You are on page 1of 10

Estimation of Electric Charge Output for

Piezoelectric Energy Harvesting

H. A. Sodano*, G. Park† and D. J. Inman*
*Center for Intelligent Material Systems and Structures, Virginia Polytechnic Institute and State University, Blacksburg, VA 24061, USA

Engineering Sciences and Applications, Weapon Response Group, Los Alamos National Laboratory, Los Alamos, NM 87545, USA

ABSTRACT: Piezoelectric materials (PZT) can be used as mechanisms to transfer mechanical

energy, usually ambient vibration, into electrical energy that can be stored and used to power other
devices. With the recent advances in wireless and micro-electro-mechanical-systems (MEMS)
technology, sensors can be placed in exotic and remote locations. As these devices are wireless it
becomes necessary that they have their own power supply. The power supply in most cases is the
conventional battery; however, problems can occur when using batteries because of their finite life
span. Because most sensors are being developed so that they can be placed in remote locations such
as structural sensors on a bridge or global positioning service (GPS) tracking devices on animals in the
wild, obtaining the sensor simply to replace the battery can become a very expensive task. Fur-
thermore, in the case of sensors located on civil structures, it is often advantageous to embed them,
making access impossible. Therefore, if a method of obtaining the untapped energy surrounding
these sensors was implemented, significant life could be added to the power supply. One method is
to use PZT materials to obtain ambient energy surrounding the test specimen. This captured energy
could then be used to prolong the power supply or in the ideal case provide endless energy for the
sensors lifespan. The goal of this study is to develop a model of the PZT power harvesting device.
This model would simplify the design procedure necessary for determining the appropriate size and
vibration levels necessary for sufficient energy to be produced and supplied to the electronic devices.
An experimental verification of the model is also performed to ensure its accuracy.

KEY WORDS: damping, piezoelectric, power harvesting, self powered

Introduction power generated when a free-falling steel ball

impacted a plate with a piezoceramic wafer attached
The idea of building portable electronic devices or to its underside. Their study used an electrical
wireless sensors that do not rely on power supplies equivalence model to simulate the energy generated
with a limited lifespan has intrigued researchers and and calculate the ability of the PZT to transform
instigated a sharp increase in research in the area of mechanical impact energy into electrical power. It
power harvesting. One method of power harvesting is was found that a significant amount of the impact
to use piezoelectric materials (PZT), which form energy was returned to the ball in the form of kin-
transducers that are able to interchange electrical etic energy during which the balls rebound off of
energy and mechanical strain or force. Therefore, the plate; however, it is stated that if the ball
these materials can be used as mechanisms to transfer impacted the plate an efficiency of 52% could be
ambient motion (usually vibration) into electrical achieved. In a later paper, Umeda et al. [2] investi-
energy that may be stored and used to power other gated the energy storage characteristics of a power
devices. By implementing power harvesting devices, harvesting system consisting of a PZT, full-bridge
portable systems can be developed that do not rectifier and a capacitor. Their work discussed the
depend on traditional methods for providing power, effect of various parameters on the efficiency of the
such as the battery, which has a limited operating life. storage circuit. Following their analytic investiga-
A significant amount of research has been devoted tion, a prototype was developed which is stated to
to developing and understanding power harvesting have an efficiency of over 35%, more than three
systems. These studies, demonstrate the feasibility of times that of a solar cell. Starner [3] examined the
using PZT devices as power sources. One early study possible location for power harvesting devices
performed by Umeda et al. [1] investigated the around the human body and surveyed the energy

 2004 Blackwell Publishing Ltd j Strain (2004) 40, 49–58 49

Estimation of Electric Charge Output for Piezoelectric Energy Harvesting : H. A. Sodano, D. J. Inman and G. Park

available from sources of mechanical energy inclu- impedance of the PZT and the load maximises the
ding blood pressure, walking, and upper limb power flow from the PZT. Their model was experi-
motion of a human being. The author claims that mentally verified using a 1-D beam structure with
8.4 W of useable power can be achieved from a PZT peak power efficiencies of c. 20%.
mounted in a shoe. Kymissis et al. [4] examined Most of the previous studies found out that the
using a piezofilm in addition to a Thunder actuator energy generated by the PZT material must be
[5, 6], to charge a capacitor and power a radio fre- accumulated before it can be used to power other
quency identification (RFID) transmitter from the electronic devices. Rather than use the traditional
energy lost to the shoe during walking. The poly- capacitor that most other studies used, Sodano et al.
vinylidene fluoride (PVDF) stave was located in the [12] investigated the use of rechargeable batteries to
sole to absorb the bending energy of the shoe, and accumulate the generated energy. The goal of this
the piezoceramic thunder actuator was located in study was to show that the small amounts of
the heel to harvest the impact energy. Their work ambient vibration found on a typical system could
showed that the power generated by the PZT devices be used to charge the battery from its discharged
was sufficient for powering functional wireless de- state and demonstrated the compatibility of
vices and were able to transmit a 12-bit signal five to rechargeable batteries and the power generated by
six times every few seconds. Following the work of PZT materials. To do this, the vibration of the air
Kymssis et al. [4], the research involving wireless compressor of a typical automobile was measured
sensors began to grow, and in 1998, Kimura [7] and a similar signal was applied to an aluminum
received a US Patent that centered on the use of plate with a PZT patch attached. It was found that
a vibrating PZT plate to generate energy sufficient to the random signal from the engine compartment of
run a small transmitter fixed to migratory birds for a car could charge the battery in only a couple of
the purpose of transmitting their identification code hours and that a resonant signal could charge the
and location. The effectiveness of the power har- battery in under an hour. Ottman et al. [13] found
vesting system is also compared with existing bat- out that if circuitry was used to maximise the energy
tery technology. Goldfarb and Jones [8] presented a generated then these storage devices could be
linearised model of a PZT stack and analysed its charged with greater efficiency. Therefore, they
efficiency as a power generation device. It was investigated the effects of utilising a DC–DC step-
shown that the maximum efficiency occurs in a low down converter with an adaptive control algorithm
frequency region, much lower than the structural to maximise the power output of the PZT material.
resonance of the stack. It is also stated that the They found out that when using the adaptive cir-
efficiency is also related to the amplitude of the cuit, energy was harvested at over four times the
input force due to hysteresis of the PZT. In addition rate of direct charging without a converter.
to the force applied in the poling direction (d33 This study concentrates on developing an analytic
mode), Clark and Ramsay [9] have investigated and model of a beam with attached PZT elements that will
compared it with the transverse force (d31 mode) for provide an accurate estimate of the power generated
a PZT generator. Their work showed that the d31 through the PZT effect. It has been found in previous
mode has a mechanical advantage in converting studies that PZT material attached to a beam with
applied pressure to working stress for power gen- cantilever boundary conditions provides an effective
eration. They concluded that a 1-cm2 piezoceramic configuration for capturing transverse vibrations and
wafer can power a MEMS device in the microwatt converting them into useful electrical power. This
range. Elvin et al. [10] theoretically and experi- configuration has proven to be effective in several
mentally investigated a self-powered wireless sensors experiments carried out by Sodano et al. [12, 14]. The
using PVDF. The power harvesting system used the model detailed in this paper is based on a more gen-
energy generated by the PVDF to charge a capacitor eral one developed by Hagood et al. [15] to estimate
and power a transmitter that could send informa- the performance of PZT shunt damping circuits for
tion regarding the strain of the beam, a distance of passive vibration control. In addition, the model
2 m. Kasyap et al. [11] formulated a lumped element developed by Crawley et al. [16] was used to develop
model to represent the dynamic behaviour of PZT in the actuation equations for PZT devices and the
multiple energy domains using an equivalent cir- constitutive equations of bimorph actuators were
cuit. Additionally, this work dealt with the con- obtained from Smits et al. [17]. However, an import-
struction of a flyback converter which allows the ant addition is made to the combination of these
circuit’s impedance to be adjusted to match that of models to accommodate power harvesting, which was
the PZT material. This concept of matching the neglected in the previous models, was to add material

50  2004 Blackwell Publishing Ltd j Strain (2004) 40, 49–58

H. A. Sodano, D. J. Inman and G. Park : Estimation of Electric Charge Output for Piezoelectric Energy Harvesting

damping; if excluded, the model can predict signifi- where c is the modulus of elasticity, e is the dielectric
cantly more energy generation that actually devel- constant and the superscript, ()S, signifies that the
oped in the real system. The following sections parameter was measured at constant strain and the
describe the development of a model of the PZT power superscript, ()E, indicates that the parameter was
harvesting device. This model would simplify the measured at constant electric field (short circuit).
design necessary for determining the appropriate size These constitutive equations relate the electrical and
and extent of vibration needed for sufficient energy to mechanical properties of the PZT element. The spe-
be produced and supplied to the desired electronic cification of these relationships allow electrome-
devices. An experimental verification of the model is chanical interaction to be included in the model. The
also performed to ensure its accuracy. Following the term e is the PZT coupling coefficient and relates the
verification of the model, the effects of power har- stress to the applied electric field. The PZT coupling
vesting on the dynamics of a structure are compared coefficient can be written as shown in Equation 6 in
with those brought on from shunt damping. terms of the more commonly specified coupling
coefficient d by:

e ¼ dij cE (6)
Model of Piezoelectric Power
Harvesting Beam where c is the modulus of elasticity and dij is the PZT
coupling coefficient with the subscripts ‘i’ and ‘j’
The following derivation uses energy methods to referring to the direction of the applied field and the
develop the constitutive equations of a bimorph PZT poling, respectively. Now we can incorporate the
cantilever beam for power harvesting. To begin the PZT properties in the potential energy function:
deviation we will start with the general form of Z Z Z
Hamilton’s principle. This states that the variational U¼ ST cs S dVs þ ST cE S dVp:  ST eT E dVp
2 Vs Vp Vp
indicator (VI) must be zero all the time, as shown
below in Equation (1): Z Z 
Z t2  ET eS dVp  ET eS E dVp (7)
Vp Vp
VI ¼ ½dT  dU þ f dxdt ¼ 0 (1)
Taking the variation of the kinetic energy from
where T, U and fdx terms are defined by: Equation 3, and the potential energy term containing
1 1 the PZT properties of Equation 7, yields:
U¼ ST T dVs þ ST T dVp  ET D dVp (2)
2 Vs 2 Vp Vp Z Z Z
Z Z dU ¼ dS cs S dVs þ dS c S dVp  dST eT E dVp
1 1
T¼ qs u_ T u_ dVs þ qs u_ T u_ dVp (3) Vs Vp Vp
2 Vs 2 Vp
X nq
X  dE eS dVp  dET eS E dVp (8)
f dx ¼ duðxi Þ fi ðxi Þ  dv qj (4) Vp Vp
i¼1 j¼1 Z Z
dT ¼ qs du_ T u_ dVs þ qp du_ T u_ dVp (9)
where U is the potential energy, T is the kinetic Vs Vp
energy, fdx is the external work applied to the system,
and superscripts S, T and E representing the strain, The variations found in Equations 4, 8 and 9 can be
stress, and electric field, respectively, D the electric substituted into Equation 1 to obtain the variational
displacement, V the volume, u the displacement, x indicator:
the position along the beam, v the applied voltage, q Z t 2 Z Z
the charge, q the density, f the applied force and the VI ¼ qs du_ T u_ dVs þ qp du_ T u_ dVp :
t1 Vs Vp
subscripts ‘p’ and ‘s’, represent the PZT material and Z Z
the substrate, respectively. Before the variational  dST cs S dVs  dST cE S dVp
Vs Vp
indicator can be used to solve for the equations of Z Z Z
motion, the PZT constitutive equations need to be þ dST eT E dVp þ dET eS dVp þ dET eS E dVp
Vp Vp Vp
introduced into the potential energy term and the
X nq
variation of both the potential and kinetic energy
þ duðxi Þ f ðxi Þ  dv qj (10)
must be found. First the PZT constitutive equations i¼1 j¼1
will be introduced, which are:
   E   This equation can now be used to solve for the
T c eT S equations of motion of any mechanical system
¼ (5)
D e eS E containing PZT elements. In order to solve

 2004 Blackwell Publishing Ltd j Strain (2004) 40, 49–58 51

Estimation of Electric Charge Output for Piezoelectric Energy Harvesting : H. A. Sodano, D. J. Inman and G. Park

Equation 10 for the cantilever beam with bimorph

PZT elements, some assumptions must be made. The
first assumption follows the Rayleigh–Ritz procedure,
which says that the displacement of the beam can be
written as the summation of modes in the beam and
a temporal coordinate [18]:
uðx; t Þ ¼ /i ðxÞri ðt Þ ¼ /ðxÞr ðt Þ (11)

where /i(x) is the assumed mode shapes of the

structure which can be set to satisfy any combination
of boundary conditions, r(t) is the temporal coordi-
nate of the displacement and N is the number of Figure 1: Schematic of beam describing the variables
modes to be included in the analysis. The second
assumption made is to apply the Euler–Bernoulli
beam theory. This allows the strain in the beam to be T
Ks ¼ y 2 / ðxÞ00 cs /ðxÞ00 dVs
the product of the distance from the neutral axis and Vs
Z (15)
the second derivative of displacement with respect Kp ¼
y2 / ðxÞ00 cE /ðxÞ00 dVp
to the position along the beam. Once the strain is Vp

defined in this way Equation 11 can be used define The electromechanical coupling matrix, Q, and the
the strain as follows: capacitance matrix, Cp, are defined by:
@ 2 uðu; t Þ
S ¼ y ¼ y/ðxÞ00 ðt Þ (12) H¼ y/T ðxÞ00 eT wðyÞdVp
@x2 Vp
Z (16)
The third and last assumption is that the electric Cp ¼ wT ðy ÞeS wðy ÞdVp
potential across the PZT element is constant. This Vp

assumption also indicates that no field is applied to The parameters defined in Equations 14, 15 and 16
the beam, which in latter equations designates the can be substituted into variational indicator of
beam to be inactive material: Equation 10. This substitution allows the variational
8 indicator to be written as:
< v=tp t=2 < y < t=2 þ tp
E ¼ wðy Þvðt Þ ¼ 0 t=2 < y < t=2 Z t2 h

: VI ¼ dr_ T ðt Þ Ms þ Mp r_ ðt Þ  dr T ðt Þ Ks þ Kp r ðt Þ
v=tp t=2  tp < y < t=2

where w(y) defines the field over the thickness of the þ dr T ðt Þhvðt Þ þ dv ðt ÞHT r ðt Þ þ dv ðt ÞCp vðt Þ
PZT, which is assumed to be constant. The previous nf
X nq
X i
þ dr ðt Þ/ðxi ÞT fi ðt Þ  dvqj ðt Þ dt ¼ 0 (17)
assumption is for a beam with bimorph PZT elements i¼1 j¼1
on the top and bottom of the beam as shown
in Figure 1. The beam in Figure 1 also shows the where d(Æ) indicates the variation of the correspond-
notation for the geometry of the beam that is used ing variable. Taking the integral of the variational
throughout the derivation. indicator leaves two coupled equations. The two
Using the previous assumptions we can simplify equations shown below are coupled by the previously
the variational indicator to include terms that defined electromechanical coupling matrix Q. The
represent physical parameters. By doing this, the first equation defines the mechanical motion and the
equations describing the system become more second equation defines the electrical properties of
recognisable when compared with those of a typical the system:
system and help give physical meaning to the nf

parameters in the equations of motion. The mass Ms þMp €r ðt Þþ Ks þKp r ðt ÞHvðt Þ¼ /ðxi ÞT fi ðt Þ
i¼1 (18)
matrices for the system can be written as:
HT r ðt ÞþCp vðt Þ¼qðt Þ
Ms ¼ qs /T ðxÞ/ðxÞdVs These equations now represent the electro-mechan-
Z (14) ical system and can be used to determine the motion
Mp ¼ qp /T ðxÞ/ðxÞdVp of the beam, however this system of equations does
not contain any energy dissipation. Because the
The stiffness matrices can be written as: model is intended to represent a power harvesting

52  2004 Blackwell Publishing Ltd j Strain (2004) 40, 49–58

H. A. Sodano, D. J. Inman and G. Park : Estimation of Electric Charge Output for Piezoelectric Energy Harvesting

system that must be removing energy, this form is through its own inertia. The standard boundary
not suitable for our needs, as it does not account for conditions of the clamped end of the beam say that
energy lost through the structure. In addition, the the slope and displacement are zero at all time. For
energy removed from the system through energy the condition of base motion, the zero displacement
harvesting must be accounted for. To incorporate condition would not be held and a new set of mode
energy dissipation into the equation one can use shapes would need to be generated. Rather than do-
Ohm’s law and add a resistive element between the ing this, it was decided that a force corresponding to
positive and negative electrodes of the PZT. The the inertia of the beam when subjected to the base
resistive element will provide a means of removing motion could be used and the clamped-free mode
energy from the system. Then electrical boundary shapes would still be valid. The forcing function used
condition becomes: to model the inertia of the beam is:
vi ðt Þ ¼ Rq_ ðt Þ (19) L b t
f ðt Þ ¼ qAx2 sinðxt Þ dz dy dx (23)
0 0 0
In addition, the system should have some type of
additional mechanical damping that needs to be With the forcing function defined, everything
accounted for. If only the electrical damping is necessary to simulate the power harvesting system
accounted for, the model will over predict the actual has been included in the model. The following sec-
amount of power generated. The amount of tions of this paper will describe the experimental
mechanical damping added to the model was procedures and results, in order to demonstrate the
determined from experimental results. This is done accuracy of this model.
using proportional damping methods and the
damping ratio that is predicted from the measured
frequency response function. With the damping Experimental Setup for Model Verification
ratio known, proportional damping can be found
from [18]: The accuracy of the model will be tested on a Quick

Pack model QP40N (Midé Technology Corporation,
C ¼ a Ms þ Mp þ b Ks þ Kp (20)
Medford, MA, USA), although it could be used to
where a and b are determined from: model any beam with attached PZT. The QP40N is a
bimorph actuator with dimensions and properties
a bxi shown in Figure 2. The Quick Pack actuator is con-
fi ¼ þ i ¼ 1; 2; . . . ; n (21)
2xi 2 structed from four piezoceramic wafers embedding in
where fi is the damping ratio found from the fre- a Kapton and epoxy matrix. Experiments were per-
quency response of the structure. Incorporating formed to verify the accuracy of the models ability to
Equations 19 and 20 into Equation 18, results in the predict the amount of power generated from this
final model of the power harvesting system: device when subjected to transverse vibrations of

varying frequency and amplitude. As mentioned
Ms þ Mp €r ðt Þ þ Cr_ ðt Þ þ Ks þ Kp r ðt Þ  HC1
p qð t Þ

¼ / ð xi Þ T f i ð t Þ (22)

Rq_ ðt Þ  C1 1
p H r ðt Þ þ Cp qðt Þ ¼ 0:

Device size (mm): 100.6 × 25.4 × 0.762

Equation 22 provides an accurate model of the power
Device weight (g): 9.52
harvesting system. The q_ ðt Þ term provides the current Active elements: two stacks of two piezos
output of the PZT element and can be directly related Piezo wafer size (mm): 45.974 × 20.574 × 0.254
to the power output of the PZT through the load Full-scale voltage range (V): ±200
resistance R.
The last portion of our model left undefined is the
forcing function. The system that will be investigated
is a cantilever beam that is excited by transverse
vibrations of the structure that it is clamped to;
therefore, no force is directly applied to the beam.
Instead the clamped end of the beam is experiencing Figure 2: Midé Technology Corporation Quick Pack model
base motion and transferring that energy to the beam QP40N (from Midé Technology Corporation)

 2004 Blackwell Publishing Ltd j Strain (2004) 40, 49–58 53

Estimation of Electric Charge Output for Piezoelectric Energy Harvesting : H. A. Sodano, D. J. Inman and G. Park

Table 1: Properties of the Quick Pack

Property Symbol Value

Dielectric constant k3T 1800

Piezoelectric strain coefficient d13 )179 · 10)12mV)1
Modulus of piezoelectric cE 63 GPa
Modulus of Kapton–epoxy cs 2.5 Gpa
Modulus of Quick Pack cb 35.17 GPa
Density of piezo material q 7700 kg m)3
Density composite matrix q 2150 kg m)3

corresponding deflection needed to be measured. The

experimental setup consists of a Transducer Tech-
niques 100 gram load cell model GSO-100C and a
Polytec laser vibrometer. The load cell was mounted
Figure 3: Quick Pack QP40N attached to the shaker and
dimensions of beam when one end is clamped on a lead screw to allow a steady force to be applied at
the tip of the beam. The results of this test found the
modulus of the beam to be 2.5 GPa. The reason for
previously we were interested in the Quick Pack
this value being so low is due to the area at the mid-
being mounted with cantilever boundary conditions.
span of the beam that consisted of only the Kapton
To provide the transverse vibration, the Quick Pack
and epoxy. When the static tests were performed on
was mounted actuator to an electromagnetic shaker
the beam it was apparent that the majority of the
as shown in Figure 3.
bending was occurring at this location. Therefore, it
One complication that arose when modelling the
was concluded that the experimental tests performed
Quick Pack actuator was due to its composite struc-
had actually measured the modulus of elasticity cor-
ture and the PZT wafers not spanning the entire
responding to the Kapton and epoxy portion of the
length of the beam, which can be seen in Figure 2.
beam. This still left the overall modulus of the beam
Because the area of the beam with no PZT wafer
unknown. The value calculated for the overall the
consisted of only Kapton and epoxy, it contained a
modulus of the beam was found by simply averaging
localised area with a lower modulus of elasticity. The
the modulus of the PZT material, that was supplied
manufacturer did not specify a value for the effective
by the manufacturer and the experimentally found
modulus of the complete beam. Therefore, the setup
modulus for the Kapton and epoxy matrix according
shown in Figure 4 was constructed to measure the
to their individual per cent of the cross sectional area.
stiffness of the Quick Pack. To obtain a value for
The resulting modulus of elasticity and the other PZT
the stiffness the force applied to the beam and its
properties used are shown in Table 1.

Model Verification
The accuracy of the model was compared against
experimental results to demonstrate the ability of the
model to accurately predict the amount of power
produced by the Quick Pack when subjected to
transverse vibration. To ensure that the model and
experimental tests were subjected to the same exci-
tation force an accelerometer was used to calculate
the amplitude of the sinusoidal force applied to the
beam through:
a ¼ Ax2 sinðxt Þ ) Amax ¼ (24)
where a is the acceleration of the clamped end of the
Figure 4: Experimental setup used to find the elastic modulus beam. The beam was excited by a sinusoidal input and
of the Kapton–epoxy matrix the steady state power output was measured across

54  2004 Blackwell Publishing Ltd j Strain (2004) 40, 49–58

H. A. Sodano, D. J. Inman and G. Park : Estimation of Electric Charge Output for Piezoelectric Energy Harvesting

–20 0.05
Predicted Model
Measured 0.04 Experimental

–60 0.02

Current (mA)
Mode shapes


–120 –0.02


–160 –0.05
0 200 400 600 800 1000 1200 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Frequency (Hz) Time (s)

Figure 5: Frequency response of the model and the Figure 7: Output current predicted by model and measured
experimental data across a 100-kX resistor at 25 Hz

several different resistors. The frequency response of The measured current generated by the Quick
the model and the experimentally tested Quick Pack Pack was compared to the predicted current from
are shown in Figure 5. The differences in the two the model for various frequencies and load resi-
responses are attributed to the Quick Pack’s compos- stances. The output current across a 100 kX resistor
ite structure resulting in coupled modes and the for an excitation frequency of 25 Hz of the model
nonlinear properties of the Kapton material, especi- and the measured current obtained through experi-
ally its modulus of elasticity that varies nonlinearly ments are shown in Figure 7. The predicted response
with frequency. Additionally, looking at the mode shown in these figures shows a transient response
shapes of a cantilever beam shown in Figure 6, it can for a small period of time while the experimental
be seen that in the second mode a large amount of results do not because they were recorded at steady
bending occurs at the beam’s midsection. However state vibration. Table 2 provides an approximate list
the Quick Pack has an area of low stiffness at the of the measured and predicted currents generated
midsection but because of the use of a uniform for numerous values of frequency and load resist-
modulus of elasticity and density in the model, the ance. The values in this table demonstrate that the
stiffness is increased at this location and the predicted model provides a very accurate measurement of the
frequency of the second mode is higher than meas- power generated at various frequencies and resistive
ured. It is expected that a beam constructed of a loads. This shows that the model would be effective
homogeneous material with a PZT mounted to is as a design tool for determining the ideal size and
surface would produce a more accurate frequency excitation level necessary to provide the desired
response. power.

1 First mode Table 2: Amplitude of measured and simulated current

Second mode
Third mode
Frequency Load Simulated Measured Percent
Normalized displacement

0.5 (Hz) resistance (X) current (mA) current (mA) error

25 100 0.101 0.104 2.97

25 10 000 0.105 0.106 0.95
25 100 000 0.032 0.032 0.00
30 100 0.360 0.345 4.17
30 10 000 0.295 0.30 1.67
30 100 000 0.065 0.068 4.61
50 100 0.20 0.20 0.00
50 10 000 0.175 0.180 2.86
50 100 000 0.033 0.032 3.03
0 0.2 0.4 0.6 0.8 1
x/1 75 10 000 0.142 0.144 1.41
150 10 000 0.0132 0.0133 0.75
Figure 6: First three mode shapes of a cantilever beam

 2004 Blackwell Publishing Ltd j Strain (2004) 40, 49–58 55

Estimation of Electric Charge Output for Piezoelectric Energy Harvesting : H. A. Sodano, D. J. Inman and G. Park

Discussion of Power Harvesting and 0.1

Shunt Damping 0.08 System: SYS

Setting time (s): 0.945
In addition to providing an accurate estimate of the 0.04 15 kΩ resistance
power generated by a beam with a complicated PZT

layout and a non-homogeneous material composi-
tion, the model also demonstrates that power 0

harvesting works much like a shunt damper. When a –0.02

power harvesting system is implemented, energy is –0.04
removed from the system and supplied to the
desired electrical components. As a result of the
removal of energy from the system conservation of 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
energy says that increased damping must occur. This Time (s)
is the same principle as that used in shunt damping Figure 9: Impulse response with a 15-kX resistive load
systems. However, when implementing shunt cir-
cuits it is often advantageous to use a resistor,
inductor and capacity (RLC) circuit that allows the 0.08
circuit to be tuned to the resonant frequency of the System: SYS
0.06 Setting time (s): 1.35
system for maximum power dissipation. In the case
of the power harvesting system that has been 0.04
100 kΩ resistance
investigated in this paper, a constant load resistance 0.02

was used which will also induce damping into the

system but over a broad range of frequencies rather
than the turned frequency of a RLC circuit. Only a –0.02
load resistance has been used in this study due to –0.04
the ability to match the impedance of the circuit or
load and the PZT using circuitry such as a flyback
converter discussed in the paper by Kasyap et al. –0.08
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
[11]. The damping effect caused by power harvesting
Time (s)
on the impulse response of a beam for three differ-
ent load resistances is shown in Figures 8–10. In Figure 10: Impulse response with a 100-kX resistive load
Figure 8 the load resistance is set at a low value of
100 X, which does not dissipate a large amount of apparent in the increased settling time of the
energy, causing only a small amount of damping to response. Now Figure 10 shows that the load resist-
be added to the system. In the case of Figure 9, the ance is further increased to 100 kX, giving the sys-
load resistance is set at an ideal value of 15 kX tem the ability to dissipate a large amount of energy
allowing the maximum flow of energy from the PZT from the system. However, when the load resistance
device and, in turn, causing higher damping that is becomes very high, the ability of energy to flow
from the PZT material is reduced causing the damp-
0.1 ing induced in the system to decrease. These figures
0.08 demonstrate the effect of power harvesting on the
dynamics of a structure. It is apparent that as more
energy is removed from the system the impulse dies
0.04 100 Ω resistance
out faster until a critical level is reached, after which
0.02 the resistive load of the circuit exceeds the imped-

0 ance of the PZT network causing lower efficiency

power generation and for this example lower energy
dissipation to the beam. The critical resistance at
–0.04 which the most efficient energy is generated occurs
System: SYS
–0.06 Setting time (s): 1.6 when the load resistance is matched with the
impedance of the PZT device. Therefore, the use of a
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 flyback converter or other circuitry to tune the load
Time (s)
resistance is a necessity in order to maximise the
Figure 8: Impulse response with a 100-X resistive load energy removed from the system. This study shows

56  2004 Blackwell Publishing Ltd j Strain (2004) 40, 49–58

H. A. Sodano, D. J. Inman and G. Park : Estimation of Electric Charge Output for Piezoelectric Energy Harvesting

that the effects of power harvesting on the dynam- REFERENCES

ics of a mechanical system are very similar to those 1. Umeda, M., Nakamura, K. and Ueha, S. (1996) Analysis of
of shunt damping, with the major difference being transformation of mechanical impact energy to electrical
that the energy is stored for use rather than of energy using a piezoelectric vibrator. Japanese Journal of
Applied Physics 35, 3267–3273.
2. Umeda, M., Nakamura, K. and Ueha, S. (1997) Energy
storage characteristics of a piezo-generator using impact
induced vibration. Japanese Journal of Applied Physics 36,
Conclusions 3146–3151.
3. Starner, T. (1996) Human-powered wearable computing.
One method of performing power harvesting is to use IBM Systems Journal 35, 618–628.
PZT materials that can convert the ambient vibration 4. Kymissis, J., Kendall, C., Paradiso, J. and Gershenfeld, N.
energy surrounding them into electrical energy. This (1998) Parasitic Power Harvesting in Shoes. Second IEEE
electrical energy can then be used to power other International Symposium on wearable Computers, October
devices or stored for later use. This technology has 19–20th, Pittsburg, PA, pp. 132–139.

gained an increasing attention due to the recent 5. Dausch, D. and White, S. (1998) Composition Effects on
Electromechanical Degradation of RAINBOW Actuators. Nati-
advances in wireless and MEMS technology, allowing
onal Aeronautics and Space Administration, Hampton, VA.
sensors to be placed in remote locations and operate
6. Face International Corporation (2001) Product Information,
at very low power. The need for power harvesting
Thunder Actuators and Sensors. Face International
devices is caused by the use batteries as power sup- Corporation, Norfolk, VA.
plies for these wireless electronics. As the battery has thunder/tech/ttech.htm.
a finite lifespan, once extinguished of its energy, the 7. Kimura, M. (1998) Piezoelectric Generation Device, US
sensor must be recovered and the battery replaced for Patent Number 5801475.
the continued operation of the sensor. This practice 8. Goldfarb, M. and Jones, L. D. (1999) On the efficiency of
of obtaining sensors solely to replace the battery can electric power generation with piezoelectric ceramic.
become and expensive task, because their wireless ASME Journal of Dynamic Systems, Measurement, and
Control, 121, 566–571.
nature allows them to be placed in exotic locations.
Therefore, methods of harvesting the energy around 9. Clark, W. and Ramsay, M. J. (2001) Smart material trans-
ducers as power sources for MEMS devices. In: Proceedings
these sensors must be implemented to expand the life
of SPIE’s 8th Annual International Symposium on Smart
of the battery or ideally provide an endless supply of Structures and Materials, Vol. 4332. San Diego, CA, USA:
energy to the sensor for its lifetime. 429–438.
We have developed a model to predict the amount 10. Elvin, N. G., Elvin, A. A. and Spector, M. (2001) A self-
of power capable of being generated through the powered mechanical strain energy sensor. Smart Materials
vibration of a cantilever beam with attached PZT and Structures 10, 293–299.
elements. The derivation of the model has been 11. Kasyap, A., Lim, J., Johnson, D., Horowitz, S., Nishida, T.,
provided, allowing it to be applied to a beam with Ngo, K., Sheplak, M. and Cattafesta, L. (2002) Energy
reclamation from a vibrating piezoceramic composite
various boundary conditions or layout of PZT pat-
beam. In: Proceedings of 9th International Congress on Sound
ches. The model was verified using experimental
and Vibration, Orlando, FL, Paper No. 271.
results and proved to be very accurate independent of
12. Sodano, H. A., Magliula, E. A., Park, G. and Inman, D. J.
excitation frequency and load resistance. In addition, (2002) Electric power generation from piezoelectric
the verification of the model was performed on a materials. In: The 13th International Conference on Adaptive
structure the contained a complex PZT layout and a Structures and Technologies, 7–9 October, Potsdam, Berlin,
non-homogenous material beam, indicating that the Germany.
model is robust and can be applied to a variety of 13. Ottman, G. K., Hofmann, H., Bhatt A. C. and Lesieutre, G.
different mechanical conditions. The damping effects A. (2002) Adaptive piezoelectric energy harvesting circuit
for wireless, remote power supply. IEEE Transactions on
of power harvesting were also shown to be predicted
Power Electronics 17, 669–676.
in the model and to follow that of a resistive shunt
14. Sodano, H. A., Park, G., Leo, D. J. and Inman, D. J.
damping circuit. This model provides a design tool
(2003) Use of piezoelectric energy harvesting devices
for developing power harvesting systems by assisting for charging batteries. In: SPIE’s 10th Annual International
in determining the size and extent of vibration Symposium on Smart Structures and Materials, 2–6 March,
needed to produce the desired level of power San Diego, CA.
generation. The potential benefits of power harvest- 15. Hagwood, N. W., Chung, W. H. and Von Flowtow, A.
ing and the advances in low power electronics and (1990) Modelling of piezoelectric actuator dynamics for
active structural control. Journal of Intelligent Material
wireless sensors are making the future of this tech-
Systems and Structures 1, 327–354.
nology look very bright.

 2004 Blackwell Publishing Ltd j Strain (2004) 40, 49–58 57

Estimation of Electric Charge Output for Piezoelectric Energy Harvesting : H. A. Sodano, D. J. Inman and G. Park

16. Crawley, E. F. and Anderson, E. H. (1990) Detailed models 18. Inman, D. J. (2001) Engineering Vibration. Prentice-Hall
of piezoceramic actuation of beams. Journal of Intelligent Inc., Upper Saddle River, NJ.
Material Systems and Structures 1, 4–25.
17. Smits, J., Dalke, S. and Cooney, T. K. (1991) The con-
stituent equations of piezoelectric bimorphs. Sensors and
Actuators 28, 41–61.

One day seminar on

Industrial Applications of Optical Methods
for Deformation & Strain Measurement

4th November 2004

Airbus UK Ltd, Bristol

Full field measurements of structural tests are increasingly used to help

engineers validate complex computer models. These enhanced simulations
lead to reduced costs and product development times.

The seminar is aimed at showing a variety of optical methods which are

being developed and used for the measurement of deformation and strain in
engineering components and structures. The emphasis is on industrial or
research applications and will focus on the implementation of established
and novel techniques.

To register please contact Sally Cryer, 38 Park Road North, Bedford, MK41 7RH
Tel/fax: 0845 1668382 Email:

Or download registration form from

58  2004 Blackwell Publishing Ltd j Strain (2004) 40, 49–58