonic removalHarmonic removal

    21

    ( )( )2

    2

    2

    22

    1,|

    ==

    x

    exfy

    ( ) ( ){ }( ) ( ).sin.sin

    sincos1|

    1

    1

    Arc

    ax

    axfy

    =

    ==

    ( ) ( )[ ] 3,|

    4

    4

    =

    xE

    x

    ( ) 5.1. measmedian

    Structural Modes:

    Harmonics:

  • 8/12/2019 Brincker_1

    22/31

    Harmonic removalHarmonic removal

    22

    HarmonicRem

    oval

    Algorithm

    HarmonicRemoval

    Algorithm

    Excluding Harmonic

    Including Harmonic

    Damping RatioNatural Frequency

  • 8/12/2019 Brincker_1

    23/31

    Known Harmonic RemovalKnown Harmonic Removal

    23

    * Notation adopted from P.Mohanty & D.J.Riexen

    P.Mohanty & D.J.Riexen (2003-2005)

    Modifications of Time domain Algorithms for Removing known Harmonic components

  • 8/12/2019 Brincker_1

    24/31

    Known Harmonic RemovalKnown Harmonic Removal

    24

  • 8/12/2019 Brincker_1

    25/31

    [ ]is

    f

    f

    ff

    d

    1

    0f

    1f 2f

    f

    Automated OMAAutomated OMA

    25

    Finding modes by FDDFinding modes by FDD

    Modal coherenceModal coherence

    Modal domainModal domain

    Harmonics RemovalHarmonics Removal

    FDD automatedFDD automated

  • 8/12/2019 Brincker_1

    26/31

    Modal CoherenceModal Coherence

    26

    )()()( 01101 fffd T

    uu=

    0)()( 101 =ffE T

    uu NffVar T /1)()( 101 =uu

    Discriminator function:

    Low modal coherence: Noise

    High modal coherence: Modal dominance

    Random vectors:

    Requirements:

    . MAC level

    . AveragingNumber

    Requirements:

    . MAC level

    . AveragingNumber

  • 8/12/2019 Brincker_1

    27/31

    Modal DomainModal Domain

    27

    The modal domain is found as the Region around the peak whereThe modal domain is found as the Region around the peak where

    )()()( 0112 fffd T

    uu=

    22 d

    Discriminator function:

    Mode property defined for all modes

    Defines the frequency region dominated by

    the mode

    [ ]is

    f

    f

    ff

    d

    1

    0f

    0f

    1f 2f

    f

    Requirements:

    MAC levelRequirements:

    MAC level

  • 8/12/2019 Brincker_1

    28/31

    Example: Plate with HarmonicsExample: Plate with Harmonics

    28

  • 8/12/2019 Brincker_1

    29/31

    Example: Heritage BuildingExample: Heritage Building

    29

  • 8/12/2019 Brincker_1

    30/31

    Example: Z24 Highway BridgeExample: Z24 Highway Bridge

    30

  • 8/12/2019 Brincker_1

    31/31

    ConclusionsConclusions

    31

    Valuable tools:Modal Coherence

    Modal Domain

    Harmonic discrimination

    Dynamic headroom etc.

    Important to note:It works..

    Easier OMAApplications in health monitoring