ABSTRACT
Title of Thesis: EMBEDDED STRAIN SENSOR WITH POWER 
SCAVENGING FROM BRIDGE VIBRATION
Ming Wang, Master of Science, 2004
Thesis directed by: Professor Peter C. Chang
Department of Civil Engineering
The Objective of this thesis is to investigate an embedded strain sensor system with 
power scavenged from highway bridge vibration for structure health monitoring. A power 
scavenging scheme is developed to power the sensor system so as to eliminate the 
dependence of batteries. Three mechanisms (PZT patch, cantilever beam and PZT 
stacks/slugs) are described that are suitable for the remote sensor system to scavenge 
mechanical energy at different locations in a bridge. Calculations and experiments are taken 
to examine the feasibility of the idea, and predicted power generation is compared to 
solar/light power. A scheme of an embedded strain sensor using resonant MEMS beam 
structure is introduced along with the theoretical analysis and simulation. The last 
consideration is the signal propagation attenuation in concrete. Measurements are taken to 
determine the RF signal attenuation at 900 MHz range transmitted through a concrete slab. 
Theoretical calculations are validated by experimental results.
EMBEDDED STRAIN SENSOR WITH POWER SCAVENGING 
FROM BRIDGE VIBRATION
by
Ming Wang
Thesis submitted to the Faculty of the Graduate School of the
University of Maryland, College Park in partial fulfillment
of the requirements for the degree of
Master of Science
2004
Advisory Committee:
Professor Peter C. Chang, Chair
Professor Chung C. Fu
Professor Darryll Pines
?Copyright by
Ming Wang
2004
ii
Table of Contents
List of Figures............................................................................................................. iii
List of Tables................................................................................................................v
Introduction ..................................................................................................................1
1.1 Power scavenging from highway bridge vibration.......................................3
1.2 Embedded beam-type strain sensor ..............................................................5
1.3 Signal attenuation in concrete ......................................................................5
Embedded Sensor System for Structural Health Monitoring.......................................8
Power Scavenging from Highway Bridge Vibration..................................................11
3.1 Bridge Vibration.........................................................................................11
3.2.1 Piezoelectric Phenomenon..................................................................17
3.2.2 PZT Patch ...........................................................................................18
3.2.2.1 Power generation from truckload-induced strain ...........................21
3.2.2.2 Power generation from strain induced by free vibration................24
3.2.3 Cantilever Beam with Tip Mass .........................................................26
3.2.4 PZT Stacks/Slugs................................................................................32
3.3 Energy Storage Consideration....................................................................36
3.4 Experimental Results..................................................................................36
3.5 Comparison to other forms of available energy .........................................39
Resonant Beam-type Strain Sensor ............................................................................42
4.1 Beam Vibration Analysis ...........................................................................43
4.2 ANSYS Simulation Result .........................................................................45
4.2.1 Beam with no pre-tension...................................................................46
4.2.2 Pre-tensioned MEMS Beam...............................................................47
4.3 Capacitance Excitement of the Beam Vibration ........................................48
4.4 Capacitance Detection of the Beam Resonant Frequency..........................49
4.5 Temperature Compensation........................................................................51
Measurement of Propagation Attenuation through Concrete.....................................53
5.1 Electromagnetic Wave Propagation ...........................................................53
5.1.1 Reflection and Transmission Coefficient ...........................................53
5.1.2 Penetration Depth ...............................................................................57
5.2 Dielectric Constant of Concrete .................................................................58
5.3 Experimental Set-up and Results................................................................59
Conclusion..................................................................................................................65
6.1 power scavenging from highway bridge vibration.....................................65
6.2 Embedded MEMS beam-type strain sensor ...............................................66
6.3 Propagation Attenuation through Concrete at 900 MHz............................67
References ..................................................................................................................68
iii
List of Figures
Figure 1.1 SG-Link node..???????????..???????....??.? 2 
Figure 2.1 Block diagram of embedded sensor system................................................9
Figure 3.1 Schematic drawing of girder bridge and truckload model........................12
Figure 3.2 Truckload model used in ANSYS.............................................................12
Figure 3.3 Finite element mesh plot...........................................................................13
Figure 3.4 First four vibration mode shapes...............................................................14
Figure 3.5 Dynamic displacement of girder, (a) vertical displacement at mid-span 
and (b) vertical displacement at a point 2 feet from support..............................15
Figure 3.6 Dynamic strains at (a) mid-span and.........................................................16
Figure 3.7 Piezoelectric sensor operation modes .......................................................18
Figure 3.8 Strain changes of girder bottom at (a) mid-span and ................................20
Figure 3.9 Vibration signals after filters (mid-span location) ....................................23
Figure 3.10 Peak power output under different test resistance load...........................23
Figure 3.11 Instantaneous power under 3.85M Ohms load........................................24
Figure 3.12 Peak power output under different test resistance load (for high 
frequency vibration) ...........................................................................................25
Figure 3.13 Instantaneous power under 281 K Ohms load ........................................26
Figure 3.14 Schematic drawing of cantilever beam with tip mass.............................28
Figure 3.15 Schematic drawing of cantilever-type generator ....................................28
Figure 3.16 Beam distortion y(t) at tip and strain in PZT layer .................................31
Figure 3.17 Peak power output under different test resistance load...........................31
Figure 3.18 Power estimation with different PZT layer thickness.............................32
Figure 3.19 Vehicle-induced compressive impact stress on PZT stack. PZT stack 
locates at the mid-span point on the top of interior girder. A two-axle truck 
moves on bridge with speed 60mile/hour...........................................................34
Figure 3.20 A multilayer stack under stress ...............................................................35
Figure 3.21 Stress amplification scheme....................................................................35
Figure 3.22 PZT stack array .......................................................................................35
Figure 3.23 Parasitic patch on a free vibration beam .................................................38
Figure 3.24 Experimental and estimated peak power output .....................................38
Figure 3.25 Solar panel???????????????????????...39
Figure 4.1 System for resonant beam strain sensor....................................................43
Figure 4.2 x and w axis set up for studying the beam vibration.................................43
Figure 4.3 Schematic diagram of the vibrating MEMS beam....................................46
Figure 4.4 MEMS beam resonant frequency response with the strain change...........47
Figure 4.5 Schematic diagram of compensation beam structure................................52
Figure 5.1 Electromagnetic wave propagates from medium1 to medium 2...............55
Figure 5. 2 Schematic drawings of TE and TM waves ..............................................55
Figure 5.3 [Loulizi, 2001] Concrete dielectric constants vs. frequency.....................59
Figure 5.4 Schematic drawing of experiment setup ...................................................61
Figure 5.5 Schematic drawing of helical antenna ......................................................61
iv
Figure 5.6 Transmitting of RF energy through different material. (a) Open space. (b) 
Wood wall board. (c) Concrete slab. (d) Metal plate. For all four plots, x-axis 
frequency vary from 0.5 to 1.5 GHz, with 50 MHz per grid. y-axis is the 
20log
10
(E
OUT
/E
IN
), with 10 dB per grid. .............................................................63
v
List of Tables
Table 3.1 Modal analysis results ................................................................................13
Table 3.2 PZT (PSI-5A-S4-ENH [Piezo Systems, Inc.]) material properties ............20
Table 3.3 Power generation comparison ....................................................................40
Table 3.4 Power obtained v.s. Power Demand...........................................................41
Table 5.1 Attenuation taken at 917.5MHz .................................................................63
1
Chapter 1
Introduction
The traditional method to ensure the structural safety of civil infrastructure is to 
do visual inspection regularly. This is a simple but ineffective means since the health of 
structures can not be diagnosed adequately by its appearance alone. Critical 
deteriorations or damages often occur inside the structure, for example the corrosion of 
rebars in concrete. Corrosion initiation and growth are difficult to be perceived and 
controlled. Research on structural health monitoring for civil infrastructures and 
space/aircraft transportation system has experienced rapid growth in recent years. For 
example, accelerometers are placed on a structure, and then the vibration of the structure 
can be used to determine the condition of the structure. This is a global health 
monitoring method by observing changes of dynamic properties of the structure. Local 
monitoring method is to place sensors at some pre-determined hot spots where damage is 
expected. Both global and local methods need to install sensors in the investigated 
structure. Useful structural information (e.g.: strain/stress, acceleration, temperature etc.) 
is then obtained by monitoring those sensors.
The traditional methods of communication between sensors and data collection 
units rely on wire connections. However not only is the wiring a large number of sensors 
costly, but also large bundles of wires from different parts of a structure are difficult to 
manage and they are subjected to environmental deterioration. So, remote wireless 
technology becomes an attractive means to establish communication between sensors 
and a central unit.
2
The concept of wireless monitoring system was introduced in [Straser et al., 
1998], and a wireless modular monitoring system (WiMMS) [Lynch et al., 2000] was 
then developed using economical and commercially available components for 
implementing the wireless communication of sensor measurements in structures. A 
WiMMS unit is powered by 3 groups of 6 AA alkaline  batteries (18 total), which 
provides up to 21 days in sleep mode and up to 11 hours for continuous operation [Straser 
et al., 2000]. SG-Link? Wireless Strain Gauge System [MicroStrain, Inc.] is a 
commercially available product which combines full strain gauge conditioning with 
micro datalogging transceiver system for fulfilling wireless strain monitoring. SG-Link 
node, as shown in Figure 1.1, is powered by a 3.6 volt lithium ion AA size internal 
battery which provides up to 230 hours of lifespan.   
In this thesis, an 
embedded sensor system 
for structural health 
monitoring of highway 
bridges is described. This 
sensor system employs a 
power scavenging device 
to harvest energy from the 
bridge vibration so that the
dependence of batteries is 
eliminated, and the lifespan of the sensor system is then increased. Strain information is 
obtained by a vibrating beam-type stress sensor and the integrated transmitter sends 
Figure 1.1 SG-Link node
3
those results back to a handheld monitor via radio frequency signals. This realizes the 
operation of a wireless, self-powered and embedded sensor system for structural health 
monitoring. 
This thesis concentrates on three main aspects: power scavenging from highway 
bridge vibration, embedded beam-type strain sensor and signal attenuation in concrete.
1.1 Power scavenging from highway bridge vibration
The capability of health monitoring system relies on the performance and 
reliability of sensors. So sensors should be able to work in a sustained mode and do not 
need to be replaced or repaired over the life of the structure. Embedded sensors are
recently considered as promising means. However, powering these sensor systems is a 
significant problem since battery power offers only a limited life span, and battery 
management is inconvenient and impractical. Self-powered sensors are attractive 
because they allow the sensors to be tetherless. Moreover they make use of the energy 
from the ambient environment which is usually dissipated as heat and radiation. The 
harvested energy could be converted into a useful form upon the demand of sensors. The 
ideal case is that the harvested power is sufficient to provide a constant power supply for 
the sensor system. This depends on both the amount of power harvested from the 
environment and the power consumption of the sensors. An alternative way is to store 
the generated energy and give a power burst to the sensor system when power is needed. 
Various approaches have been studied to convert and store energy from ambient 
energy sources, including solar [Lee et al., 1995], [Warneke et al., 2002], human power 
[Starner, 1996], radio frequency [Friedman et al., 1997], [Wu et al., 2001], [Jung et al., 
4
1999], thermal gradients [Glosch et al., 1999] and mechanical vibration [Williams et al., 
1997], [Meninger et al., 2001], [Amirtharajah et al., 1998], [Glynne-Jones et al., 2001], 
[El-hami et al., 2001], [Shearwood et al., 1997]. 
A successful example of power scavenging is the ?parasitic piezoelectric shoe? 
[Shenck et al., 2001] developed by the MIT Media Laboratory. The ?piezoelectric? shoe 
utilizes PZT uniform and polyvinylidene fluoride (PVDF) stave mounted under an insole 
as the power generating element; an average of 1.3 to 8mW of power was collected at a 
0.9 Hz walking pace. The harvested energy successfully powered an RFID tag system 
and transmitted an identification signal to local surroundings. Similarly researchers in 
Sandia National Laboratory have developed piezoelectric-based self-powered sensor tags 
and proposed the idea to use the sensor tags to monitor the state-of-health of structures 
[Pfeifer et al., 2001].  The design of a MEMS transducer system that converts ambient 
mechanical vibration into electrical energy through the use of a variable capacitor was 
presented in [Meninger et al., 2001]. The feasibility of operating a digital system from 
power generated through a vibration-based power generator is demonstrated in 
[Amirtharajah et al., 1998], in which an integrated chip was designed and tested. A thick-
film piezoelectric microgenerator and vibration-based electromechamical power 
generator were introduced and experimental setups described in [Glynne-Jones et al., 
2001] and [El-hami et al., 2001], respectively.
With the development of vibration-to-electric energy conversion technology, 
researchers have begun to think of using energy harvesting means to power the sensors 
installed in highway bridges, such as accelerometers, strain gages, thermometers,
anemometers, and even video monitors. [Williams et al., 1997] presented a feasibility 
5
study of a vibration-to-electric generator designed to generate power from bridge 
vibration. However it is not clear how much power can be harvested in a highway 
bridge. This thesis attempts to estimate the power that can be generated quantitatively. 
1.2 Embedded beam-type strain sensor
The most common strain sensors are surface mounted resistance strain gauges.  
Vibrating wire spot-weldable strain gauge [Ace Instrument Co., Ltd.] is designed to be 
embedded in reinforced concrete to detect strain in concrete or strain in steel rebars. This 
kind of embedded strain gauge needs shielded cables to transmit the signals out. So the 
amount of such sensors in a structure has to be restricted in order not to influence the 
performance of the structure. 
In this thesis, a MEMS beam-type strain sensor which can be either mounted on 
the surface or embedded inside of a structure is investigated. The device is a beam fixed 
at both ends for which the resonant frequency is determined by its axial strain which is 
related to its resonant frequency. This kind of strain sensor is more stable than the 
traditional resistance strain gauges. The linear relationship between strain and frequency 
is studied by simulating a beam-type device under different axial stresses. The 
capacitance excitement and detection technologies are also investigated.
1.3 Signal attenuation in concrete
Once an embedded sensor is installed in a bridge, and it is powered by the means 
of energy scavenging, it is necessary to know if the harvested power is sufficient for the 
sensor?s demand. The attenuation of radio frequency energy is ineluctable when radio 
6
frequency wave transmits through concrete to reach the embedded sensors or when the 
sensors transmit signals out. For a scheme of power scavenging, the energy harvested by 
the sensors should be sufficient to complete sensing, data processing and data out-
transmitting. If the propagation attenuation is too big, the signal strength sent from 
transmitter has to be high enough, but sometimes this requirement is infeasible when the 
power supply is limited. Some empirical and theoretical models have been developed to 
investigate the influence of construction material such as concrete wall, wood or steel 
door, windows and so on in modern office building environment [Lafortune et al., 1990], 
[Honcharenko et al., 1992], [Honcharenko et al., 1993], [Seidel et al., 1994], [ Lawton et 
al., 1994] and [Lawton et al., 1999]. In order to determine the depth at which an 
embedded sensor can possibly be placed, the attenuation of RF energy through concrete 
needs to be studied. 
In this thesis, chapter 2 introduces the embedded sensor system for structural 
health monitoring. Chapter 3 describes the simulation of the traffic-induced vibration of 
a girder bridge model and then investigates three approaches for piezoelectric-based 
power scavenging based on the simulation results. Power storage considerations and 
some experimental results are also discussed in this chapter, and the expected power 
generation is compared to solar power and the power demand estimation of the 
embedded strains sensor system. In Chapter 4, a resonant MEMS beam strain sensor is 
studied. The capacitance excitement and detection technologies are also discussed. In 
chapter 5, the measurement of propagation attenuation through concrete at 900 MHz 
range was presented. Chapter 6 concluded this thesis with summary of results and 
insights and indicates areas of future research.
7
8
Chapter 2
Embedded Sensor System for Structural Health Monitoring
Strain sensors often need to be embedded inside the structure to measure local 
strain of the structural components such as reinforcement steel. The goal of this research 
is to investigate the possibility of using an embedded sensor system for health monitoring 
of highway bridges. The sensor system consists of a sensor module, a power scavenging 
module, a system controller module and a transceiver module. The sensor module takes 
measurements of structural information such as strain. The power scavenging module is 
to harvest energy from ambient sources such as highway bridge vibration, and the 
harvested energy is stored in a power storage device. This module provides the power 
demand for the sensor system. The system controller module controls the operation 
modes of the system including sleeping mode and working mode. The transceiver module 
has two functions: receiving command signals from a handheld monitor and transmitting 
information of structures back to the handheld monitor. The system block diagram is 
shown in Figure 2.1.
9
Figure 2.1 Block diagram of embedded sensor system
Without commands from handheld monitor, the sensor system is in ?sleeping? 
status. The power scavenging module provides a constant power for receiver and system 
controller module to be sustained in a ?waiting? status. When the receiver gets a ?wake 
up? command signal from the handheld monitor, the system controller module gives 
?working? commands to the sensor and power scavenging modules. The power 
scavenging module provides sufficient power for the demand of the sensor module and 
the transceiver module by discharging the power storage device. The embedded sensor 
starts to collect structure information and the transmitter sends those results back to the 
handheld monitor. When the handheld monitor obtains the expected information, it sends 
a ?sleeping? command out, and the sensor system goes to ?sleeping? mode. The collected 
information is either stored in the handheld device or sent to a central unit for advanced 
analysis.  
The sensor system has three main advantages:
Hand held monitor
Transceiver Module
Receiver
Transmitter
Power Scavenging Module
Power scavenging device
Power storage device
c
o
m
m
a
n
d
S
t
r
u
c
t
u
r
e
 
I
n
f
o
r
m
a
t
i
o
n
Sensor Module
Embedded sensor
System Controller Module
Power 
supply
Power supply
Power supply
?working/
sleeping?
?
w
o
r
k
i
n
g
/
s
l
e
e
p
i
n
g
?
S
t
r
u
c
t
u
r
e
 
I
n
f
o
r
m
a
t
i
o
n
c
o
m
m
a
n
d
10

 Wireless

 Embedded in structures

 Self-powered
Since the operation of the embedded sensor system relies on the power supplied 
by the power scavenging module, Chapter 3 gives the estimation of expected power 
generation from bridge vibration, and compares it to solar power. Finally, the power 
demand estimation of the sensor system is also given for comparison. 
Traditional resistance strain gauges drift significantly in even short periods of 
time, so they are not suitable for the purpose of long-term monitoring. Chapter 4 
introduces a MEMS resonant beam-type strain sensor which is more stable, and it can be 
embedded and it is wireless. Excitation and detection methods are also investigated.  
Since the embedded sensor system is designed to be placed inside of a structure, 
the signal attenuation has to be considered when the transceiver module sends a signal 
out or receives a signal from handheld monitor. Chapter 5 investigates the signal 
attenuation problem by doing measurement.                
11
Chapter 3
Power Scavenging from Highway Bridge Vibration
3.1 Bridge Vibration
Highway bridges vibrate considerably because of vehicle loading and wind. This 
mechanical energy is dissipated to the environment as heat. In order to determine the 
amount of mechanical energy that can be harvested, a typical T-shape girder bridge 
model was developed using ANSYS [ANSYS, Inc., 1998] and the dynamic 
characteristics of the bridge were obtained by modeling the passage of trucks.
A typical single span bridge is assumed. It is supported by three T-shape girders, 
and it is assumed to be simply supported at both ends. The traffic on the bridge consists 
of two lanes with a single lane in each direction as shown in Figure 1. The span length, L, 
is 88 feet and the deck thickness is 7 inches, which represents a typical highway overpass 
bridge. The height of the T-shape girder is 65 inches and the thickness is 8 inches. The 
span space, S, is 10 feet. 
The truck wheel load is considered as a three-axle truck (36KN-142KN-142KN) 
as described in AASHTO Standard (HS20-44 Loading). Figure 3.2 shows the truck 
loading model. Assume the trucks speed to be 96 km/hour (60 mph). At   this speed a 
wheel takes approximately 1 second to cross this single span. 
12
Figure 3.1 Schematic drawing of girder bridge and truckload model
Figure 3.2 Truckload model used in ANSYS
A four-node isoparametric shell (Shell 63) [ANSYS, Inc., 1998] is chosen as the 
structure element in ANSYS. Both modal and transient dynamic analyses were 
performed. Figure 3.3 shows the finite element mesh used.
T
r
a
f
f
ic
D
ir
e
c
t
io
n
T
r
a
f
f
ic
D
ir
e
c
t
io
n
L
S
S
T Girder section
16K
4K
16K
16K 16K
4K
2
0
 
f
e
e
t
1
4
 
f
e
e
t
6
 
f
e
e
t
13
The modal analysis shows that the natural frequency of the first mode is 8.51 Hz. 
Table 3.1 shows the first five modes of vibration. Figure 3.4 gives the first four vibration 
mode shapes.
Transient dynamic analysis is performed to determine the dynamic response of 
the structure under the action of the moving truck. The dynamic displacement of an 
interior girder was obtained from the transient dynamic analysis. Figure 3.5 (a) and (b) 
shows the time-history displacement at the mid-span point and a point 2 feet from the left 
support, respectively. Figure 3.6 (a) and (b) show their dynamic strains.
Table 3.1 Modal analysis results
Figure 3.3 Finite element mesh plot
1
st
 mode 2
nd
 mode 3
rd
 mode 4
th
 mode 5
th
 mode
8.51Hz 8.56Hz 12.14Hz 17.71Hz 19.35Hz
14
Figure 3.4 First four vibration mode shapes
15
0123456
-5.400
-5.200
-5.000
-4.800
-4.600
-4.400
-4.200
D
i
s
p
l
a
c
e
m
e
n
t
 
(
m
m
)
Time (t)
(a)
(b)
Figure 3.5 Dynamic displacement of girder, (a) vertical displacement at mid-span 
and (b) vertical displacement at a point 2 feet from support
0123456
-1.200
-1.000
-0.800
-0.600
-0.400
-0.200
0.000
D
i
s
p
l
a
c
e
m
e
n
t
 
(
m
m
)
Time (s)
16
(a)
(b)
Figure 3.6 Dynamic strains at (a) mid-span and 
(b) a point 2 feet from support
These results show that the dynamic responses consist of two parts. One part 
corresponds to the bridge?s free vibration which corresponds to the higher frequency 
01234
85
90
95
100
105
110
115
120
S
t
r
a
i
n
 
(
1
0
 
 
 
 
 
 
-
6
 
m
/
m
)
T im e (s)
01234
21 0
21 5
22 0
22 5
23 0
23 5
24 0
24 5
S
t
r
a
i
n
 
(
1
0
 
 
 
 
 
 
-
6
 
 
m
/
m
)
T im e  (s)
17
signals (around 8 Hz), and the other part corresponds to the truckload-induced vibration 
which appears as the higher amplitude signal with lower frequency (around 0.8 Hz). 
3.2 Approaches for Power Scavenging
The idea of energy scavenging here is to employ a piezoelectric structure as the 
power generation element. Piezoelectric materials generate an electricity when subjected 
to a mechanical deformation; and vice versa. The piezoelectric materials such as PZT 
(lead zirconate titanate), ZnO, Quartz, PVDF (polyvinylidene fluoride) etc. have been 
widely used as sensors and actuators. In [Weinberg, 1999] the solutions of working 
equations for a piezoelectric beam are derived. 
In this thesis, three types of approaches are studied: patch, cantilevered beam, 
and stacks/slugs. They are suitable for harvesting different kinds of mechanical energy in 
a bridge: structural strain, vibration, and impact compression. 
3.2.1 Piezoelectric Phenomenon
Piezoelectric phenomenon was first discovered in 1880s by Pierre and Jacque 
Curie. They found that some crystalline materials generate a voltage when compressed 
(generator effect) and change shape when an electric field is applied (motor effect). 
Although some natural materials (eg: quatz) are found to show piezoelectric effects, 
most modern piezoelectric devices are made with artificial polycrystalline ceramics such 
as PZT (lead zirconate titanate), which is considered in this thesis. Piezoelectric 
properties are directionally dependent. Figure 3.7 shows the 33 and 31 operation modes 
under generator effect. P denotes the polarization direction of the PZT material. 
18
Figure 3.7 Piezoelectric sensor operation modes
The piezoelectric behavior can be described by the following equations:
ij ij
D d X E
=+ (3.1a)
i ij i
S sX d E=+  (3.1b)
The index i indicates that the electrodes are perpendicular to axis i, and index j 
indicates that the piezoelectric induced strain or the applied stress is in the direction j. D 
is electrical displacement along polarization axis, and S is strain. d
ij
 is defined as 
piezoelectric strain coefficient. 
 is the dielectric constant of the material. E is the electric 
field generated between electrodes. X is mechanical stress developed in piezoelectric 
material. s is mechanical compliance in a constant electrical field. Equation 3.1a reflects 
the generator effect and equation 3.1b reflects the motor effect.
3.2.2 PZT Patch
The patch configuration is the most direct and simple means to convert bridge 
vibration into power. PZT patches can be placed on the surface of girders, usually on top 
+
-
P
F
F
+
-
P
F
F
(a) 33 mode
3
1
2
(b) 31 mode
19
of a flange or on the bottom of a slab, where the strain change is high relative to other 
locations. At these locations, the PZT patches are subjected to stress changes due to 
traffic loads. If the bond between the PZT patch and concrete transfers the deflection 
perfectly, the attached PZT patch has the same strain as the girder. Figure 3.8 (a) and (b) 
show the strain change at the mid-span point on the bottom of girder and at the point 
close to the support, respectively. The strain change in the PZT patch can be divided into 
two parts: (1) traffic-induced strain and (2) free vibration strain. The traffic-induced 
output has the properties of high amplitude but low frequency (26 micro strain and 0.77 
Hz). The free vibration output has the properties of low amplitude but relatively high 
frequency (4 micro strain and 8 Hz). The total power harvested by the PZT patches is 
from the combination of these two vibration sources. Here a low-pass filter is used to 
obtain the truck-induced vibration and a high-pass filter used to obtain the free vibration. 
Then the power generation from these two vibration sources is calculated. Figure 3.9 
shows the vibration signals after the filters.
If a 10cm?10cm?0.2cm PZT patch with mechanical and piezoelectric properties 
shown in Table 3.2 is used, the harvested power can be estimated.
20
Table 3.2 PZT (PSI-5A-S4-ENH [Piezo Systems, Inc.]) material properties
(a)
(b)
Figure 3.8 Strain changes of girder bottom at (a) mid-span and 
(b) the point which is 2 feet to the left support
01234
-5
0
5
10
15
20
25
30
S
t
r
a
i
n
 
c
h
a
n
g
e
 
(
1
0
 
 
 
 
 
 
 
-
6
 
 
 
m
/
m
)
Tim e (s)
d
31
(m/volts)
g
31
(voltm/N)




3
k
31


(Kg/m)
E
11
(N/m
2
)

-190E-12 -11.6E-3 1800 0.32 7800 6.6E10 0.31
d
31
: piezoelectric strain coefficient
g
31
: piezoelectric voltage coefficient


3
: relative dielectric constant between poling electrodes
k
31
: coupling coefficient

: density
E
11
: elastic modulus in longitude 
: Poisson?s ratio
01234
-5
0
5
10
15
20
25
30
S
t
r
a
i
n
 
c
h
a
n
g
e
 
(
1
0
 
 
 
 
 
-
6
 
 
m
/
m
)
Time (s)
21
3.2.2.1 Power generation from truckload-induced strain
This part of the time-history of strain change can be considered as a periodic 
signal with frequencies around 0.5 Hz which is shown in Figure 3.9. The generated open 
circuit voltage from the PZT patch can be written in terms of strain: 
11 31open s
V E g T
= (3.2)
where 

S
 is the longitudinal strain of the patch. T is the thickness of the PZT ceramic 
layer.
From Figure 3.9, the maximum strain of the truck-induced vibration 

s
=21 micro 
strain. Substituting it into (3.2), the maximum open circuit voltage is approximately 31.5 
volts. Then the average energy generated in the PZT patch can be calculated based on the 
peak voltage output:  
2
2
1
CVE
e
=
(3.3)
C is the capacitance of the PZT patch which is determined by
T
A
C
30



= (3.4)
A is the surface area of the electrode. In this case, A is equal to 100 cm
2
. 

3
 is the relative 
dielectric constant between poling electrodes and 

0
 is the permittivity of vacuum. 
(

0
=8.85?10
-12
 farads/m)
From (3.4) the capacitance of the single PZT patch is about 80 nF. The available 
energy then is calculated from (3.3): 
JJvoltsFaradE
e
?40100.4)5.31(100.8
2
1
528
=?=?=

22
The average available power in the 0.5Hz vibration is then:
5
2 2 0.5 4.0 10 40
e e
P f E Hz J W?

=?? =? ? ? =  
f is frequency in Hz, and f is equal to 0.5Hz in this case. The area of the PZT 
patch is 100cm
2
. So the power density is about 0.4?W/cm
2
.
The output power can also be determined by connecting a test resistance to the 
patch device. The output power can be written: [Kendall, 1998]
2
11 31
2
( )
(1 ( ) )
S
o
A E d R
P
CR

	
	
=
+
(3.5)
where =2
f, and R is the test resistance. 
If the maximum strain from Figure 3.9 (

s
=21 micro strain) is substituted into to 
(3.5), the peak power output can be obtained at an optimized test resistance. Figure 3.10 
illustrates the peak power output versus test resistance, and it shows that the peak power 
is around 124 ?W at 3850 K Ohms. 
If the optimized test resistance (R=3850 K Ohms) is substituted into (3.5), and use 
the time-history strain change to replace 

s
, the instantaneous power can be obtained, and 
is shown in Figure 3.11. The average power output is about 33 ?W, which is close to 40 
?W estimated by equations 3.3 and 3.4.
23
Figure 3.9 Vibration signals after filters (mid-span location)
1000 2000 3000 4000 5000 6000 7000 8000
70
80
90
100
110
120
130
P
e
a
k
 
p
o
w
e
r
 
o
u
t
p
u
t
 
(
mi
c
r
o
 
W
a
t
t
s
)
Test resistance (K Ohms)
Figure 3.10 Peak power output under different test resistance load
(for low frequency vibtation)
01234
-5
0
5
1 0
1 5
2 0
 lo w  fre q u e n c y  v ib ra tio n
 h ig h  fre q u e n c y  v ib ra tio n
S
t
r
a
i
n
 
C
h
a
n
g
e
 
(
1
0
 
 
 
 
 
 
 
 
-
6
 
 
m
/
m
)
T im e  (s )
24
Figure 3.11 Instantaneous power under 3.85M Ohms load 
(for low frequency vibration)
3.2.2.2 Power generation from strain induced by free vibration
The part of free vibration can be considered as a periodic signal with frequency 
8.51 Hz shown in Figure 3.9. Following the same procedure as described in section 
3.2.2.1, the PZT single patch generates a peak power of about 100 ?W at 281 K ohms as 
shown in Figure 3.12, and an average power of 22 ?W as shown in Figure 3.13.
The results show that both low frequency and high frequency vibrations have 
contributions to the power generation. Under optimized resistive loads, the total available 
power from the PZT device is around 55 ?W. If the PZT patch is placed on the bottom of 
the interior girder at 1/4 span, the total available power output is around 25 ?W. 
The harvested power is related to the size of the PZT patch. Making the single 
patch thicker will increase the power output. A thick PZT patch is commercially 
01234
0
20
40
60
80
100
120
140
Average Power
P
ow
e
r
 (
m
i
c
r
o W
atts
)
Time (s)
25
available. For example, Morgan Electro Ceramics Ltd. supplies PZT sheets with 
thickness ranging from 0.2mm to 15mm [Morgan Electro Ceramics].
Although the strain in the girder is relatively high (up to 300 micro strain), the 
strain change due to traffic is relatively low (<30 micro strain). That is why the PZT 
patch seems not to be an effective way to scavenge power according to the above results. 
However the patches have their own advantages in that the electrical output from them is 
proportional to the strain on the girders. This is like the function of strain gauges. The 
maximum voltage or power output is connected to the maximum strain which is 
important for health monitoring purposes. For example, once the output voltage jumps 
over a pre-determined value, it can be recorded or transmitted out to a central unit by 
discharging the stored power in the capacitor. 
Figure 3.12 Peak power output under different test resistance load (for high 
frequency vibration)
100 200 300 400 500 600 700
20
40
60
80
100
P
e
a
k
 
p
o
w
e
r
 
o
u
t
p
u
t
 
(
m
i
c
r
o
 
W
a
t
t
s
)
Test resistance (K Ohms)
26
Figure 3.13 Instantaneous power under 281 K Ohms load 
(for high frequency vibration)
3.2.3 Cantilever Beam with Tip Mass
Figure 3.14 (a) is the schematic drawing of a cantilever beam structure with a tip 
mass. The tip mass changes the natural frequency of the beam structure.
The model of the beam structure can be simplified since the weight of the beam is 
small enough to be neglected compared to the tip mass. Figure 3.14 (b) and (c) show that 
the clamped end is subject to a motion u(t), and y(t) is the displacement of the free end. 
The equilibrium equation of the beam system is:    
( )my cy ky mu P t++= =

 
 

 (3.6)
m is the mass at the free end; c is the mechanical damping coefficient and k is the 
stiffness coefficient.
01234
0
20
40
60
80
100
Average power
P
o
w
e
r
 
(
m
i
c
r
o
 
W
a
t
t
s
)
Time (s)
27
Equation 3.6 shows that the problem of the cantilever beam structure vibration 
due to the excitation of the fixed end can be considered as the cantilever beam excited by 
a external force )()( tumtP 

= applied at the free end, see Figure 3.14 (c).
The excitation force can be obtained from the data of structural dynamic response. 
This is a second-order continuous-time system and it can be described in the canonical 
differential equation form:
2
2
n n
y y y u	 	++=

 
 

 (3.7)
where mk
n
=	  and 
n
mc 	 2= . u(t) can be obtained from a finite element 
simulation described previously and Figure 4 shows u(t) at the mid-span point and the 
point near the support. All the data are the nodal displacement in the vertical direction.   
One example implementation of the cantilever-type generator is shown in Figure 
3.15. The substrate (including the tip mass) is made of steel and a PZT layer is deposited 
on it. The fixed end is attached to a bridge girder and the dynamic displacement of the 
bridge makes it vibrate. So the PZT ceramic layer on the beam is subjected to dynamic 
strains. The strain in the PZT layer can be higher than directly mounting the PZT patch 
onto the girder as the tip mass may amplify the strain caused by the girder vibration. This 
kind of cantilever beam structure has been studied in [Kov?cs et al., 2001] as the 
prototype for an accelerometer.  
28
Figure 3.14 Schematic drawing of cantilever beam with tip mass
Figure 3.15 Schematic drawing of cantilever-type generator
As an example, a cantilever beam device with L
1
=50mm, L
2
=30mm, T
1
=0.2mm, 
T
2
=0.3mm, T
3
=23mm and b=30mm is considered. Its stiffness coefficient is:
3
2
1
3
( )
2
EI
k
L
L
=
+
(3.8)
where E and I are the Young?s modulus and the moment of inertia of the two-layer 
cantilever beam. 
(a)
(b)
y(t)
m
u(t)
u(t)+y(t)
P(t)
y(t)
(c)
T2
T1
b
L1 L2
PZT (E1)
Steel (E2)
T3
x
0 L1 L1+L2
Z
29
1 1 2 2
1 2
E T E T
E
T T
+
=
+
(3.9)
3 3 2 2
1 2 1 2 1 2 2
( ) ( / 2 ) ( / 2)
12
b
I T T bT T T z bT z T== + + +  +  (3.10)
z denotes the distance from the bottom of the beam to its neutral axis. In order to 
determine E and I, the position of the neutral axis of the two-layer beam needs to be 
determined. z can be written as [Weinberg, 1999]:
1 1 2 1 2 2 2
1 1 2 2
( / 2) ( / 2)E T T T E T T
z
E T E T
++
=
+
(3.11)
The natural frequency of the cantilever beam is given by [Kov?cs et al., 2001]: 
3
2 3 1 2
1 3
2 2 ( / 2)
n
n
s
EI
f
L bT L L
	



==
+
(3.12)
where 

s
 is the density of the substrate layer. For the dimension chosen for the example 
cantilever, the stiffness coefficient is found to be 572 N/m, and its natural frequency is 
about 11Hz.
In [Williams et al., 1997], it is mentioned that u(t) can be transformed into a sum 
of sinusoids by using a Discrete Fourier Transform algorithm, then average power is 
derived by solving the differential equation 3.7. Here the differential equation 3.7 is 
solved with the help of MATLAB [MathWorks, Inc.]. u(t) is the displacement of the mid-
span point from 0 to 1.5 second shown in Figure 3.5(a). The tip mass displacement y(t) 
due to beam distortion was calculated and the result is shown in Figure 3.16. 
The strain in the PZT layer is given by:
1 2 1 2
( )( )
( ) ( )
2
s
k L x L T T
x y t
EI


 ++
= (3.13)
30
where x is the distance from the fixed end. The average strain is taken as the strain at 
x=L
1
/2 in the PZT layer as shown in Figure 3.15. The average energy generation can be 
written in terms of average strain 

s
:
2 2
31 31 1 2 1e n s
P f g d L bT E 
= (3.14)
For a 5cm?3cm?0.02cm PZT ceramic layer deposited on the steel cantilever 
beam, the average generated power is:
3 12 2 2 2
10 2 4 2
11.5 11.6 10 190 10 5 10 3 10 0.02 10
(6.6 10 ) (100 10 ) 330
e
P
W?
 

=???? ??????
?? ? ? =
 
The size of the cantilever beam is about 8cm by 3cm. So the power density is 
around 14 ?W/cm
2
. Figure 3.17 illustrates that the peak power is 1.1 mW at 120 KOhms. 
Thicker PZT ceramic layer generates more power. But if the deposited PZT layer 
is too thick, the stiffness of the cantilever beam becomes high so that it will not develop 
much strain under excited vibration. To obtain the optimal thickness of PZT ceramic 
layer T
3
 for energy harvesting, the power obtained through the cantilever vibration is 
expressed in T
3
. Figure 3.18 shows the power scavenged as a function of the thickness of 
PZT layer. The maximum power is about 330 ?W, and the optimal thickness is 0.2mm. 
31
Figure 3.16 Beam distortion y(t) at tip and strain in PZT layer
Figure 3.17 Peak power output under different test resistance load
0.0 0.5 1.0 1.5
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
Time (s)
D
i
s
t
o
r
t
i
o
n
 
(
m
m
)
-150
-100
-50
0
50
100
150
A
v
e
r
a
g
e
 
s
t
r
a
i
n
 
i
n
 
P
Z
T
 
l
a
y
e
r
 
(
1
0
-
6
m
/
m
)
0 5 0 1 0 0 1 5 0 2 0 0 2 50 3 0 0
0
2 0 0
4 0 0
6 0 0
8 0 0
1 0 0 0
1 2 0 0
P
e
a
k
 
p
o
w
e
r
 
o
u
t
p
u
t
 
(
m
i
c
r
o
 
W
a
t
t
s
)
T e st re sista nc e  (K O h m s)
32
Figure 3.18 Power estimation with different PZT layer thickness
3.2.4 PZT Stacks/Slugs
Vehicle wheels give high impact forces on the bridge deck when they move on a 
bridge. This results in a compressive stress change between the decking and the girders. 
PZT stacks or slugs can be placed on the top of girders to sense this vehicle-induced 
compressive stress. Every time a vehicle moves over the point where a stack/slug of 
piezoelectric material is embedded, the stack/slug is subjected to an impact compressive 
force and generates a power impulse. Figure 3.19 shows the compressive stress change at 
the mid-span point when a 4.8-ton two-axle truck moves over the girder bridge. 
Single-layer piezoelectric stacks/slugs are usually used as elements for generating 
high voltage [Morgan Electro Ceramics, 2000]. One application is the ?Igniter? [Sensor 
Technology Ltd.] which generates a high voltage due to an impact, and this high voltage 
is applied to an air gap to generate an arc. The operation of the stack/slug generator 
considered here is similar to the ?Igniter.? Vehicle wheels push PZT stack/slugs to 
generate electric power. However high charge, not high voltage, is more desirable in 
generators. Multilayer PZT stacks are suitable for high stress environments and for 
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25
0
50
100
150
200
250
300
350
P
ow
er
 
es
t
i
m
at
i
on
 
(
m
i
c
r
o 
W
at
t
s
)
PZT layer thickness (m m)
33
harvesting large volumes of charge. Figure 3.20 shows a multilayer stack. The open 
voltage and capacitance are [Piezo Systems, Inc.]:
33
_
s s
s open
T g
V
N

= (3.15)
0 3 s
s
s
A
C N
t



= (3.16)
T
s
 is the total thickness of the stack. N is the number of layers. g
33
 is the 
piezoelectric voltage coefficient in the thickness direction. 
s
 is the stress in the thickness
direction, and 
s
=F/A
s
 . A
s
 is the cross section area of the stack. t
s
 is the thickness of a 
single layer.
The values in Figure 3.19 show that the compressive stresses on the top of the 
girder are not high. Higher impact force on a PZT stack/slug generates more electric 
energy. So a high stress on the PZT stack/slug generator is expected. A way to amplify 
the stress is to increase the vertical stress by using a piston type stress concentrator as 
shown in Figure 3.21. The stress on the stack/slug can be calculated by:
amp
s c
s
A
A


= (3.17)
A
amp
 is the top area of stress amplification piston. 
c
 is the traffic-induced vertical stress 
on the piston.
A 1cm diameter?2cm long 10-layerPZT stack with g
33
=24?10
-3
m/Volt (PSI-5A-
S4-ENH [Piezo Systems, Inc.]) is considered. Assume A
amp
=100cm
2
(10cm?10cm), 

c
=92KPa from simulated results. The open voltage and stack capacitance can be 
calculated from equation 3.15 and 3.16: V
s_open
=563Volts and C
s
=6.25pF. So the 
available energy is obtained from equation 3.3: E
s
=1mJ. Assuming the vertical 
34
compressive stress on the stack reaches its peak value every 2 seconds, the available 
power generated from the PZT stack is 1 mW.
The results show that the power generated in the PZT stack is relatively high 
when a stress amplification scheme is used, because when cars or trucks move across a 
bridge, the impact force between wheels and pavement is high. The high impact force on 
the PZT stacks placed in pavement can generate a relatively high power for a short 
duration. However, the high impact force may break the PZT stacks (compressive 
strength=0.5GPa for PSI-5A-S4-ENH [Piezo Systems, Inc.]). Figure 3.22 shows a 
scheme of stack array. This PZT stack array maybe also be used to sense the wheel 
impact force and the displacement of the stacks can be controlled by restraining the 
movement of the piston. 
Figure 3.19 Vehicle-induced compressive impact stress on PZT stack. PZT stack 
locates at the mid-span point on the top of interior girder. A two-axle truck moves 
on bridge with speed 60mile/hour.
0
10000
20000
30000
40000
50000
60000
70000
80000
90000
100000
0 0.2 0.4 0.6 0.8 1
time (s)
s
t
re
s
s
 
(
P
a
)
35
Figure 3.20 A multilayer stack under stress
Figure 3.21 Stress amplification scheme
Figure 3.22 PZT stack array
PZT stack
Array
PZT stacks
piston
housing
wheel impact force
PZT Stack
stress
amplification
piston
housing
Traffic-induced
compressive stress
-
+
F
 As
ts
Ts=Nts
36
3.3 Energy Storage Consideration
The generated power from a piezoelectric device has to be converted into a useful 
form. As just seen, although the output voltage from a piezoelectric transducer can be 
high, the output energy and frequency are quite low. Usually the generated charge passes 
a rectifier circuit, and then is stored in a capacitor bank. The capacitor is connected with a 
linear regulator (e.g. MAX666) [Maxim Integrated Products] which provides a specified 
stable voltage if the charge from the piezoelectric device can be generated continuously. 
This is the most direct and simple means of energy storing and discharging. In [Paradiso 
et al., 2001], the power harvested upon a single button push was stored and translated on 
the order of 0.5 mJ at 3 volts in 30 ms. However, charging a capacitor directly results in a 
substantial loss of the total available energy. More efficient means need to be developed. 
A self-powered digital signal processing system was tested and the generator subsystem 
and regulator architecture were discussed in [Amirtharajah et al., 1998]. [Shenck et al., 
2001] introduced a forward-switching power-conditioning system which was more 
efficient for converting raw electric energy into a useful form. This scheme could be 
applied to the similar low-frequency piezoelectric source from bridge vibration. 
3.4 Experimental Results
A simple experiment was conducted to verify the concepts of power scavenging 
described in the previous sections. A plastic beam (31.6cm?2.7cm?0.38cm) is used to 
generate vibration. The test setup is shown in Figure 3.23. The test sample is a single-
layer PZT sheet (6.3cm?3.1cm?0.02cm). The PZT sheet is attached to a plastic beam. 
The plastic beam vibrates under its natural frequency (around 8 Hz). The output voltage 
37
from electrodes on the PZT sheet was observed in an oscillograph. This is a validation 
testing for the patch-type generator. The output of PZT sheet subjected to mechanical 
vibration is measured, and the strain on the PZT sheet is also obtained from the strain 
gauge. The estimated values are calculated based on the strain value by applying the 
equations introduced in section 3.2.2. So the experimental results are compared to the 
estimated values. Moreover, various load resistances are connected to the patch 
generator, and the generated power is compared to the estimated values which are 
calculated by equation 3.5.
As shown in Figure 3.23, the single PZT sheet was attached close to the fixed end 
of the plastic beam, and a strain gauge was also installed at the same location of the 
plastic beam. The free end of the plastic beam was pushed to deflect 4cm then released. 
When the free end was pushed down to deflect 4cm, the strain value obtained from the 
strain gauge was 103 micro strain. Results show that the output is a periodic signal with 
frequency around 8 Hz which is close to the natural frequency of the plastic beam. With 
the effect of mechanical damping, the amplitude of the output signal decreased to zero 
after 5 seconds. The maximum peak-to-peak open voltage was measured to be around 25 
volts. This is close to the estimated open voltage 30 volts, which is calculated based on 
the strain value obtained from the strain gauge. The power generation from the patch-type 
generator is about 210?W (302?W, estimated value). The peak power output under 
various load resistances is shown in Figure 3.24. 
The experimental results show good agreement with the estimated values, and the 
patch-type generator performed properly as expected, and a remarkable amount of raw 
power is obtained from mechanical vibration. 
38
Figure 3.23 Parasitic patch on a free vibration beam
Figure 3.24 Experimental and estimated peak power output 
under different test resistance load
PZT patch
strain gauge
plastic beam
free vibration
0 50 100 150 200 250 300
200
400
600
800
1000
Test resistance (K Ohms)
P
e
a
k
 
p
o
w
e
r
 
o
u
t
p
u
t
 
(
m
i
c
r
o
 
W
a
t
t
s
)
 Estimated Values
 Experimental Data
PZT patch
plastic beam
Strain gauge
39
3.5 Comparison to other forms of available energy
Solar radiation is an attractive source for energy scavenging as its technology is 
well developed. A highway bridge has large amounts of surface exposed to sun light. No 
doubt that the solar energy collection transducer can 
provide abundant electric power on a sunny day. The 
available energy is decreased to roughly one 
hundredth of that on a cloudy day, and to one 
thousandth for indoor lighting conditions.  When 
solar cells are placed on bridges, they are not always 
exposed directly to the sun even on sunny days. At 
locations of interior girders, the illumination is quite 
low and it is similar to indoor lighting conditions. 
Many types of solar cells have been fabricated and studied. New solar cells have 
efficiency as high as 30%. Reference [Lee et al., 1995] introduces a miniaturized high-
voltage solar array (100 cells in 1 cm
2
) and it was utilized as a power supply for MEMS. 
SmartDust [Warneke et al., 2002] was designed to be powered by a SOI solar cell array 
which could harvest around 1 mW/mm
2
 in full sunlight or 1 ?W/mm
2
 under bright indoor 
illumination with 10 to 12% efficiency. 
Flexible solar panels are now commercially available. Figure 3.25 shows a 5cm 
?3.8cm PowerFilm 3V 22mA solar panel supplied by Iowa Thin Film Technologies 
[Iowa Thin Film Technologies, Inc]. The power output is around 66 mW for direct sun 
(100%) and 0.1mW under low illumination conditions (0.1%). So the power generation 
density is in the range of 3.4mW/cm
2
 to 5?W/cm
2
.  
Figure 3.25 Solar panel
40
Table 3.3 Power generation comparison
    The piezoelectric based power generation density is in the range of 0.0004 to 
1mW/cm
2
. This is quite low compared to that of solar cells in sunny conditions. Table 3.3 
shows the comparison. Even so, it is hard to say that solar power generation is the best 
choice to supply power for sensors in a bridge. The vibration-induced power generation 
has its competitive advantages. It is totally independent of climate and weather. At this 
point, solar cells are preferred in sunny conditions. Anytime the bridge is in use, 
piezoelectric transducers can generate power continuously both in daytime and nighttime. 
Moreover vibration-induced power generation is the most feasible means for powering 
embedded sensors. So a good idea would be to combine a vibration-based generator and 
solar cells together to power sensors in a bridge. 
For a wireless sensor system, various operations consume different amount of 
power. Usually the operations include sensing/data processing, data transmission and 
system sleeping. The operation of sensing/data processing is to acquire the changes (e.g.: 
resistance change) of sensors, and process the raw data (e.g.: delete some unwanted or 
unimportant data). The operation of data transmission is to send out useful data as radio 
frequency signals, and the operation of system sleeping is to maintain the mode of 
sleeping, in which the system terminates other operations temporarily and waits for 
activation commands. Table 3.4 shows the power requirement for the operations of a 
sunny cloudy nighttime
Solar 3~10mW/cm
2
0.01~0.1mW/cm
2
0
Vibration 0.0004~1mW/cm
2
0.0004~1mW/cm
2
0.0004~1mW/cm
2
41
wireless sensor system. The estimation of power scavenging from both solar and bridge 
vibration is based on a 10 cm
2
 generator and solar cell. In sunny days, the harvested 
power is sufficient for the operations of data transmission, sensing and data processing. 
In nighttime, the harvested power from bridge vibration is still enough for sensing, data 
processing and limited data transmission. 
Table 3.4 Power obtained v.s. Power Demand
Power Obtained (based on 10 cm
2
) Power Demand [Amirtharajah et al, 1998], 
[Paradiso et al., 2001], [Arms et al., 
2003], [Vittoz, 1994]
Daytime 
(sunny)
Nighttime Sensing/data 
procesing
Data 
transmission
System 
sleeping
Solar 30~100mW N/A
Vibration 0.004~10mW 0.004~10mW
Sloar+Vibration 30~110mW 0.004~10mW
0.1~10mW 5~50mW 0.01~0.1mW
42
Chapter 4
Resonant Beam-type Strain Sensor 
Because traditional resistance strain gauges drift significantly over even short 
periods of time, they are not suitable for the purpose of ?embedding?. A more stable 
strain sensor that is easily integrated with transceivers and system circuits is more 
desirable. In this chapter, a resonant beam strain sensor using standard CMOS MEMS 
technology [The MOSIS Service] is presented. Standard CMOS has been widely used to 
fabricate various sensors [Baltes, 1993], [Eyre et al., 1998]. The device in this chapter is 
a beam fixed at both ends for which the resonant frequency is determined by its axial 
stress. Changes in its axial stress give changes in its resonant frequency. The beams can 
be formed from metals or poly-silicon used in standard CMOS processing by etching 
thick field oxides away from underneath a beam. 
The system of the vibrating beam strain sensor is shown in Figure 4.1. A current 
pulse generating circuit creates a positive pulse to charge up the capacitance with large 
charge Q
1
(to induce the beam vibration) and the discharge of the charge Q
2
 < Q
1
(to
leave a small amount of charge Q for detecting the beam vibration). The reason this 
scheme is chosen is explained in Section 4.3. The beam resonant frequency reflects the 
axial strain on the beam. 
43
Current 
pulse 
generating 
circuit
Current 
pulse
+    +    +  +    +
-----
Deposited charge
Vibrating beam
Beam resonant 
frequency 
voltage signal
Voltage 
signal 
detecting 
circuit
Q
1
Q
2
Figure 4.1 System for resonant beam strain sensor
The vibrating beam strain sensor can be embedded into a structure. For example, 
it can be mounted onto the steel rebar in a reinforced concrete structure, which is similar 
to the usage of vibrating wire spot-weldable strain gauge [Ace Instrument Co., Ltd.]. The 
axial strain in the vibrating beam is changed with the stress change in the steel rebar. So 
the strain of the rebar can be obtained by detecting the frequency change of the vibrating 
beam. 
4.1 Beam Vibration Analysis
To study the beam vibration eigenvalues and eigenfunctions, the x and w axes are 
set up, as shown in Figure 4.2. The length of beam is l, the width is b, and the thickness is 
h.
d
0
d(x,t)
w(x,t)
x
w
0
h
Figure 4.2 x and w axis set up for studying the beam vibration
44
In general, the differential equation of motion for beam structure is given by 
equation 4.1 [Sniegowski et al., 1999]:
2 2 2 2
2 2 2 2
( , ) ( , ) ( , )
( , )
w x t w x t w x t
EI A P q x t
x x t x

  
+  =


 (4.1)
where w(x,t) is the vertical deformation function of position x and time t. E is the 
Young?s module, I is moment of inertia,  
 is density of the beam, A is the cross sectional 
area of the beam. P is the internal axial load applied on the beam and P=EA, where  is 
axial strain. q(x, t) is the external load on the beam. When there is neither axial nor 
perpendicular external load on the beam, the equation of motion can be written as 
equation 4.2.
2 2 2
2 2 2
( , ) ( , )
0
w x t w x t
EI A
x x t

 
+=


 (4.2)
To obtain analytical solutions, separation of variables can be applied as shown in 
equation 4.3 [Shames et al., 1991].
( , ) ( ) ( )w x t W x T t= (4.3)
Under the boundary conditions of both ends of the beam being clamped, the 
eigenfunctions can be solved as shown in equations 4.4 and 4.5. 
( )( ) cos
n n n
T t W t	=+ (4.4)
() ()
( ) ( )
() ()
() ()
cos cosh
( ) cosh cos sin sinh
sin sinh
n n
n n n n n
n n
k l k l
W x k x k x k x k x
k l k l

=  + 

(4.5)
in which 
n
 and k
n
 are:
( ) ( )cos cosh 1
n n
k l k l = (4.6)
45
2 4
n n
EI
k
A
	

= n=1,2,3? (4.7)
When an axial force P is applied uniformly on the beam and the perpendicular 
external load on the beam is zero, the equation of motion of the beam can be written as 
equation 4.8. 
2 2 2 2
2 2 2 2
0
w w w
EI A P
x x t x

  
+  =

 

 (4.8)
It is quite difficult to solve equation 4.8 analytically. One way to solve for the 
resonant frequency is the Rayleigh-Ritz method, and the solution has the form as shown 
in equation 4.9: 
( )( , ) ( )cosw x t W x t	=+ (4.9)
with W(x) as a trial function which satisfies the boundary conditions. Then ( , )w x t  can be 
solved by this method, and the beam resonant frequency  is a function of the strain 
applied on the beam. In this thesis, finite element code ANSYS is applied to simulate a 
vibrating beam under axial forces.
4.2 ANSYS Simulation Result
In this section, the resonant frequency variation with respect to the strain is 
presented. First, the situation when there is no pre-tension added on the beam was 
studied. Later, a pre-tensioned beam scheme is considered because the strain gauge 
should work under tension forces and compression forces. The simulation is done by 
using ANSYS [ANSYS, Inc., 1998].
46
4.2.1 Beam with no pre-tension
The schematic diagram of the vibrating MEMS beam is shown in Figure 4.3. The 
studied model is an fixed-end beam with length l of 100 micron, width of b=15 micron, 
and thickness of h=2 micron. The resonant frequencies of the beam structure under 
various situations of longitudinal axial stress are calculated in ANSYS. The composite 
material consists of SiO
2
 and Metal layer (Aluminum). The Young?s Modulus E and 
density 
 are given: for SiO2, 
 = 2200 Kg/m
3
, E=57 to 85 GPa; for Aluminum, 
 = 2702 
Kg/m
3
, E=70 GPa [Gardner et al., 2001])
While applying a tensile stress  along the beam, the strain in the beam is =/E. 
The resonant frequency change can be observed when the strain in the beam is changed. 
The ratio of the loaded resonant frequency and unloaded one, f
r
/f
0,
 is used to measure the 
resonant frequency change with respect to strain change. Figure 4.4 shows that the ratio 
f
r
/f
0
 has approximately a linear behavior with strain change. 
Figure 4.3 Schematic diagram of the vibrating MEMS beam
47
0 500 1000 1500 2000
1.0
1.1
1.2
1.3
1.4
1.5
f
r
 
/
 
f
0
Strain (10
-6
 m/m)
Figure 4.4 MEMS beam resonant frequency response with the strain change.
With zero longitudinal axial stress in the beam, its resonant frequency is 1133.08 
KHz. Under an application of 500 micro strains in tension, the resonant frequency is 
shifted to 1300.83 KHz. The calculation result shows that the resonant beam is sensitive 
to the axial strain change as low as 1 micro strain with an increase of resonant frequency 
of about 0.5 KHz for 1 micro strain in tension.
4.2.2 Pre-tensioned MEMS Beam
A strain gauge is used to measure both tensile and compressive strain change in a 
structure. For the MEMS resonant beam strain gauge presented here, a small compressive 
stress will easily make the thin beam structure buckle. There have been studies and 
reported results on growing pre-stressed layers [Ferman, 1999]. MOS fabrication is also 
adopting the technology of growing pre-strained layers in MOS transistor fabrication to
improve the performance and to look for new devices [Ernst et al., 2002].The making of 
pre-tensioned beams is not well developed, but it is possible.
48
For example, when the beam is pre-tensioned under 1000 micro strains, the initial 
resonant frequency f
0
 will be 1447.33 KHz. If an external compressive force is applied on 
the strain gauge, let?s say, 500 micro strains generated in compression, there is still 500 
micro strains in tension in the beam. The resonant frequency is shifted to f
r
 =1300.83KHz 
comparing to1447.33KHz.
4.3 Capacitance Excitement of the Beam Vibration
To excite the beam vibration, a pulse of current is used to charge the capacitance 
plates, with a large charge Q
1
. Then the Coulumb force on the beam bends the beam and 
excites the vibration. The current-pulse driven electrostatic actuators are examined in 
[Casta?er et al., 2001].
The Coulumb force, q
0
, per unit length on the bridge is given in equation 4.10:
2
0 2
0
1
2
Q
q
l b

=  (4.10)
Under a uniform force, q
0
, per unit length perpendicular to the beam, the deformation 
equation of the resonant beam is:
()
4
0
4 4
1
2
1 cos
8
n
n
q l l
w x
EI n





=




=

(4.11)
The mid point of the beam has the maximum deformation:
( )
4 4
0 0
4 4 4
1
1 cos 41
2 8 162
n
nq l q l
l
w
EI n EI






=


= 



(4.12)
To be able to excite the beam, a charge Q1 has to be relatively large so the 
Coulumb force could give the beam enough deformation. Assuming the mid point 
49
deflection is 0.1 micro meters, the required charge on the beam is about 8.4?10
-13
Coulumbs based on equations 4.10 to 4.12. This is much smaller than the charge 
generated from a PZT patch described in section 3.2.2. The generated charge from the 
PZT patch is in the order of 10
-6
:
-8 -6
0
Q=C U=8?10 Farads?32volts=2.6?10 Coulumbs (4.13)
So the harvested power from bridge vibration, as described in section 3, is sufficient to 
excite the beam considered here. 
Since q(x,t) plays an important role in equation 4.1 which describe the motion of 
the beam, solution of the eigenvalues cannot be usually done in closed-form. Also, the 
resonant frequency of the beam is also charge dependant, something that leads to 
difficulty for calibration. 
The scheme considered to solve this problem here is to use a positive current 
pulse, followed by a negative pulse. Therefore, the capacitance is charged up with a large 
Q
1
 to excite the beam vibration. Then the capacitance is discharged by a slightly smaller 
charge, Q
2
. Only a small amount of charge, Q
1
-Q
2
, is left on the capacitance. Since the 
charge is small, the remaining Coulumb force on the beam is negligible in the equation of 
motion. Equations 4.2 and 4.8 can be used to describe the motion of the beam without 
and with axial force.
4.4 Capacitance Detection of the Beam Resonant Frequency
The beam resonant frequency can be determined by measuring the output AC 
voltage frequency. The output voltage is related to the capacitance between the vibrating 
50
beam and bottom plate. To calculate the capacitance, as shown in Figure 4.2, first look at 
the infinitesimally small capacitance, which is expressed by equation 4.14, 
0
( , )
( , )
b
C x t dx
d x t


 = (4.14)
in which 

0
=8.85?10
-12
 farads/m, the permittivity of vacuum,
and 
0
( , ) ( , )d x t w x t d=+ (4.15)
For 
0
( , )w x t d<< ,
0 0
0 0 0
0
1 ( , )
( , ) 1
( , )
1
b b w x t
C x t
w x t
d d d
d



 = 


+
(4.16)
Then the capacitance between the two plates can be expressed as equation 4.17:
0
0 00 0
( , )
( ) ( , ) 1
l l
b w x t
C t C x t dx
d d

 
=  = 



 (4.17)
Equation 4.17 can be rewritten as:
0
0 0
1
( ) 1 ( , )
l
C t C w x t dx
d l

= 



(4.18)
in which C
0
 is the capacitance of the beam with no deformation: 
0
0
0
bl
C
d


= (4.19)
where d
0
 is the gap distance between the vibrating beam and the substrate.
The current pulse generation circuit charges up the capacitance with fixed charge 
quantantity Q, then the voltage between the two plates can be expressed as equation 4.19.
51
( )
( )
Q
V t
C t
= (4.20)
For 
0
),( dt xw << , and with 
00
/ CQV = ,
0
0 0
( )
1
1 ( , )
l
V
V t
w x t dx
d l
=





(4.21)
Normally, when the beam is excited to vibrate, many modes exist. When the 
uniformly distributed force is added to excite the beam vibration, the dominant mode is 
the fundamental mode. To simplify the analysis, only the first mode is considered in this 
case. The capacitance and voltage are expressed by equations 4.18 and 4.21.
As shown in equation 4.21, V(t) has the same frequency as w(x,t), so the small 
AC voltage signal has the same frequency as the resonant frequency of the beam. 
Therefore, by detecting the AC voltage frequency, the strain on the beam can be 
obtained.
4.5 Temperature Compensation
Temperature compensation is a significant issue for a strain sensor. The ambient 
temperature effect can be compensated by measuring the temperature with a strain-
insensitive resonant beam. The strain-insensitive beam can be a cantilevered beam with 
the same width as the strain-sensitive beam, but it is detached from one support so that its 
vibration is not affected by the strain, as shown in Figure 4.5. 
52
Figure 4.5 Schematic diagram of compensation beam structure
The signal out from the compensation cantilever beam can be compared to its 
calibration values so as to obtain the temperature effect. So the signal out from the fix-
end resonant beam can be adjusted.
In this chapter, a scheme of using resonant beam to detecting the strain on the 
beam has been demonstrated. The theoretical analysis is given along with the ANSYS 
simulation results. Also, a temperature compensation scheme is provided. The results 
show that detection of the AC voltage frequency can be used to detect strain.
53
Chapter 5
Measurement of Propagation Attenuation through Concrete
Embedded sensors are designed to be mounted onto steel rebars of a reinforced 
concrete structure, and the strain information can be sent out via a radio frequency 
transceiver module as described in Chapter 2. In this way, the reliance of wires for 
communicating between sensors and central units can be eliminated. Chapter 3 shows 
that the harvested power from bridge vibration and solar cells can provide sufficient 
power supply for the operations of the embedded sensor system. However, an important 
issue has to be considered: signal attenuation in concrete. The higher the attenuation is, 
the more power a transceiver module consumes in order to ensure that the received signal 
strength can be sustained in a detectable level. The investigation on the signal attenuation 
in concrete provides a reference for transceiver design and embedded sensors placement. 
In this chapter, a theoretical study is presented and validates experimental 
measurements. The measurements are taken at the 900 MHz radio frequency range which 
is an unlicensed Industrial, Scientific, Medical (ISM) radio band widely used in civil 
applications.
5.1 Electromagnetic Wave Propagation
5.1.1 Reflection and Transmission Coefficient
Consider a uniform plane wave impinging from medium 1 (

1
, ?
1
) to medium 2 
(

2
, ?
2
). 
, ? are the electric permittivity and magnetic permittivity of the materials. At the 
54
boundary of two different materials, part of the incident energy is reflected back into 
medium 1 and the rest of the energy is transmitted into medium 2. The normal of the 
boundary and the incident wave propagation vector form the plane of incidence, as shown 
in Figure 5.1 The angle of incidence is denoted as 
1
, and the angle of transmission is 
denoted as 
2
. 
2
 can be obtained from Snell?s law:
1 1 1
2 1
2 2
sin ( sin )
? 


? 


=                                             (5.1)
Air and construction material such as concrete, wood etc. are low conducting 
medium, and their relative magnetic permittivity are approximately equal to 1. 
1 2 0
???== (5.2)
?
0
 is the magnetic permittivity of free space and equal to 
7
4 10


? Henry/meter.
So equation 5.1 can be written as: 
1 1
2 1
2
sin ( sin )






= (5.3)
55
Figure 5.1 Electromagnetic wave propagates from medium1 to medium 2
Figure 5. 2 Schematic drawings of TE and TM waves
Figure 5.2 shows two types of electromagnetic waves with same propagation 
vector. E
i
, H
i
, 
i
 represent electric field, magnetic field and propagation vector, 
respectively. Figure 5.2(a) shows that the electric field vector is perpendicular to the 
plane of incidence and this type of wave is called Transverse Electric (TE) wave. Figure 
H
i
E
i


i
H
i
E
i


i
(a) (b)

1

1

2
Medium 1
Medium 2
1

 
56
5.2(b) shows that the electric field vector is parallel to the plane of incidence and this 
type of wave is called Transverse Magnetic (TM) wave.
When a handheld device is designed to communicate with embedded sensor, it is 
necessary to know how much energy of the signal is transmitted into concrete and how 
much can be received by sensors. So the reflection coefficient and transmission 
coefficient are considered.
The reflection coefficient of TE wave is [Rao, 1999]:
1 1 2 2
1 1 2 2
cos cos
cos cos
TE







 =
+
(5.4)
The transmission coefficient of TE wave is [Rao, 1999]:
1 1
1 1 2 2
2 cos
cos cos
TE


 



=
+
(5.5)
The reflection coefficient of TM wave is [Rao, 1999]:
1 2 2 1
1 21 2 1
cos cos
cos cos
TM




 


 =
+
(5.6)
The transmission coefficient of TM wave is [Rao, 1999]:
1 1
1 2 2 1
2 cos
cos cos
TM


 



=
+
(5.7)
When a normal wave impinges to the interface (
1
=0), the reflection and 
transmission coefficient can be written:
1 2
1 2







 =
+
(5.8)
57
1
1 2
2 

 



=
+
(5.9)
 =1 means all wave energy is transmitted from medium 1 to medium 2, and  =0 
means all wave energy is reflected at the interface. 
5.1.2 Penetration Depth
Consider a plane wave with electric field of the form: 
*
( 2 )
0
jk j ft
E E e

 +
= (5.10)
where 
*
k  is complex wave number in direction  , and   is position vector. 
*
k has real 
part 
r
k  and imaginary part 
i
k  (
*
r i
k k jk=  ). The plane wave can be written:
( ( ) 2 ) ( 2 )
0 0
r i i r
j k jk j ft k jk j ft
E E e E e e

  
 + +
== (5.11)
It shows that the magnitude of the wave is changed to
0
i
k
E e

. This attenuation is 
due to the wave propagation in dielectric material in direction  . Attenuation factor can 
be written as:
20log( )
i
k
e! = (5.12)
The magnitude of the wave can be written in terms of propagation depth 
pg
d :
0
i pg
k d
mag
E E e

= (5.13)
For dielectric material with low conductivity, k
i
 can be written in terms of 

r
? and 


r
? which are the real and imaginary parts of the relative dielectric constant:
0 0
( )
i r
r
k f

 ?





""=
"
(5.14)
58
The relative complex dielectric constant is defined as:
r r r
j

 

"""
=+  (5.15)
The real part 
r


"
is related to the energy stored in dielectric material and the imaginary 
part 
r


""
is related to the energy dissipated.  
Penetration depth is defined as the distance at which wave field strength falls to 
1/e (36.8%) of its initial value:  
0 0
1
r
p
i
r
d
k
f




? 

"
==
""
(5.16)
f is the frequency in Hz of incident wave. 
5.2 Dielectric Constant of Concrete
The complex dielectric constant of concrete is dependent on the frequency in low 
radio frequencies range, and this dependency is significantly reduced in the microwave 
frequencies. This is shown in Figure 5.3. In [Loulizi, 2001], concrete slabs with different 
mix constituents were tested to obtain the complex dielectric constant of concrete in the 
frequency range of 0.5 to 1.0GHz. Reference [Rhim et al., 1998] also studied the 
variation of dielectric constant of concrete with different moisture content in the 
frequency range of 0.1 to 20GHz. With the knowledge of dielectric constant, the 
transmission coefficient and penetration depth can be obtained from equations 5.4 to 5.7 
and 5.16.
Figure 5.3 shows that the complex dielectric constant of concrete is about 9+2j. 
Substitute 
r


"
=9,  
r


""
=2 and f =917.5MHz into equation 5.15, the penetration depth is: 
59
6 12 7
9
15.6
3.14 917.5 10 8.85 10 4 3.14 10 2
p
d cm

==
? ?? ? ????
.
This means that the power of a signal (917.5MHz) falls to 36.8% of its initial value when 
it penetrates a depth of 15.6cm in concrete.   
Figure 5.3 [Loulizi, 2001] Concrete dielectric constants vs. frequency
5.3 Experimental Set-up and Results
Figure 5.4 shows the experimental set-up. Two antennas are placed facing each 
other directly, and the samples are placed between them. Network analyzer is used to take 
the measurement of the two antennas. Antennas are connected to the ports of the network 
analyzer. The diffraction effects at sample edges have been minimized by increasing 
sample area and by making antennas axially perpendicular to sample plane. The S
12 
reading (in dB) of the network analyzer gives the ratio of the energy received by the 
receiving antenna and the energy sent by the transmitting one, S
12
=20log
10
(E
OUT
/E
IN
).
60
The Two helical antennas are designed with the working range of 0.8 GHz to 1.1 
GHz and one functions as the transmitting antenna and the other as the receiving one. 
Helical antenna is considered because of its high directivity when it is operating in the 
axial mode [Balanis, 1982] as shown in Figure 5.5. Total length L=30cm, diameter 
D=10cm, spacing s=7.5cm and total turns N=4. Helical antenna is formed by screwing a 
conducting wire. The radiation characteristics of antenna can be changed by controlling 
the dimensions. Axial mode is the most practical operation of helical antenna. In this 
mode, the majority of radiation intensity is along the helix axis. In order to excite this 
mode, the ratio 
D/# must be between 3/4 and 4/3. Usually the helical antenna is used 
with a ground plane whose diameter is at least #/2. For 917MHz wave, #/2 is equal to 
16.4cm. A 20cm?20cm copper plate is used in the experiment. 
The nominal impedance of the helical antenna in axial mode is in the range of 100 
to 200 Ohms. A thin metal strip is attached to the wire close to the feed point, and the 
antenna impedance is then adjusted to 50 Ohms by changing the size of metal strip in 
order to match the impedance of transmission lines.  
61
Figure 5.4 Schematic drawing of experiment setup
Figure 5.5 Schematic drawing of helical antenna
In order to compare the influence of concrete with other materials, the 
measurements were taken under four situations. (a) There is no blocking between the two 
S
A
M
P
L
E
S
L
D
c
o
p
p
e
r
p
l
a
t
e
62
antennas; (b) Wall board is inserted between the two antennas; (c) Concrete slab with 4.5 
inch thickness is inserted between the two antennas. (d) Metal plate is inserted between 
the two antennas. The output of the network analyzer gives the S
12 
reading (in dB) with 
respect to wave frequency, as shown in Figure 5.6. X-axis denotes the frequency that 
varies from 0.5 to 1.5 GHz. Since the antennas are designed to operate in 0.8 to 1.1 GHz, 
Figure 5.6(a) shows that the antenna has less noise in this range. 917.5MHz is marked as 
the tracking point and results of measurement taken at this point are presented in Table 
5.1.
63
(a)
(b)
(c) (d)
Figure 5.6 Transmitting of RF energy through different material. (a) Open space. (b) 
Wood wall board. (c) Concrete slab. (d) Metal plate. For all four plots, x-axis 
frequency vary from 0.5 to 1.5 GHz, with 50 MHz per grid. y-axis is the 
20log
10
(E
OUT
/E
IN
), with 10 dB per grid.
Table 5.1 Attenuation taken at 917.5MHz
Tested Sample Propagation Attenuation
(a) Open Space -14dB
(b) Wood Wall Board -14dB
(c) Concrete Slab -19dB
(d) Metal Plate -30dB
Results show that the wood wall board barely had any influence on the 
measurement. With the concrete slab, the reading dropped about 5 to 6 dB
64
(
21
( / 20)
/ 10 56.2% 50.1%
S
OUT IN
E E to== ) at 917.5 MHz. That means there is about 50% 
power lost. To get an idea about energy level of the scattering field, a metal plate was 
inserted between the two antennas to block the direct transmission between them. The
reading dropped about 30 dB ( / 3.2%
OUT IN
E E = ), which is 11 dB more than the loss due 
to the concrete slab.
A 5 to 6 dB power loss due to a concrete slab with 4.5 inch (11.4cm) thickness 
was found in the experiment. The experimental results also show that the signal 
attenuation due to wood wall board is small enough to be neglected. Equation 5.15 shows
that every time the signal (917.5MHz) penetrates a depth of 15.6cm in concrete, the 
power of the signal fall to 36.8%. So if the signal penetrates a depth of 4.5 inch (11.4cm), 
the power falls to 48.2% (
0.73
e

, from equation 5.13) of its initial value. This is very close 
to the experimental results (50.1% to 56.2%). 
65
Chapter 6
Summary and Conclusion
In this thesis, an embedded sensor system for structural health monitoring is 
described. The sensor system collects structural information by detecting embedded 
strain sensors. A power scavenging device is employed to harvest energy from highway 
bridge vibration for the power demand of the sensor system. The structural information is 
sent back to a handheld monitor wirelessly. This thesis concentrates on three topics:
6.1 power scavenging from highway bridge vibration
Scavenging energy from the ambient is an attractive means to power sensors in 
structures. This eliminates the dependence of batteries and allows the sensors to be 
tetherless. Power scavenging from traffic-induced vibration of highway bridge vibration 
is investigated and PZT (lead zirconate titanate) material is studied as the major 
piezoelectric element for power generation.  Three mechanisms (Patch, cantilever beam 
and stacks) are described that are suitable for sensors to scavenge mechanical energy at 
different locations. The PZT patches can also be used to sense the strain changes in 
bridge girders. The harvested power is relatively low due to the low strain changes. The 
generated power level is about 0.5?W/cm
2
. Cantilever beam device is design to sense 
bridge vibration and generates high strain. The power generation from this mechanism is 
about 20?W/cm
2
 and its dimension can be optimized in order to maximize the generated 
power in different vibration environments. PZT stacks and stack arrays have the biggest 
potential (1mW/cm
2
) of the three schemes investigated. By applying the stress 
66
amplification scheme, a multilayer stack generates around 1mW power under traffic-
induced vertical compressive stress at the top of girders. A stack array scheme is 
designed to be placed in pavement and it makes use of the wheel impact force on 
pavement. All calculations are based on the vibration data of a single span girder bridge 
using the finite element code ANSYS. The dynamic performance is studied by simulating 
trucks moving across the bridge. The results show that the power generation density of 
0.0004mW/cm
2
 (for patches) to 1mW/cm
2
 (for stacks) can be achieved. Solar power 
harvested in solar cells is 3 to 10mW/cm
2
 in sunny day, and 0.01 to 0.1mW/ cm
2
 in 
cloudy day. Solar cells have the limitation in nighttime and low light conditions. 
However the vibration-induced power generation is independent of climate and weather. 
So combing vibration-induced power generator and solar cells together is an effective 
way to provide power supply for the operation an embedded sensor system, including 
sensing, data processing and signal transmission. 
6.2 Embedded MEMS beam-type strain sensor
A scheme of using resonant beam to detect the axial strain of the MEMS beam is 
demonstrated. The resonant frequency of the vibrating beam is determined by its axial 
stress. This allows for detection of under one micro strain (10
-6
 m/m) change in a 100 
micron long beam both for compression and tension. The strain on the beam can be 
sampled remotely by observing the resonant frequency of beam while compensation for 
such factors as temperature change can be performed. The theoretical analysis is given
along with the ANSYS simulation results and the capacitance excitement and detection 
67
technologies are provided. The results show that detection of the AC voltage frequency 
provides beam strain information.
6.3 Propagation Attenuation through Concrete at 900 MHz
The propagation of radio frequency signal through concrete at 900 MHz range is 
discussed. Measurements are taken to determine the radio frequency energy attenuation 
while transmitting through a 4.5 inch concrete slab, a wood wallboard and a metal sheet. 
The results show that the wood wallboard has very little effect on signal propagation. A 5 
to 6 dB loss due to the concrete slab with 4.5 inch thickness is found.  The experimental 
results have a good agreement to the theoretical calculations. The theoretical study and 
experimental measurement provide a reference for determining embedded depth of 
sensors in concrete and designing the transceiver module of an embedded sensor system.
68
References
Ace Instrument Co., Ltd., http://my.ecplaza.net/aceins/3.asp
Amirtharajah R and Chandrakasan A.P., 1998, ?Self-Powered Signal Processing Using Vibration-
Based Power Generation,? IEEE Journal of Solid-State Circuits, Vol. 33, No. 5, pp. 687-
695, May 
ANSYS, Inc., 1998, Tenth Edition ANSYS Manuals Release 5.5, September
Arms S.W. and Townsend C. P., 2003, ?Wireless Strain Measurement Systems 
Applications & Solutions,? Presented at NSF-ESF Joint Conference on 
Structural Health Monitoring, Strasbourg, France, October
Balanis C. A., 1982, Antenna Theory Analysis and Design, pp. 386, Harper & Row Publishers
Baltes H. P., 1993, ?CMOS as sensor technology,? Sensors and Actuaters, A37-38, 
pp.51-56, June-August
Casta?er L., Pons J., Nadal-Guardia R. and Rodr?guez A., 2001, ?Analysis of the 
extended operation range of electrostatic actuators by current-pulse drive,? 
Sensors and Actuators A 90, pp.181-190
El-hami M., Glynne-Jones P., White N. M., Hill M., Beeby S., James E., Brown A. D., Ross J. 
N., 2001, ?Design and fabrication of a new vibration-based electromechanical power 
generator,? Sensors and Actuators Vol. 92, pp.335-342
Ernst T., Hartmann J.-M., Loup V., Ducroquet F., Dollfus P., Guegan G., Lafond D., Holliger P., 
Toffoli A., and Deleonibus S., 2002, ?Fabrication of a Novel Strained SiGe: C-Channel 
Planar 55nm nMOSFET for High-Perfomance CMOS,? 2002 Symposium on VLSI 
Technol., Hilton Hawaiian Village, Honolulu, HI, June 11-15
Eyre B. and Miller L., 1998, ?MEMS Magnetic Sensor in Standard CMOS,? Science 
Closure and Enabling Technologies for Constellation Class Missions, edited by V. 
Angelopoulos and P. V. Panetta, pp. 99-102, UC Berkeley, Calif.
Ferman M. A. 1999, ?Silicon-Based Heterostructures: Strained-Layer Growth by Molecular 
Beam Epitaxy,? Cryst. Res. Technol. Vol. 34, pp. 583-595
Friedman D., Heinreich H., and Duan D.W., 1997, ?A low-power CMOS integrated circuit for 
field-powered radio frequency identification tags,? ISSCC Dig. Tech. Papers, Feb. 1997, 
pp. 294-295
69
Gardner J. W., Varadan V. K., Osama A. O., 2001, Microsensors MEMS and Smart Devices. 
John Wiley & Sons Ltd, Chichester, England
Glosch H., Ashauer M., Pfeiffer U., Lang W., 1999, ?A thermoelectric converter for energy 
supply,? Sensors and Actuators Vol. 74, pp.246-250
Glynne-Jones P., Beeby S.P. and White N.M., 2001, ?Towards a piezoelectric vibration-powered 
microgenerator,? IEE Proc. ?Sci. Meas. Technol., Vol. 148, No. 2, pp.68-72, March 
Honcharenko W., Bertoni H. L., Dailing J. L., Qian J., and Yee H. D., 1992, ?Mechanisms 
Governing UHF Propagation on Single Floors in Modern Office Buildings,? IEEE 
Transaction on Vehicular Technology, Vol. 41, No. 4, November 
Honcharenko W., Bertoni H. L., and Dailing J., 1993, ?Mechanism Governing Propagation 
Between Different Floors in Buildings,? IEEE Transaction on Antennas and Propagation, 
Vol. 41, No. 6, June 
Iowa Thin Film Technologies, Inc., http://www.iowathinfilm.com/
Jung K., Bredow J. W., Joshi S. P., 1999, ?Electromagnetically coupled embedded sensors,? 
Proceedings of SPIE Vol. 3674, Industrial and Commercial Applications of Smart 
Structures Technologies
Kendall C. J., 1998, ?Parasitic Power Collection in Shoe Mounted Devices,? Thesis of Bachelor 
Degree, Massachusetts Institute of Technology, June 
Kov?cs ?d?m and V?zv?ry Zsolt, 2001, ?Structural parameter sensitivity analysis of cantilever-
and bridge-type accelerometers,? Sensors and Actuators A 89 (2001) 197-205
Lafortune J. F. and Lecours M., 1990, ?Measurement and Modeling of Propagation Losses in a 
Building at 900 MHz,? IEEE Transaction on Vehicular Technology, Vol. 39, No. 2, May 
Lawton M. C. and McGeehan J. P., 1994, ?The Application of a Deterministic Ray Launching 
Algorithm for the Prediction of Radio Channel Characteristics in Small-Cell 
Environments,? IEEE Transaction on Vehicular Technology, Vol. 43, No. 4, November 
Lee J. B., Chen Z., Allen M. G., Rohatgi A. and Arya R., 1995, ?A Miniaturized High-Voltage 
Solar Cell Array as an Electrostatic MEMS Power Supply,? IEEE Journal of MEMES, 
Vol. 4, No. 3, September 
Loulizi Amara, 2001, ?Development of Ground Penetrating Radar Signal Modeling and 
Implementation for Transportation Infrastructure,? Ph.D. Dissertation, Virginia 
Polytechnic Institute and State University, February 
Lynch J. P., Law K. H., Straser E. G., Kiremidjian A. S. and Kenny T. W., 2000, ?The 
Development of a Wireless Modular Health Monitoring System for Civil Structures,?  
70
Proceedings of the MCEER Mitigation of Earthquake Disaster by Advanced 
Technologies (MEDAT-2) Workshop, Las Vegas, NV, USA, November
MathWorks, Inc., http://www.mathworks.com
Maxim Integrated Products, http://www.maxim-ic.com/quick_view2.cfm/qv_pk/1330
Meninger S.,Mur-Miranda J.O., Amirtharajah R, Chandrakasan A.P., Lang J.H., 2001, 
"Vibration-to-Electric Energy Conversion," ISPLED99, pp. 48-53
MicroStrain, Inc., http://www.microstrain.com/SG-Link.htm
Morgan Electro Ceramics, http://www.morganelectroceramics.com/
Morgan Electro Ceramics, 2000, ?Piezoelectricity High Voltage Generators,? Technical 
Publication TP-219
Paradiso J. A. and Feldmeier M, 2001, ?A Compact, Wireless, Self-Powered Pushbutton 
Controller,? UBICOMP Conference Proceedings, Atlanta GA, Sept. 2001, Springer-
Verlag Berlin Heidelberg, 
Pfeifer K. B., Leming S. K., and Rumpf A. N., 2001, ?Embedded Self-Powered MicroSensors for 
Monitoring the Surety of Critical Buildings and Infrastructures,? Sandia National 
Laboratories, SANDIA REPORT, SAND2001-3619, Unlimited Release, Printed 
November 
Piezo Systems, Inc., Piezo Generator/Sensor Kit Manual, 186 Massachusetts Ave. Cambridge, 
MA 02139 USA
Rao Nannapaneni N., 1999, ?Elements of Engineering Electromagnetics,? Ftfth Edition, 
PRENTICE HALL, Upper Saddle River, NJ 07458
Rhim H. C. and B?y?k?zt?rk R., 1998, ?Electromagnetic Properties of Concrete at Microwave 
Frequency Range,? ACI Material Journal, pp. 262-271, May-June
Seidel S. Y., and Rappaport T. S., 1994, ?Site-Specific Propagation Prediction for Wireless In-
Building Personal Communication System Design,? IEEE Transaction on Vehicular 
Technology, Vol. 43, No. 4, November 
Sensor Technology Ltd., http://www.sensortech.ca/igniter.html
Shames I. H., and Dym C. L., 1991, Energy and Finite Element Methods in Structural Mechanics. 
Taylor & Francis, Levittown, PA 
Shearwood C. and Yates R. B., 1997, ?Development of an Electromagnetic Microgenerator,? 
Electronics Letters, Vol. 33 No. 2, October
71
Shenck N. S. and Paradiso J. A., 2001, ?Energy Scavenging With Shoe-Mounted Piezoelectrics,? 
IEEE 0272-1732/01 
Sniegowski J.J., Guckel H. and Christenson T. R., 1999, ?Performance Characteristic of Second 
Generation Polysilicon Resonanting Beam Force Transducers,? Solid-State Sensor and 
Actuator Workshop, 4th Technical Digest, Hilton Head Island, SC, IEEE, pp. 9-12
Starner T., 1996, ?Human powered wearable computing,? IBM Syst. J., Vol. 35, pp. 618-629
Straser E. G., Kiremidjian A S., Law K. H. and Blume J. A., 2000, ?A Framework for 
Health Monitoring of Structures,? Structures Congress and Exposition, ASCE, 
Philadephia, PA, May 8-10
Straser E. G., Kiremidjian A. S., Meng T. H., and Redlefsen L., 1998, ?Modular, wireless 
network platform for monitoring structures,? Proceedings of the International Modal 
Analysis Conference ? IMAC, pp. 450-456
The MOSIS Service, http://www.mosis.org/
Vittoz E. A., 1994, ?Low-power design: Ways to approach the limits,? ISSCC 1994 Dig. 
Tech. Papers, Feb., pp.14-18
Warneke B. A., Scott M. D., Leibowitz B. S., Zhou L., Bellew C. L., Chediak J. A., Kahn J. M., 
Boser B. E., Pister K. S. J., 2002, ?An Autonomous 16 mm
3
 Solar-Powered Node for 
Distributed Wireless Sensor Networks,? IEEE 0-7803-7454-1/02  
Weinberg M. S., 1999, ?Working Equations for Piezoelectric Actuators and Sensors,? IEEE 
Journal of MEMS, Vol. 8, No. 4
Williams C.B., Pavic A., Crouch R.S., and Woods R.C., 1997, ?Feasibility Study of Vibration-
Electric Generator for Bridge Vibration Sensors,? IMAC-XVI Proceeding
Wu J., Quinn V., Bernstein G. H., 2001, ?A simple, wireless powering scheme for MEMS 
devices,? Proceedings of SPIE Vol. 4559 (2001), MEMS Components and Applications 
for Industry, Automobiles, Aerospace, and Communication
Zhang H., Hayashida T., Takashi Yoshino, Shiro Ito and Yoji Nagasawa, 1999, ?A Deterministic 
Model for UFH Radio Wave Propagation through Building Windows in Cellular 
Environments,? IEICE Transaction on Communication, Vol. E82-B, No. 6 June 
72