56813979 Modal Analysis Report Main With MATLAB

Embed Size (px)

Citation preview

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    1/51

    Modal Analysis ProjectL7005A Senior Sound Design Project

    Andre Lundkvist

    [email protected]

    Lulea, January 28, 2010

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    2/51

    Abstract

    This report is about a modal analysis project done in the course L7005A Senior Sound DesignProject. It is based upon modal measurements from a project in the course L7001A ExperimentalAcoustics and Dynamics, which were previously analysed in I-DEAS Test.

    The measurements were imported into Matlab. Functions to find poles (frequency, damping andamplitude) were constructed using the Complex Mode Indicator Function (CMIF) as a base, andmode shapes were extracted from the measurements.

    From these parameters, a multiple degrees of freedom (MDOF) modal-parameter model was con-structed by summation of many single degree of freedom (SDOF) models. The modal-parametermodel was analysed by different methods, including mathematical versus measured FRF compar-

    isons, MAC matrix and visual interpretation of the mathematical mode shapes.

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    3/51

    Contents

    1 Theory 2

    1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

    1.1.1 Analytical mathematical models . . . . . . . . . . . . . . . . . . . . . . . . 3

    1.1.2 Experimental mathematical models . . . . . . . . . . . . . . . . . . . . . . . 3

    1.2 Modal parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

    1.2.1 Mode shape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

    1.2.2 Modal frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

    1.2.3 Modal damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

    1.3 Mathematical theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

    1.3.1 Degrees of Freedom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

    1.3.2 Mobility matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

    1.3.3 Mathematical models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

    1.4 Parameter estimation methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

    1.4.1 Complex Mode Indicator Function (CMIF) . . . . . . . . . . . . . . . . . . . 8

    1.4.2 Least Squares Frequency Domain method (LSFD) . . . . . . . . . . . . . . . 9

    1.5 Modal parameter model validation . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

    1.5.1 Modal Assurance Critereon (MAC) matrix . . . . . . . . . . . . . . . . . . . 10

    1.6 Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

    1.6.1 Measurement estimation and validation . . . . . . . . . . . . . . . . . . . . 11

    1.6.2 Measurement errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

    1.6.3 Excitation and response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

    1.6.4 Response transducers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

    2 Measurements 15

    2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

    2.2 Measured FRFs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

    1

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    4/51

    3 Method 20

    3.1 Complex Mode Indicator Function (CMIF) . . . . . . . . . . . . . . . . . . . . . . . 20

    3.2 Mode shape estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

    3.3 Modal-parameter model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

    4 Results 22

    4.1 Comparison between I-Deas Test and Matlab poles . . . . . . . . . . . . . . . . . . 22

    4.2 Complex Mode Indicator Function (CMIF) . . . . . . . . . . . . . . . . . . . . . . . 23

    4.3 Mathematical vs. measured FRFs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

    4.4 Mathematical vs. measured mode shapes . . . . . . . . . . . . . . . . . . . . . . . 28

    4.5 Modal Assurance Criterion Matrix (MAC) . . . . . . . . . . . . . . . . . . . . . . . . 31

    4.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

    A I-Deas Test stability diagram 34

    B Lists 35

    B.1 Target modal parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

    B.2 List of detected frequencies and dampings . . . . . . . . . . . . . . . . . . . . . . . 36

    C Matlab code 37

    C.1 Import FRFs (impfiles.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

    C.2 Process FRFs (process frfs.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

    C.3 Complex Mode Indicator Function (cmif.m) . . . . . . . . . . . . . . . . . . . . . . 42

    C.4 Find frequencies (find freqs.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

    C.5 Find dampings (find damps.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

    C.6 SDOF modal model (sdof.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

    C.7 MAC matrix (mac.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

    C.8 Animate mode (animate mode.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

    References 48

    2

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    5/51

    Chapter 1

    Theory

    1.1 Introduction

    Modal analysis is a method to determine modal parameters of a system. The system can bedefined as a simple structure or a more complex system involving several subsystems [1]. For thepurpose of this project, the system will be a simple structure.

    Mathematical dynamic models may be used for a number of reasons.

    Gain understanding how structures behave under dynamic loads

    Curve fitting, to smoothen and reduce data

    Simulation or prediction when external forces are existent

    Physical modifications to the structure, simulate the change of dynamic characteristics

    Mathematical models describe generally the dynamic behavior, not the structure itself. Theyare often constrained by assumptions and b oundary conditions. The modal parameters thatcompletely describe the dynamics of a system are:

    Modal frequency

    Modal damping

    Mode shape

    Any forced dynamic deflection of a structure can be represented as a weighted sum of its modeshapes. Each mode can be described by an Single Degree of Freedom (SDOF) model.

    There are mainly two approaches for the study of system vibrations, analytical and experimental.The analytical starts out with knowledge about the structure geometry, boundary conditions andmaterial characteristics (mass, stiffness and damping). The experimental utilize measurements ofdynamic input forces and output responses on the structure or a prototype of the structure.

    3

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    6/51

    Modal analysis requires some assumptions of the system to work. These are:

    Linearity: The system is linear, which means the response is proportional to the force. Also, theexcitation level should not affect the behavior.

    Reciprocity: Derives from the Linearity property, and implies that the force and response canswitch place and still produce the same result.

    Time Invariant: The properties of the system should not change over time.

    Causal: The structure should not be influenced by vibration before excitation.

    Stable: The vibrations should die out, and not continue to oscillate and converge to infinity whenthe force is removed.

    1.1.1 Analytical mathematical modelsAnalytical models are often created by using FEM1 based on calculated mass and stiffness underspecific boundary conditions. The model will contain a number of differential equations, whichare coupled [2].

    1.1.2 Experimental mathematical models

    Experimental modal analysis consists of five different phases. These span from setup, data ac-quisition, system identification, validation and using the obtained information for improving thesystem in a systematic way [2].

    Usually measurements are observed in the frequency domain as Frequency Response Functions(FRFs), H(). Which is the ratio between output and input as a function of frequency. Seefigure 1.1.

    The FRFs can be described by the modal parameters, and the modal parameters can therefore beextracted from the FRFs.

    Figure 1.1: Frequency Response Function (FRF) example [3]

    1Finite Element Method

    4

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    7/51

    1.2 Modal parameters

    1.2.1 Mode shape

    The mode shape is as the name suggests the shape of the mode. It is connected to the modalfrequency and therefore the pole location of the mode, and expresses a deflection pattern in thestructure [1].

    From measurements, a mode shape is a vector containing sampled values. The amount ofsamples is directly linked to the degrees of freedom (DOF) used. Often these vectors containcomplex numbers that describe both magnitude and phase, and these describe modes that can beconsidered as propagating waves. Standing waves is considered as a normal mode, meaning isa real number vector.

    Unreliable measurements can wrongly indicate complex modes instead of normal modes.

    1.2.2 Modal frequency

    The resonance frequency of the specific mode is called modal frequency. These can be determinedby observing the maximum magnitudes of the measured or calculated FRF.

    1.2.3 Modal damping

    Modal damping is the damping to the specific mode, and can be determined from an FRF by findingthe 3 dB bandwidths of the modes. This can be a problem on a lightly dampened structure, asthe resonance peaks are very narrow.

    5

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    8/51

    1.3 Mathematical theory

    1.3.1 Degrees of Freedom

    Specifying the Degrees of Freedom (DOF) is based on the purpose of the test, geometry and thefrequency range. A free point in space has 6 DOFs, both rotation and translation on axes x, y & z.There are not many rotational transducers available, but three transitional is usually enough todescribe the the displacement and motion [1].

    The number of possible input/output (force/response) combinations will be n2, where n is theselected DOF.

    1.3.2 Mobility matrix

    The mobility matrix contains all different combinations of the FRFs. For a three-DOF system, themobility matrix would look like equation 1.1. Each row contains FRFs with a common responseDOF, and each column have a common excitation DOF. The diagonal contains the driving pointFRF [1].

    H =

    H1,1 H1,2 H1,3H2,1 H2,2 H2,3

    H3,1 H3,2 H3,3

    (1.1)

    1.3.3 Mathematical models

    SDOF modal-parameter model

    Below is described a Modal-Parameter Model FRF, see equation 1.2, figures 1.2 and 1.3, where (equation 1.4) is the pole location and A (equation 1.3) is the residue. Equation 1.2 describes anSDOF system, but can be expanded to an MDOF system [2].

    H(j) =A1

    j 1+

    A1j

    1

    (1.2)

    with

    A1 = j1

    2M 1

    A1 = j1

    2M 1(1.3)

    1 = +j1

    1

    = j1 (1.4)

    In the pole equation, is the damping factor and 1 is the damped natural resonance frequency.The residue describes the mode strength, and is related to the mode shape. Because of thecomplex conjugate part is almost negligible around resonance, the model can be approximated toequation 1.5.

    H(j) A1

    j 1(1.5)

    6

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    9/51

    Figure 1.2: Example of a magnitude plot of the FRF for a SDOF system [2]

    Figure 1.3: Example of a phase plot of the FRF for a SDOF system [2]

    The impulse response is described in equation 1.6, and it can look similar to figure 1.4. The

    impulse response is the Fourier transform of the FRF [2].

    h(t) = A1e1t + A

    1e

    1t = e1t

    A1e

    j1t + A1

    ej1t

    (1.6)

    Figure 1.4: Example of an impulse response for a SDOF system [2]

    7

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    10/51

    MDOF modal-parameter model

    The MDOF model can be seen in equation 1.7 [2].

    H(j) =

    Nr=1

    Ar

    j r+

    Arj r

    (1.7)

    where the poles can be expressed in matrix form as in equation 1.8.

    r =

    1 +j1. . . 0

    N +jN1 j1

    0. . .

    NjN

    (1.8)

    If the poles are considered as eigenvalues of a system, its eigenvectors can be interpreted as themodal shapes, where each eigenvector corresponds with a specific eigenvalue. See equation 1.9.

    =

    1{}1 . . . N{}N

    1{}1 . . .

    N{}

    N

    {}1 . . . {}N {}1 . . . {}

    N

    (1.9)

    The residues, Ar is described by equation 1.10, where Qr is a scaling factor, and {} is the modalshape matrix for the specific mode.

    Ar = Qr{}r{}tr (1.10)

    By using the notation above, the FRF can be expanded to the form described in equation 1.11.

    H(j) =

    Nr=1

    Qr{}r{}

    tr

    (j r)+

    Qr{}

    r{}tr

    (j r)

    (1.11)

    Figure 1.5: Example of a FRF magnitude plot constructed from two SDOF FRFs [2]

    8

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    11/51

    1.4 Parameter estimation methods

    1.4.1 Complex Mode Indicator Function (CMIF)

    The CMIF is based upon the singular value decomposition of the FRF matrix, containing all possibleinput output FRF combinations. The diagonal Singular Matrix ([ ]) is used as the CMIF. Itindicates the existence of real (normal) or complex modes and also gives the relative magnitudeof each mode. It is also capable of yielding the corresponding mode shape and/or participationvector [2].

    The Singular Value Decomposition (SVD) of FRF matrix [ H ] is expressed as,

    [ H ] = [ U ] [ ] [ V ] (1.12)

    where,

    [ H ] is the Frequency Response Function Matrix[ U ] is the Left Singular Matrix (unitary)[ ] is the Singular Matrix (diagonal)[ V ] is the Right Singular Matrix (unitary)

    (1.13)

    500 1000 1500 2000 2500 3000

    20

    40

    60

    80

    100

    120

    Complex Mode Indicator Function (CMIF) of FRF Matrix [H]

    Frequency (Hz)

    Amplitude(dB)

    Figure 1.6: Example of a CMIF for a FRF matrix

    9

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    12/51

    1.4.2 Least Squares Frequency Domain method (LSFD)

    The non-linear Least Squares Frequency Domain (LSFD) method estimates poles, mode shapes andmodal participation factors of a system. The modal participation factors can only be estimatedif there are multiple inputs. The LSFD method is mostly used for model updating, because of thecomplexity and amount of parameters that needs to be determined [2].

    This method is based on the frequency domain modal-parameter model described in section 1.3.3on page 7. If equation 1.7 is expanded, including the modal participation factors, the form will bedescribed as

    Hij(j) =Nr=1

    irLrj

    j r+

    irL

    rj

    j r

    + U Rij

    LRij

    2(1.14)

    where i denotes a response position and j denotes an input station. U Rij and LRij is the upperand lower residual terms. These terms approximates the modes below and above the frequency

    band of interest.Hij(j) is the measured FRF at point (i, j). The method is about iteratively determine theunknown parameters r, ir, Lrj , U Rij and LRij by minimizing the error between the measuredFRF and the model. The amount of parameters is Nu.

    The difference between the measured FRF and the model can be expressed as,

    eij(j ) = Hij(j) Gij(j,r, ir, Lrj , U Rij , LRij)|r=1,Nm (1.15)

    and the total squared error over the frequency range of interest (0 Nf) becomes,

    Eij =

    Nf

    f=0

    eij(jf) e

    ij(jf) (1.16)

    The total error between all inputs and outputs can be expressed as,

    E =

    Noi=1

    Nij=1

    Eij (1.17)

    To minimize the global error, the derivatives of each parameter from the total error E shouldconverge to 0, i.e for parameter rk,

    dE

    dr1= 0 (1.18)

    dE

    dr2= 0 (1.19)

    ... (1.20)

    dE

    drNu= 0 (1.21)

    10

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    13/51

    1.5 Modal parameter model validation

    1.5.1 Modal Assurance Critereon (MAC) matrix

    The Modal Assurance Criterion (MAC) matrix is a mathematical tool to compare two vectors toeach other. It can be used to investigate the validity of estimated modes [2].

    The MAC between two mode shape vectors {}r and {}s is defined as,

    MAC({}r, {}s) =

    {}Tr {}s

    2({}Tr {}r) ({}

    Ts {}s)

    (1.22)

    The MAC will approach the value 1 if{}r and {}s are the same mode shape. It {}r and {}sare different mode shapes, the MAC value should be low, due to the orthogonality condition of the

    mode shapes.

    Figure 1.7: Example of a MAC matrix on 5 randomly generated vectors (Gaussian)

    11

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    14/51

    1.6 Measurements

    1.6.1 Measurement estimation and validation

    There are a couple of estimators used to calculate the FRF. For noise at the output, a usefulestimator is H1, see equation 1.23. The function H1 is derived by using least squares method, andis the cross spectrum divided by the autospectrum of the force F. If averaging the FRFs measuredwith H1, the random noise will suppress and H1 will converge towards the true H [1].

    H1() GFX()

    GFF()(1.23)

    For noise at the input H2 is a useful estimator for the FRF. It is derived from the same principle asthe H1 estimator, and is defined as the autospectrum of the response divided by the cross spectrum.

    The noise at the input is removed more and more from the cross spectrum with increased averages.

    H2() GXX ()

    GXF()(1.24)

    A way to validate a measurement is to observe the coherence function, seen in equation 1.26. It isderived from the cross spectrum inequality, equation 1.25 which states that if the autospectrumcontains non-coherent noise, the magnitude of the squared cross spectrum is smaller than theproduct of the autospectrum.

    The coherence function have the boundaries described in equation 1.27. For the value 1, themeasurement contains no noise, and for the value 0, there is pure noise in the measurement. Thecoherence function also indicates the linearity between the input and output signal.

    When the coherence is 1, estimator H1 and H2 will yield the same result, therefore the estimatorsare overcompensated, and the true FRF will be somewhere in between.

    |QXF()|2 GXX () GFF() (1.25)

    ()2 |GFX()|

    2

    GXX () GFF()(1.26)

    0 ()2 1 (1.27)

    To calculate the autospectrum, the spectrum is multiplied with its complex conjugate. The com-plex conjugate is the same spectrum but with opposite sign for the imaginary part. The autospec-trum is always real.

    The cross spectrum is calculated by multiplying a spectrum with the complex conjugate of adifferent spectrum. For instance the spectrum of the force and the response. Cross spectrum is acomplex entity which describes the phase shift between the different spectrums. The magnitudeof the cross spectrum describes the coherent product of power in the spectrums.[1]

    12

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    15/51

    1.6.2 Measurement errors

    There are some error that can occur when performing mobility measurements. These are listedin table 1.1, and its abbreviations in table 1.2. The different errors can be divided into differentclasses, depending on the nature of the cause. The classes are Random & Bias errors [1].

    Random errors are caused by noise, and can occur both at the input and the output of the signalchain. Random noise can be minimized by averaging over many measurements, because the noiseis uncorrelated.

    Systematic bias error are the same for every measurement (both magnitude and phase). The onlyway to minimize bias errors is to select another estimator.

    Table 1.1: Possible errors for mobility measurementsType of Error H1 H2

    2

    Noise at the output R B +Noise at the input B R +

    Random excitation / non-linearity B/R B/R +

    Deterministic excitation B B 0

    Scatter of impact (point / direction) R R +

    Random excitation B (B) +

    Deterministic impact B B 0

    Table 1.2: Error table abbreviationsAbbreviation Description

    B Systematic Bias errorR Random error

    + Coherence Function can indicate error

    0 Coherence Function can not indicate error

    13

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    16/51

    1.6.3 Excitation and response

    Waveform

    The choice of excitation waveform depends on which type of measurement is to be performed.There is a list of certain parameters to consider, see table 1.3 [1].

    The crest factor is the ratio between the peak and the standard deviation RMS of the signal. Formodal analysis, a high crest factor in the excitation waveform is undesired, as it contributes to acouple of factors. Noise will be introduced because of the sharp peaks, which decreases the signalto noise ratio SNR. The other factor is that nonlinearity might be introduced due to the high peakforces.

    Because modal analysis assumes that the system is linear and time invariant, this can be a problem,although there are ways of making linear approximations of a non-linear system.

    Table 1.3: Excitation conditional parametersParameter Description

    Application What is the targeted system, which type of measurement

    Spectrum Control Capability to control the frequency range

    Crest Factor Ratio between the peak and RMS in the signal

    Linearity Linear and time invariant system

    Speed of Test How much time is needed for a measurement

    Table 1.4: Excitation waveform comparisonWaveform

    Factor Sinusoid Random Pseudo Random Impact Multiple Impact

    Analysis Speed Very Slow Slow Fast Fastest Slow

    Leakage Error Yes Yes No No Yes

    Approximation No Yes No No Some

    Crest Factor Good Fair Fair Poor Poor-Fair

    Spectrum Control High High High Limited Limited

    Zoom Analysis Yes Yes Yes No No

    Detect Linearity No Yes No No (Yes)

    Detect Leakage No Yes No No Yes

    Excitation sources

    There are many different exciters to choose from, such as shakers, hammers, pendulums etc. Theseare categorized into two fastening categories, attached and non-attached. The choice of exciterdepends on the excitation conditional parameters, the structure etc [1].

    There are advantages and disadvantages for both categories. The attached exciters affects thestructure quite a lot, whereas the non-attached does not. However, the attached exciters give amore controllable excitation than a non-attached.

    14

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    17/51

    1.6.4 Response transducers

    Accelerometers

    To be able to measure the structural response, the need for a transducer is imminent. The mostcommonly used transducer is an accelerometer. The standard measurement for an accelerometeris acceleration, but velocity or displacement can be calculated by integration [1].

    The piezoelectric accelerometer has good linearity, broad dynamic range, wide frequency rangeand a strong construction. The dynamic range is typically around 160 dB and ranges in frequencybetween 0.2 Hz to at least 10 kHz, with a linearity of 5 %.

    Accelerometers can also be very light, about 1 g, and are easy to mount by glue, wax, screwetc. See a list of selected mounting techniques below. Beeswax is a commonly used method, seefigure 1.8. The resonance frequency of the accelerometer is dependent on its weight though, sothis have to be considered, and the mounting also affects the applicable frequency range.

    Steel stud

    Beeswax

    Cement stud

    Thin tape

    Thick tape

    Magnet

    For measurements in the driving point, an impedance head may be used if possible by the setup.The impedance head is a transducer with integrated force and response accelerometers. Anothertypical setup is by placing a standard accelerometer transducer on the opposite side of the structureto where the force is applied to.

    Figure 1.8: Example of a few accelerometers mounted with beeswax

    15

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    18/51

    Chapter 2

    Measurements

    2.1 Introduction

    This chapter discusses the measured data of the metal plate showed to the left in figure 2.1. It ishung by springs to reduce external excitation and to avoid influence on the structure.

    The measurement grid is showed to the right in figure 2.1, where the point (1,1) would be thelower left hand corner, and it is also the excitation point used. The reference to the measurementgrid is counting from bottom left (row from bottom, column from left), throughout the report.

    Figure 2.1: Left: The metal plate used for measurements, mounted with springs. Right: Themeasurement grid used for measuring the metal plate

    16

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    19/51

    The settings for the measurements were as listed below in table 2.1. An impact hammer was usedfor excitation, and it had a transient window applied to avoid multiple excitations.

    Accelerometer 2 was used as a trigger to start the measurement. The hold-off was set to the samelength as the measurement time to avoid the measurement to restart before completion.

    Table 2.1: Measurement settingsSetting Parameter Value

    Base Resolution 6400 lines

    Break Frequency 3200 Hz

    Measure Time 2 s

    Means 3, linearly weighted

    Overload Reject

    Accelerometer Window Exponential

    0.3 s

    Excitation Window Transient Start 0.9 ms Stop 1.3 ms

    Trigger Signal Accelerometer 2

    Level 5 %

    Hold-Off 2 s

    17

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    20/51

    2.2 Measured FRFs

    This section presents some chosen FRFs measured on the plate described in figure 2.1.The driving point FRF is showed in figure 2.2 below. There are clear resonance peaks, and overallthey are quite separated. The phase response has a weird behavior in the low frequency range,probably due to measurement problems.

    500 1000 1500 2000 2500 3000

    3

    2

    1

    0

    1

    2

    3

    Imaginary Response Values (1,1)

    Radians

    Frequency [Hz]

    500 1000 1500 2000 2500 3000

    50

    0

    50

    Real Response Values (1,1)

    M

    agnitude[dB(m/s2/N)]

    Frequency [Hz]

    Figure 2.2: Example of a phase and magnitude plot of the FRF associated with the point (1,1)

    The measurement in point (3,4), according to the grid in figure 2.1), is showed in figure 2.3.In comparison to the driving point FRF, this response has fewer clear resonance peaks, and theapparent ones has lower amplitude which might suggest higher damping.

    Figure 2.4 shows the FRF at point (5,1) which is located at the left edge, just above the middle

    of the plate. This FRF has clear separated resonances, except at about 680 Hz, where there couldbe a double pole.

    18

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    21/51

    500 1000 1500 2000 2500 3000

    3

    2

    1

    0

    1

    2

    3

    Imaginary Response Values (3,4)

    Radians

    Frequency [Hz]

    500 1000 1500 2000 2500 3000

    25

    20

    15

    10

    5

    0

    5

    10

    Real Response Values (3,4)

    Magnitude[dB(m/s2/N)]

    Frequency [Hz]

    Figure 2.3: Example of a phase and magnitude plot of the FRF associated with the point (3,4)

    500 1000 1500 2000 2500 3000

    3

    2

    1

    0

    1

    2

    3

    Imaginary Response Values (5,1)

    Radians

    Frequency [Hz]

    500 1000 1500 2000 2500 3000

    40

    20

    0

    20

    40

    Real Response Values (5,1)

    Magnitude[dB(m/s2/N)]

    Frequency [Hz]

    Figure 2.4: Example of a phase and magnitude plot of the FRF associated with the point (5,1)

    19

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    22/51

    500 1000 1500 2000 2500 3000

    3

    2

    1

    0

    1

    2

    3

    Imaginary Response Values (7,2)

    Radians

    Frequency [Hz]

    500 1000 1500 2000 2500 3000

    20

    10

    0

    10

    20

    Real Response Values (7,2)

    Magnitude[dB(m/s2/N)]

    Frequency [Hz]

    Figure 2.5: Example of a phase and magnitude plot of the FRF associated with the point (7,2)

    500 1000 1500 2000 2500 3000

    3

    2

    1

    0

    1

    2

    3

    Imaginary Response Values (9,3)

    Radians

    Frequency [Hz]

    500 1000 1500 2000 2500 300030

    20

    10

    0

    10

    20

    30

    Real Response Values (9,3)

    Magnitude[dB(m/s2/N)]

    Frequency [Hz]

    Figure 2.6: Example of a phase and magnitude plot of the FRF associated with the point (9,3)

    20

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    23/51

    Chapter 3

    Method

    3.1 Complex Mode Indicator Function (CMIF)

    The complex mode indicator function (CMIF) (see section 1.4.1, page 8) is used to evaluate theresonate frequencies and get an approximation for relative amplitude and damping. The Matlabcode used to calculate the CMIF can be found in appendix C.3 on page 42. The curve is similarto the average FRF used in I-DEAS Test to calculate the stability diagram (see appendix A atpage 34).

    The algorithm to search for frequencies is an iterative method. The code for it can be found inappendix C.4 on page 43. It searches for peaks by stepping through the curve, and when the curvehas reached a maximum, it will indicate a new pole by a red line in the CMIF, see figure 3.1.

    500 1000 1500 2000 2500 30000

    20

    40

    60

    80

    100

    120

    Complex Mode Indicator Function (CMIF) of FRF Matrix [H]

    Frequency (Hz)

    Amplitude(dB)

    Figure 3.1: CMIF for all FRFs

    The algorithm for finding damping ratios is similar, but it will start out from the detected fre-quencies and iteratively search for the -3 dB decay limits. The code can be found in appendix C.5

    on page 44. The -3 dB limits will be indicated on the CMIF as green lines, see figure 3.1.

    21

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    24/51

    It tries to approximate the damping by taking the closest limit for both the left and right limit.Meaning, that if for example the left limit will diverge, but the right will not, it will approximatethe left limit to FP - abs(d2), where FP is the pole location and d2 is the right limit, both in

    samples.

    Figure 3.2 shows a close up display of four detected poles, indicating resonance frequency anddamping for each pole.

    250 260 270 280 290 300 310 320 330 340 350

    10

    20

    30

    40

    50

    60

    70

    Complex Mode Indicator Function (CMIF) of FRF Matrix [H]

    Frequency (Hz)

    Amplitude

    (dB)

    Figure 3.2: Detailed CMIF for all FRFs

    3.2 Mode shape estimation

    The mode shapes are estimated from the imaginary part of the measured FRF response matrix{H} at each system pole detected by the CMIF [4]. This produces a mode shape matrix for eachpole frequency, with the dimensions of the measurement grid (figure 2.1).

    3.3 Modal-parameter model

    The modal model was created by using equation 1.7 (page 7). To use the equation in Matlab, itwas rewritten like

    Hi,j(j) Nr=1

    Qr{i,j}r{i,j}Tr

    j r jr+

    Qr{i,j}

    r{i,j}Tr

    j r +jr(3.1)

    where Qr is a scaling factor and {i,j}r is the modal shape amplitude and direction informationfor the specific mode and node position. r is the damping factor and r is the damped naturalresonance frequency for the specific mode.

    The amplitude of the mode was used as scaling. The Matlab code can be found in appendix C.6

    on page 45. After creating all SDOF models, these are summed together to create the MDOF model.

    22

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    25/51

    Chapter 4

    Results

    4.1 Comparison between I-Deas Test and Matlab poles

    A comparison between the detected poles in I-DEAS Test and Matlab can be found in table 4.1below. The frequencies detected in Matlab is quite similar to I-DEAS Test, with values bothhigher and lower to I-DEAS Test. The damping is relatively the same for some poles, but candiffer a bit on some. The difference indicates the difference (I-DEAS Test - Matlab).

    Poles number 7, 14 and 15 was not detected by I-DEAS Test. Note on pole number 18 and 19,both the frequency and damping is quite close to each other. However, this might be a coincidence.

    Table 4.1: Comparison between detected frequencies and damping ratios

    Pole Frequency Frequency Frequency Damping Damping Damping

    Nr I-DEAS Test Matlab Difference I-DEAS Test Matlab Difference

    1 272.544 270.458 2.0860 1.437 0.238 1.199

    2 335.348 331.448 3.9000 0.665 0.719 -0.054

    3 620.261 621.403 -1.1420 0.202 0.110 0.092

    4 644.843 641.900 2.9430 0.090 0.042 0.048

    5 783.756 787.377 -3.6210 0.127 0.056 0.071

    6 962.906 965.849 -2.9430 0.232 0.158 0.074

    7 973.848 0.181 8 1221.286 1227.31 -6.0240 0.214 0.032 0.182

    9 1294.681 1296.80 -2.1190 0.214 0.069 0.145

    10 1660.698 1664.74 -4.0420 0.336 0.187 0.149

    11 1853.714 1852.21 1.5040 0.280 0.135 0.145

    12 1999.744 1998.19 1.5540 0.287 0.113 0.174

    13 2089.418 2090.67 -1.2520 0.355 0.219 0.136

    14 2110.17 0.187 15 2373.13 0.382 16 2401.034 2399.63 1.4040 0.451 0.222 0.229

    17 2901.708 2901.55 0.1580 0.268 0.093 0.175

    18 3023.607 3023.53 0.0770 0.212 0.273 0.061

    19 3093.047 3094.02 -0.9730 0.435 0.452 -0.017

    23

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    26/51

    4.2 Complex Mode Indicator Function (CMIF)

    To evaluate the generated mathematical FRFs, first the CMIF function was used. A comparisonbetween the measurement CMIF and the mathematical FRF CMIF can be seen in figure 4.1 below.

    Overall, the mathematical poles seem quite close to the measurements. The first pole, locatedat about 270 Hz, which has almost double amplitude. The overall trend seems to be that lowerfrequency poles has higher amplitude than the measured, and higher frequency poles has loweramplitude than measured. This error could probably have something to do with the differencedescribed in the previous section 4.1.

    500 1000 1500 2000 2500 3000

    20

    40

    60

    80

    100

    120

    Reconstructed FRF CMIF from SDOF Data

    Frequency (Hz)

    Amplitude

    Reconstructed

    CMIF

    Figure 4.1: CMIF for all mathematical FRFs

    24

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    27/51

    4.3 Mathematical vs. measured FRFs

    Another evaluation of the generated mathematical FRFs is to compare them to the measured FRFs.The first comparison done is in the driving point (position (1,1)) seen in figure 4.2 below.

    The mathematical does not correspond well to the measurement, which indicates that the modelis not accurate at this position of the plate. The mathematical phase is not coherent with themeasurement phase either.

    500 1000 1500 2000 2500 30003

    2

    1

    0

    1

    2

    3

    Comparison Reconstructed FRF and Real FRF (1,1)

    Frequency (Hz)

    P

    hase[rad]

    Math

    Real

    500 1000 1500 2000 2500 300050

    0

    50

    Comparison Reconstructed FRF and Real FRF (1,1)

    Frequency (Hz)

    Magnitude[dB(m/s2/N)]

    Math

    Real

    Figure 4.2: Comparison of the mathematical FRF against the measured FRF in point (1,1)

    The comparison for position (3,4) can be seen in figure 4.3. It correspond much more with themeasured response at this position, especially between 800 2300 Hz. Also the phase is betterthan in the driving point comparison, but still not good.

    More comparisons can be seen in figures 4.4 to 4.7. An overall conclusion is that the mathematicalFRFs is mostly accurate for the center positions of the plate. Why this is the case has not beenthoroughly investigated.

    25

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    28/51

    500 1000 1500 2000 2500 30003

    2

    1

    0

    1

    2

    3

    Comparison Reconstructed FRF and Real FRF (3,4)

    Frequency (Hz)

    Phase[rad]

    500 1000 1500 2000 2500 3000

    20

    10

    0

    10

    20

    Comparison Reconstructed FRF and Real FRF (3,4)

    Frequency (Hz)

    Magnitude[dB(m/s2/N)]

    Math

    Real

    Math

    Real

    Figure 4.3: Comparison of the mathematical FRF against the measured FRF in point (3,4)

    500 1000 1500 2000 2500 30003

    2

    1

    0

    1

    2

    3Comparison Reconstructed FRF and Real FRF (5,1)

    Frequency (Hz)

    Phase[rad]

    500 1000 1500 2000 2500 3000

    40

    20

    0

    20

    40

    Comparison Reconstructed FRF and Real FRF (5,1)

    Frequency (Hz)

    Magnitude[dB(m/s2/N)]

    Math

    Real

    Math

    Real

    Figure 4.4: Comparison of the mathematical FRF against the measured FRF in point (5,1)

    26

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    29/51

    500 1000 1500 2000 2500 30003

    2

    1

    0

    1

    2

    3

    Comparison Reconstructed FRF and Real FRF (7,2)

    Frequency (Hz)

    Phase[rad]

    500 1000 1500 2000 2500 3000

    15

    10

    5

    0

    5

    10

    15

    Comparison Reconstructed FRF and Real FRF (7,2)

    Frequency (Hz)

    Magnitude[dB(m/s2/N)]

    Math

    Real

    Math

    Real

    Figure 4.5: Comparison of the mathematical FRF against the measured FRF in point (7,2)

    500 1000 1500 2000 2500 30003

    2

    1

    0

    1

    2

    3Comparison Reconstructed FRF and Real FRF (8,2)

    Frequency (Hz)

    Phase[rad]

    500 1000 1500 2000 2500 3000

    20

    15

    10

    5

    0

    5

    10

    15

    Comparison Reconstructed FRF and Real FRF (8,2)

    Frequency (Hz)

    Magnitude[dB(m/s2/N)]

    Math

    Real

    Math

    Real

    Figure 4.6: Comparison of the mathematical FRF against the measured FRF in point (8,2)

    27

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    30/51

    500 1000 1500 2000 2500 30003

    2

    1

    0

    1

    2

    3

    Comparison Reconstructed FRF and Real FRF (9,3)

    Frequency (Hz)

    Phase[rad]

    500 1000 1500 2000 2500 3000

    20

    10

    0

    10

    20

    30

    Comparison Reconstructed FRF and Real FRF (9,3)

    Frequency (Hz)

    Magnitude[dB(m/s2/N)]

    Math

    Real

    MathReal

    Figure 4.7: Comparison of the mathematical FRF against the measured FRF in point (9,3)

    28

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    31/51

    4.4 Mathematical vs. measured mode shapes

    This section presents the shapes of some of the detected modes. To the left, the mode shapefrom the measurement data is presented, and to the right is the mathematical mode shape. Themodal-parameter model used to create the mathematical FRFs uses an estimate of the mode shapefrom the measurements (section 3.2).

    Overall the mode shapes for the lower frequency band looks reasonable, but the top and bot-tom of the plate indicates possible errors. Intuitively, these rows appear to be 180 degrees out ofphase. The reason for this has not been investigated.

    The mode shapes indicates that there are torsional and bending waves present in the modes.

    1

    2

    3

    4

    5

    1

    2

    3

    4

    5

    6

    7

    8

    9

    1

    0.5

    0

    0.5

    1

    Nodes X

    Mode Animation: 270.458 Hz

    Nodes Y

    1

    2

    3

    4

    5

    1

    2

    3

    4

    5

    6

    7

    8

    9

    1

    0.5

    0

    0.5

    1

    Nodes X

    Mode Animation: 270.458 Hz

    Nodes Y

    Figure 4.8: Measured (left) and Mathematical (right) mode shape for 270 Hz

    1

    2

    3

    4

    5

    1

    2

    3

    4

    5

    6

    7

    8

    9

    1

    0.5

    0

    0.5

    1

    Nodes X

    Mode Animation: 331.448 Hz

    Nodes Y

    1

    2

    3

    4

    5

    1

    2

    3

    4

    5

    6

    7

    8

    9

    1

    0.5

    0

    0.5

    1

    Nodes X

    Mode Animation: 331.448 Hz

    Nodes Y

    Figure 4.9: Measured (left) and Mathematical (right) mode shape for 331 Hz

    29

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    32/51

    1

    2

    3

    4

    5

    1

    2

    3

    4

    5

    6

    7

    8

    9

    1

    0.5

    0

    0.5

    1

    Nodes X

    Mode Animation: 621.403 Hz

    Nodes Y

    1

    2

    3

    4

    5

    1

    2

    3

    4

    5

    6

    7

    8

    9

    1

    0.5

    0

    0.5

    1

    Nodes X

    Mode Animation: 621.403 Hz

    Nodes Y

    Figure 4.10: Measured (left) and Mathematical (right) mode shape for 621 Hz

    1

    2

    3

    4

    5

    1

    2

    3

    4

    5

    6

    7

    8

    9

    1

    0.5

    0

    0.5

    1

    Nodes X

    Mode Animation: 641.9 Hz

    Nodes Y

    1

    2

    3

    4

    5

    1

    2

    3

    4

    5

    6

    7

    8

    9

    1

    0.5

    0

    0.5

    1

    Nodes X

    Mode Animation: 641.9 Hz

    Nodes Y

    Figure 4.11: Measured (left) and Mathematical (right) mode shape for 642 Hz

    1

    2

    3

    4

    5

    1

    2

    3

    4

    5

    6

    7

    8

    9

    1

    0.5

    0

    0.5

    1

    Nodes X

    Mode Animation: 973.848 Hz

    Nodes Y

    1

    2

    3

    4

    5

    1

    2

    3

    4

    5

    6

    7

    8

    9

    1

    0.5

    0

    0.5

    1

    Nodes X

    Mode Animation: 973.848 Hz

    Nodes Y

    Figure 4.12: Measured (left) and Mathematical (right) mode shape for 974 Hz

    30

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    33/51

    1

    2

    3

    4

    5

    1

    2

    3

    4

    5

    6

    7

    8

    9

    1

    0.5

    0

    0.5

    1

    Nodes X

    Mode Animation: 1227.31 Hz

    Nodes Y

    1

    2

    3

    4

    5

    1

    2

    3

    4

    5

    6

    7

    8

    9

    1

    0.5

    0

    0.5

    1

    Nodes X

    Mode Animation: 1227.31 Hz

    Nodes Y

    Figure 4.13: Measured (left) and Mathematical (right) mode shape for 1227 Hz

    1

    2

    3

    4

    5

    1

    2

    3

    4

    5

    6

    7

    8

    9

    1

    0.5

    0

    0.5

    1

    Nodes X

    Mode Animation: 1664.74 Hz

    Nodes Y

    1

    2

    3

    4

    5

    1

    2

    3

    4

    5

    6

    7

    8

    9

    1

    0.5

    0

    0.5

    1

    Nodes X

    Mode Animation: 1664.74 Hz

    Nodes Y

    Figure 4.14: Measured (left) and Mathematical (right) mode shape for 1665 Hz

    1

    2

    3

    4

    5

    1

    2

    3

    4

    5

    6

    7

    8

    9

    1

    0.5

    0

    0.5

    1

    Nodes X

    Mode Animation: 3023.53 Hz

    Nodes Y

    1

    2

    3

    4

    5

    1

    2

    3

    4

    5

    6

    7

    8

    9

    1

    0.5

    0

    0.5

    1

    Nodes X

    Mode Animation: 3023.53 Hz

    Nodes Y

    Figure 4.15: Measured (left) and Mathematical (right) mode shape for 3024 Hz

    31

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    34/51

    4.5 Modal Assurance Criterion Matrix (MAC)

    The MAC matrix of the mathematical model created withI-DEAS Test

    can be seen below infigure 4.16. It indicates that the modes are not similar to other modes except for themselves. Thisis the wanted result.

    For the Matlab created mathematical modes, the MAC matrix can be found in figure 4.17. Thecode to create this MAC matrix can be found in appendix C.7 on page 46.

    Compared to the I-DEAS Test MAC matrix, the model created in Matlab is not optimal. Manymodes are too similar to be trusted. The four most similar modes are listed in table 4.2. Onlymodes 7, 14 and 15 are not detected by I-DEAS Test. If mode 7 is a faulty mode, this couldexplain the similarity between mode 6 and 7.

    Figure 4.16: MAC for all mathematical FRFs from I-DEAS Test

    Table 4.2: Most similar modesModes Frequency Similarity

    Mode 1 Mode 2 Mode 1 Mode 2 MAC Value

    4 13 642 2090 0.7367

    6 7 966 974 0.3144

    9 17 1297 2902 0.2886

    5 12 787 1998 0.2226

    32

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    35/51

    Figure 4.17: MAC for all mathematical FRFs from Matlab

    33

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    36/51

    4.6 Discussion

    By comparing the detected modes (frequency and damping), theMatlab

    model is fairly close tothe model created with I-DEAS Test. The mathematical FRFs, MAC matrix and the mode shapesall indicate however, that the model is not very good. A reason for this could be that the modeshape estimation process is not very good.

    Another reason could be that some of the modes detected are propagating waves through thesurface. In the model used, all modes are handled as standing waves, and might cause problemsif this is the case.

    A way to enhance the model is to use the Least Squares Frequency Domain method, describedin section 1.4.2, to iteratively improve the parameters. This process has not been done, or ex-plored further. The method is referred to as hard to perform successfully, and requires a lot ofexperience to make the iteration to converge.

    There are other approaches available to update the model [5].

    34

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    37/51

    Appendix A

    I-Deas Test stability diagram

    Figure A.1: Stability diagram from I-DEAS Test

    35

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    38/51

    Appendix B

    Lists

    B.1 Target modal parameters

    PARM SHAPE FREQUENCY DAMPING AMPLITUDE PHASE MCF REF,RES

    LABEL REC (HERTZ) (%) (RAD)

    1 1 272.544 1.437 1.2707E+00 1.571 1.000 1Z-,28Z+

    2 2 335.348 0.665 5.1796E+01 1.571 1.000 1Z-,28Z+

    3 3 620.261 0.202 2.0356E+02 1.571 1.000 1Z-,28Z+

    4 4 644.843 0.090 4.9071E+01 1.571 1.000 1Z-,28Z+

    5 5 783.756 0.127 6.0770E+00 1.571 0.999 1Z-,28Z+

    6 6 962.906 0.232 1.7722E+01 1.571 1.000 1Z-,28Z+

    7 7 1221.286 0.214 5.5194E+01 1.571 1.000 1Z-,28Z+8 8 1294.681 0.214 5.1247E+02 -1.571 1.000 1Z-,28Z+

    9 9 1660.698 0.336 6.8301E-01 1.571 1.000 1Z-,28Z+

    10 10 1853.714 0.280 5 .8477E+02 - 1.571 1 .000 1Z-,28Z+

    11 11 1999.744 0.287 5 .5141E+01 - 1.571 1 .000 1Z-,28Z+

    12 12 2089.418 0.355 4 .1940E+01 - 1.571 1 .000 1Z-,28Z+

    13 13 2401.034 0.451 8 .8062E+01 - 1.571 1 .000 1Z-,28Z+

    14 14 2901.708 0.268 6 .4988E+02 1.571 1 .000 1Z-,28Z+

    15 15 3023.607 0.212 1 .5767E+02 1.571 1 .000 1Z-,28Z+

    16 16 3093.047 0.435 1 .5229E+02 1.571 1 .000 1Z-,28Z+

    36

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    39/51

    B.2 List of detected frequencies and dampings

    Pole Frequency Damping Amplitude

    (Nr) (Hz) (Rel. dB)

    01 270.458 0.2380 37.814

    02 331.448 0.7185 22.264

    03 621.403 0.1098 81.923

    04 641.900 0.0419 119.320

    05 787.377 0.0561 89.111

    06 965.849 0.1584 75.765

    07 973.848 0.1814 77.186

    08 1227.31 0.0321 124.489

    09 1296.80 0.0687 131.048

    10 1664.74 0.1872 74.787

    11 1852.21 0.1351 103.584

    12 1998.19 0.1134 97.024

    13 2090.67 0.2188 68.53314 2110.17 0.1867 69.634

    15 2373.13 0.3822 68.017

    16 2399.63 0.2222 89.982

    17 2901.55 0.0932 107.287

    18 3023.53 0.2727 54.992

    19 3094.02 0.4519 57.526

    37

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    40/51

    Appendix C

    Matlab code

    C.1 Import FRFs (impfiles.m)

    1 % I m po r t a l l f r eq u e nc y f i le s t o w o rk s p ac e2 importfile( c o n v / F R F _ C 1 _ A 1 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 1 _ A 2 . t x t );3 importfile( c o n v / F R F _ C 1 _ A 3 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 1 _ A 4 . t x t );4 importfile( c o n v / F R F _ C 1 _ A 5 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 2 _ A 1 . t x t );5 importfile( c o n v / F R F _ C 2 _ A 2 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 2 _ A 3 . t x t );6 importfile( c o n v / F R F _ C 2 _ A 4 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 2 _ A 5 . t x t );7 importfile( c o n v / F R F _ C 3 _ A 1 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 3 _ A 2 . t x t );8 importfile( c o n v / F R F _ C 3 _ A 3 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 3 _ A 4 . t x t );9 importfile( c o n v / F R F _ C 3 _ A 5 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 4 _ A 1 . t x t );10 importfile( c o n v / F R F _ C 4 _ A 2 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 4 _ A 3 . t x t );11 importfile( c o n v / F R F _ C 4 _ A 4 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 4 _ A 5 . t x t );12 importfile( c o n v / F R F _ C 5 _ A 1 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 5 _ A 2 . t x t );13 importfile( c o n v / F R F _ C 5 _ A 3 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 5 _ A 4 . t x t );

    14 importfile( c o n v / F R F _ C 5 _ A 5 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 6 _ A 1 . t x t );15 importfile( c o n v / F R F _ C 6 _ A 2 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 6 _ A 3 . t x t );16 importfile( c o n v / F R F _ C 6 _ A 4 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 6 _ A 5 . t x t );17 importfile( c o n v / F R F _ C 7 _ A 1 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 7 _ A 2 . t x t );18 importfile( c o n v / F R F _ C 7 _ A 3 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 7 _ A 4 . t x t );19 importfile( c o n v / F R F _ C 7 _ A 5 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 8 _ A 1 . t x t );20 importfile( c o n v / F R F _ C 8 _ A 2 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 8 _ A 3 . t x t );21 importfile( c o n v / F R F _ C 8 _ A 4 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 8 _ A 5 . t x t );22 importfile( c o n v / F R F _ C 9 _ A 1 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 9 _ A 2 . t x t );23 importfile( c o n v / F R F _ C 9 _ A 3 . t x t ) ; i m p o r t f i l e ( c o n v / F R F _ C 9 _ A 4 . t x t );24 importfile( c o n v / F R F _ C 9 _ A 5 . t x t );25

    26 % R e co ns t ru ct t he F RF s b y ( 1, 3 , . . . , n - 1 = R EA L ), ( 2, 4 , . . . , n = I MA G )27 F R F _ C 1 _A 1 = 1 i * ( d o w n s a m pl e ( F R F _ C 1 _A 1 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 1 _A 1 ,2 ) ;28 F R F _ C 1 _A 2 = 1 i * ( d o w n s a m pl e ( F R F _ C 1 _A 2 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 1 _A 2 ,2 ) ;29 F R F _ C 1 _A 3 = 1 i * ( d o w n s a m pl e ( F R F _ C 1 _A 3 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 1 _A 3 ,2 ) ;30 F R F _ C 1 _A 4 = 1 i * ( d o w n s a m pl e ( F R F _ C 1 _A 4 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 1 _A 4 ,2 ) ;31 F R F _ C 1 _A 5 = 1 i * ( d o w n s a m pl e ( F R F _ C 1 _A 5 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 1 _A 5 ,2 ) ;32

    F R F _ C 2 _A 1 = 1 i * ( d o w n s a m pl e ( F R F _ C 2 _A 1 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 2 _A 1 ,2 ) ;33 F R F _ C 2 _A 2 = 1 i * ( d o w n s a m pl e ( F R F _ C 2 _A 2 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 2 _A 2 ,2 ) ;34 F R F _ C 2 _A 3 = 1 i * ( d o w n s a m pl e ( F R F _ C 2 _A 3 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 2 _A 3 ,2 ) ;35 F R F _ C 2 _A 4 = 1 i * ( d o w n s a m pl e ( F R F _ C 2 _A 4 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 2 _A 4 ,2 ) ;36 F R F _ C 2 _A 5 = 1 i * ( d o w n s a m pl e ( F R F _ C 2 _A 5 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 2 _ A5 , 2) ;37 F R F _ C 3 _A 1 = 1 i * ( d o w n s a m pl e ( F R F _ C 3 _A 1 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 3 _ A1 , 2) ;38 F R F _ C 3 _A 2 = 1 i * ( d o w n s a m pl e ( F R F _ C 3 _A 2 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 3 _ A2 , 2) ;39 F R F _ C 3_ A 3 = 1 i * ( d o w n s a m pl e ( F R F _ C 3 _A 3 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 3 _ A3 , 2) ;40 F R F _ C 3_ A 4 = 1 i * ( d o w n s a m pl e ( F R F _ C 3 _A 4 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 3 _ A4 , 2) ;41 F R F _ C 3_ A 5 = 1 i * ( d o w n s a m pl e ( F R F _ C 3 _A 5 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 3 _ A5 , 2) ;42 F R F _ C 4_ A 1 = 1 i * ( d o w n s a m pl e ( F R F _ C 4 _A 1 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 4 _ A1 , 2) ;43 F R F _ C 4_ A 2 = 1 i * ( d o w n s a m pl e ( F R F _ C 4 _A 2 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 4 _ A2 , 2) ;44 F R F _ C 4_ A 3 = 1 i * ( d o w n s a m pl e ( F R F _ C 4 _A 3 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 4 _ A3 , 2) ;45 F R F _ C 4_ A 4 = 1 i * ( d o w n s a m pl e ( F R F _ C 4 _A 4 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 4 _ A4 , 2) ;46 F R F _ C 4_ A 5 = 1 i * ( d o w n s a m pl e ( F R F _ C 4 _A 5 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 4 _ A5 , 2) ;47 F R F _ C 5_ A 1 = 1 i * ( d o w n s a m pl e ( F R F _ C 5 _A 1 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 5 _ A1 , 2) ;48 F R F _ C 5_ A 2 = 1 i * ( d o w n s a m pl e ( F R F _ C 5 _A 2 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 5 _ A2 , 2) ;49 F R F _ C 5_ A 3 = 1 i * ( d o w n s a m pl e ( F R F _ C 5 _A 3 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 5 _ A3 , 2) ;50 F R F _ C 5_ A 4 = 1 i * ( d o w n s a m pl e ( F R F _ C 5 _A 4 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 5 _ A4 , 2) ;

    51 F R F _ C 5_ A 5 = 1 i * ( d o w n s a m pl e ( F R F _ C 5 _A 5 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 5 _A 5 ,2 ) ;52 F R F _ C 6_ A 1 = 1 i * ( d o w n s a m pl e ( F R F _ C 6 _A 1 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 6 _A 1 ,2 ) ;

    38

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    41/51

    53 F R F _ C 6 _A 2 = 1 i * ( d o w n s a m pl e ( F R F _ C 6 _A 2 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 6 _A 2 ,2 ) ;54 F R F _ C 6 _A 3 = 1 i * ( d o w n s a m pl e ( F R F _ C 6 _A 3 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 6 _A 3 ,2 ) ;55 F R F _ C 6 _A 4 = 1 i * ( d o w n s a m pl e ( F R F _ C 6 _A 4 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 6 _A 4 ,2 ) ;56 F R F _ C 6 _A 5 = 1 i * ( d o w n s a m pl e ( F R F _ C 6 _A 5 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 6 _A 5 ,2 ) ;

    57 F R F _ C 7 _A 1 = 1 i * ( d o w n s a m pl e ( F R F _ C 7 _A 1 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 7 _A 1 ,2 ) ;58 F R F _ C 7 _A 2 = 1 i * ( d o w n s a m pl e ( F R F _ C 7 _A 2 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 7 _A 2 ,2 ) ;59 F R F _ C 7 _A 3 = 1 i * ( d o w n s a m pl e ( F R F _ C 7 _A 3 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 7 _A 3 ,2 ) ;60 F R F _ C 7 _A 4 = 1 i * ( d o w n s a m pl e ( F R F _ C 7 _A 4 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 7 _A 4 ,2 ) ;61 F R F _ C 7 _A 5 = 1 i * ( d o w n s a m pl e ( F R F _ C 7 _A 5 ,2 , 1 ) ) + d o w n s a m pl e ( F R F _ C 7 _A 5 ,2 ) ;62 F R F _ C 8 _A 1 = 1 i * ( d o w n s a m pl e ( F R F _ C 8 _A 1 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 8 _ A1 , 2) ;63 F R F _ C 8 _A 2 = 1 i * ( d o w n s a m pl e ( F R F _ C 8 _A 2 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 8 _ A2 , 2) ;64 F R F _ C 8 _A 3 = 1 i * ( d o w n s a m pl e ( F R F _ C 8 _A 3 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 8 _ A3 , 2) ;65 F R F _ C 8_ A 4 = 1 i * ( d o w n s a m pl e ( F R F _ C 8 _A 4 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 8 _ A4 , 2) ;66 F R F _ C 8_ A 5 = 1 i * ( d o w n s a m pl e ( F R F _ C 8 _A 5 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 8 _ A5 , 2) ;67 F R F _ C 9_ A 1 = 1 i * ( d o w n s a m pl e ( F R F _ C 9 _A 1 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 9 _ A1 , 2) ;68 F R F _ C 9_ A 2 = 1 i * ( d o w n s a m pl e ( F R F _ C 9 _A 2 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 9 _ A2 , 2) ;69 F R F _ C 9_ A 3 = 1 i * ( d o w n s a m pl e ( F R F _ C 9 _A 3 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 9 _ A3 , 2) ;70 F R F _ C 9_ A 4 = 1 i * ( d o w n s a m pl e ( F R F _ C 9 _A 4 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 9 _ A4 , 2) ;71 F R F _ C 9_ A 5 = 1 i * ( d o w n s a m pl e ( F R F _ C 9 _A 5 ,2 , 1 ) ) + d o w n s a m p le ( F R F _ C 9 _ A5 , 2) ;72

    73 % C o n st r uc t F R F M a t r ix H74 H = [ .. .75 FRF_C1_A1 FRF_C1_A2 FRF_C1_A3 FRF_C1_A4 FRF_C1_A5 .. .76 FRF_C2_A1 FRF_C2_A2 FRF_C2_A3 FRF_C2_A4 FRF_C2_A5 .. .77 FRF_C3_A1 FRF_C3_A2 FRF_C3_A3 FRF_C3_A4 FRF_C3_A5 .. .78 FRF_C4_A1 FRF_C4_A2 FRF_C4_A3 FRF_C4_A4 FRF_C4_A5 .. .79 FRF_C5_A1 FRF_C5_A2 FRF_C5_A3 FRF_C5_A4 FRF_C5_A5 .. .80 FRF_C6_A1 FRF_C6_A2 FRF_C6_A3 FRF_C6_A4 FRF_C6_A5 .. .81 FRF_C7_A1 FRF_C7_A2 FRF_C7_A3 FRF_C7_A4 FRF_C7_A5 .. .82 FRF_C8_A1 FRF_C8_A2 FRF_C8_A3 FRF_C8_A4 FRF_C8_A5 .. .83 FRF_C9_A1 FRF_C9_A2 FRF_C9_A3 FRF_C9_A4 FRF_C9_A5 .. .84 ];85

    86 % H = H ( 1: r ou nd ( e nd / 3) , :) ; % U se o nl y u p t o a pp ro x 1 k Hz87

    88 % S e t u p t h e f r e q ue n c y s c a l e89 t = 1: length( F R F _ C 1 _ A 1 ) ;90 t = t / t (end) * 3 2 0 0 ;91 % t = d o w ns a m pl e ( t , 2) ;92 % t = t ( 1: r ou nd ( e nd / 3) ) ; % U se o nl y u p t o a pp ro x 1 k Hz

    39

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    42/51

    C.2 Process FRFs (process frfs.m)

    1 % = = = = = = = = = = = = = = = = = = = = = = == = = = = = = = = = = = = = == = = = = = = = = = = = =2 % MODAL FRE QUENCY AND DAMPING RATIO3 % = = = = = = = = = = = = = = = = = = = = = = == = = = = = = = = = = = = = == = = = = = = = = = = = =4

    5 format long6

    7 % S t ar t c l oc k8 t i c ;9

    10 c lo s e a l l ;11 clear y;12 clear y2 ;13

    14 a n im a t io n = 0 ; % S et t o 1 t o a ni ma te m od e s ha pe s15

    16 % C r ea t e C o m p l ex M o de I n di c a to r F u nc t io n f r om a ll F R Fs i n H17 h = c mi f (H ) ;18

    19 % S m oo t he n t h e C M I F b y a 2 5 s a m p l e l o n g R u n n i ng A v er a ge f i lt e r

    20 h s = s m oo t h (h , 2 5) ;21

    22 % F i n d F r e q u e n c i e s i t e r a t i v e l y23 [ f r e qz , f _ i , f _ l , f _ a ] = f i n d _ f re q s ( h s , t ) ;24

    25 % F i nd D a mp i ng r a ti o s i t e ra t iv e l y26 [ d , f _ i , d _ i1 , d _ i 2 ] = f i n d _ d am p s ( h s , t , f _ l ) ;27 d _e rr = 5 ; % D a mp i ng e rr or t h re s h ol d28

    29 % P r in t o u t t h e f i nd i ng s30 f p r i n t f ( Nr Fre quency Damping Ampl itude \n );31 lf = length( f r e q z ) ; % W il l b e u pd at ed i n l oo p32 k = 1; % R e se t l o op c o un t er33 while k d _ er r % I f w r on g p o le ( d a m pi n g f a ct o r u n r ea l i st i c )35 % R e mo v e t h i s p o le f r om l i st36 d ( k) = [ ];37 f r eq z ( k ) = [ ] ;38 f _a ( k ) = [ ] ;

    39 f _l ( k ) = [ ] ;40 lf = length( f r e q z ) ; % U p da t e l e ng t h41 e lse

    42 f p r i n t f ( %02 g % 7 g % 7 g % 7g \ n , k , f r e q z ( k ) , d ( k ) , f _ a ( k ) ) ;43 k = k + 1;44 end;45 end;46

    47 % P l ot48 figure ;49 plot ( t , h s , b - ); % P lo t C MI F50 hold on ;51 stem( t , f _ i , r - . ); % P l ot P o le I n d ic a to r L i ne s52 stem( t , d _ i 1 , g - - . ); % P l ot L e ft D a mp i ng L i mi t I n d ic a to r53 stem( t , d _ i 2 , g - - . ); % P l ot R i gh t D a mp i ng L i mi t I n di c a to r54 hold of f ;55 t i t l e ( C o mp l e x M o de I n di c at o r F u n ct i on ( C M IF ) o f F RF M a tr i x [ H ] );56 xlabel( F r e q u e n cy ( H z ) );57 ylabel( A m p l i t u de ( d B ) );58 axis tight;59

    60 o u t f i le = . . / r e p o r t / f i g u r e s / c m i f ;61 print ( - d e p s c , o u t f i l e )62

    63

    64

    65

    66

    67

    68

    69

    70

    71

    72

    73

    74

    75

    7677

    40

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    43/51

    78 % = = = = = = = = = = = = = = = = = = = = = = == = = = = = = = = = = = = = == = = = = = = = = = = = =79 % MODE SHAPE VECTORS80 % = = = = = = = = = = = = = = = = = = = = = = == = = = = = = = = = = = = = == = = = = = = = = = = = =81

    82 % S t ep t h ro u gh a l l d e t e c te d p o le s83 fo r k = 1: length( f _ l )84 u = 1;85 % C y cl e t h ro u gh a l l m e a su r em e n t p o s it i on s86 fo r j = 1: 987 fo r l = 1 :588 % S a ve t h e m o d e s h a p e i n f o r ma t i on t o P s i89 P s i ( k , j , l ) = H ( f _ l ( k ) , u ) ;90 u = u + 1 ;91 end;92 end;93 i f( a ni ma t io n = = 1 )94 % A n im a te t he m o de ( e x p or t t o . a v i - f il e a n d s a v e i m a g e )95 a n i m a t e _ m o d e (imag( P s i ( k , : , : ) ) , 4 , f r e q z ( k ) ,sp rin t f ( a n i m s / r e a l / t e s t % g . a v i ,k));96 a n i m a t e _ m o d e (imag( P s i ( k , : , : ) ) , 3 . 5 , f r e q z ( k ) ) ;97 o ut fi le = [ . . / r e p o r t / f i g u r e s / m o d e s h a p e _ num2str(round( f r e q z ( k ) ) ) ] ;98 print ( - d e p s c , o u t f i l e )99 end;100 end

    101

    102 % = = = = = = = = = = = = = = = = = = = = = = == = = = = = = = = = = = = = == = = = = = = = = = = = =103 % M A TH EM A TI C AL F RE QU E NC Y R ES PO NS E F UN CT I ON S104 % = = = = = = = = = = = = = = = = = = = = = = == = = = = = = = = = = = = = == = = = = = = = = = = = =105

    106 y = zeros(length( f r e q z ) , 4 5 ,length( t ) ) ;107

    108 fo r k = 1: length( f r e q z )109 R e s i d = s q u e e ze ( P s i ( k , : , : ) ) ;110 u = 1;111 fo r m = 1: 9112 fo r n = 1 :5113 % C a l cu l at e t h e m o d e s h ap e s c al a r f or t h e s p e c i fi c p o si t io n114 P s i L = R e s i d ( : , n ) * R e si d ( m , : ) ;115 % C r ea t e t h e S D OF m o da l m o de l f or t h is p o si t io n a n d m o d e116 y ( k , u , : ) = s d o f ( t , f r eq z ( k ) , 2 0 * d ( k ) , P s i ( k ,m , n ) , 1 / f _ a ( k ) ) ;117 u = u +1 ;118 end

    119 end120 end

    121

    122 y s = s q ue e ze (sum( y , 1 ) ) ; % S D OF s u mm e d123

    124 % S o me h ow t he r e al p a rt o f t h e g e n er a te d f r f i s t h e i m a g in a r y ..125 % L et s f l ip t h em . .126 fo r k = 1 :4 5127 y s _ t e mp ( k , : ) = re al ( y s ( k , : ) ) ;128 y s (k , : ) = imag( y s ( k , : ) ) ;129 y s (k , : ) = y s (k , : ) - 1 i . * y s_ t em p ( k ,: ) ;130 end;131

    132 % = = = = = = = = = = = = = = = = = = = = = = == = = = = = = = = = = = = = == = = = = = = = = = = = =133 % MATH EMA TICA L MODE SHAPE VECTORS134 % = = = = = = = = = = = = = = = = = = = = = = == = = = = = = = = = = = = = == = = = = = = = = = = = =135 fo r k = 1: length( f _ l )136 u = 1;137 fo r j = 1: 9

    138 fo r l = 1 :5139 P s i _ M at h ( k , j , l ) = y s ( u , f _ l ( k ) ) ;140 P s iG ( k , u ) = P s i (k , j , l) ; % I n l on g v ec to r f or m141 u = u + 1 ;142 end;143 end;144 i f( a ni ma t io n = = 1 )145 a n i m a t e _ m o d e (imag( P s i _ M a t h ( k , : , : ) ) , 4 , f r e q z ( k ) ,.. .146 sp rin t f ( a n i m s / m a t h / t e s t _ m a t h % g . a v i , k ) ) ;147 a n i m a t e _ m o d e (imag( P s i _ M a t h ( k , : , : ) ) , 3 . 5 , f r e q z ( k ) ) ;148 o ut fi le = [ . . / r e p o r t / f i g u r e s / m o d e s h a p e _ m a t h _ num2str(round( f r e q z ( k ) ) ) ] ;149 print ( - d e p s c , o u t f i l e )150 end;151 end

    152

    153

    154

    155

    156

    157

    41

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    44/51

    158 % = = = = = = = = = = = = = = = = = = = = = = == = = = = = = = = = = = = = == = = = = = = = = = = = =159 % PLOTTING160 % = = = = = = = = = = = = = = = = = = = = = = == = = = = = = = = = = = = = == = = = = = = = = = = = =161

    162 % C r ea t e t h e C M IF f o r t h e g e ne r at e d F RF m a tr i x y s ( t h e s u mm e d s d o f s )163 figure ;164 y c = c m if ( y s );165

    166 % S m oo t he n t h e C M I F b y a 2 5 s a m p l e l o n g R u n n i ng A v er a ge f i lt e r167 y c m i f = s m o o t h ( yc , 2 5 ) ;168

    169 % P lo t t he C MI F170 plot ( t , y c m i f . /max(abs( y c m i f ) ) . *max(abs( h s ) ) , t , h s ) ;171 t i t l e ( R e c on s t ru c t ed F R F C M I F f r om S D OF D a ta );172 xlabel( F r e q u e n cy ( H z ) );173 ylabel( A m p l i t u d e );174 axis tight;175 legend ( R e c o n s t r u c t e d , C M I F , L o c a t i o n , B e s t );176 o ut fi le = [ . . / r e p o r t / f i g u r e s / c m i f _ g e n e r a t e d ];177 % p r i n t ( - d e p s c , o u t f i l e )178

    179 % P l ot a ll g e n er a te d F R Fs a g ai n st t h e m e as u re d F RF s , i n c lu d i ng p h as e r e sp o ns e180 l = 1;181 fo r m = 1 : 9182 fo r n = 1: 5183 clear y184 y = s q ue e ze ( y s ( l ,: ) ) ;185

    186 figure (5);187 subplot ( 2 , 1 , 1 ) ;188 plot (t ,angle( y ) , b - ,t ,angle ( H ( : , l ) ) , r - ); % . * 2 + p i189 t i t l e ([ C o m pa r i so n R e c on s t ru c t ed F RF a n d R e a l F R F ( num2str(m ) , num2str(n ) ) ]) ;190 xlabel( F r e q u e n cy ( H z ) );191 ylabel( P h a s e [ r a d ] );192 axis tight;193 legend( M a t h , R e a l , L o c a t i o n , B e s t );194

    195 subplot ( 2 , 1 , 2 ) ;196 plot (t , real ( y ) . /max(abs( real ( y ) ) ) . *max(abs ( re al ( H ( : , l ) ) ) ) , b - ,t , re al ( H ( : , l ) ) , r - );197 t i t l e ([ C o m pa r i so n R e c on s t ru c t ed F RF a n d R e a l F R F ( num2str(m ) , num2str(n ) ) ]) ;198 xlabel( F r e q u e n cy ( H z ) );

    199 ylabel( M a g n i t u de [ d B ( m / s ^ 2 / N ) ] );200 axis tight;201 legend( M a t h , R e a l , L o c a t i o n , B e s t );202 o ut fi le = [ . . / r e p o r t / f i g u r e s / f r f _ g e n e r a t e d _ num2str( l ) ] ;203 % p r i n t ( - d e p s c , o u t f i l e )204 l = l + 1;205 end

    206 end

    207

    208 % G e ne r at e M A C M a t r i x209 M A C = m a c (P si G , f r eq z ) ;210

    211 % H ow m uc h t im e i s e la ps ed ?212 toc ;

    42

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    45/51

    C.3 Complex Mode Indicator Function (cmif.m)

    1 function h = c mi f (H )2 % F u nc t io n t o c o mp u te t h e C o mp l ex M o de I n di c at o r F u n ct i on ( C M IF )3 %4 % h = c mi f (H ) ;5 %6 % I t w i l l c o mp u te f o r t h e F RF m a tr i x H , w i th d i m en s i on s7 % n u m b e r _ o f_ l i n e x x n u m b e r _ o f _r e s p o n s e s x n u m b e r _ o f _r e f e r e n c e s8 %9

    10 % E x tr a ct t he r e s ol u t io n o f t h e g i ve n F R F m a t r ix11 n u m b e r _ o f_ l i n e s = siz e ( H , 1 ) ;12 n u m b e r _ o f _r e s p o n s e s = siz e ( H , 2 ) ;13 n u m b e r _ o f _r e f e r e n c e s = siz e ( H , 3 ) ;14

    15 % A l lo c at e m e mo r y f o r t h e r e s u lt16 h = zeros(number_of_references ,number_of_lines);17

    18 % S t ep t h ro u gh a l l v a l u es19 fo r i = 1 : n u mb e r _o f _ li n e s

    20 % C a lc u la t e t he S i ng u l ar V a lu e D e co m p os i ti o n o f t h e s q ue e ze d m a tr i x H21 % S q ue e ze r e mo v es s i ng l et o n d i m en s i on s ( i f o n l y m x n m a tr i x , k d i m r e m . )22 h ( : , i) = svd(abs( s q u e e z e ( H ( i , : , : ) ) ) ) ;23 end

    43

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    46/51

    C.4 Find frequencies (find freqs.m)

    1 function [ f , f _i , f _ l , f _ a ] = f i n d _ f r eq s ( H , F ) ;2 %3 % F u nc t io n t o i t e r a t iv e l y f i n d p o le s o f a s a mp l ed f r f .4 % S y nt a x :5 % [ f] = f in d_ fr eq s (H ,F )6 % w h er e :7 % f : O u tp ut v e ct or c o nt a in in g t he f o un d f re qe nc y p ol es8 % f _i : P ol e i nd ic at o r li ne s o ut pu t9 % f _l : P o le i n di ca to r l i ne s s a mp le p os it i on o ut pu t10 % f _a : A mp li tu de o f th e po le11 %12 % H : Input FRF data13 % F : In pu t Fr eq ue nc y li ne da ta14 %15

    16 k = 1;17 u = 1;18

    19 % P o le i n di c at o r l i ne s f or p l ot

    20 f _i = zeros(length( H ) , 1 ) ;21

    22 fo r i = 1 + u : length( H ) - u23 i f H ( i+ u ) < H ( i)24 i f H ( i- u ) < H ( i)25 f ( k) = F ( i) ;26 f _ i (i ) = max( H ) ;27 f _l ( k) = i ;28 f _ a (k ) = H ( i ) ;29 k = k + 1 ;30 end

    31 end

    32 end

    44

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    47/51

    C.5 Find dampings (find damps.m)

    1 function [ d , f _i , d _ i 1 , d _ i 2 ] = f i n d _ d am p s ( H , F , F P )2 %3 % F u nc t io n t o i t e r a t iv e l y f i n d d a mp i ng a nd p o le s o f a s a mp l ed f r f .4 % S y nt a x :5 % [ d , f_ i , d_ i1 , d _ i2 ] = f i nd _d a mp s (H , F)6 % w h er e :7 % d : O ut pu t ve ct or c on ta in in g th e fo un d da mp in gs8 % f _i : P ol e i nd ic at or l in es o u tp ut9 % d _i 1: L ef t fr eq l in e fo r po le10 % d _i 2: R ig ht f re q li ne f or p ol e11 %12 % H : Input FRF data13 % F : Input Frequen cy line data14 % FP : I npu t P ol e L oc at io n da ta15

    16 p = 1;17 t hr = 1 0;18

    19 % P o le i n di c at o r l i ne s f or p l ot

    20 f_i = zeros(length( H ) , 1 ) ;21 d _ i1 = zeros(length( H ) , 1 ) ;22 d _ i2 = zeros(length( H ) , 1 ) ;23

    24 % S t ep t h ro u gh a l l d e te c te d f r e qu e n ci e s25 fo r i = 1 : length( F P )26 % R e se t p a ra m e te r s27 k = F P( i ); % S t ar t s a mp l e p o si t i on f or t h is f r eq u en c y28 d 1 = 0 ; % F r e qu e nc y f o r d a m p i ng l i mi t l e ft29 d 2 = 0 ; % F r e qu e nc y f o r d a mp i ng l i mi t r i gh t30 f _ i (k ) = max( H ) ;31 while ( (H ( k) > H ( FP ( i) ) - th r ) & & ( (H ( k) ) > 0 ) & & ( k > 1 ) ) % F i nd l i mi t s32 k = k - 1 ;33 end;34 k 1 = k ; % S a ve l e ft l i mi t s a mp l e p o si t io n35 d 1 = F ( k) ; % S a ve f r eq u e nc y o f t h is p o si t io n36

    37 k = F P( i );38 while ( (H ( k) > H ( FP ( i) ) - th r ) & & ( (H ( k) ) > 0 ) & & ( k < length( H ) - 1 ) )

    39 k = k + 1 ;40 end;41 k 2 = k ;42 d 2 = F ( k) ;43

    44 % T ry t o c or re ct i f t he re a re m an y p ol es c lo se t o e ac h o th er b y45 % a p pr o x im a ti n g t h e c o r r e ct d a mp i ng . T h is i s d o ne b y u s in g t h e46 % p o le a s c e n t e r l o c a ti o n a n d a d ju s t in g t h e l i m it t h at i s f a r a w ay47 % t o b e t he s a me l e ng t h f r om t h e p o l e a s t he c or r ec t l i mi t .48 i f( abs( F P ( i) - k 2 ) > abs( F P ( i) - k 1 ) )49 d 2 = F ( FP ( i) + abs( F P ( i ) - k 1 ) ) ;50 d _ i 2 ( F P ( i ) + abs( F P ( i) - k 1 )) = max( H ) ;51 k 2 = F P( i) + abs( F P ( i ) - k 1 ) ;52 e lse

    53 d _ i2 ( k 2 ) = max( H ) ;54 end

    55 i f( abs( F P ( i) - k 1 ) > abs( F P ( i) - k 2 ) )56 d 1 = F ( FP ( i) - abs( k 2 - F P ( i ) ) ) ;57 d _ i1 ( F P ( i) - abs( F P ( i) - k 2 )) = max( H ) ;58 k 1 = F P( i) - abs( F P ( i ) - k 2 ) ;59 e lse

    60 d _ i1 ( k 1 ) = max( H ) ;61 end

    62

    63 % C a lc u la t e d a mp i ng r a ti o f o r t h e s p ec i fi e d p o le64 d ( p ) = ( d 2 - d 1 ) /H ( F P( i ) ) ;65

    66 % G o t o n ex t p ol e67 p = p +1 ;68

    69 end

    45

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    48/51

    C.6 SDOF modal model (sdof.m)

    1 function r et = s do f (w , w d , s , P si , Q )2 %3 % F u nc t io n t o c r e a t e a S D OF m o da l m o de l .4 % S y nt a x :5 % ret = s do f( w, wd , s, P si , Q )6 % w h er e :7 % w : Freque ncy vector8 % w d : D am pe d n at ur al r e so na nc e f re qu en cy9 % s : Damping factor10 % Psi : M od e sha pe v alu e11 % Q : Scaling factor12

    13 % C a l c u l at e r e s i d u al s14 A = Q* Psi ;15 Ac = conj ( Q ) *conj( P s i ) ;16

    17 L = s + i* wd ;18 L c = s - i * w d ;19

    20 fo r k = 1: length(w )21 r et ( k) = ( A /( i *w ( k) - L ) ) + ( A c/ ( i* w (k ) - L c) ) ;22 end;

    46

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    49/51

    C.7 MAC matrix (mac.m)

    1 function M A C = m a c ( P si , f r e q z , R E S )2 %3 % M o da l A s s ur a nc e C r i te r eo n M a tr i x F u nc t io n4 % T ak es a m od al s ha pe m at ri x i n f or m :5 % P si = [ F re qu en cy , M o d e S h ap e V e c to r ]6 % O p ti o na l p a r am e te r f re q z , c o n ta i n in g t he f r e qu e nc y i n f or m a ti o n f o r a x is .7 % O p ti o na l p a r am e te r R ES , f or p l o tt i ng r e so l u ti o n ( i f < = 0 , n o p l o t i s g e n e r at e d )8 %9 % S y nt a x :10 % M AC = m ac ( Ps i , fr eq z , RE S) ;11 %12

    13 % F i nd o ut h ow m a ny f r eq u e nc i e s14 F r eq s = siz e ( P s i , 1 ) ;15

    16 % I f t h e o p ti o na l f r eq z a r gu m e nt i s n o t e x is t a nt17 if n argin < 218 f r eq z = 1 : F re q s ;19 end;

    2021 if n argin < 322 R ES = 1 00 ; % S t an d a rd r e s ol u t io n f o r p l o t23 end;24

    25 % A l lo c at e m e mo r y26 M AC = zeros( F r e q s ) ;27

    28 % C o mp a re e a ch f r eq u e nc y29 fo r u = 1 : Fr eq s30 fo r l = 1 : Fr eq s31 P s i R e d1 = s q u e e ze ( P s i ( u , : ) ) ; % C r ea t e t e mo r ar y v e ct o rs32 P s i R e d2 = s q u e e ze ( P s i ( l , : ) ) ;33 % C a lc u l at e M A C v a l u e34 M A C ( u , l ) = ( P s i R e d 1 * P s iR e d 2 ) ^ 2 / ( ( P s i R e d1 * P s i R e d1 ) * ( P s i R e d 2 * P s iR e d 2 ) ) ;35 end

    36 end

    37

    38 M AC = abs( M A C ) ;

    3940 i f( R ES > 0 )41 % F or a n ic er s te m 3 d p lo t .. .42 M A C _ P LO T = zeros( F r e q s * R E S ) ;43 fo r u = 0 : F r eq s - 144 fo r l = 0 : F r eq s - 145 M A C _ P LO T ( u * R E S + 1 : ( u + 1 ) * RE S - ( R E S / 2 ) , l * R E S + 1 : ( l + 1) * R E S - ( R E S / 2 ) ) = M A C ( u + 1 , l + 1) ;46 end;47 end;48

    49 f igu re;50 mesh( M A C _ P L O T ) ;51 axis square;52 axis tight;53 se t (gca , X T i c k ,round( R E S / 2 ) : R E S : F r e q s * R E S +round( R E S / 2 ) )54 se t (gca , Y T i c k ,round( R E S / 2 ) : R E S : F r e q s * R E S +round( R E S / 2 ) )55 se t (gca , X T i c k L a b e l , f r e q z )56 se t (gca , Y T i c k L a b e l , f r e q z )57 colormap(hsv)58 view(57,43);59 light( P o s i t i o n ,[8,-8,20])60 l i g h t in g g o u r a ud61

    62 t i t l e ( M o da l A s s ur a nc e C r it e r eo n ( M AC ) M a tr i x );63 xlabel( F r e q u e n cy ( H z ) );64 ylabel( F r e q u e n cy ( H z ) );65 zlabel ( V a l u e );66 o ut fi le = [ . . / r e p o r t / f i g u r e s / m a c _ g e n e r a t e d ];67 print ( - d e p s c , o u t f i l e )68

    69 end;

    47

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    50/51

    C.8 Animate mode (animate mode.m)

    1 function y = a n i ma t e _m o d e ( Ps i , M ax t im e , F re q , N a me )2 %3 % F u nc t io n t o a n i m a te a m o de s ha pe , g i ve n i t s m a tr i x f o r m4 % S y nt a x :5 % y = a n im a te _ mo de ( Ps i , M a xt im e , F re q , N a m e );6 % w h er e :7 % P si : Th e mo de s ha pe m at ri x in t he s iz e of t he m ea su re me nt g ri d8 % M ax ti me : An im at i on ti me9 % F re q : F re qu en cy t o a ni ma te ( j us t fo r pr in ti ng t he t it le )10 % N am e : Fo r e xp or ti ng to .avi - fi le11 %12

    13 if n argin < 314 F r eq = NaN;15 end;16

    17 if n argin < 418 e = 0;19 N a me = N a N ;

    20 e lse21 a v i o b j = a v i f i le ( N a m e , f p s ,30);22 e = 1;23 end

    24

    25 % R e m o v e s i n g l e t o n d i m e n s i o n s26 P s i = s q ue e ze ( P s i );27 % N o r m a l iz e28 P si = P si . /max(max(abs(abs ( P s i ) ) ) ) ;29

    30 la = length(- pi : . 1 : M a x t i m e *pi );31

    32 k = 0;33 f p r i n t f ([ sp rin t f ( E x p or t in g M o de A n im a ti o n % 6 g H z t o , F r e q ) N a m e . ]) ;34 fo r x= - pi : . 1 : M a x t i m e *pi35 f igu re (1 )36 surf(Psi.* si n ( x ) ) ;37 colormap hot;38 axis equal;

    39 z l i m ( [ - 1 1 ] ) ;40 t i t l e ( sp rin t f ( M od e A n im a t io n : % 6 g H z , F r e q ) ) ;41 xlabel( N o de s X );42 ylabel( N o de s Y );43 i f e = = 144 f r am e = getframe(gca);45 a v i o b j = a d d f r am e ( a v i o bj , f r a m e ) ;46 end;47 k = k +1 ;48 i f ( m o d (l a , k / la ) = = 0 )49 f p r i n t f ( . );50 end;51 end

    52

    53 f p r i n t f ( \ n \ n );54

    55 i f e = = 156 a v io b j = close ( a v i o b j ) ;57 end;58

    59 % D on t r e tu r n a n yt h in g u s ef u l60 y = 0;

    48

  • 8/6/2019 56813979 Modal Analysis Report Main With MATLAB

    51/51

    Bibliography

    [1] O. Dossing and B. & Kjaer, Structural testing, April 1988.

    [2] P. Sas and W. Heylen, eds., Theory and Practice, International Seminar on Modal Analysis,(Heverlee), September 1993.

    [3] Vibration. http://en.wikipedia.org/wiki/Vibration, October 2009.

    [4] P. Avitabile, Modal space - in our little world (sem experimental techniques), August 1999.

    [5] R. J. Allemang and D. L. Brown, A Unified Matrix Polynomial Approach to Modal Identi-fication. Journal of Sound and Vibration 211(3), 301-322: Academic Press Limited, March1998.