Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

3338 IEEE SENSORS JOURNAL, VOL. 15, NO.

6, JUNE 2015

Performance Enhancement of Piezoelectric


Energy Harvesters Using Multilayer and
Multistep Beam Configurations
Rammohan Sriramdas, Sanketh Chiplunkar, Ramya M. Cuduvally, and Rudra Pratap, Member, IEEE

Abstract— Low-power requirements of contemporary sensing stress and strain, thermal energy, indoor sun light, results
technology attract research on alternate power sources that can in μW to mW level of power that is suitable for replac-
replace batteries. Energy harvesters absorb ambient energy and ing conventional batteries [1] for low power applications.
function as power sources for sensors and other low-power
devices. Piezoelectric bimorphs have been demonstrating the The structural health monitoring systems and continuous
preeminence in converting the mechanical energy in ambient monitoring for catastrophe prevention systems require unin-
vibrations into electrical energy. Improving the performance of terrupted power supply. Periodic replacement of batteries in
these harvesters is pivotal as the energy in ambient vibrations these sensing systems can be difficult as these systems are
is innately low. In this paper, we focus on enhancing the usually installed in inaccessible locations. Integration of a
performance of piezoelectric harvesters through a multilayer
and, in particular, a multistep configuration. Partial coverage harvester with a sensor results in a self-powered system
of piezoelectric material in steps along the length of a cantilever making the sensing node autonomous. In typical sensing
beam results in a multistep piezoelectric energy harvester. We also environments, the available energy in the form of ambient
discuss obtaining an approximate deformation curve for the mechanical vibrations is considerable, and usually sufficient
beam with multiple steps in a computationally efficient manner. to power wireless sensor nodes [2], [3]. The primary meth-
We find that the power generated by a multistep beam is
almost 90% more than that by a multilayer harvester made out ods to convert energy in mechanical vibrations to electrical
of the same volume of polyvinylidinefluoride (PVDF), further energy are based on piezoelectric, electromagnetic or electro-
corroborated experimentally. Improvements observed in the static principles. Piezoelectric harvesters are smaller, lighter,
power generated prove to be a boon for weakly coupled low and have as much as three times higher energy densities
profile piezoelectric materials. Thus, in spite of the weak compared to electromagnetic and electrostatic counterparts [2].
piezoelectric coupling observed in PVDF, its energy harvesting
capability can be improved significantly using it in a multistep Moreover, they provide the necessary voltages directly without
piezoelectric beam configuration. the need of a separate voltage source as required in electro-
static conversion process [4]. These methods have been exten-
Index Terms— Energy harvesting, multilayer beam, multistep
beam, piezoelectric coupling, PVDF harvesters. sively compared by several researchers and the appropriate
method is largely governed by the system under consideration
I. I NTRODUCTION and the environmental conditions [5]. Nevertheless, piezoelec-

E NERGY harvesters are central for powering sensors


without batteries. Batteries have limited operational life
and have to be recharged after every usage. The use of
tric harvesters have higher energy conversion capability at low
frequencies compared to electromagnetic harvesters [6] for
device volume < 1 cm3 and hence, are best suited to harvest
energy harvesters allows us to power sensors without batteries energy from low frequency vibrations, typically below 100 Hz.
from the unused ambient energy. Unlike high levels of power The most ubiquitous configuration for harvesting energy using
witnessed in renewable energy harvesting, the energy from the piezoelectric principle is a bimorph.
ambient sources such as mechanical vibration, mechanical Efforts to improve the power generated by a bimorph have
been made by many researchers. It has been verified that a tri-
Manuscript received December 10, 2014; accepted December 22, 2014. Date
of publication January 6, 2015; date of current version April 22, 2015. This angular bimorph has higher tolerable excitation amplitude than
work was supported in part by the National Program on Micro and Smart a rectangular bimorph [7]. Harvester shape optimization can
Systems (NPMASS) Grant and in part by the NPMASS Project through result in significant increase in the harvested energy [8], [9].
the Centre for Nano Science and Engineering, Indian Institute of Science,
Bangalore, India. The associate editor coordinating the review of this paper Apart from the shape of the bimorph, it is noticed that the
and approving it for publication was Dr. Hsiao-Wen Zan. electromechanical coupling factor (ke ), the proof mass, and
R. Sriramdas is with the Department of Mechanical Engineer- the quality factor of the device greatly influence the generated
ing, Indian Institute of Science, Bangalore 560012, India (e-mail:
rammohan@mecheng.iisc.ernet.in). power [10], [11]. In addition to the physical and geometrical
S. Chiplunkar and R. M. Cuduvally are with the Centre for Nano Science parameters, it was observed that the load circuit impedance is
and Engineering, Indian Institute of Science, Bangalore 560012, India (e-mail: an important performance parameter [11]. Recent advances in
sanketh_chiplunkar@cense.iisc.ernet.in; ramya.m@cense.iisc.ernet.in).
R. Pratap is with the Centre for Nano Science and Engineering, Department improving the performance of the harvester are in optimizing
of Mechanical Engineering, Indian Institute of Science, Bangalore 560012, the processing electronics [12]–[14]. Multilayer harvesters
India (e-mail: pratap@mecheng.iisc.ernet.in). have been used with single supply pre-biasing (SSPB) for
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. improved power generation [15]. Triple layer harvesters have
Digital Object Identifier 10.1109/JSEN.2014.2387882 been compared with a bimorph against the same total thickness
1530-437X © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
SRIRAMDAS et al.: PERFORMANCE ENHANCEMENT OF PIEZOELECTRIC ENERGY HARVESTERS 3339

Fig. 1. Perspective view of a multilayer harvester showing three layers of


piezoelectric material partially covered on each side of the passive substructure
with end mass.

Fig. 2. Cross sectional view of the multilayer harvester showing the directions
in [16] assuming that the piezoelectric material is covered of polarization in different layers and the respective terminal combinations.
on the beam completely. Partial coverage of the piezoelectric
material on a beam was analyzed using the pin force model
in [17] and equivalent beam approach in [18]. In the current
work, we present a harvester analysis with no such approxima-
tions in modeling the multilayer and multistep configurations.
Coupling index G, and total capacitance C p for partially
covered multilayer and multistep beams are given in closed
form. We corroborate experimentally that a harvester with
multistep configuration yields almost 90% more power than
that from a multilayer configuration when both are made out Fig. 3. Perspective view of a multistep harvester showing three layers
of the same volume of piezoelectric material. Moreover, we of piezoelectric material gradually decreasing in length on each side of the
derive an approximate deflection shape for the multistep beam passive substructure with end mass.
which is relatively simple for design optimization compared
to a transcendental form. The proposed shape function can
the generated power. A similar poling configuration is adopted
be used conveniently to model piezoelectric patch on a beam,
in the multistep harvester where the length of coverage of
particularly when piezoelectric layers are not built into the
piezoelectric material on either side of the beam decreases with
clamped end. Rayleigh quotient is then used for the beam to
each step as shown in Fig. 3. The relatively shorter lengths of
obtain the multistep beam resonance.
piezoelectric material towards the extreme layers along the
The piezoelectric material used to fabricate harvesters
thickness have lower capacitance at higher strains. Similarly,
for our study is Polyvinylidinefluoride (PVDF) in β–form.
the layers closer to the substrate have lower strains with
This phase of PVDF is obtained by stretching the α–form
relatively larger capacitance. Both conditions are favorable for
films developed through solvent cast method. Low frequency
higher power generation. Moreover, it would be observed that
harvesters can be easily designed with PVDF due to the
the multistep harvester has lower natural frequency than the
inherent low elastic modulus. Despite the weak coupling index
multilayer harvester when constructed from the same volume
G of PVDF, performance of the harvester can be improved
of piezoelectric material. This lowering of natural frequency
significantly by selecting the proposed multistep design.
is advantageous because, as we show later, a resonator driven
at a lower frequency generates more power than the one oscil-
II. M ODELING OF M ULTILAYER AND lating at a higher frequency (with everything else remaining
M ULTISTEP H ARVESTERS the same). However, modeling a multistep harvester requires
A. Harvester Configurations analyzing the beam with different cross sections along the
length. Such multistep electromechanical beams are easily
A harvester with a single layer of piezoelectric material
examined with the extended Hamilton’s principle by consid-
on the substrate is referred to as a unimorph. Piezoelectric
ering the associated energies.
bimorph has a layer of piezoelectric material on both sides
of the substructure. A typical configuration of a multilayer
harvester is shown in Fig. 1. The multilayer harvester has B. Modeling Methodology
several layers of piezoelectric material positioned one over the The partially covered multilayer piezoelectric harvester is
other on both sides of the passive substructure. In the current modeled as a two segment beam with constant cross section
work, only parallel configuration of the layers is studied as in either segment. Equations of motion for the harvester
this is observed to be superior to the series arrangement [19]. are derived employing the extended Hamilton’s principle.
In parallel configuration the layers are placed in such a This principle applied to piezoelectric material has been
manner that the direction of poling in every adjacent layer used to model piezoelectric actuators [20], and it has been
is opposite to each other on either side of the beam as employed to describe the behavior of a piezoelectric bimorph
shown in Fig. 2. In this method of arrangement, layers are harvester [21]. The parallel configuration of the multilayer
combined in parallel as every layer contributes additively to harvester has different directions of poling in different layers
3340 IEEE SENSORS JOURNAL, VOL. 15, NO. 6, JUNE 2015

Fig. 4. Schematic view of the multilayer harvester showing the length of the
piezoelectric material l, covering the beam, and length L, at which the end
mass is attached.

as shown in Fig. 2. This difference in the direction of poling is


appropriately considered while writing the energies associated
with the respective layers. Fig. 5. Elemental length of multilayer piezoelectric beam showing the
thickness of each layer, and the position of the neutral axis.
C. Equations of Motion
The equations of motion for the multilayer and multistep
presented here can easily be extended to analyze multistep har-
piezoelectric energy harvester are derived by formulating the
vesters by including appropriate modifications in (1), and (2).
action integral. The first variation of the action integral δ A in
The multistep harvester with m segments and n m layers in the
terms of the variations in kinetic energy K e , potential energy
respective segment has the coupling index G s given by,
He of the electromechanical beam, and the virtual work of ⎛⎛ ⎞ ⎞
non-conservative forces Wnc is given by  m  nm  i
t
bi e31i ⎝⎝ t j − z̄ ⎠ − ⎠
i
t2 Gs =
2
j =1 i=1 j =1
δA = (δK e − δ He + δWnc )dt = 0. (1)


× φ j,x l j − φ j,x l j −1 . (6)
t1

A typical configuration of the multilayer harvester is shown where l j indicates the length of the beam up to, and including
in Fig. 4. A portion of the segment containing both active and the j th segment, measured from the clamped end. The coupling
passive layers of the multilayer harvester is shown in Fig. 5. index G s is proportional to the difference in slopes of the two
The transverse displacement w is written separately in the two ends in a given segment. By arranging layers in several steps,
segments as the effective stiffness of the beam reduces resulting in a lower
 natural frequency. Obtaining the expressions for the natural
φ1 (x) W (t) , 0 ≤ x ≤ l frequency and mode shape for a beam with multiple steps in
w (x, t) = (2)
φ2 (x) W (t) , l ≤ x ≤ L analytical form is both lengthy and laborious particularly with
in terms of the shape functions φi (x) in the respective segment increasing number of segments. We present a fairly accurate
and the temporal displacement function W (t). yet simple method of obtaining the resonant frequency and
Substituting (2) into (1) and invoking the variations in shape functions for a multistep beam with a tip mass.
displacement and voltage as described in Appendix 1, the
governing equations of motion for temporal displacement W , D. Resonant Frequency and Shape Function
and voltage V of the electromechanical beam connected to a The first natural frequency of the harvester is determined
load resistance R are obtained as by computing Rayleigh’s quotient. The shape function for the
M W,t t + C W,t + K W − GV = F harvester is assumed in the form of an appropriate polynomial
satisfying the boundary conditions. The segment of the beam
RC p V,t + RGW,t + V = 0, (3)
that has the end mass is assumed to have a higher degree
where M, C, K, and F denote effective mass, effective damp- polynomial than the others. This assumption simultaneously
ing coefficient, effective stiffness, and effective force of the allows us to satisfy the boundary conditions, and determine
multilayer harvester. The electromechanical coupling index G, the coefficients in the polynomial. The generic form of shape
and harvester capacitance C p for the multilayer harvester are functions for a beam with m segments is
given as:
⎛⎛ ⎞ ⎞ 
k

 n  i φi = Ci j x j, (7)
ti
G= bi e31i ⎝⎝ t j − z̄ ⎠ − ⎠φ1,x (l), (4) j =0
2
i=1 j =1
where i denotes the segment number from 1 to m, and
n S
bi lε33 Ci j are the coefficients of the polynomial. The value of k is 3
Cp = . (5) for all the segments except for the last segment having the end
ti
i=1
mass. The m th segment with end mass has k = 4, facilitating
The piezoelectric stress coefficient in the k th layer is denoted the determination of all the coefficients in terms of any one
by e31k . The derivation of the equations of motion and the coefficient. The shape functions for a beam with single step
associated boundary conditions are described in Appendix 1. and no end mass can be found in [23]. The coefficients of
These boundary conditions are employed to obtain shape shape functions, for a beam with 3 segments and more, are
functions in both the segments of the beam. The procedure determined by satisfying the additional boundary conditions
SRIRAMDAS et al.: PERFORMANCE ENHANCEMENT OF PIEZOELECTRIC ENERGY HARVESTERS 3341

imposed at the segment ends. These coefficients and hence, for optimum load resistance Ropt at resonance as
the shape functions can be readily obtained from the relations  
described in Appendix 2. 1 2ζ
Ropt =  . (14)
These shape functions are employed in determining the C p ωn ke4 + 4ζ 2
natural frequency of the multistep harvester using Rayleigh’s
The displacement, voltage, and power corresponding to this
quotient. The effort required in calculating the natural
optimum load resistance are obtained by substituting (14) in
frequency by solving the transcendental characteristic equa-
appropriate form into (9), (10), and P = Vrms 2 /R as
tion for a multistep harvester increases enormously with the
  
number of segments. In the proposed method, a sixth order F ke4 + 8ζ 2
polynomial in frequency, ω, has to be solved to obtain the W =  , (15)
Mωn2 2ζ 4ζ 2 + γ 2
fundamental natural frequency ωn irrespective of the number  
of segments. The natural frequency is obtained by solving the F ke2
equation: V =  , and (16)
G 4ζ 2 + γ 2
 2 
l1 lm F2 ke
P = , (17)
E I1 φ1,x
2
xdx + ···+ E Im φm,x
2
xdx Mωn 8ζ γ

0

lm−1 where γ = ke2 + ke4 + 4ζ 2 .
l1 lm The expressions in square brackets in (15), (16), and (17)
2⎜
− ω ⎝ m̄ 1 φ1 d x + · · · +
2
m̄ m φm
2
d x + mφm
2
(L) are the dimensionless factors determining the limits for the
0
displacement, voltage, and the power generated, respectively.
lm−1
⎞ Evidently, when a harvester is operated at the optimum load
resistance, the maximized power given by (17) greatly depends
2 ⎟
+ J φm,x (L) ⎠ = 0, (8) on ke and ζ besides excitation F, equivalent mass M, and
natural frequency ωn . Equations (13) and (17) indicate that
if G increases and ωn decreases, the harvester has larger
where m̄ p denotes the equivalent mass per unit length of the ke and hence, delivers more power. This scenario is noticed
respective segment, E I p is obtained from (A.10) with the in multistep harvester. Hence, for a given volume of the
appropriate value for n in the pth segment. After obtaining piezoelectric material and the mechanical damping ratio, the
the natural frequency and shape functions, the equations of power developed from a multistep harvester is more than that
motion (3) are solved for harmonic excitation. from a multilayer harvester. We verify these observations by
experimenting on multilayer and multistep harvesters fabri-
cated out of the same volume of piezoelectric material and
E. Performance of Harvesters
report the results in Section IV.
The steady state displacement and voltage are obtained
by solving the equations of motion assuming a complex III. S IMULATION OF H ARVESTER P ERFORMANCE
harmonic base excitation. The non-dimensional displacement
W ∗ and voltage V ∗ are written in terms of the dimensionless The performance of multilayer and multistep harvesters
parameters α, β, damping ratio ζ , and ke as with PVDF as the piezoelectric material is simulated. The
 arrangement of the piezoelectric layers in parallel configu-
∗ 1 + α2 β 2 ration results in equal capacitances for both the harvesters.
W = 
2

2 By arranging the layers in multiple steps the effective stiffness
1 − β 2 − 2ζ αβ 2 + β 2 2ζ + α 1 − β 2 + ke2 of the beam reduces (compared to having uniform layers
(9) of the same volume), resulting in higher ke . Under these
ke αβ
2 conditions the normalized power at the optimum load resis-
V ∗ = 
2

2 , tance is solely governed by ke and ζ . Moreover, these para-
1 − β − 2ζ αβ
2 2 + β 2ζ + α 1 − β + ke
2 2 2
meters influence even voltage and displacement. However,
(10) from (15) and (16) it can be noted that the increase in ke ,
increases the normalized voltage but decreases the normalized
where displacement amplitude (Fig. 6).
The normalized displacement increases with decreasing
α = RC p ωn , (11)
damping ratio ζ irrespective of the coupling factor ke .
β = ω/ωn , and (12)
  However, for a given damping ratio, increase in the coupling
ke2 = G 2 C p Mωn2 . (13) factor results in lower displacement values. This trend is
predominant for ke up to 0.25. The additional decrease in the
The power developed by a harvester is measured across a amplitude is due to the piezoelectric damping.
load resistance R. In the process of reducing the number of The contour plot of the normalized voltage is shown in
variables, the power delivered by the harvester is optimized. Fig. 7. It can be noticed that with the increase in the coupling
The power P = Vrms2 /R is maximized to obtain an expression factor, the normalized voltage increases irrespective of the
3342 IEEE SENSORS JOURNAL, VOL. 15, NO. 6, JUNE 2015

Fig. 8. The normalized power of a harvester as a function of the damping


Fig. 6. The normalized displacement of a harvester as a function of the ratio and the coupling factor. For very low damping factors, maximum power
damping ratio ζ and the coupling factor ke at resonance. For a given mechan- is attained with no further improvement in ke beyond a certain value. There
ical damping ratio, the improvement in ke would decrease the displacement exists an upper limit on the generated power as damping ratio and coupling
amplitude. factor increase.

TABLE I
G EOMETRY OF M ULTILAYER AND M ULTISTEP
H ARVESTERS FOR S IMULATION

and multistep harvesters used in the simulation are listed


Fig. 7. The normalized voltage of the harvester as a function of the damping in Table I.
ratio and the coupling factor. If the damping ratio is large, higher voltages Harvesters with six layers of 60 μm thick PVDF film on
are attained by designing the harvesters to have higher coupling factors.
However, at low damping ratio, small increase in the coupling factor up to
100 μm thick copper substructure as shown in Figs. 1 and 3
0.2 corresponds to the same significant increase in the voltage. are analyzed. The material density for copper substrate and
PVDF used in the simulations are 8600 kg/m3 and 1750 kg/m3
respectively. The elastic modulus for the substrate and the
damping ratio. However, if the damping ratio is less than 0.02, piezoelectric material are taken as 125 GPa and 2.55 GPa
the increase is larger for ke < 0.1. The normalized power as respectively. The piezoelectric strain coefficient and the
a function of the damping ratio and the coupling factor is relative permittivity used in the simulation are 9.8 pC/N and
depicted in Fig. 8. It can be observed from the plot that for 15 respectively.
a given mechanical damping ratio, the increase in power with The shape functions of harvesters in the approximate form
ke is significant only up to a certain value of ke . Irrespective defined in (7) are obtained from the relations given in
of the damping ratio, any increase in the coupling factor Appendix 2. The classical exact forms of shape functions are
up to 0.1 improves the generated power. This increase is also determined by satisfying the boundary conditions in each
significant at low damping ratios. Nevertheless, for ke < 0.1 segment. The exact and approximate functions for multilayer
and at all damping ratios, the rate of increase in power is and multistep harvesters are shown in Fig. 9. The performance
considerable with the increase in coupling factor ke . Harvesters of these harvesters is verified by using both exact and the
made of PVDF have coupling factor less than 0.1 and hence, approximate shape functions.
improvement in power developed with ke is expected to be The first natural frequency is determined using both tran-
considerable. The geometric parameters for both multilayer scendental characteristic equation and the Rayleigh’s quotient
SRIRAMDAS et al.: PERFORMANCE ENHANCEMENT OF PIEZOELECTRIC ENERGY HARVESTERS 3343

Fig. 10. The plot shows variation of displacement with excitation frequency
and load resistance. Decrease in the amplitude at optimum load resistance is
not perceivable in low ke harvesters.

Fig. 9. The plot shows the overlap of transcendental and approximate


shape functions over the entire length of the beam in two cases. (a) The Fig. 11. The voltage developed by the harvester at frequencies close to
shape functions of the multilayer harvester with two segments. (b) The shape resonance with different load resistances. The plot includes the voltages
functions in the multistep harvester with three segments. A smooth transition developed for 500 k
, 1 M
, and 10 M
load resistances in the order shown.
from one segment to the other can be noticed at 0.025 m in (a) and at 0.01 m
and 0.025 m in (b).

Hence, the variation in the displacement with load resistance


TABLE II
is marginal.
P ERFORMANCE OF M ULTILAYER AND M ULTISTEP H ARVESTERS
The potential developed by the harvester increases with
increasing load resistance. Typical voltage response of the
harvester near resonance, for different values of the load resis-
tance, is shown in Fig. 11. For low ke harvesters, the difference
between the short circuit and open circuit frequencies is
insignificant. Hence, the resonance frequency shift in a PVDF
harvester with increasing load resistance is not considerable.
As the current generated decreases with load resistance, the
power delivered is maximized at the optimum load resistance.
Hence, the power developed is higher at resonance and at a
specific optimum load resistance. A typical plot of the power
developed by the harvester as a function of the excitation
frequency and load resistance is shown in Fig. 12. Some
researchers have observed that harvesters with high ke have
with approximate polynomials given in (7). The comparison two identical peaks of maximum power near resonance [24].
of values is shown in Table II. The first natural frequency in However, this behavior is not significant in harvesters with
multistep harvester is lower than that in multilayer harvester, low ke . The peak power corresponds to the optimum load
resulting in increased value of the coupling factor ke . For the resistance in weakly coupled harvesters.
same damping ratio and input acceleration, the tip displace- The optimum load resistance given by (14) is expected to
ment, voltage, and power are higher in multistep harvester. be lower in a multistep harvester but the increased influence
The expected displacement from the multistep harvester is of the natural frequency over the coupling factor ke makes
shown in Fig. 10 as a function of forcing frequency and load it larger than that in the multilayer harvester. Nonetheless,
resistance. The displacement at the optimum load resistance simulations reveal that the power from the multistep harvester
(1.3M
) is expected to be less than that at the other loads. The is atleast 30% more than that from the multilayer harvester
decrease is not appreciable in the harvesters that have low ke . for the geometry and parameters described in Tables I and II.
3344 IEEE SENSORS JOURNAL, VOL. 15, NO. 6, JUNE 2015

Fig. 15. The plot shows the displacement at 10 mm away from the tip of
Fig. 12. The power developed by the harvester for various load resistances the multistep harvester as a function of frequency for 0.5 g input acceleration.
and forcing frequencies. The maximum power is delivered at the resonance Standard deviation in the experimental measurements is plotted over the
and at the optimum load resistance which is 1.3 M
. simulated displacement curve.

TABLE III
S IMULATION OF THE FABRICATED H ARVESTERS

Fig. 13. The multilayer and multistep harvesters fabricated using the same
volume of the piezoelectric material - PVDF are shown in (a) and (b),
respectively.
as a passive layer in obtaining electromechanical coupling
index G. The fabricated harvesters and schematic of the test
setup are shown in Fig. 13 and 14 respectively. The harvesters
are subjected to sinusoidal vibration using a shaker (LDS,
V406/8-PA100E, Royston, U.K.) at 0.5 g acceleration (g =
9.81 m/s2 ) from 25 Hz to 100 Hz. The developed potential is
then measured across a set of load resistances. The optimum
load resistance of the harvester is experimentally determined
by noting the maximum power transfer.
The displacement of the multistep harvester at a point that is
10 mm away from the end mass for an excitation acceleration
of 0.5 g is measured experimentally using a laser vibrome-
ter (Polytec GmbH, PDV-100, Germany). The displacement
Fig. 14. A schematic of the experimental setup for evaluating the performance data extracted at selected frequencies are compared with the
of the harvesters. The desired acceleration input is given to the shaker through simulated displacement in Fig. 15. However, the damping
a controller, and the displacement is monitored using a laser vibrometer and ratio is computed using half-power bandwidth method applied
voltage is measured across a resistance using an oscilloscope.
to the response from the harvester. The FFT and curve fit
options available in the polytec software - VibSoft 5.0 (Polytec
These results are verified by experimenting on the fabricated GmbH, Germany) are used to obtain the half-power points that
harvesters having close resemblance with those studied here facilitate the determination of the damping ratio. The measured
analytically. values of the damping ratio are used for simulation and the
results are listed in Table III. The simulation results are in good
agreement with the measured values (Figs. 15–19), however,
IV. E XPERIMENTAL R ESULTS AND D ISCUSSION minor difference between simulated and measured parameters
Multilayer and multistep harvesters with the attributes is attributed to the measurement uncertainties in the estimation
described above are fabricated and tested. The layers are glued of the damping ratio and the coupling factor.
using epoxy adhesive. The presence of glue is appropriately The voltages developed by the multistep harvester at
included in determining the segment flexural rigidity and various frequencies for selected load resistances are recorded.
SRIRAMDAS et al.: PERFORMANCE ENHANCEMENT OF PIEZOELECTRIC ENERGY HARVESTERS 3345

Fig. 16. The plot shows the voltage frequency response of the multistep
harvester for 0.5 g input acceleration. The standard deviations in the measured
voltages for two load resistances are shown overlaid on the simulated Fig. 18. The voltage developed by the multilayer and multistep harvesters
responses. for various load resistances at their respective resonant frequencies. The
experimentally measured values are overlaid on the simulated voltage curves.

Fig. 17. The power developed by the multistep harvester for various load
resistances at resonance. Markers for measured power are overlaid on the
simulated power. The maximum power is delivered at the optimum load Fig. 19. The plot shows the measured and simulated values of power
resistance which is 1.3 M
. developed by multistep and multilayer harvesters at various resistances when
operated at their respective resonances 30.8 Hz and 33.5 Hz for 0.5 g input
acceleration. The power from multistep harvester is more than 90% that
from multilayer harvester, where both are made from the same volume of
The measured values are compared with the voltage frequency piezoelectric material - PVDF.
response as shown in Fig. 16.
The power delivered by the multistep harvester for 0.5 g
input acceleration as a function of load resistance is shown in two harvesters is 8.1 μW and 4.3 μW respectively.
Fig. 17. Power is measured across 14 load resistances from Nonetheless, the performance in terms of the developed power
38 k
to 10 M
. The harvester is operated at its resonant is higher in the multistep harvester by as much as 90%.
frequency of 30.8 Hz during the measurements. Maximum
power of 8.1 μW is obtained with a load resistance of 1 M
. V. C ONCLUSION
The voltages developed by the multilayer and multistep We have shown that a multistep, multilayer cantilever
harvesters as a function of load resistance for an input of 0.5 g beam configuration is much better than a uniform multilayer
acceleration at the respective resonant frequencies of 33.5 Hz configuration for harvesting energy from vibrations using
and 30.8 Hz are measured and compared in Fig. 18. The piezoelectric materials. We have discussed the method of
measured voltages at 10.4 M
for the multistep and multilayer analyzing such configurations using extended Hamilton’s prin-
harvesters are 7.2 V and 4.6 V respectively. However, the ciple. This method facilitates the analysis of harvesters with
voltages near the optimum load resistance (1 M
) are partial coverage of piezoelectric material with relative ease.
3.96 and 2.98 V. Evidently, the multistep harvester develops We have also described a computationally efficient approach
higher voltage at the given load resistance. The power from for obtaining the fundamental mode frequency, and approxi-
these harvesters is computed from the measured voltages and is mate shape functions for a beam with any number of segments
plotted in Fig. 19. Invariably, power developed by the multistep with any number of arbitrary layers. Our analysis gives a range
harvester is more than that by the multilayer harvester. of coupling factor ke and damping factor ζ, which influence
The developed power corresponding to 1 M
load in the the performance of the harvester significantly. It is noticed
3346 IEEE SENSORS JOURNAL, VOL. 15, NO. 6, JUNE 2015

that at low damping ratios below 0.02, any improvement in kinetic energy K e is
the coupling factor ke up to 0.1 considerably improves the ⎡ l
power generated. Having verified that the coupling factor in a   n L
1
multistep harvester is greater than that in a multilayer Ke = ⎣ m̄ i w,t2 d x + m̄ s w,t2 d x + mw,t (L)2
2
harvester, the two types of harvesters were fabricated using 0 i=1 l
PVDF as the piezoelectric material, and tested. We observed ⎤
experimentally that the power developed by a multistep
+J w,xt (L)2 ⎦, (A.4)
harvester is superior to that developed by a multilayer
harvester. The decrease in the beam stiffness and increase
in the coupling index G result in increased ke , and hence, where n is the total number of layers (both active and passive
improve the generated power. Moreover, the comparison is adhesive layers including the substrate), m̄ is the mass per unit
rational because the volume of piezoelectric material used length of the respective layers, m and J are the proof mass
in both the harvesters is the same. The realized multistep and the polar moment of inertia respectively, and subscript s
and multilayer harvesters resonate at 30.8 Hz and 33.5 Hz stands for the substrate. Besides, the strain in x direction and
respectively. The power developed by them at the respective the electric field in z direction contribute significantly to the
optimum load resistances are 8.59 μW and 4.53 μW. Further- potential energy He of the electromechanical beam. Strain
more, the configuration induced performance improvements in x direction neglecting the axial stretch is S1 = −zw,x x .
for harvesters with coupling factors up to 0.1, boost the power Denoting the elastic modulus, piezoelectric stress coefficient,
per unit volume considerably. Almost 90% improvement in the stresses and strains in contracted form; the total potential
developed power strongly suggests the use of weakly coupled energy is
low profile piezoelectric materials like PVDF for large scale ⎡ ⎤

energy harvesting. 1⎣  E 2 
He = c11 S1 − 2e31 E 3 S1 − ε33
S 2
E 3 d
⎦. (A.5)
A PPENDIX I 2

The equations of motion for the multilayer and multistep


harvester are derived using the extended Hamilton’s principle. The domain
represents the entire volume of the harvester.
The kinetic energy K e , and electrical enthalpy density H̄e , Assuming that the beam is excited by ground acceleration
of an elastic dielectric are used to formulate the action wg,t t in z direction, the first variation of virtual work of
integral [22]. Electrical enthalpy density for a piezoelectric non-conservative forces is
material subjected to stresses Ti j , the electric displacement Di ,  l n

the corresponding strains Si j , and the electric field E i (where δWnc = − m̄ i wg,t t + r1 w,t δwd x
subscripts i , j , k, and l vary from 1 to 3) is
0 i=1
1 1
H̄e = Ti j Si j − Di E i . (A.1) L
2 2
− m̄ s wg,t t + r2 w,t δwd x −mwg,t t δw(L)+ qδV,
The first variation of the action integral δ A in terms of
l
the variations in kinetic energy K e , potential energy He (A.6)
of the electromechanical beam, and the virtual work of
non-conservative forces Wnc is given by where r1 and r2 are the velocity independent coefficients of
t2 friction in the two segments, δw and δV represent the variation
in displacement, and voltage respectively. The displacement w
δA = (δK e − δ He + δWnc )dt = 0. (A.2)
is written separately in the two segments as
t1

The piezoelectric constitutive relations with strain and φ1 (x) W (t) , 0 ≤ x ≤ l
w (x, t) = (A.7)
electric field as the independent variables are φ2 (x) W (t) , l ≤ x ≤ L
Ti j = ciEj kl Skl − eki j E k in terms of the shape functions φi (x) in the respective segment
Dk = eki j Si j + εkl
S
El , (A.3) and the displacement temporal function W (t).
Substituting (A.7) into the first variations of (A.4) and (A.5),
where ciEj kl is the elastic modulus at constant electric field, and subsequently from (A.6) and (A.2), the governing equa-
eki j is the piezoelectric stress coefficient, and εkl
S
is the relative tions of motion for the electromechanical beam connected to
permittivity at constant strain. a load resistance R are obtained as
Depending on the chosen material with the intrinsic crystal
symmetry and the electrode location for the beam, unknowns M W,t t + C W,t + K W − GV = F
in (A.3) reduce considerably and hence, result in fewer terms RC p V,t + RGW,t + V = 0, (A.8)
contributing to the energy in (A.1).
Assuming the displacement of the beam w, to be where M, C, K, and F denote effective mass, effective
purely in z direction as a function of x and time t, the damping coefficient, effective stiffness, and effective force
SRIRAMDAS et al.: PERFORMANCE ENHANCEMENT OF PIEZOELECTRIC ENERGY HARVESTERS 3347

   
−1 j +1 ( j + 1) L m L m−1 −1 j +1 ( j + 1) L m−1
4− j 4− j
P1 P2
Cm j = j −2+ C13 + 3 − j + C22 , where, j = 2 to 3.
( j − 1) (L m − L m−1 )2 Pm 3 ( j − 1) (L m − L m−1 )2 Pm
(A.17)
   

i−1
1 1
4− j
( j + 1) L m L m−1  k −l
m−2
j +1 3− j
Ci j = −1 ( j + 2) P1 − Lk + C13 +
k=1
Pk Pk+1 (4 − j) Pm (L m − L m−1 )2 l=1 (m − 2)!
   

i−1
1 1 ( j + 1) L
4− j  k −l
m−2
−1 j +1 ( j + 1) P2
2− j m−1
Lk − + C22 ,
Pk Pk+1 6Pm (L m − L m−1 )2 (m − 2)!
k=1 l=1
where, i = 2 to m and j = 0 to 1. (A.18)

C22 = C13
 !  #
"  3 1
m−1 ! "  2 1
m−1 !
P1 L m
6P1 LL mm +L m−1
−L m−1 + m̂ω 2 2P
1 L k Pk − 1
Pk+1 − 3L P
m 1 L k Pk − 1
Pk+1 + Pm 2 (L m − L m−1 ) 2
− 2L 2
m
k=1 k=1
×   m−1  ! #
−4 P2 " " 
m−1 !
L m −L m−1 − m̂ω 2 P
2 L 2 1 − 1
k Pk Pk+1 − 2L m P2 L k
1
Pk − 1
Pk+1 + PPm2 16 (L m − L m−1 )2 − 6L 2m
k=1 k=1
where, m̂ is the proof mass. (A.19)

given by: clamped beam consisting of two segments with an


end mass
l 
n L
M = m̄ i φ12 d x + m̄ s φ22 d x + mφ22 (L) + J φ2,x
2
(L) φ1 (l) = φ2 (l) ; φ1,x (l) = φ2,x (l) (A.12)
0 i=1 l E I1 φ1,x x (l) = E I2 φ2,x x (l)
l L E I1 φ1,x x x (l) = E I2 φ2,x x x (l) (A.13)
C = r1 φ12 d x + r2 φ22 d x mφ2 (L) W,t t − E I2 W φ2,x x x (L) = 0
0 l J φ2,x (L) W,t t + E I2 W φ2,x x (L) = 0. (A.14)
l L
The shape function in each segment has to satisfy four
K = E I1 φ1,x
2
xdx + E I2 φ2,x
2
xdx conditions. These boundary conditions can be extended to
0 l any number of segments noting that first segment has zero
l 
n L displacement and slope at the clamped end, and the last
F =− m̄ i φ1 wg,t t d x − m̄ s φ2 wg,t t d x −mwg,t t φ2 (L). segment has proof mass at the free end.
0 i=1 l A PPENDIX II
(A.9)
The approach for finding approximate shape functions for
Flexural rigidity in the pth
segment is denoted by E I p , and is a beam with m segments (>2) is described here. These are
computed using the expressions: derived by satisfying 4m boundary conditions along the beam.
⎛ ⎛ ⎞⎛ ⎞ ⎞ The shape functions used are given in (7). As the last segment
n i 
i−1
ti3 uses a fourth order polynomial, the total number of coefficients
EI = bi c11i ⎝ti ⎝ t j −z̄ ⎠ ⎝ t j −z̄ ⎠ + ⎠ for an m segment beam is 4m + 1. While satisfying the
3 boundary conditions every coefficient is expressed in terms
i j =1 j =1

(A.10) of C13 . As a matter of convenience every coefficient is initially


⎛ ⎞$ written as a combination of C13 and C22 . However, C22 is
n 
i 
n
ti 
n finally expressed in terms of C13 . Denoting the flexural rigidity
z̄ = ⎝ c11i Ai tj− c11i Ai ⎠ c11i Ai . of the k th segment by Pk , the relations for all the coefficients
2
i=1 j =1 i=1 i=1 are given as follows.
(A.11)
Pi+1
Ci j = Ci+1 j (A.15)
The parameters bk , c11k , tk , and Ak indicate width, elastic Pi
modulus, thickness, and the area of cross section of the where, i = 1 to m − 1; and j = 2 to 3.
k th layer. In addition to (A.8), additional conditions are
obtained at the boundaries after solving (A.2). These −1

Cm4 = L m P1 C13 + 13 P2 C22 (A.16)
are the boundary conditions for a partially covered 2Pm (L m − L m−1 ) 2
3348 IEEE SENSORS JOURNAL, VOL. 15, NO. 6, JUNE 2015

It may be noted that C10 and C11 are zero. These relations [20] H. Tanaka, “Generalized basic equations for bending motions of piezo-
(from A.15 to A.19) can be used to obtain the shape function electric bars formulated from Hamilton’s principle,” J. Acoust. Soc.
Amer., vol. 95, no. 4, p. 1768, 1994.
for an m segment beam in terms of one coefficient-C13 . [21] H. A. Sodano, G. Park, and D. J. Inman, “Estimation of electric charge
The function is then plugged into (8) to obtain the fundamental output for piezoelectric energy harvesting,” J. Strain, vol. 40, no. 2,
natural frequency, and subsequently into (A.9) to determine the pp. 49–58, 2004.
[22] R. G. Ballas, “Linear theory of piezoelectric materials,” in Piezoelectric
harvester response. Multilayer Beam Bending Actuators. Berlin, Germany: Springer-Verlag,
2007, ch. 3, pp. 31–43.
ACKNOWLEDGMENT [23] S. Rammohan, S. Chiplunkar, C. M. Ramya, S. J. Kumar, A. Jain,
The authors are grateful to Dr. A. Jain and and R. Pratap, “Multi-layer piezoelectric energy harvesters for improved
power generation,” presented at the 7th Int. Conf. Smart Mater., Struct.,
S. J. Kumar, NAL, Bengaluru, for providing the required Syst., Bangalore, India, Jul. 8–11, 2014.
PVDF films for the experimental evaluation of the harvesters. [24] Y. C. Shu and I. C. Lien, “Efficiency of energy conversion for a piezo-
electric power harvesting system,” J. Micromech. Microeng., vol. 16,
R EFERENCES no. 11, pp. 2429–2438, 2006.
[1] H. S. Kim, J.-H. Kim, and J. Kim, “A review of piezoelectric energy
harvesting based on vibration,” Int. J. Precis. Eng. Manuf., vol. 12, no. 6,
Rammohan Sriramdas received the B.Tech. degree
pp. 1129–1141, Dec. 2011.
in mechanical engineering from Jawaharlal Nehru
[2] S. Priya, “Advances in energy harvesting using low profile piezoelectric
Technological University, Hyderabad, India, in 2005.
transducers,” J. Electroceram., vol. 19, no. 1, pp. 167–184, Sep. 2007.
[3] H. A. Sodano, G. Park, and D. J. Inman, “Generation and storage of He is currently pursuing the Ph.D. degree in mechan-
electricity from power harvesting devices,” J. Intell. Mater. Syst. Struct., ical engineering at the Indian Institute of Science,
vol. 16, no. 1, pp. 67–75, Jan. 2005. Bangalore. He served as a Design Engineer with
[4] S. Roundy, P. K. Wright, and J. Rabaey, “A study of low level vibrations Hindustan Aeronautics Ltd., Bangalore, India, in
as a power source for wireless sensor nodes,” Comput. Commun., vol. 26, 2005, and a Scientist/Engineer with the Inertial
no. 11, pp. 1131–1144, Jul. 2002. Systems Unit, Indian Space Research Organiza-
[5] S. R. Anton and H. A. Sodano, “A review of power harvesting using tion, Bangalore, from 2006 to 2011. His research
piezoelectric materials (2003–2006),” Smart Mater. Struct., vol. 16, interests include vibration isolation, vibration
no. 3, pp. R1–R21, May 2007. control, design and analysis of electro-mechanical sensors, actuators, and
[6] P. D. Mitcheson, E. K. Reilly, T. Toh, P. K. Wright, and E. M. Yeatman, systems.
“Transduction mechanisms and power density for MEMS inertial energy
scavengers,” in Proc. Power MEMS, vol. 6, 2006, pp. 275–278.
[7] F. Goldschmidtboeing and P. Woias, “Characterization of different beam Sanketh Chiplunkar received the B.E. degree in
shapes for piezoelectric energy harvesting,” J. Micromech. Microeng., mechanical engineering from the National Insti-
vol. 18, no. 10, p. 104013, Sep. 2008. tute of Engineering, Mysore, India, in 2012. He is
[8] M. I. Friswell and S. Adhikari, “Sensor shape design for piezoelectric currently a Project Assistant with the Centre for
cantilever beams to harvest vibration energy,” J. Appl. Phys., vol. 108, Nano Science and Engineering, Indian Institute of
no. 1, p. 014901, Jul. 2010. Science, Bangalore. His research interests include
[9] D. Benasciutti, L. Moro, S. Zelenika, and E. Brusa, “Vibration energy dynamics and vibrations.
scavenging via piezoelectric bimorphs of optimized shapes,” Microsyst.
Technol., vol. 16, no. 5, pp. 657–668, May 2010.
[10] S. Roundy, “On the effectiveness of vibration-based energy harvesting,”
J. Intell. Mater. Syst. Struct., vol. 16, no. 10, pp. 809–823, Oct. 2005.
[11] S. Jiang, X. Li, S. Guo, Y. Hu, J. Yang, and Q. Jiang, “Performance of
a piezoelectric bimorph for scavenging vibration energy,” Smart Mater. Ramya M. Cuduvally received the bachelor’s
Struct., vol. 14, no. 4, pp. 769–774, Jul. 2005. degree in electrical and electronics engineering from
[12] Y. K. Ramadass and A. P. Chandrakasan, “An efficient piezoelectric the M.S. Ramaiah Institute of Technology, Ban-
energy harvesting interface circuit using a bias-flip rectifier and shared galore, India, in 2012. She is currently a Project
inductor,” IEEE J. Solid-State Circuits, vol. 45, no. 1, pp. 189–204, Assistant with the Centre for Nano Science and
Jan. 2010. Engineering, Indian Institute of Science, Bangalore.
[13] J. Dicken, P. D. Mitcheson, I. Stoianov, and E. M. Yeatman, “Power- Her current research is on developing power man-
extraction circuits for piezoelectric energy harvesters in miniature and agement circuits for energy harvesting applications.
low-power applications,” IEEE Trans. Power Electron., vol. 27, no. 11, Her research interests also include microelectronics
pp. 4514–4529, Nov. 2012. and semiconductor device physics.
[14] W. K. Kim, “Design and analysis of switching circuits for energy
harvesting in piezostrutures,” Ph.D. dissertation, Dept. Mech. Eng.,
Virginia Polytech. Inst. State Univ., Blacksburg, VA, USA, 2012.
[15] A. D. T. Elliott and P. D. Mitcheson, “Power density improvement of a Rudra Pratap (M’02) received the B.Tech. degree
piezoelectric energy harvester through use of a micropower switch-mode from IIT Kharagpur, Kharagpur, India, in 1985,
interface,” in Proc. 11th IEEE Sensors Conf., Taipei, Taiwan, Oct. 2012, the M.S. degree in mechanics from the University
pp. 1–4. of Arizona, Tucson, AZ, USA, in 1987, and the
[16] D. Zhu, A. Almusallam, S. Beeby, J. Tudor, and N. Harris, Ph.D. degree in theoretical and applied mechanics
“A bimorph multi-layer piezoelectric vibration energy harvester,” in from Cornell University, Ithaca, NY, USA, in 1993.
Proc. PowerMEMS, Leuven, Belgium, 2010, pp. 335–338. He is currently a Professor and the Chairperson of
[17] T. Eggborn, “Analytical models to predict power harvesting with piezo- the Center for Nano Science and Engineering with
electric materials,” M.S. thesis, Dept. Mech. Eng., Virginia Polytech. the Indian Institute of Science, Bangalore, India.
Inst. State Univ., Blacksburg, VA, USA, 2003. He is also an Associate Faculty Member with the
[18] R. Shukla, L. C. Lim, and P. Gandhi, “Piezoelectric single crystal Department of Mechanical Engineering. He taught
power generator for low frequency vibrating machines and struc- at the Sibley School of Mechanical and Aerospace Engineering, Cornell
tures,” presented at the 18th IEEE Int. Symp. Appl. Ferroelectr., University, from 1993 to 1996. Since 1996, he has been with the Department
Aug. 2009, pp. 1–9. [Online]. Available: http://ieeexplore.ieee.org/ of Mechanical Engineering, Indian Institute of Science. He was an Invited
stamp/stamp.jsp?tp=&arnumber=5307554 Professor with the École Polytechnique Fédérale de Lausanne, Lausanne,
[19] S. Rammohan, C. M. Ramya, S. Jayanth Kumar, A. Jain, and R. Pratap, Switzerland, from 2004 to 2005. His current research interests include
“Low frequency vibration energy harvesting using arrays of PVDF MEMS design, computational mechanics, nonlinear dynamics, mechanobi-
piezoelectric bimorphs,” J. Inst. Smart Struct. Syst., vol. 3, no. 1, ology, and vibroacoustics. He is a Fellow of the Indian National Academy of
pp. 18–27, Mar. 2014. Engineering.

You might also like