Taguchi-based Experimental Investigation of Vortex-Induced

Vibration (VIV) Energy Converter Under a Controlled Environment

Jonas D. Gampon *, Daniel R. Ngo *, John Andrew V. Pamintuan *, Chester Jericho O. Ramos *

*School of Mechanical and Manufacturing Engineering,

Mapúa University, Muralla St., Intramuros, Manila, 1002 Philippines

*Corresponding author E-mail: [email protected]; [email protected];

[email protected]; [email protected]

Abstract
Globally, most of the energy consumption rely on non-renewable energy sources which is inevitably going to run out for the future
generations. To promote the use of renewable energy, the researchers aim to fabricate and test a prototype vortex-induced vibration
(VIV) energy harvester by integrating it with the spring-loaded mechanism and linear generator. The resulting electrical power output
was measured by varying input parameters such as spring width, spring length, and flow velocity. Taguchi’s parametric design of
Taguchi method was used to facilitate the design of experiment. Minitab 17 was used for the statistical software to analyze the signal-
to-noise ratio and perform the analysis of variance (ANOVA) while Fusion 360 software was used to create the drawing model. The
cylinders were bought, incorporated, and tested with the fabricated VIV energy harvester. Having the prototype worked; it was known
that linear generators can be used in replacement for piezoelectric for kinetic to electric energy conversion. The experimental results
show the behavior of the combined parameters and revealed that the best combination is A 1B3C3 which optimized the power output to
71.41mW with a 90.03% accuracy when compared to the Taguchi estimation. The ANOVA results revealed that all the parameters
were significant, and the most influential parameter is stream velocity.
Keywords: Vortex-induced Vibration, Power Output, Taguchi Method, Linear Generator, Spring-loaded Mechanism

INTRODUCTION an oscillatory lift force on the body, causing it to vibrate up and

down. These oscillations can then be harvested as electrical
The production of clean, renewable energy is one of the power. VIV has several advantages over other hydropower
biggest challenges faced by the world today. Non-renewable methods since it can be used in flow speeds ranging from 0.2 to
sources make up 85% of worldwide energy usage causing it to 2.5 m/s [4]. It has potential to be used in low current
be drained at an unsustainable rate leading to environmental applications and along coastlines. In addition, VIV devices are
problems such as pollution and global warming. In recent more environment-friendly because they cause minimal
decades, geothermal, wind, solar, biomass, and hydropower damage or alterations to ecosystems and do not interrupt marine
have been explored as renewable energy sources. However, traffic.
these energy sources still only make up about 12% of the
While VIV has many benefits, it is currently not
world’s energy consumption because they are more expensive
commercially utilized as a form of power generation. One of the
and require vast areas of land than their non-renewable
main challenges of utilizing VIV is determining how to
counterparts [1].
maximize the power that can be harvested from the oscillations.
According to the International Energy Agency [2], Significant studies have been done to address this issue,
hydropower currently represents about 17% of the total focusing mainly on the mechanical and electrical components
electricity production. Hydropower is a particularly promising to maximize the power that can be harvested from the
field because rivers and oceans make up about 71% of Earth’s oscillations [5]-[7]. By testing different values in the
surface area. Hydropower dams are the most commonly used influencing factors, system optimization is possible.
method due to the amount of power they can produce.
This study is being conducted to investigate the effects of
Accessibility is limited since they can only be built on large
selected input factors to amplify the oscillation of the cylinders
rivers with large flow rates, dams can also have significantly
to generate more power. The Taguchi method is used in order
negative impacts on river ecosystems and wildlife.
to minimize the variation in the electrical power output of a VIV
An alternative type of hydropower energy is by harnessing energy harvester. This study aims to design and assemble the
the power of vortex induced vibrations (VIV) [3]. VIV is a VIV energy harvester by integrating the spring-loaded
phenomenon that occurs when a bluff body is placed in a aluminum pipe cylinder mechanism and the linear generator in
flowing fluid; a shear layer forms on either side of the body, and an open channel water tank Such is done by application of
as the shear layers separate, they curl back behind the body Taguchi’s parametric design as a method of sampling by
forming a pattern of alternating vortices. These vortices exert varying the plate spring width, plate spring length, and stream

1
velocity. To evaluate the results, the signal-to-noise ratio and the output variable which is done by Quality Loss Function
ANOVA techniques of Taguchi’s method [8] is used to wherein the Noise Factor and Signal to Noise Ratio are
determine the optimal build combinations and identify the most considered. A robust output characteristic of the data can be
influential factors on the power output. obtaining with the optimum result taking the noise factor into
consideration. In addition, Orthogonal Arrays and Analysis of
This study also purposes to develop a vortex induced Variance (ANOVA) are also used in this method.
vibration generator that is more efficient and generates more
power to promote the use of hydropower as a source of energy. METHODOLOGY
With this, the use of non-renewable resources will be lessened,
and pollutant emission will be suppressed. This study can also Experimental Design Phase
contribute to the existing literature through the proper selection
of levels for the influencing factors and optimizing the power The researchers determine the factors that affects the
output through the optimization technique of the Taguchi oscillation of cylinder of the VIV generator system. As shown
method. Best combinations of the factors and its levels will also in Fig. 2, spring width, spring length and stream velocity were
improve the variability on the power output. The data and chosen as the controlled factors that greatly affects the VIV
results of this study can be used for the future researches related generator system based on the results provided in the previous
to the vortex induced vibration for its development. studies [5]-[7].

Vortex Induced Vibration

Vortex-Induced Vibration (VIV) is a phenomenon that
occurs when a fluid flows freely over a bluff body, such as a
cylinder, and causes it to oscillate. This is due to vortex
shedding as a result from the separation of the flow that then
forms boundary layers behind the bluff body. The fluid rotates
inwards behind the bluff body, forming the boundary layers or
distinct vortices. For instance, Reynolds number range from
300 to 1.5 ×10^5 (laminar flow) is the criterion for VIV in
which the bluff body can have stable oscillation.

Figure 2. The Controlled and uncontrolled parameters.

In obtaining the most robust output data,

uncontrolled/noise factors are also identified with respect to the
nature of the environment and the materials used. The
temperature of fluid is considered since it affects the viscosity
of the fluid. As the temperature rises, the viscosity decreases
due to excitation of the particles which breaks the bond and
makes the molecule free to move. Another uncontrolled
parameter that occur in the duration of the study is the bending
stress that the spring experiences during the experiment. The
Figure 1. Vortex Shedding of a cylinder.
flat springs are affected in a way that their properties start to
wear down as bending stress are applied [9]. Air pressure is also
This discussion follows Blevins [3] and Bearman [4].
considered as an uncontrolled parameter since having different
Taguchi Method atmospheric pressure in a certain area can affect the flow of
water.
It is a method created by Genichi Taguchi in order to
enhance and improve experimental strategies wherein a more The researchers determine value of levels of control
systematic, modified and standardized type of design of parameters, as shown in the Table 1, relative to the result of the
experiment is used [8]. Compared to other methods for design previous studies. The values of the stream velocity are based on
of experiment, Taguchi method is more efficient and economic the calibrated pump setups that will be discussed in the latter
since it reduces the number of experiments and variability on part of the phase.

2
 Table 1. Design of Experiment Parameters and their levels. researchers can visualize the motion that the components can
Level Level Level make as well as their motion extents.
Factor Parameter Unit
1 2 3
Spring
A mm 10 20 -
Width
Spring
B mm 250 275 300
Length
C Velocity m/s 0.33 0.43 0.58

Based on the table above, there are three parameters used,

and for one parameter, the number of levels is two; and for the
other two parameters, the number levels of three.
The system of factors falls under the L18 array with a
maximum number of eight factors with one of them having only
two levels and the other seven having three levels each. Thus,
the orthogonal size DOE is 𝐿 18 (21 × 32 ). The experimental Figure 3. CAD modeled Isometric View of the Experimental
layout for the VIV energy harvester is shown in Table 2. Each Set-up.
row of this table refers to a run with various combination of
factors and their specific levels. Fabrication Phase
Table 2. Experimental Plan using an L 18 Orthogonal Array. The first part of the fabrication phase is the fabrication of
Vortex Induced Vibration the open channel water tank and its components.
Parameters Level
NUMBER OF The water tank 50" x 20" x 15" with thickness 3/8" made
A B C
EXPERIMENT of pure glass is a scaled down concept presented by Zahari et.
Spring Spring
Velocity al. [10]. It is designed to withstand the waterpower of the
Width Length
1 1 1 1 pumps, the weight of water and other equipment.
2 1 1 2
3 1 1 3
4 1 2 1
5 1 2 2
6 1 2 3
7 1 3 1
8 1 3 2
9 1 3 3
10 2 1 1
11 2 1 2
12 2 1 3 Figure 4. Water Tank
13 2 2 1
14 2 2 2 For continuous water circulation, a partition is fabricated
15 2 2 3 by cutting two acrylic glasses into 37" x 14" w/ thickness of 45
16 2 3 1 mm each, as shown in Fig. 5.
17 2 3 2
18 2 3 3

Using this Table, the naming of each run can easily be

identified by coding. For example, taking experiment number
3, it can be easily be identified by A1B1C3.
Concept Model Design Phase
The CAD model was drawn using Fusion 360 to act as the
blueprint of the whole VIV system. The Fusion 360 software Figure 5. Cutting the Acrylic Partition using Mapua
University's Laser Cutter Machine
can also simulate the movement of the components in the
system through its motion study function. Through this, the

3
 A GI sheets 30" x 14" serves as a flow guide and a For the last part of the fabrication phase, the electrical
honeycomb-like screen (made of plastic straws placed into a components. These are necessary for data gathering of output
window type acrylic glass) is placed after the flow guides for values, especially the output power, and measuring the accuracy
the water flow to be laminar as shown in Fig. 6. of the stream velocity.
The linear generator shown in Fig. 9 works by moving a
magnet in a linear/vertical motion inside a coil producing
electrical energy. Two neodymium bar magnet (2 grams each)
and ten ferrite magnetic rings (1 gram each) are used. A total of
10 grams was mounted on the prime mover since two ferrite
magnetic rings above and below are attached to the casing to
repulse the magnets at a certain distance with respect to the
coils.
Figure 6. Flow Guides (right) and Screen (left).

Two Kawasaki Model: JSW-100 submersible pumps

(Gray, 1/6HP, 100L/min) and one Ingco SPD7401-5
submersible pump (Yellow, 1HP, 216L/min) are procured. To
maximize the power of the pump, flexible hoses were used with
length 40” relative to the length of the water tank as shown in
Fig. 7.

Figure 9. Permanent Linear Generator

A circuit arrangement, shown in Fig. 10, is needed to

convert the AC produced by the Linear generator into DC with
minimal losses. This will stabilize the power output, lessen the
Figure 7. Submersible Pumps with attached Flexible Hoses. fluctuations when gathering data, and avoid expensive
equipment for reading erratic AC values.
The second part of the fabrication phase is the VIV
Assembly. The VIV assembly shown in Fig. 8, is the heart of
the whole study where most of our observation and data
gathering took place. It is composed of the box type framework
made of aluminum flat bars for the vertical columns and L- bars
for the upper and lower bases. A hollow aluminum cylinder
with a diameter of 1" (1/6" thickness), length of 10". Two pairs
of springs (10 mm x 310 mm and 20 mm x 310 mm, both having
a thickness of 1mm) made up of aluminum flat bars. And lastly
a clamping mechanism, wherein acrylic glass setup is attached
to the slider and a roller clamp attached to the plate spring.

Figure 8. VIV Device Assembly (right) and Two pairs of flat Figure 10. Breadboard (top) and Schematic (bottom)
springs. Diagrams of the Linear Generator Circuit Using Fritzing

4
 For data gathering, total voltage generated is measured by The graphed values, shown in Fig. 12, verifies that the
tapping the multimeter into the legs of the capacitor. For water flow is laminar since there is only minimal variations of
measuring of the total current, another multimeter was put into the velocities at any given point in each sample.
series at the end of the full bridge rectifier.
For the calibration of the flow measurement, a YF-S201
flow meter powered by ARDUINO UNO, as shown in Fig. 11.
This concept of flow measuring device is adapted from the
study conducted by Nasution et. al. [11].

Figure 13. Assembled Experimental Setup.

Testing and Data Gathering Phase

For the power output test, all 18 runs in Table 2 were
performed. Two multimeters are used, to obtain voltage and
Figure 11. Breadboard Diagram of the Water Velocity Sensor current simultaneously as shown in Fig. 14. The obtained
Circuit using Fritzing. voltage and current are multiplied to obtain power for each run.
The researchers have calibrated three pump setups. First
was using the two (2) Kawasaki pumps. Second using the Ingco
pump only. And for the third setup, all the pumps will be
working together. In each pump setup, the flow meter is placed
at the top (20 cm) and the middle (10 cm) with respect to the
flow sensor (note: water level 22 cm). 39 samples for both the
top and middle levels were gathered. The values obtained where
used for the levels of stream velocity.
Figure 14. Actual Setup for gathering of the output Voltage
and output Current.
Voltage and current readings are obtained every 3 seconds,
having a total of 20 readings per run (total of 1-minute
duration). Average of these readings are used and are computed
to power for each run.
For the Amplitude and Frequency Analysis, oscillation of
cylinder for each run is recorded using a video recorded for 10
seconds and was analyzed using software called Tracker shown
in Fig. 15. A ruler is placed closed to cylinder which then is
used for reference points.

Figure 12. Sample Data for Stream Velocity Measurement. Figure 15. Video Analysis via Tracker software.

5
 For the Rate of Voltage increase, the multimeter is Relating to power output, A1B3C3 (run order 9) yielded the
connected to capacitor while the switch is off. The rate of largest amplitude produced the highest power output due to
voltage increase readings are obtained per three seconds in 1 Faraday’s law. Having the largest vertical displacement of the
minute, which totals to 20 readings per run. Data was analyzed magnet, more coils experience a change in their magnetic
by graph using Excel. environment; therefore, producing the highest power output.
For the Statistical Analysis Minitab 17 software was used For the change in frequency, the highest frequency
to give further analyze the data using Taguchi analysis. The obtained was 4.02 Hz which is from designated run coded
types data obtained in the software includes the S/N Ratios and A1B1C3 (run order 3). The run consists of having a spring width
Analysis of Variance (ANOVA). The larger-is-better is used of 10 mm, spring length of 250 mm, and a stream velocity of
since the optimum power output is aimed in this study. Larger 0.58 m/s. On the other hand, the lowest frequency obtained was
signal means it overcomes the noise/uncontrolled factors. 1.4 Hz which is from designated run coded A 1B3C3 (run order
Response Table for S/N Ratio for each parameter and Main 9). The run consists of having a spring width of 10 mm, spring
effects plot are included as results for our research, along with length of 300 mm, and a stream velocity of 0.58 m/s. This can
the General Linear Model for ANOVA and Taguchi Prediction be seen in Fig. 18.
for S/N Ratio and Power Output.

RESULTS AND DISCUSSION

Experimental Results (Mechanical)

For the change in amplitude, the largest amplitude obtained
was 30.13 mm which is from designated run coded A1B3C3 (run
order 9). The run consists of the following conditions: 10 mm
spring width, 300 mm spring length and 0.58 m/s stream
velocity. The time series for this run is illustrated in Fig. 16a.
The smallest amplitude obtained was 5.34 mm which is from Figure 17. Amplitude vs Time graph for Run 3.
designated run coded A1B1C1 (run order 10). The run consists
of the following conditions: 10 mm spring width, 250 mm
spring length, and 0.33m/s stream velocity. The time series for
this run is illustrated in Fig. 16b.

Figure 18. Boxplot of Frequencies for Runs 1-9.

Aside from showing the frequency values, the box plot

shows the precision of the data. The vertical line shows the
maximum and minimum value attained while the box indicates
that the value gathered are most likely to hit that range. The
thickness of the box depends on the standard deviation of the
data; thus, the thinner box shows more precision. The horizontal
line in the middle of the box represents the median.
Based on the data, the frequency increases with smaller
spring width, shorter spring length and faster stream velocity.
Figure 16. Amplitude vs Time graph for (a) Run 9 and (b)
Run 10. Relating to power output, A1B3C3 (run order 9) which has
Based on the data, the amplitude increases with smaller the lowest frequency but produces the highest power output.
spring width, longer spring length, and faster stream velocity. This is achieved because even though it is slowly oscillating, all

6
the magnet hits all the coils inside the linear generator since it For A2B1C1 (run order 10), the equation of the line graph
has the highest amplitude producing the largest power. shows that the value of the slope is 0.0014 which is significantly
smaller than the previous run discussed. This shows that this
Effect on Stored Voltage particular run has a very slow rate of increase since it only adds
The fastest rate of voltage increase was 3.14 V/min which approximately 0.0014 V in the initial value of 0.0373 V as the
is from A1B2C3 (run order 6) with the following conditions: 10 time is increased per second. With respect to the trendline, the
mm spring width, 275 mm spring length, and 0.58 m/s stream rate of voltage increase is 0.1213 V/min and comparing it with
velocity. The time series for this is illustrated in Fig. 19a. The the actual reading of 0.12 V/min produces a percentage
slowest rate of voltage increase was 0.12 V/min which is from difference of 1.08%.
run A2B1C1 (run order 10) with the following conditions: 20 Optimization of Power Output
mm spring width, 250 mm spring length, and 0.33 m/s stream
velocity. The time series for this is illustrated in Fig. 19b. The output power is obtained by computations of the
average voltage and average current obtained in each
experiment run. The power obtained in the experiments was
analyzed using the Taguchi analysis function of the Minitab
software. Table 3 shows the experiment results for the average
voltage, average current, power output, and the S/N ratio.
Table 3. Experiment Results for Average Voltage, Current, and Power, and
S/N Ratio.
A B C
Average Average S/N
Run Spring Spring Stream Power
Voltage Current Ratio
Order Width Length Velocity (mW)
(V) (A) (dB)
(mm) (mm) (m/s)
1 10 250 0.33 0.17 0.0008 1.14 1.1381
2 10 250 0.43 1.59 0.0216 34.30 30.7061
3 10 250 0.58 1.64 0.0273 44.76 33.0178
4 10 275 0.33 0.51 0.0047 2.37 7.5005
5 10 275 0.43 1.70 0.0391 66.51 36.4576
6 10 275 0.58 1.75 0.0171 29.85 29.4977
7 10 300 0.33 1.48 0.0133 19.64 25.8613
8 10 300 0.43 1.65 0.0182 29.91 29.5157
9 10 300 0.58 1.68 0.0425 71.41 37.0752
10 20 250 0.33 0.04 0.0052 1.02 0.1720
11 20 250 0.43 0.82 0.0035 2.82 8.9902
12 20 250 0.58 1.20 0.0047 5.58 14.9291
13 20 275 0.33 0.39 0.0032 1.23 1.8208
14 20 275 0.43 0.76 0.0104 7.95 18.0082
15 20 275 0.58 1.21 0.0038 4.61 13.2765
16 20 300 0.33 0.75 0.0102 7.60 17.6210
17 20 300 0.43 1.03 0.0195 20.08 26.0562
18 20 300 0.58 1.39 0.0103 14.20 23.0466

As shown in the table, designated run code A1B3C3 (run

order 9) produces the highest output power out of all the
experiment runs. The run consists of having a spring width of
Figure 19. Voltage increase over time for (a) Run 6 and (b) 10 mm, spring length of 300 mm and a stream velocity of 0.58
Run 10. m/s.
Fig. 20. shows the main effects plot of the S/N ratio for
The solid line presents the actual increase in the experiment
power.
while the broken line shows how it is formed when the actual
form is presented through an equation representing the
trendline. The slope, which is generally known as ‘rise’ over
‘run’, represents how the voltage increases over an amount of
time.
For A1B2C3 (run order 6), the slope of the line in this
particular run is 2.064 V/min. This means that the initial value,
where in this case is the y-intercept which has a value of 1.2161
V, is being increased approximately by the slope value as time
also increases per second. With respect to the trendline, the rate
of voltage increase is 3.28 V/min and comparing it with the
actual reading of 3.14 V/min produces a percentage difference
of 4.36%. Figure 20. Main Effects Plot of S/N Ratios for Power.

7
 Based on the table, all three parameters have p-values that
The graph shows how each parameter and their levels can are less than 0.05. This means that the three factors all have
be influential to the output by observing the slope of their lines. major relevance and are largely significant on influencing the
The steepness of the line determines the robustness of a power output being robust as a response. Spring width and
parameter. When the line under a parameter is steeper, then the velocity both have the lowest of p-values, while spring length
range within the selected levels are considered more influential has the highest p-value.
to the output. It is shown that a change in velocity from 0.33
m/s to 0.43 m/s produces a large increase of mean of S/N ratios, The Minitab software can also predict the Taguchi results
but when the change is from 0.43 m/s to 0.58 m/s, the increase based on the Taguchi analysis that was made. As previously
of S/N ratio is minimal. This can also be observed in spring discussed, the run combination A1B3C3 was found to be the
length where from 250 mm to 275 mm, the difference of S/N levels of parameters that maximize the power output. Table 6.
ratios is comparatively smaller than when it changes from 275 shows the predicted values based from the ANOVA results for
mm to 300 mm. As for the spring width, the change in S/N ratio power and S/N ratio as processed by the software. The absolute
is large for the two levels. Table 4 shows the response table for error found by comparing the actual and Taguchi-based
S/N ratio of power. prediction in the power output is 7.12 mW. By comparing the
Taguchi estimation with the experimental result, it is found that
Table 4. Response Table for S/N Ratios of Power (Larger-is-better). the method is 90.03% accurate.
Spring Spring
Level Velocity Table 6. Predicted Power and S/N Ratio based on ANOVA.
Width Length
1 25.641 14.826 9.019 Combination Power Output S/N Ratio (dB)
2 13.769 17.760 24.956 A1B3C3 78.53 mW 37.9008
3 26.529 25.140
Delta 11.872 11.704 16.122
Rank 2 3 1 CONCLUSION AND RECOMMENDATION

The plot in Figure 20 are shown in numeric values, along This research study evaluated the influence of the VIV
with the delta and rank of each parameter. It can also be seen parameters to the power output. In this scope, the spring width,
that the combination A1B3C3 is the combination that presents spring length, and flow velocity are the parameters selected
the largest S/N Ratio. Therefore, this combination of which were observed. This was achieved by using the designed
parameters, which has the spring width at 10 mm, the spring and assembled VIV energy harvester through integrating the
length at 300 mm, and velocity at 0.58 m/s, can minimize the spring-loaded aluminum pipe cylinder mechanism and the
variability of the power output the greatest. It is also shown that linear generator in an open channel water tank. Given that the
the most influential parameter is velocity, as its increase of S/N prototype was able to run properly, the researchers found out
ratio from 9.019 dB (for 0.33 m/s) to 25.140 dB (0.58 m/s) is that it is possible to use a linear generator instead of using
found to be the largest. It is followed by the spring width, then piezoelectric. Moreover, the comparison of the range of
the spring length. frequency on VIV using linear generator having a range of
which has 1.4 - 4.02 Hz and the range of frequency on the
For the ANOVA, the model that was used is a General previous study with VIV using piezoelectric which has 2.5 to
Linear Model. The p-value in ANOVA is crucial as it 4.5 Hz clearly shows that the frequency range of linear
determines what parameters are significantly relevant to the generator is wider at lower values compared to that of
robustness of the power output to variability. In order to identify piezoelectric. Thus, the linear generator would be a much more
this, the value of p-value for each parameter should be less than effective compared to piezoelectric generator in terms of power
0.05. Table 5. shows the results of ANOVA for power. production at cases where low frequency oscillations are
produced.
Table 5. Results of ANOVA for Power.
Variation Adj F- p- The researchers were able to employ Taguchi’s parametric
DoF Adj SS
Source MS value value design in the experiment of varying the plate spring width, plate
Spring spring length, and stream velocity. In varying the plate spring
1 643.3 634.27 17.44 0.001
Width width, the amplitude and frequency of the oscillation and the
Spring stored voltage vary inversely with it. In varying the plate spring
2 445.0 222.49 6.12 0.015
Length length, the amplitude of the oscillation along with the stored
Velocity 2 1027.8 513.92 14.13 0.001 voltage vary directly, while the frequency varies inversely. In
Error 12 436.3 36.36 varying the stream velocity, the amplitude and frequency of the
Total 17 oscillation and the stored voltage vary directly; taking into
consideration that there is a limit as to how fast the velocity can
be as increasing it further would cause turbulence. The highest
power output obtained is from designated run coded A 1B3C3

8
(spring width = 10 mm, spring length = 300 mm, and stream placed into irrigation canals near the agricultural fields and
velocity = 0.58 m/s) which is a combination of the smaller landscapes with which can help the farmers.
spring width, the longest spring length, and the fastest stream
velocity; all of which exhibit characteristics for obtaining a
higher power output. ACKNOWLEDGEMENT
The researchers were also able to evaluate the results using The researchers would like to thank Engr. Emmanuelle R.
the Taguchi Method by analyzing the signal-to-noise ratios and Biglete for his willingness to support and guidance throughout
doing analysis of variance (ANOVA). Statistically, it was also our research.
determined that the combination A1B3C3 (spring width = 10
mm, spring length = 300 mm, and stream velocity = 0.58 m/s)
presents the largest S/N ratio of 37.0752dB. Regardless of the
quality characteristic selected, a larger S/N ratio corresponds to REFERENCES
a smaller variance of the power output around the target value.
Thus, this run provides the most optimum power output [1] American Geosciences Institute. “How much of world
considering the effects of the uncontrollable factors, out of all energy consumption and electricity generation is from
the 18 runs. By using ANOVA, the results show that all three renewable energy” https://www.americangeosciences.org
parameters are significantly influential to the response as their (accessed March 27, 2019)
p-values are all below 0.05. Also, the predicted power output
was determined to be 78.53 mW, and if compared to the actual [2] U.S. Geological Survey. “Hydroelectric power water use”.
power output, Taguchi prediction presents a 90.03% accuracy. http://water.usgs.gov/edu/wuhy.html (accessed March 27,
According to the response table, stream velocity is found to be 2019)
the most influential factor out of the three parameters, in terms [3] R. Blevins. "Flow-Induced Vibration". New York: Van
of the change in S/N ratio between the highest and the lowest Nostrand Reinhold. 1990.
level.
[4] P. W. Bearman, “Vortex Shedding From Oscillating Bluff
This research contains additional learnings with regards to Bodies", Annual review of fluid mechanics, 16(1), 1984.
each parameter that was used. The highest level of stream pp. 195–222.
velocity in this research is 0.58 m/s, which is close to the
maximum value of velocity that a system should have to [5] H. Kong, V. Roussinova, and V. Stoilov, “Renewable
achieve lock-in. Exceeding this particular number would make Energy Harvesting from Water Flow.” International
the water flow turbulent, and thus would not achieve proper Journal of Environmental Studies 76 (1), 2019, pp. 84–
vibration to the VIV system. In addition to this is the 101, doi:10.1080/00207233.2018.1494928
relationship of the spring stiffness to the concept of moment of
force. In this study, the spring is fixed to one end, and at the [6] M. Griffith, and J. Leontini, “Energy-Harvesting From
other end is where the weight or force is located. It is related to Flow-Induced Vibration of Circular and Elliptical
the spring stiffness by integrating the formula of moment which Cylinders.” XXIV ICTAM, 21-26 August 2016, Montreal,
is the force multiplied by the length of the lever arm. In this Canada.
case, the length changes since one of the parameters is spring [7] M. Hirao and K. Hosotani, “Output Characteristics of the
length. The spring then experiences different moment Flicking Piezo Electric Power Generator for Vortex
depending on what is the level of spring length that is to be used. Induced Vibration Driven Single Circular Cylinder,”
The following recommendations can be taken for Power 4, 2015, pp. 7-38,
conducting a research related to the study. For further research, doi:10.13140/RG.2.1.4999.9845.
it is recommend using air as a medium for optimization of the [8] A. Freddi, and M Salmon, “Introduction to the Taguchi
present study. Conduct a screening experiment in order to filter Method”, Design Principles and Methodologies, 2019, pp.
the most effective controlled parameter. Furthermore, 159-180, doi: 10.1007/978-3-319-95342-7
considering interaction effects between the parameters is also
recommended, for deeper understanding and representation of [9] P. Childs, "Springs: Leaf Springs" Mechanical Design
the relationship between the input, and output parameters. Use Engineering Handbook, 2018, pp. 757-761,
Computational Fluid Dynamics (CFD) in which generates a doi:10.1016/B978-0-08-102367-9.00015-9
model through computer simulations. This simulation can prove
that the data gathered is efficient and with accordance to the [10] M. Zahari, and S. Dol, “Design and Development of Low-
actual experiments done related to this study. Lastly, it is Cost Water Tunnel for Educational Purpose.” IOP
recommended to scale up the dimensions of the prototype as it Conference Series: Materials Science and Engineering
would produce more power which would be enough to lighten Vol. 78, no. 1, p. 012040, IOP Publishing, 2015.
up a bulb or other electrical components. The prototype can be doi:10.1088/1757-899X/78/1/012040

9
[11] T. Nasution, E. Siagian, and K. Tanjung, “Design of River
Height and Speed Monitoring System by Using Arduino.”
IOP Conference Series: Materials Science and
Engineering Vol. 308, no. 1, p. 01203, IOP Publishing,
2018. doi:10.1088/1757-899X/308/1/012031.

10