You are on page 1of 12

EXPERIMENTAL SHIP HYDRODYNAMICS

-SIGNAL PROCESSING-

XU CHENG
MARTINUS PUTRA WIDJAJA

FLICIEN BONNEFOY

EMSHIP
EUROPEAN MASTERS COURSE
IN INTEGRATED ADVANCED SHIP DESIGN
COLE CENTRALE DE NANTES
2015

EXPERIMENTAL SHIP HYDRODYNAMICS 2015

List of Contents
INTRODUCTION ................................................................................................................ 3
REGULAR WAVES ............................................................................................................. 3
Time Domain Analysis ...................................................................................................... 3
Preliminary Frequency Analysis ........................................................................................ 4
Frequency Analysis of the steady regime ........................................................................... 5
IRREGULAR WAVES ......................................................................................................... 8

List of Figure & Table


FIGURE 1. REGULAR WAVE FREE SURFACE ELEVATION .......................................................... 3
FIGURE 2. FAST FOURIER TRANSFORM OF REGULAR WAVE SIGNAL ........................................ 4
FIGURE 3. THE ENLARGED RESULTS FROM FFT OF REGULAR WAVE SIGNAL ........................... 5
FIGURE 4. THE COMPARISON BETWEEN "FFT.M" AND "FOURIER.M" ......................................... 6
FIGURE 5. THE COMPARISON OF TIME SELECTION ON "FOURIER.M"......................................... 6
FIGURE 6. IRREGULAR FREE SURFACE WAVE ELEVATION ....................................................... 8
FIGURE 7. IRREGULAR WAVE SIGNAL PROCESSING BY "FOURIER.M" ....................................... 9
FIGURE 8. THE COMPARISON OF SPECTRAL DENSITY WITH DIFFERENT SEGMENT OF PERIOD.. 10

TABLE 1. THE OBVIOUS COMPARISON OF SPECTRAL DENSITY ............................................... 10


TABLE 2. THE STATISTICAL RESULTS OF DIFFERENCE TSEGMENT ......................................... 11

EMSHIP 5th COHORT- COLE CENTRALE DE NANTES

EXPERIMENTAL SHIP HYDRODYNAMICS 2015


INTRODUCTION
This report analyses about the signal processing in terms of wave signal used on the wave
basin of ECN/LHEEA. The main objective of this study is to understand how to analyze the
regular and irregular waves by applying the Fourier transform in MATLAB. A difference of
input parameter is applied in order to understand how it will effected the behavior of the
results.

REGULAR WAVES
The three different parts are introduced in order to understand how to analyze the signal.
These three parts are based on a binary file named regular.mat which contain 3 variables :
1. The sampling frequency (f_samp) = 60 Hz
2. The time which stored in vector, (time in seconds)
3. The wave elevation which measured by resistive probe, (wave in meters)

Time Domain Analysis

Figure 1. Regular Wave Free Surface Elevation

EMSHIP 5th COHORT- COLE CENTRALE DE NANTES

EXPERIMENTAL SHIP HYDRODYNAMICS 2015


After loading "regular.mat", we will obtain 2 vectors which are time, wave and the sampling
frequency. Then we can type time(end) and length(wave)

in order to obtain the signal

duration and number of data points which is 51.083 s and 3066 data points. By observing the
data inside the time vector, a constant step is observed as dt and this value is 0.0167 s.
Therefore, the relation between dt and the sampling frequency (60 Hz) is

There are two regions observed inside the wave signal by plotting the time and wave as in
figure 1. The low and developing region is called transient region and the harmonic part
commonly known as the steady region. There is a low noise existing at both regions, this
might happen due to electrical noise, etc. From around 6 s, the wave-maker starts to generate
the wave and at around 13th s the steady wave occurs until the end. The wave amplitude is
around 0.32 m and the wave period is around 2 second. Since the noise detected in the
transient region is only between -0.005 to 0.005 m, the influence of the noise is relatively
small compared to the amplitude of the steady wave signal. So this noise will not have a big
effect to the analysis.

Preliminary Frequency Analysis

Figure 2. Fast Fourier Transform of Regular Wave Signal

The result in figure 2 has a unique shape which creating a symmetric value about the xplane=30 Hz which means both two sides have the same information. So, the frequency can

EMSHIP 5th COHORT- COLE CENTRALE DE NANTES

EXPERIMENTAL SHIP HYDRODYNAMICS 2015


be divided by using the Shannon or Nyquist criterion ( f < f_samp/2) in order to simplify the
computation. Then a Fourier transform could be done as depicted in figure 3.

Figure 3. The Enlarged Results from FFT of Regular Wave Signal

After zooming in at the main peak, we are able to see one obvious peak where the value is
160 mm at 0.5 Hz. Also a little peak is observed at twice the first's peak frequency which the
value is around 20mm. The frequency which has the highest peak is the frequency of the
wave (1/2second= 0.5 Hz) and the highest amplitude of the FFT results is half of the wave
amplitude (approximate 0.16 m).

Frequency Analysis of the steady regime


The access to gain the information about the frequency for each amplitude of the initial wave
signal could be obtained by zooming in into the graph. This should be done in order to have a
better result by using a Fourier series code that had been developed by LHEEA. We need to
choose a signal from the steady region and also from peak to peak or trough to trough. In this
case, t_begin=13 second and t_end=48.8667 second which chosen from peak to peak.
The corresponding signal selection based on this command
selection1=wave(time>t_begin & time<t_end);

The Fourier series command developed by LHEEA and provide Fx (m) and freq (Hz)
[Fx1,freq1]=Fourier(selection1,f_samp);

EMSHIP 5th COHORT- COLE CENTRALE DE NANTES

EXPERIMENTAL SHIP HYDRODYNAMICS 2015

Figure 4. The Comparison between "fft.m" and "Fourier.m"

There is a significant difference between the approaches of FFT function and Fourier
transform from LHEEA on the peak signal as could be observed in figure 4. In the first case,
it has two peak and unsteady signal around the hill of the peak, while by using the Fourier.m
from ECN/LHEEA, better information could be observed due to a steady signal which allow
the main and secondary peak to be seen clearer.

Figure 5. The Comparison of Time Selection on "Fourier.m"

EMSHIP 5th COHORT- COLE CENTRALE DE NANTES

EXPERIMENTAL SHIP HYDRODYNAMICS 2015


From the above Figure 5, we can see two different color of results, the red one shows the
result with integer period, from peak to peak, while the blue one presents the result with the
integer plus half more of the period. As can be seen, there is only one higher peak in the red
one, which means the highest amplitude of the steady part of the wave at 0.5 Hz. In the other
hand, two points occur at the peak of the blue result, the reason is that the time selection starts
from one peak to another trough of the wave which will create two different amplitude.
Finally, the time selection based on the integer value (peak to peak, trough to trough) should
be used in order to know the maximum wave amplitude and at which frequency it will occur.
This is also due to the fact that from the practical point of view, the biggest amplitude will be
used for the further analysis.

Bonus Question (just present what we understand from studying this example)
For twice of the peak frequency,

, water depth for ECN Wave Basin is 5m. Then,

we can calculate the wavenumber k, which is equal to 4.02


The formula of 2nd Stokes solution for elevation is:

here,

. for

So, the above equation can be simplified as:

So the final amplitude for 2nd order item is calculated as :

In the other hand, the amplitude of the second order Stokes Solution has the following
relation:

After running the Fourier Function, a new parameter,

will be created. In this part,

From the above Figure 5, it is observed that the amplitude of twice the peak frequency is
around 0.0163m.

EMSHIP 5th COHORT- COLE CENTRALE DE NANTES

EXPERIMENTAL SHIP HYDRODYNAMICS 2015


IRREGULAR WAVES
After loading the file irreg.mat, we know that this time the signal is quite different from the
previous regular wave as can be observed from Figure 6 below. The wave elevation is quite
unstable and there is a large fluctuation between the range of around -0.05m to 0.05m. The
period of the wave is a little less than 2 seconds, around 1.8s.

Figure 6. Irregular Free Surface Wave Elevation

We get the Figure 6 via Fourier transform on the whole signal. It is not too easy to understand
this spectrum and the points on the spectrum fluctuate too strong along frequency axis. The
reason is that the original wave signal is an irregular wave in time domain and different
frequencies of the signal have different amplitudes.

EMSHIP 5th COHORT- COLE CENTRALE DE NANTES

EXPERIMENTAL SHIP HYDRODYNAMICS 2015

Figure 7. Irregular Wave Signal Processing by "Fourier.m"

In order to analyze the irregular waves, a Fourier transform which had been developed by
ECN could be used as in regular waves. But, as can be seen in figure 7, the result is still
unclear and too fluctuates. Therefore another built in function in Matlab called pwelch.m
should be used in order to estimate the power spectral density which could help to gain the
useful information from the above result. Before using this function, the different lengths of
time segment must be set. In this case, a set of different time segment had been chosen as
follow, (Seg1=7s, Seg2=17s, Seg3=23S and Seg4=37s) to investigate the effect on the results.
For instance, the syntax of first segment, 7 seconds, can be expressed as following:
[S1,freq1] = pwelch (signal, [], [], seg1*f_samp, f_samp);

EMSHIP 5th COHORT- COLE CENTRALE DE NANTES

EXPERIMENTAL SHIP HYDRODYNAMICS 2015

Figure 8. The Comparison of Spectral Density with Different Segment of Period

The final Figure 8 can be obtained after superimposing the above four spectra. As we can see,
there are four different situations of the spectral density distribution and both the longest and
shortest time segments are not good because of the reasons shown in Table 1:
Table 1. The Obvious Comparison of Spectral Density

T_segment Frequency Interval Numbers of Points

Problem

7s

short

large

large

less

few information

37s

long

small

small

much

large fluctuation

Based on the figure 8, a qualitative conclusion could be done in terms of the proper selection
of the time segment. Thus the reasonable lengths of the time segments in this situation are the
middle between two values (17-23 seconds). But the 23 seconds of the time segment is the
best one from all the four time lengths, which is marked as blue line in the figure 8.
After developing such kind of graph, a further analysis could be done in order to obtain the
information about the wave such as peak period (Tp), significant wave height (Hs), mean
period (Tm), up-crossing period (Tz), and the analytical (Tp). This could be done as follow,
For each segment length, evaluate following statistical values

10

EMSHIP 5th COHORT- COLE CENTRALE DE NANTES

EXPERIMENTAL SHIP HYDRODYNAMICS 2015

Tp directly from graph

For each segment length, we calculate the

based on the following formula:

We can employ the MatLab expression:


Trapz (freq, S.* freq.^n), with n=-2, 0, 1, 2
In this part, because we know that f (1) = 0, so 1/f will be infinite, so for every frequency in
this part will start from the second value to the end. Then, we get the values of

for each segment. Finally, the different parameters of period are obtained based on these
spectrum moments. The final results are listed in the following Table 2
Table 2. The Statistical Results of Difference Tsegment

Relative

T_seg

Hs

Tm

Tz

Tp

Tp / Tz

7s

0.1033m

1.7392s

1.6435s

2.1247s

1.293

8.89%

17s

0.0952m

1.6828s

1.5818s

2.1223s

1.342

4.91%

23s

0.099 m

1.6946s

1.5972s

2.1129s

1.323

6.42%

37s

0.096 m

1.7089s

1.6033s

2.1694s

1.353

4.06%

Error

The relation between Tp and Tz had been calculated as shown in the above table. The relative
error between the relationship of Tp/Tz in Bretschneider spectrum (1.408) with the calculated
results also shown in the table 2. The lowest relative error is occur for time segment 17 and
37 second, therefore the proper time segment should be located around this range. But for 23
second of time segment has an increase of relative error which means another time segment
should be analyzed except 23 second which has the good correlation with the Bretschneider

11

EMSHIP 5th COHORT- COLE CENTRALE DE NANTES

EXPERIMENTAL SHIP HYDRODYNAMICS 2015


spectrum relationship. The 37 seconds of time segment has the lowest relative error, thus this
one has the best correlation to obtain proper information of the wave. Nevertheless, a
narrower time segment analysis should be done in order to get more accurate value since
there is still possibility between 17-37 seconds of time segment.

Bonus Question is already explained in the previous paragraph which is about the accuracy
of the different time segment to obtain the good information from the irregular wave.

CONCLUSION
The parameters required in the Fourier analysis for irregular wave are:

Time segment in second, the length of this parameter should be neither too long nor
too short

The irregular wave data

The f_samp.

The parameters required for the Fourier analysis in regular wave are:

The sampling frequency

The time window will be depend on the selection of the t_begin and t_end.

The Regular Wave Data

Through this Lab work, we learn a useful tool to link the relation between the input or
measured regular & irregular wave data and the reaction of the ship modal. The operator
between these is called RAO (Response Amplitude Operator)

Here,

is the spectral density and

is the amplitude of input

is the spectral density and

is the amplitude of input

Once we know the RAO of the system studied, we will be able to study its response to a
given input or measured wave, or vice versa.

12

EMSHIP 5th COHORT- COLE CENTRALE DE NANTES

You might also like