Abstract

This paper investigates the analysis and design of a vibration-based energy harvester for
rotary motion applications. The energy harvester consists of a cantilever beam with a tip
mass and a piezoelectric ceramic attached along the beam that is mounted on a rotating shaft.
Using this system, mechanical vibration energy is induced in the flexible beam due to the
gravitational force applied to the tip mass while the hub is rotating. The piezoelectric
transducer is used to convert the induced mechanical vibration energy into electricity. The
equations of motion of the flexible structure are utilized along with the physical
characteristics of the piezoelectric transducer to derive expressions for the electrical power.
Furthermore, expressions for the optimum load resistance and maximum output power are
obtained and validated experimentally using PVDF and PZT transducers. The results indicate
that a maximum power can be extracted by using a PZT transducer .This amount of power is
sufficient to provide power for typical wireless sensors such as accelerometers and strain
gauges.
Literature

The piezoelectric materials, in special PZTs, have been largely used as mechanisms to
convert ambient motion, usually vibration, into electrical energy that may be stored or used
directly to provide power to other devices, e.g. mobiles, portable electronics or wireless
sensors networks. These examples represent a huge potential for commercial applications in
different areas as presented in recent market researches. Structural health monitoring
applications are one of the most benefited areas with devices for power harvesting. An
interesting application is given by Starner and Paradiso (2004), who discuss the possibility of
using alternative sources of vibration, such as the vibration of human breath captured
through PZT in the human chest, and even recovery energy of blood pressure or provided by
vibration when a person walks with shoes bonded with PZTs patches. Another very
interesting study is the use of mechanical vibration caused by machine.

An important stage to develop a practical design in an energy harvesting device is to model

correctly the dynamical actions of the integrated system composed of mechanical structure,
in general a clamped beam, electromechanical coupling between the PZTs and the
mechanical system, and, finally, the electrical load attached to the device. However, the
multidisciplinary nature of this field has caused some modelling problems. Erturk and Inman
(2008) presented some considerations about the oversimplified, incorrect physical
assumptions and mistakes in analytical modelling of piezoelectric energy harvesters.

The authors clarified through improved models with lumped and distributed parameters,
besides presenting a good overview of the numerical and analytical modelling of
electromechanical systems for power harvesting. The spot light in this strand of papers about
piezoelectric energy harvesting models to study the maximum power that can be dissipated
in a simple resistor or combination of linear electrical elements. The most part of these
models focused on simplifying the energy harvesting circuit by a simple resistive load, but in
the real-world applications, the energy harvesting circuit attached is more complex than a
simple resistor. Thus, few information about the interaction between practical rectifiers
circuits, used to transform alternating current (AC) into direct current (DC), and the
electromechanical devices attached are discussed clearly.
On the other hand, the members of the research community in power electronics focus on the
developing non-linear electronic models by using diodes, transistors, synchronized switch,
etc. Provided an analytical and graphical analysis equation relating the output power with the
efficiency of the rectifier circuit, which shows how important is the step of rectifying and
storing the electrical charge. Wickenheiser and Garcia (2010) observed that the full-wave
rectifier has a smoothing capacitor to provide a tension approximately constant over the load.
A Synchronized Switch Harvesting (SSH) has also been analyzed in the area of energy
harvesting with PZT sensors. Also, a Synchronized Switch Harvesting on Inductor (SSHI)
has been developed, taking up to 160% efficiency over standard rectifier. Other studies show
the possibility of using inductors for switching. These techniques of circuit switched consider
that the frequency is constant with sinusoidal signal. However, there is a lack of studies
concerning circuit switched with electrical noise, where the non-controlled rectifier circuits
have advantages.

The goal, in general, is to study the optimal conditions to control the power flow and to
charge an electrochemical battery or supercapacitors or directly feed an electronic system.
The most part of these papers employed simplest models of mechanical resonator (spring,
mass, damper) coupled with the electrical circuit, normally, with a single degree of freedom.
The simplicity of the mechanical model can give good results close to the resonance
frequencies. However, it is well known in the literature that to predict accurately the
electromechanical performance of piezoelectric energy harvester it is necessary to use a
distributed parameter model so that considers multiple vibration modes.
Introduction
Nowadays energy is one of the most important issues around the world. Especially in
Bangladesh energy crisis is a big problem. Renewable energy sources can be a great media to
solve this energy crisis problem in Bangladesh. As we know natural resources will finish one
day. That’s why researchers are trying to introduce substitute energy sources from nature.
That must be green and not harmful for the environment. Energy harvesting is defined as
capturing minute amounts of energy from one or more of the surrounding energy sources.
Human beings have already started to use energy harvesting technology in the form of
windmill, geothermal and solar energy. The energy came from natural sources, termed as
renewable energy. Renewable energy harvesting plants generate kW or MW level power; it
is called macro energy harvesting technology. Moreover, micro energy also can produce
from those natural sources that are called micro energy harvesting. Micro energy harvesting
technology is based on mechanical vibration, mechanical stress and strain, thermal energy
from furnace, heaters and friction sources, sun light or room light, human body, chemical or
biological sources, which can generate mW or μW level power. Micro power supply needs is
increasing greatly with time as our technology is moving to the micro and nano fabrication
levels. Our discussion on this is based on generating micro energy from vibration and
pressure using piezoelectric material.

The aim of this paper is to evaluate and compare the experimental results and the results
provided by a theoretical model of a full-wave diode bridge attached directly to the
electromechanical cantilever beam with base excitation to cause transverse vibrations. A
very simple and comprehensive description for choosing the capacitance and resistance loads
is provided seeking practical application.

All numerical tests and results are described and detailed by using the Matlab and the
SimPower System toolbox of the Simulink. The experimental setup is driven through the use
of a DSpace 1104 data acquisition with Control Desk. In order to show some features, it is
provided a number of simulations to illustrate the approaches. At the end, the final remark
presents suggestions for further research.

Nomenclature,
Ec = energy stored, J

C = capacitance, F

i = current, A

ID = diode current, A

Lb = length of the beam, mm

Lp1 = length of the PZT1, mm

Lp2 = length of the PZT2, mm

rd = internal resistance of the diode, ohms

R = resistance, ohms

P = power, W

q = electric charge, C

tb = thickness of the beam, mm

tp = thickness of the PZT, mm

v0 = voltage in steady state, V

w = width of the beam or PZT, mm

Greek Symbols

= permittivity of PZT, C²/Nm²] = time constant, s

Subscripts

Avg = relative to average

AC = relative to alternate current

b = relative to beam

DC = relative to direct current

L = relative to load

p= relative to PZT

RMS = relative to root means square

Block diagram
Circuit diagram
Non-Parametric Model of Piezoelectric Energy Harvesting Beam
The cantilever beam with bimorph PZT patches used in the present paper is shown in Fig. 1.
The geometry and details are also presented in Fig. 1(b) with Lb = 78 mm, Lx = 20 mm, tb
=0.91 mm, Lp1 = 45 mm, Lp2 = 46 mm, tp1 = 0.191 mm and tp2 =0.191 mm. The width
of the beam is 21 mm, the PZT1 15 mm and the PZT2 17 mm. The PZTs used are the model
PSI-5H4E manufactured by Piezo Systems. It is important to observe that the ideal
placement for the piezoceramics should be attached to the cantilevered end of the
electromechanical beam in order to maximize the power output. However, three reasons are
used to choice the value of Lx = 20 mm; first of all, this is a beam already existing in the
laboratory. Secondly, the attaching of the beam with the shaker is difficult if the PZT is
bonded near to the cantilevered end of the electromechanical beam. Finally, the goal in this
paper is not relative to maximize power in function of the geometry and placement of PZT,
but to show the best performance for the appropriate choice of topology and components of
the resistive and capacitive rectifier circuit used.

The goal is only to describe it by using experimental data, a non-parametric model which
is representative of the dynamics of the beam. Thus, a non-parametric model, the frequency
response function (FRF), is obtained from spectral analysis. In order to determine the FRF of
the system, an input chirp signal with frequency ranging from 0 to 2 kHz of lower PZT
(actuator) was applied. The response on top PZT (sensor) of the beam connected in open
circuit for measurement was recorded. The sampling rate used was 8 kHz and all acquisition
was performed by using the DSpace 1104 board controlled by the Control Desk software.
Figure 2 presents the result of the experimental FRF calculated by using the Welch method
with 3200 samples, 50% of overlapping and rectangular window.

It is worth noting that at 58 Hz the system has a high peak gain, significantly greater than the
other frequencies. Experimentally when a sinusoidal frequency with 58 Hz is applied in the
shaker as input the level of vibration is high stating and proving this is the first natural
frequency of the beam. For monitoring the level of force applied and displacement in the
beam, two additional transducers were employed: a load cell, model TF-YD-312 TMC with
sensitivity of 3.02 pC/1 N, and an IEPE accelerometer TA-YD-193 TMC with sensitivity of
1.007 mV/ms-2. The measurements performed with this sensor were only qualitative.
Model of Energy Storage Circuit

Usually, in real-world application of vibration power harvesting, a rectifier circuit is

necessary to convert AC to DC in order to charge a battery or to feed directly an electronic
device. A very simple and common non-controlled rectifier circuit is the full-wave rectifier
with Diode Bridge. A schematic diagram is shown in Fig. 3.

The non-controlled rectifier can transform an alternating signal continuously, but if a

capacitor in parallel to the load is not employed, the output signal could have a considerable
ripple, which may be unacceptable for powering loads with voltage DC. So, the circuit to
supply a resistive load uses a capacitive filter which smooths the effect of variation in load.
The ripple was defined by the reason of the component AC to DC. The component DC of the
signal is illustrated by the following equation.
Components List

Piezoelectric Sensor

This day most of the research in the energy field is to develop sources of energy for future.
It is time to find renewable surceases of energy for the future. Piezoelectric materials are
being more and more studied as they turn out to be very unusual materials with very specific
and interesting properties. In fact, there materials have the ability to produce electrical
energy from mechanical energy for example they can convert mechanical behaviour like
vibrations in to electricity.

Such devices are commonly referred to as energy harvesters and can be used in applications
where outside power is unavailable and batteries are not a feasible option. While recent
experiments have shown that these materials could be used as power generators, the amount
of energy produced is still very low, hence the necessity to optimize them.

Piezoelectric materials have two properties that are define as direct and converse effect.
Direct effect is the property of some materials to develop electric change on their surface
when mechanical stress is exerted on them, while converse effect is the property of some
materials to develop mechanical stress when an electric charge is induced.
 1. Full-Wave Bridge Rectifier

Full-bridge rectifier is commonly used as rectifier circuits to convert the AC output of a

piezoelectric into a DC voltage. The rectifying circuits consist of 4 diodes. The voltage needs
to rectify due to the need for constant supply of voltage light up the series of LED placed in
parallel

A diode bridge is an arrangement of four (or more) diodes in a bridge circuit configuration
that provides the same polarity of output for either polarity of input.
When used in its most common application, for conversion of an alternating-current (AC)
input into a direct-current (DC) output, it is known as a bridge rectifier. A bridge rectifier
provides full-wave rectification from a two-wire AC input, resulting in lower cost and weight
as compared to a rectifier with a 3-wire input from a transformer with a centre-
tapped secondary winding.
Smoothing Capacitor

 A Smoothing capacitor is a capacitor that acts to smooth or even out fluctations in a

signal.
 The most common and used application for smoothing capacitors is after a power
supply voltage or a rectifier.
 Power supply voltage can sometimes supply erratic and unsmooth voltages that
fluctuate greatly.
 When a steady DC signal is needed and is necessary, a smoothing capacitor is the right
component needed in order to smooth out the fluctuating signal to make it more
steady.
 We'll go over an example of this now.
 A prime example of when a smoothing capacitor is used is in conjunction with a
rectifier circuit.
 If you place a resistor in series with a diode and then input an AC signal into the
circuit:
Resistors

Ar esisto
r is
ap assiv
et wo-

terminal electrical component that implements electrical resistance as a circuit element. In

electronic circuits, resistors are used to reduce current flow, adjust signal levels, to divide
voltages, bias active elements, and terminate transmission lines, among other uses. High-
power resistors that can dissipate many watts of electrical power as heat may be used as part
of motor controls, in power distribution systems, or as test loads for generators. Fixed
resistors have resistances that only change slightly with temperature, time or operating
voltage. Variable resistors can be used to adjust circuit elements (such as a volume control or
a lamp dimmer), or as sensing devices for heat, light, humidity, force, or chemical activity.
Resistors are common elements of electrical networks and electronic circuits and are
ubiquitous in electronic equipment. Practical resistors as discrete components can be
composed of various compounds and forms. Resistors are also implemented
within integrated circuits.
The electrical function of a resistor is specified by its resistance: common commercial
resistors are manufactured over a range of more than nine orders of magnitude. The nominal
value of the resistance falls within the manufacturing tolerance, indicated on the component.
Multimeter

A multimeter or a multitester, also known as a VOM (volt-ohm-milliammeter), is

an electronic measuring instrument that combines several measurement functions in one unit.
A typical multimeter can measure voltage, current, and resistance. Analog multimeters use
a microammeter with a moving pointer to display readings. Digital multimeters (DMM,
DVOM) have a numeric display, and may also show a graphical bar representing the
measured value. Digital multimeters are now far more common due to their cost and
precision, but analog multimeters are still preferable in some cases, for example when
monitoring a rapidly varying value.

A multimeter can be a hand-held device useful for basic fault finding and field service work,
or a bench instrument which can measure to a very high degree of accuracy. They can be
used to troubleshoot electrical problems in a wide array of industrial and household devices
such as electronic equipment, motor controls, domestic appliances, power supplies, and
wiring systems.
Design of the load Capacitance load and Resistance Load

The design of the capacitance load of full-wave rectifier circuit is based on the criterion of
time charging, level of voltage provided on the load and stored energy in the form of electric
field. It is also assumed that the resistor is an open circuit (infinity value). A sinusoidal signal
of excitation of amplitude 0.5 V was applied on the Thevenin source voltage. Figure 6(a)
shows the experimental rectifier circuit implemented on a protoboard. The electromechanical
cantilever beam is numerically simulated trough a Thevenin source voltage with a PZT
capacitance PZT calculated by Eq. (9). The Simulink software with the Simulation Power
System toolbox is used to perform numerical simulations, Fig. 6(b). The frequencies used in
the test were set based on Fig. 2, 58 Hz and 437 Hz. The 58 Hz is the first natural frequency
with the greatest intensity of vibration in the beam. The natural frequency of 437 Hz is used
only to compare the results. Figure 7 presents the assessment of capacitive load in the diode
bridge for both frequencies.
As one can observe in Fig. 7 for both frequencies the load capacitance of CL = 1 mF takes
longer to reach the steady state, but without completing the load. The other capacitances
reach the steady state before with almost the same voltage in both. Since both capacitors
reach the steady state with the same voltage level, one can use Eq. (7) to choose the
capacitance to provide the greater energy stored in the form of electric field. Thus, it is
determined that the load capacitor with CL = 1 µF is a right choice for both frequencies of 58
and 437 Hz, even where the voltage is lower. For the frequency of 437 Hz and when the CL
= 1 µF, the steady voltage amplitude is = 0.4 V and = 80 nJ calculated by Eq. (9). For CL = 1
nF the steady voltage amplitude is = 0.419 V and = 7.6 pJ. Finally, for CL = 1 pF, the steady
amplitude is = 0.42 V and = 0.088 pJ. Thus, the capacitance of CL = 1 µF is a viable choice,
even with the low level of voltage amplitude, the level of energy and the current provided is
bigger when compared to the other capacitance values tested.

Another important task is to choose an appropriate resistive load to consume the active
power. This passive element represents an electronic device and is designed to maintain a
certain amount of ripple and voltage level measured at its terminals. It is also necessary to
consider that the power consumed is directly proportional to the square of the voltage and
inversely proportional to resistance. The same voltage input signal was applied to the source,
but now the resistor is not an open circuit. Table 1 presents the average values of voltage,
RMS voltage of AC ripple and resistance to different values. The frequencies of 58 Hz and
437 Hz were assumed, as well as the same capacitance CL and AC voltage source.
v v P
Freq. RL avg rms Rip. Lavg

AC

[Hz] [Ω] [V] [V] [%] [µW]

58 1 kΩ 0.020 0.012 61.175 0.417

58 1 MΩ 0.387 0.028 7.353 0.150

58 1 GΩ 0.396 0.029 7.432 0.000

58 1 TΩ 0.396 0.029 7.433 0.000

437 1 kΩ 0.103 0.016 15.553 10.552

437 1 MΩ 0.395 0.011 2.838 0.156

437 1 GΩ 0.399 0.011 2.860 0.000

437 1 TΩ 0.399 0.011 2.860 0.000

Analyzing the results in Table 1, it is observed that the higher the resistance value RL, the
ripple voltage is lower and the load voltage is greater. The increase of the voltage load is
caused by the increase of RL, because the implementation of CL and RL parallel increases too,
and is still the same, causing a greater voltage on the load. The ripple decreases because
increases as the resistance increases, causing the discharge signal to be smooth. For both
frequencies, the resistors with resistance of 1 MΩ, 1 GΩ and 1 TΩ have approximately the
same ripple and average voltage, but as of the resistor 1 MΩ is bigger, this value is the
chosen one. Therefore, the resistance chosen is 1 MΩ for the frequencies of 58 Hz and 437
Hz with a capacitive load of CL = 1 µF in parallel to RL = 1 MΩ. It is worth noting that the
Table 1 shows only same resistance candidates for the best value by considering ripple v avg
and PL avg power consumed by the load. The values RL = 500 kΩ could be refined to find
better parameters, e.g., by simulating a resistance of RL = 500 kΩ and observing if the
ripple, V avg and PL avg improve as a function of the parameter concerned. The same
consideration can be performed to choose the CL.

Experimental Results

The experimental test of the power harvesting device designed is performed by using a
modal shaker TJ 50 TMC, where the cantilever beam is attached, a power amplifier TE
5874A TMC, sweeping and sinusoidal signal generator TE 1311E TMC, Displace 1104 data
acquisition board controlled by Simulink and Control Desk software and a protoboard to
attach the circuit designed. Figure 8 shows the complete experimental setup utilized in the
tests with a schematic diagram.

The beam has two PZTs coupled, and the ceramics have been connected in parallel and the
layers were coupled with the poling direction out-of-phase. The full-wave rectifier circuit
was composed of four 1N5817 diodes, resistor, and electrolytic capacitor, as described
before.

In all tests of this section, the output voltage data in PZT and in the load were recorded with
a sampling rate of 8 kHz and with 16001 samples in each file, being measured during two
seconds. In order to monitor the level of displacement and force applied in the beam an
oscilloscope was employed to measure these parameters.

Figure 9 presents the comparison between the numerical and experimental test. The
simulation was performed with Simulink, Fig. 6(b), and by using the peak value of sinusoidal
voltage source of 0.43 V, measured experimentally. In the Simulink, the ode23t (Mod.
stiff/Trapezoidal) command is used as solver with variable time step.

Now using the excitation frequency of 437 Hz, the same test was done. The value of force
was set by using the gain in the power amplifier to supply the shaker in order to reach a level
of 0.47 V in the PZT. The voltage magnitude across the PZT to the frequency of 58 Hz and
437 Hz are different because the highest possible gain has been avoided before the
accelerometer signal becomes noisy. Figure 10 presents the comparison between the
numerical and experimental test.

Figures 11(a) and 11(b) indicate the experimental PZT and load voltage for the steady
condition of the frequencies of 58 Hz and 437 Hz, respectively.

Finally, a direct comparison between the load voltages obtained experimentally and
numerically is shown in Fig. 12 for both frequencies. The level of voltage and the general
behaviour are well simulated with the mathematical model provided by Simulink. However,
it is clear that the experimental ripple is greater. The diode is simulated by static
representation, as a resistance in series with a DC voltage source, where the real circuit
highlights the dynamic characteristics of the diode, including non-linearity of this
component. As the curve shows different points of slope (and thus resistance), the static
modelling loses some information. The ripple is presented as something dynamic that is not
to be represented in static form, but the DC voltage level had good representation. So, a
better model updating is wished in the next tests. Another point is relative to compare other
numerical methods to solve the non-linear equations in the rectifier circuits.

Application

The power harvesting device is tested for a possible application as an indicator of the level of
charging in a battery. A Light-Emitting Diode (LED) is installed parallel to a capacitor to
indicate when the charge in the capacitor reaches the adequate value of voltage. Figure 13
shows the experimental setup used.
 (a) Resonance Frequency Comparison.

(b) Simple Frequency Comparison.

Figure 14. Time of loading and the voltage level on the capacitive load for different
frequency and level of force amplitude.

If the LED is on, the capacitor voltage reaches a minimum level expected and, consequently,
the stored charge can be calculated by Eq. (8). Two sets of test were realized to lighting the
LED. For both tests, the load is = 3300 , is the LED and a sampling rate of 2 kHz is used.
These tests were recorded with a sampling rate of 2 kHz and with 200001 samples per file,
being measured by one hundred seconds.
The aim of the first analysis is to compare the time of loading and the voltage level on the
capacitive load when the excitation signal has the frequency equal to the first natural
frequency of 58 Hz and to the natural frequency of 437 Hz. The gain in the power amplifier
to supply the modal shaker is maintained constant in both cases.

Figure 14(a) shows the voltage across the LED (and capacitor) for different frequency
values, but keeping the fixed applied force. The displacement in the clamped of the beam for
58 Hz was higher than the frequency of 437 Hz, and the LED was on for about 120 seconds.
For the frequency of 437 Hz, after a wait of 6 minutes the LED is still not lit.

The second analysis performed the same comparison, but by considering a resonance
frequency of 437 Hz and any one frequency of 300 Hz. Figure 14(b) presents the results of
the load voltage for both frequencies, where the same level of force was applied
corresponding to approximately three times the first analysis.

For this value of force at 437 Hz, the LED is on for about 50 seconds. Clearly, the first
resonance frequency is better to charge the load. Now, for the frequency of 300 Hz, after a
wait of 6 minutes the LED is not lit.
Advantages:

 Very high frequency response.

 Self-generating, so no need of external source.
 Simple to use as they have small dimensions and large measuring range.
 Barium quartz can be made in any desired shape and form. It also has a large dielectric
constant. The crystal axis is selectable by orienting the direction of orientation.

Disadvantages

 It is not suitable for measurement in static condition.

 Since the device operates with the small electric charge, they need high impedance
cable for electrical interface.
 The output may vary according to the temperature variation of the crystal.
  The relative humidity rises above 85% or falls below 35%, its output will be affected.
If so, it has to be coated with wax or polymer material.

Final Remarks
Vibration power harvesting represents a huge potential for commercial applications in
different areas, as presented in recent researches, mainly in networks for structural health
monitoring (SHM). In this sense, this paper illustrated the applicability and some aspects for
a fast and simple design of a device for practical application. It was shown through several
experimental tests that the higher the load resistance, the greater the voltage magnitude.
However, as the higher the resistive load, the lower the power consumed by the load, thus the
performance for a practical application that requires a level of considerable power will be
difficult. The greater the capacitance of the load, the more time is required to charge the
capacitor.

However, the ripple decreases and is shorter, causing a faster discharge. Another important
point is to choose an adequate value of excitation frequency, mainly in the first natural
frequencies where the level of vibration amplitude can be bigger. Usually, the mechanical
beam design is based on the information about the real-world mechanical source, e.g. a
machine or a structural frame. Further research is being developed by the authors to reach a
mathematical model of the mechanical vibration of the beam with multiple modes attached to
the non-controlled rectifier circuits in order to provide the information needed for this
design.

Conclusion

The project is successfully tested which is the best economical, affordable energy solution to
common people. This can be used for many applications in city areas where want more
power. Bangladesh is a developing country where energy management is a big challenge for
huge population. By using this project. I can drive D.C loads according to the force I
applied on the piezo electric sensor. Although the theory developed in this report justifies the
use of switching techniques in efficiently converting that energy to a usable form, there are
obviously some practical limitations to the systems presented.

The final prototype design does fulfil the objective of generating electricity from
piezoelectric disk. Due to the low cost design of the piezoelectric system it is a practical
product which could increase the operating period of most common products. The data
collected is capable of extending the operational lifespan per charge of portable electronic
devices.
Although the theory developed in this report justifies the use of switching techniques in
efficiently converting that energy to a usable form, there are obviously some practical
limitations to the systems presented. Measurements of source current into the primary and
load current transferred from the secondary reveal that very little current gain truly occurs
between the input and output ports of the switch in the forward converter hybrid.

Further, similar results were encountered when one examines the energy transferred through
the series switch and inductor in the buck converter. In addition, based on the results
gathered in this investigation, the final prototype design does fulfil the objective of
generating electricity from piezoelectric disk. Due to the low cost design of the piezoelectric
system it is a practical product which could increase the operating period of most common
products. The data collected is capable of extending the operational lifespan per charge of
portable electronic devices.

References

1. W. Wang, F. Ismail, F. Golnaraghi, A neuro-fuzzy approach to gear system monitoring

2. M. Bhardwaj, T. Garnett, A. P. Chandrakasan, "Upper bounds on the lifetime of sensor

3. C. B. Williams, R. B. Yates, "Analysis of a micro-electric generator for microsystems", Show

4. S. Roundy, P. Wright, J. Rabaey, "A study of low level vibrations as a power source for wireless
sensor nodes",

5. S. Priya, "Advances in energy harvesting using low profile piezoelectric transducers", J. Electro

6. H. A. Sodano, D. J. Inman, G. Park, "Comparison of piezoelectric energy harvesting devices for

recharging batteries", J. Intell. Mater. Syst. Struct