You are on page 1of 23

REPORT

DYNAMIC TEST 1

Friday, 28 March 2008


Full scale Building
489 Drummond Street, Carlton
INTRODUCTION

Location of Accelerometer

1036 1038
South 1039 1037 North

This Project is the collaboration between Swinburne University of Technology and University of Melbourne.
The ambient test was held in the site on 28 March 2008. The analyses are divided into two parts. First part
shows the identification of natural frequencies using manual method. And the second part investigates all
modal parameters (natural frequency, modal shape and damping ratio) by using ARTeMIS program. Both
analyses show high similarity in the result.

2
PART I. BASIC FREQUENCY DOMAIN / PEAK PICKING TECHNIQUE

The technique used in this paper is the classical frequency domain approach often referred to as the basic
frequency domain technique, or the peak picking technique. Analysis has been done using DAQ Program to
calculate Fourier Fast Transformation.

The FRF is the result of taking an FFT of the time domain data. It is frequently interchanged by calling the
FRF the FFT.

The FRF is calculated from the time domain data by calculating the cross and auto power spectral densities
of the input and output time domain data (this involves taking the FFT of the time domain data). Assuming
that x and y are the input and output signals respectively the H1 and H2 FRF estimators are calculated as: H 1
= Sxy/Syy and H2 = Sxx/Syx. S denotes the auto (Sxx and Syy) and cross (Sxy and Syx) power spectral density of the
input and output signals.

The large peaks on the FRF will be the modes. These peaks also contain contributions of all other modes in
your structure. How much this contribution is depends on the structure under consideration. Generally, if the
peaks in the FRF are not too close then the actual mode shape can be estimated reasonably well directly from
the amplitude and phase of the FRF. The difficulty is determining when peaks are too close.

Impact Testing
The selection of impact force takes important role in finding good measurement. On laboratory model, the
selection of hammer tip will affect the measurement of data. If the tip is too soft, then all modes will not be
excited sufficiently to get comprehensive measurement. The harder the tip, the wider of frequency
measurement we can get. In full scale field test, of course, it’s harder to get non destructive hammer. Wind
force can be used to excite the building.

Indicator Function.
To indicate the natural frequencies and modes, and also to reduce the data, there are some different mode
Indicator Functions (MIF) that frequently used in experimental modal analysis. Some of them:

SUM : summation function


MIF : mode indicator functions
MMIF : multivariate MIF
CMIF : complex mode indicator function
Stability diagram

In this project, SUM is used as a tool to formulate the data. This is a very simple formulation. Basically, it is
the sum of all of the FRFs measured. The SUM will reach a peak in the region of a mode of the system. The
idea is that if all the FRFs are considered, then all of the modes will be seen in the majority of the
measurements. As more and more FRFs are included, there is a greater chance that all of the modes will be
seen in the collection of FRFs summed together. This is obviously better than one particular measurement
where all the modes may not be present. The SUM function will identify modes reasonably well especially if
the modes are well separated. In the figure, there are five peaks observed, which indicates that there are at
least five modes in the frequency band shown. Another important feature of the SUM function is that each of
the peaks is generally fairly wide and if closely spaced modes exist, then this may not show the entire modes
well. While the SUM function is useful, it is not always very clear when modes are closely spaced.

The archetype MIF can solve the problem of closely spaced modes. The real part of FRF is divided by the
magnitude of FRF, so the MIF usually have more abrupt change across mode. But if the measurements have
some distortion, then these tools may not be able to draw the modes.

3
Minimizing Leakage Error (Choosing of Window)

In order to minimize leakage error in converting from time based to frequency based analysis, the use of
windows as weighting function is needed. This weighting function is applied to force the data to better
satisfy the periodicity requirements of the Fourier Transform process. Even though in the other hand, it can
distort the data itself, particularly in the degradation of peak amplitude of the function and addition of more
damping than it in actual condition. Fortunately, the consideration of this project is not in finding the
amplitude but find natural frequency and mode shape of the building, so we can avoid this consideration.
The most common windows are the exponentially decay, Uniform, Hanning and Flat Top windows. The
exponentially decay windows is suitable for impact testing. Uniform window, which is also denoted as No
Window, is used if the measurement data can satisfy the periodicity requirement of the FFT process. As an
example, both of excitation signal of shaker testing, burst random excitation and sine chirp excitation, can
fulfill this purposes. Hanning windows is used on random excitations and general field signal category. Flat
Top is applied on sinusoidal signal that do not fulfill the periodicity requirement of FFT process.
In this project, No Window is used to minimize the distortion. Beside that, to reduce leakage error and also
the using of window, the bandwidth can be narrowed or the number of time resolution can be increased. Two
thousand resolutions within twenty second total time is used in the measurement test.

Averaging Data

There are some averaging mode to improve signal-to-noise ratio :


Vector : minimizes noise from synchronous signals
RMS : reduce signal fluctuations by finding the mean value of the absolute values of the data at each
time or frequency
Peak Hold : only the peak value not arithmetic mean
Weighting modes :
Linear : find the mean value in a set of numbers.
Exponential : used for time varying characteristic, and not used in data used for experimental modal
analysis.
To average data, RMS Averaging mode and Linear Weighting mode is chosen in High Speed DAQ VI
program.

Prediction of Movement based on the trend of PSD Graphs


(+) (-) (-)
(+)
1039 1039 1037
1037 1039 + 1037 = max
1039 – 1037 = min
Both in the same sign

(-) (+) (+) (-)


1036 1038 1036 1038
1036 + 1038 = min
1036 – 1038 = max
Both in the same sign

1039 (+) (-) 1037


1038
1039 – 1037 = max
1036
(-) (+) 1039 + 1037 = min
Both in the same sign

1036 – 1038 = max


1036 + 1038 = min
Both in the same sign

4
FFT Spectrum as the Result of Fast Fourier Transform :

Picture 1. Fast Fourier Transforms Spectrum of East-West Excitation

Picture 2. Fast Fourier Transforms Spectrum of North-South Excitation

5
ANALYSIS OF NATURAL FREQUENCY

A. Accelerometer on Strong Direction


1. Excitation on the strong direction

South 1039 1037 North


Excitation
Stair 2 Stair 1

Characteristic of movement of building due to the excitation on stair 1 and 2 can be categorized as:

(a) First Mode : displacement (b) Further mode: rotation

Mode (a) describes displacement movement that can be found by summation of 1039 + 1037. Mode (b) that
describes the rotation can be found by: 1039 – 1037.
If both transducers face the same direction, the positive sign of amplitude they record will indicate that they
have same direction of movement. In contrary, if both transducers face the opposite direction, then the
positive magnitude sign of measurement point the opposite direction of movement. So, first frequency can be
found by adding two transducers (that located in the same direction) to get the biggest amplitude or by
subtract one to another if they placed in the opposite direction.
With the same rules, rotation of plane which shows the reverse behavior from first mode can be determined.
Two transducers with the same direction will record the opposite sign of measurement due to the same
rotation. And the opposite direction of transducer will read same sign of amplitude.

Picture 3. FFT Spectrums of the summation of both accelerometer 1039 and 1037 due to EW excitation

6
Picture 4. FFT Spectrums of both accelerometer 1039 and 1037 due to EW excitation

It is obvious from picture 3, that the pattern of summation graph (1039+1037) differs from that of subtraction
graph (1039-1037). On summation graph, the biggest amplitude is at frequency 2.6 Hz, while on subtraction
graph both of 2.9 Hz and 3.1 Hz have the biggest magnitude. At frequency 2.6Hz, we can draw that when we
add amplitude of 1039 and 1037 then the result is bigger than that when we make a subtraction of them. But
the reverse trend is shown at frequencies 2.9Hz and 3.1Hz. When we subtract the magnitude 1039 by 1037 at
both frequencies, the result is bigger than that if we add both amplitudes. Meanwhile, we can see from table
4, both of graph have the same trend, and doesn’t show the opposite trend. It implies that at frequency 2.6Hz,
the amplitude of both 1039 and 1037 have the same sign of value, but at frequencies 2.9Hz and 3.1Hz, they
have the opposite sign of value for the same movement. So it can be concluded that 2.6Hz is the first mode
(displacement), and 2.9Hz and 3.1Hz are the rotation modes.

2. Excitation on the weak direction

South 1039 Excitation 1037 North


Stair 2 Stair 1

The direction of excitation is perpendicular to transducer, so the dominant mode


will be the rotation not the displacement. And as the case which both transducers
face in the same direction, they will read the opposite sign of measurement due to
the same rotation. So the form that can be used: 1039 – 1037 for NS-direction of
excitation.

7
Picture 5. FFT Spectrums of the summation of both accelerometer 1039 and 1037 due to NS excitation

Picture 5 shows there are peaks of amplitude at frequencies 2.6Hz and 3.1Hz, in which apparent in both
summation and subtraction graphs. It can be hypothesized that 2.6Hz and 3.1Hz are the main frequencies.
And because the amplitude value of 1037 are higher than 1039, then the building will move rotationally with
north part of building moves further than south part.
At frequency 3.4Hz, maximum amplitude is occurred just in summation graph. Both transducers, 1039 and
1037, have similar amount of absolute amplitude value. And when they are added, it provides maximum
value. But on contrary it has minimum value when subtracted. It implies that both transducers have the same
sign at frequency 3.4Hz, in which they both move together in the same direction, which is on EW direction
displacement (perpendicular to direction of excitation).

Picture 6. FFT Spectrums of both accelerometer 1039 and 1037 due to NS excitation

B. Accelerometer on Weak Direction

1. Excitation on the weak direction

8
1036 1038

South Excitation North


Stair 2 Stair 1

Even though excitation was not located in the middle of the area of building, we can presume that the
dominant mode will be the first mode (displacement) instead of rotation, because of the same direction
between transducers and excitation.
Since transducer 1036 and 1038 are headed in the opposite direction, to find frequencies due to NS
excitation, one of transducer have to be subtracted to another to get maximum amplitude.

Picture 7. FFT Spectrums of the summation of both accelerometer 1036 and 1038 due to NS excitation

Picture 8. FFT Spectrums of both accelerometer 1036 and 1038 due to NS excitation

9
As can be seen in picture 7, the frequency 2.2Hz is very dominant in the subtraction graph (1036-1038), but
eliminated in the summation graph (1036+1038). It is occurred, because both transducers read the similar but
with opposite sign of amplitude value. And the fact that the transducer is faced in opposite direction, it means
that they displace in the same direction.
Accelerometer 1036 1038
(+) - (-) = maximum
(+) + (-) = minimum

By subtract 1036 by 1038 we can have maximum amplitude. And by adding them, the magnitude will be
disappeared. It can be summarized that the first mode (displacement) takes place at the frequency 2.2Hz.
The opposite behavior is occurred at the frequency 4Hz. When adding both transducers 1036 + 1038, the
maximum amplitude is found. And it is eliminated when 1036 is subtracted by 1038. Considering both
transducers are heading to opposite direction, but record the same sign of value, it implies that both
transducer move in the opposite direction movement. That means the plane is either elongated and/or
shortened. Because of the lack of transducer position, it is rather difficult to determine whether it is rotation
mode or not, but because 4Hz is not the first frequency, we can assume that at 4Hz the mode shape is rotation
(with elongation and shortening)

2. Excitation on the strong direction

1036 1038
South North
Excitation
Stair 2 Stair 1

The excitation that applied separately will rotate the plane even though the
direction of impact is perpendicular to the direction of accelerometer. Both accelerometers face the opposite
direction, so they will read the opposite sign of value for the same rotation. To get maximum amplitude, the
form of subtraction (1036-1038) can be used for EW excitation.

Picture 9. FFT Spectrums of the summation of both accelerometer 1036 and 1038 due to EW excitation

10
Picture 10. FFT Spectrums of both accelerometer 1036 and 1039 due to EW excitation

Frequency 2.2 Hz has peak amplitude in the subtraction graph (1036 – 1038), but eliminated in the
summation graph (1036+1038). It implies that both transducers record the similar amount of amplitude, as
can be seen in picture 10, but with the opposite sign. It means they move together in the same directions, as
can be the result of pure displacement or combination between displacement and rotation. But since the
impact is perpendicular to the direction of accelerometer, the combination displacement+rotation would be
the case.

On the other hand, at frequency 3.9Hz, the reverse trend is obvious. While both records have similar
amplitude, when they are added, the result is maximum value in summation graph, but disappears in
subtraction graph. It can be concluded that both point move in the opposite direction. That means the
building get elongation or shortening at that frequency. And also because of the impact is perpendicular to
the direction of accelerometer, the possibility mode shape is rotation with elongation or shortening.

From the summation graph in picture 9, there are three peaks on 2.6 Hz, 2.9 Hz, and 3.1 Hz. It can be seen
that the value of that peaks in subtraction graph is larger than summation graph, in which give the possibility
that rotation are the mode shapes of the three frequencies.

11
Summary

Frequencies where peaks of magnitude can be found.

1. Based on combination between direction of transducers and direction of impact

Main Indicator Frequencies on


Analysis
Functions peak magnitude
A.
1039 + 1037 EW excitation f1 = 2.6Hz Displacement (EW)
S 1039 1037 N 1039 – 1037 EW excitation f2 = 2.9Hz & f3 = 3.1Hz rotation
Excitation

B.
1039 – 1037 NS excitation f1 = 2.6Hz Rotation
S 1039 1037 N f2 = 3.1Hz Rotation
Excitation
f3 = 3.4Hz Displacement (EW)

C.
1036 – 1038 NS excitation f1 = 2.2Hz Displacement (NS)
1036 1038
Excitation
S N f2 = 4Hz Rotation (with
elongation or
shortening)

D. f1 = 2.2 Hz Displ + rotation


1036 + 1038 EW excitation
S 1036 1038 N f2 = 2.6Hz , f3 = 2.9Hz Rotation
Excitation & f4 = 3.1Hz Rotation
f5 = 3.9 Hz Rotation (with
elongation or
shortening)

Determining natural frequencies

 From (A), we can find that for strong direction of excitation, the first natural frequency is 2.6 Hz
with translation as the mode shape.
 From (C), translation mode shape can be found at frequency 2.2 Hz as the natural frequency for
weak direction excitation
 And natural frequency for rotation is 3.1 Hz because this frequency appears on both (B) and (G).
 Frequency 2.9Hz just appears at strong direction excitation and can not be found on weak direction.
Whereas rotation mode should be found on both direction of excitation. So 2.9Hz is not natural
frequency.
 Excitation on frequency 3.4Hz produces translation that perpendicular to the direction of impact.
This translation can be avoided, and the frequency as well.
 Both frequency 3.9Hz and 4Hz result in rotation with elongation or shortening of the building. The
possibility of this movement is that the building consists two part blocks that relatively move to each
other. Both frequencies are not the natural frequencies.
 Mode shape at frequency 2.2Hz and 2.6Hz at (B) and (C) appear as a rotation, because the directions
of excitations are perpendicular to the direction of accelerometers, so that the rotation of building is
recorded on accelerometers.

12
PART II. ARTeMIS TESTOR & EXTRACTOR ver 3.5

On the previous report of experimental modal analysis, manual method is used to extract and determine
frequencies of the building based on the direction of the impact.

In this report, the extraction of frequencies and modal shapes is done using ARTeMIS programs in order to
recheck and complete the previous report. ARTeMIS Testor v3.1 is used to build input file, and ARTeMIS
Extractor Handy v4.1 identify the modal parameters from experimental data written on ARTeMIS Testor
output file.

ANALYSIS
Analyses of modal parameters are divided into complex and simple model. Complex model is built to
analyse natural frequencies and modal shapes, while the simple one is used to make an animation of modal
shapes.

Figure 1. Note Editor of Input file on ARTeMIS Testor.

13
Figure 2. Complex Geometry Description of the Building

Figure 3. Simple Geometry Description of the Building

14
I. East-West Impact (Strong Direction)

ARTeMIS Model for EW Impact

Power Spectral Density

15
Frequency Domain Decomposition Analysis

Enhanced Frequency Domian Decomposition Analysis

16
Mode Shapes of Excitation on Strong Direction

Mode 1

Mode 2

Mode 3

17
Mode 4

Mode 5

18
I. North-South Impact (Strong Direction)

ARTeMIS Model for NS Impact

Power Spectral Density

19
Frequency Domain Decomposition Analysis

Enhanced Frequency Domian Decomposition Analysis

20
Mode Shapes of Excitation on Weak Direction

Mode 1

Mode 2

Mode 3

21
Mode 4

Mode 5

22
Analysis :

1. Strong Direction (East-West) Excitation


From FFD and EFDD graphs of east-west excitation, there are five peaks on five frequencies, i.e. 2.6Hz,
2.9Hz, 3.1Hz, 3.8Hz and 6Hz. In accordance with that, mode shapes of the five frequencies are
translation (east-west direction), and twisting for the rest four frequencies respectively. It is clear for the
first natural frequency is 2.6Hz with translation mode shape. Because all other four frequency have
similar mode shapes, they have to be compared with frequencies form weak direction excitation.
Frequency with the related rotation mode shapes has to be appeared on both impact directions.

2. Weak Direction (North-South) Excitation


Both FFD and EFDD graphs show five peaks, ie. 2.2Hz, 2.6Hz, 3.1Hz, 3.4Hz, and 3.9Hz. The related
mode shapes are translation (north-south direction), twisting, twisting, translation (east-west direction)
and rotation (with elongation) respectively. It is also clear for the first natural frequency is 2.2Hz with
translation mode shape. It is obvious that frequency 3.1Hz is found on both impact directions, so this
frequency is the natural frequency for rotation mode.

Conclusion
 First natural frequency mode for strong direction excitation is 2.6Hz with translation mode shape.
 First natural frequency mode for weak direction excitation is 2.2Hz with translation mode shape.
 Natural frequency for rotation is 3.1Hz, because it appears on both direction of excitation.

23

You might also like