Use of Piezoelectric Materials for Strain Measurements and Wave Propagation Analysis Alexandra Woldman – Undergraduate Researcher Dr. Haichang Gu – Postdoctoral Mentor Dr. Gangbing Song – Faculty Mentor University of Houston, Houston, TX
Use of Piezoelectric Materials for Strain
Measurements and Wave Propagation Analysis
Alexandra Woldman – Undergraduate Researcher
Dr. Haichang Gu – Postdoctoral Mentor
Dr. Gangbing Song – Faculty Mentor
University of Houston, Houston, TX
1
Table of Contents:
Abstract 2
Introduction 2
Experimental Setup 4
Results 7
Conclusions 12
Future Work 14
Acknowledgements 14
References 15
Appendices 16
2
Abstract
This paper covers several uses for piezoelectric material. Piezoelectric materials deform when a
voltage is applied to them and inversely will produce a voltage when they are deformed. For this
reason, they can be used as both sensors and actuators. Piezoelectric sensors should theoretically
be able to give an accurate measurement of the stress and strain in an object if the output voltage
is measured accurately and the relationship of output voltage to strain is known. This paper
explains an attempt at calibrating a charge amplifier so the voltage output can be used to
determine the strain in the material. The experiment succeeded on an aluminum beam but failed
to work similarly for sensors embedded in concrete. A different type of piezosensor is necessary
for implementation in concrete. This paper also discusses the use of piezoceramic sensors and
actuators to test the propagation of waves in concrete under static loading. Two concrete
cylinders with embedded piezoceramics were tested in compression. One cylinder was tested
until failure. The voltage at the sensor was converted to energy to show the change in the amount
of energy that propagates to the sensor. There is a dramatic drop in energy during the loading of
the first 10% of the load. The change in energy at higher loads is small. When the concrete
cylinder is broken, the energy begins to decrease as small fractures form in the cylinder. There is
a tremendous energy drop-off once the cylinder breaks.
Introduction
Concrete is a popular and widely used material in structural engineering. Throughout the life of a
structure, concrete often becomes damaged from repeated loading. Currently, the most
widespread concrete health monitoring tests are destructive tests. One method of testing concrete
is by casting cylinders of the same concrete as used in a structure and periodically destructively
3
testing these samples to monitor the strength of the concrete. Since the cylinders and structural
members are made from the same mix and cast at the same time, they should have the same
properties over time. However, casting cylinders does not accurately portray to state of the
concrete, since structural members are of a different shape and therefore cure differently than a
cylinder with the same concrete mix (Limaye, 2002). Moreover, the structural members are
experiencing stresses that the cylinders are not, which may have a large effect on the strength of
the concrete. A more accurate method of destructive testing is coring. This process removes a
cylinder of concrete from the actual structure (Limaye, 2002). While more accurate, this method
directly damages a structure. Piezoelectric sensors are the newest method for monitoring the
health of concrete structures throughout their lifetime. Piezoelectric materials produce a voltage
when they are stressed and also deform when a voltage is applied to them. These materials are
made by a process called poling. During poling, an electric field is applied to the material,
aligning the random electric dipoles of the atoms. Once the material is removed from the electric
field, the dipoles retain some their alignment. When a voltage is applied to the material, the poles
realign, causing a shift in the shape of the material. Likewise, when the shape is changed, a
voltage is produced because of the induced realignment of the atoms (Piezo, 2008). The
piezoceramic used in this specific study is lead zirconate titanate (PZT), a fairly common and
inexpensive piezoelectric material. Patches of PZT are much more durable than strain gauges and
can be cast into concrete and used to nondestructively monitor the health of a structure.
Moreover, piezoelectric sensors allow for continuous structural monitoring, rather than
monitoring at one arbitrary point in time. Piezoelectric sensors can detect damage before it is
significant enough to detect visually. This allows maintenance on structures before they become
dangerous and must be closed for repairs.
4
1. Previous studies have been done on impact detection through the use of PZT sensors and
actuators (Song, Gu, Mo, 2008). However, these studies do not give any specific
numerical data about the stress and strain in the structure. Numerical data regarding the
stress and strain within structural members will indicate whether any one member is
under an unexpected amount of stress, signaling a problem with the distribution of the
load in the structure. This can prevent the catastrophic failure of a structure. This project
aims to calibrate a charge amplifier to get accurate measurements for the stress and strain
in concrete under static loading.
2. Wave propagation is a nondestructive method for testing concrete. This method uses the
velocity of the waves through the concrete to determine its strength. It can also detect the
presence of cracks and voids before they are visible. Most research using wave
propagation has focused on isolated structural members (Gassman & Tawhed, 2004).
This experiment analyzes the change in wave propagation under static loading. The static
loading conditions mimic the load conditions the member would experience when in is
embedded in a structure. The results should indicate whether loading of a structural
member may lead to an inaccurate judgment of the health of the member or possibly be
confused with a damaged sensor (Overly, 2007).
Experimental Setup
Two different experimental setups existed in this experiment. One was for the calibration of the
charge amplifier while the other aimed to check wave propagation in concrete. In future
experiments, the results of the two should be combined for accurate numerical results. Both
experiments use the piezoceramic material PSI-5A4E. This material can be used as both a sensor
5
and an actuator. The first experiment focuses on sensors, while the second utilizes both
capabilities of the material.
1. The first experimental setup aimed to
calibrate the charge amplifier so that it
would generate a quantitative measure of the
stress and strain in concrete. This test
features a thin aluminum beam (5cm x 61
cm x .08 cm), with a PZT patch mounted on
one side 11 cm from the end, and a strain
gauge mounted on the other side 11 cm from the end. This placement assured that the
strain on the two sides was equal but opposite in sign. Then the beam was fixed by a
clamp at one end. The strain gauge was connected to Vishay Signal Conditioning
Amplifier System (Model A2). The piezoceramic was connected to the Kistler Charge
Amplifier (Type 5073). The output voltage from both devices was fed into a computer,
saved and graphed. This was done to get develop a conversion from the voltage read by
the charge amplifier to the voltage read by the strain
gauge. The signal conditioner attached to the strain gauge
was set so that 1 mV was equivalent to 1 microstain.
After this initial setup, the charge amplifier was
connected to various samples of piezoceramics embedded
in concrete and the voltage readings were taken manually
with a multimeter. Many different samples were tested (Figure 2).
Figure 2: Several different samples
of concrete with embedded
piezoceramic sensors.
Figure 1: Cantilever beam used to calibrate the charge
amplifier. The piezoceramic patch and strain gauge were
mounted at the same spot on opposite sides of the beam.
6
Figure 4: The same PZT patches were used as sensors
and actuators in the cylinders, on top and on bottom.
Figure 3: Hydraulic load on
cylinder on testing surface.
2. The wave propagation tests were set up as shown in Figure 3. A concrete cylinder of
height 12” and diameter 6” was placed on the testing surface. The tests were performed
on two different cylinders cast at the same time. Each cylinder had two piezoceramic
patches embedded inside of it, each at the center of the
circular cross section 1” from the bottom and top of the
cylinder (Figure 4). The hydraulic load was lowered onto
the cylinder and increased as the test proceeded. The first
measurement for the test was always taken at zero load. At
least 10 measurement, and sometimes several more, were
taken as the load on the cylinder increased from 0 to just
over 40 kips. Each data set recorded the voltage at the
sensor from the signal sent by the actuator. The signal from
the actuator was the same of every data point of every test.
3 tests were performed for each cylinder in each configuration drawn in Figure 4.
Piezoceramic patches were used
for the sensor and the actuator.
When the tests with the actuator on
top and the sensor on bottom were
complete, the cylinder was turned
upside down so the actuator would
be on the bottom and the sensor on
7
the top. This was done so the piezoceramic patch that acted as the actuator and sensor
remained the same, for consistent results. After all 6 tests were completed on the first
cylinder, it was loaded to failure. This test was done with the PZT on actuator on top.
Results
1. Charge Amplifier Tests:
The first step in the charge amplifier calibration was
to work with a vibrating aluminum beam with a
strain gauge on one side and a piezoceramic sensor
on the other side. The strain on both sides is of the
same magnitude, but of different signs. In order to
simplify visual comparison, the data from the
charge amplifier was immediately multiplied by -1.
Data was collected while the beam was manually
moved over a 30 second time period, allowing both
devices to record voltage measurements
simultaneously. Figure 5(a) shows the voltage
measured by the two devices before any transformation had been applied for a given test.
Figure 5(b) shows the two data sets after the charge amplifier data is transformed to
match the strain gauge data. The linear transformation used for this data is:
Vstain gauge = 1.577341 * Vcharge amp + 0.132023. (1)
Figure 5: (a) The voltage from the charge
amplifier and strain gauge before
transformations are applied and (b) after
transformation.
8
Since the charge amplifier can never be completely reset to zero, the first value (when the
aluminum beam was under zero strain) was afterwards subtracted from every data point,
making the complete transformation, which was used to make Figure 5(b)
Vstain gauge = 1.577341 * Vcharge amp + 0.132023-V0,charge amp . (2)
Once the charge amplifier voltage is converted into the equivalent strain gauge data, the
known value of 1 mVstrain gauge = 1µε is used to determine the strain at any point on the
graph. A slight drift was observed in the charge amplifier data, as seen by the difference
in endpoints of Figure 5(b). Matlab was used to load, transform and graph data. The code
used can be found in Appendix 1.
The next stage of the experiment was to implement the charge amplifier transformation
data to determine the strain in concrete. Several concrete samples with embedded
piezoceramics (a variety of which are shown in Figure 2) were used to test out the
transformation determined in Equation 2. Unfortunately, there was a dramatic drift in the
voltage every time an attempt was made to gather data. The drift in the data was so
dramatic, that no accurate data collection was possible. Using an oscilloscope as a visual
aide, the charge amplifier did give an obvious indication of when an impact occurred.
However, this is no improvement over the work done by Song, Gu and Mo, which did not
use a charge amplifier, but measured voltage directly (2008). The application of a static
load was too small to be registered by the charge amplifier as compared to its own drift.
Several different insulation methods were used to try to minimize the drift. These
methods include a water insulation coating and spray on electrical tape insulation coating.
The drift with these insulated samples was less dramatic than the drift noticed in samples
with no insulation. However, the insulation was not enough stop the drift from occurring.
9
2. Wave Propagation:
The data collected for wave
propagation in a concrete cylinder
under static loading gave the voltage
at the sensor from the signal output
by the actuator. Figure 6 shows a
typically graph for one test. 40 000
voltage measurements were taken
over the span of 10 second at each
given load, approximately 4 – 5 kips
apart. Theoretically, the voltage measurements should be centered at zero. However, due
to interference, all of the measurements had an offset. The data was detrended before any
calculations were done (see Appendix 2 for applicable Matlab code). Each voltage vector
was then converted into the energy at the sensor with the equation
Energy = [x] * [x] T . (3)
Figure 7 and 8 show the energy at different loads for the 6 tests done on each cylinder.
The plots indicate that the energy at the sensor decreases drastically as soon as the load
changes from zero. However, for loads larger than approximately 10 kip, the energy level
does not change much with increased load. For both cylinders, the tests with the sensor
on top lead to slightly higher energy levels than the analogous tests with the actuator on
top. Noting the scale on the two figures, one can see that the energy at the sensor depends
dramatically on the cylinder itself. Each cylinder has its own trend. The energy levels at
the sensors for destructive test of the first cylinder are shown in Figure 9. After the 45
Figure 6: A typical data sample for the wave propagation test
at a specific load. The plot consists of 40 000 individual data
points of the voltage at the sensor at some time during the 10
second sampling time.
10
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
x 104
50
60
70
80
90
100
110
120
130
Load Applied (lbs)
Energy
Act-Top 1
Act-Top 2
Act-Top 3
Sen-Top 1
Sen-Top 2
Sen-Top 3
Figure 7: The energy for different load for the first test cylinder. The blue data points are the samples with the sensor on top
while the warm colors represent the tests with the actuator on top.
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
x 104
10
15
20
25
30
35
40
45
50
55
60
Load Applied (lbs)
Energy
Act-Top 1
Act-Top 2
Act-Top 3
Sen-Top 1
Sen-Top 2
Sen-Top 3
Figure 8: The energy for different load for the second test cylinder. The blue data points are the samples with the sensor on top
while the warm colors represent the tests with the actuator on top.
11
Figure 9: The energy at different loads for the destructive test of the first cylinder. The energy dropped as soon as the load
changed from zero, and stayed relatively steady as more load was applied. The load began to decrease at 90 kips until failure.
kip mark used as the cutoff for all the previous tests, the energy at the sensor increases
slightly until it reaches a maximum at approximately 90 kips. The energy then decreases
until failure. Failure occurred at 129 kips. The primary crack formed in between the
sensor and the actuator (the images from this test can be found in Appendix 3). When the
cylinder broke, the hydraulic jack immediately stopped adding to the load. Since the
material had rearranged, the load was no longer 129 kips. A reading was taken before the
hydraulic jack was moved up, removing all load from the cylinder. The load read 6 kips
when the final reading was made (labeled as "Broken” in Figure 9). The energy at the
sensor from the signal sent by the actuator is substantially smaller after the cylinder broke.
12
Conclusions
1. Although the calibration of the charge amplifier only succeeded for the aluminum, the
experiment showed a need for better sensors to be used in conjunction with the charge
amplifier. Even in the aluminum experiment, where the PZT was mounted on the surface
of the metal, there was still some drift. The drift is most likely the effect of low insulation
resistance, which is a common characteristic of ceramic piezosensors (Fialkowski, 2008).
With a better sensor, the charge amplifier voltage output can be used to give stress and
strain measurements for concrete used in a structure. This knowledge would allow much
more accurate quantitative monitoring of structures for fatigue and wear. Instead of
monitoring purely for impact, this monitoring method would allow for analysis of gradual
changes and shifting of weight in a structure over time.
2. The propagation of waves through concrete changes dramatically as soon as any load is
applied to the cylinder. This is due to the change in boundary condition. When the load is
zero, the boundary condition at the top can essentially be modeled as a free end. Once
any load is applied, the top of the cylinder can no longer move freely. This suppresses the
vibrations in the concrete, leading to a significant drop in energy that propagates through
the concrete. As compared to the ultimate load on the concrete cylinders (129 kips), the
dramatic drop in energy continues until a load of approximately 10% of the ultimate load
is reached. This could be considered the settling range for the samples. Taking wave
propagation measurements in concrete that is loaded to 0 – 10% of its ultimate load,
especially completely unloaded, could cause misleading results. Once the sample is
loaded to a significant portion of its ultimate load (>10%), the energy at the sensor
remains fairly constant. The use of wave propagation testing should therefore not be
13
affected by static loading as long as the initial measurements are taken after the load has
been applied and the load is a significant percentage of the ultimate load.
3. When the load on the cylinder comes close to the ultimate load, the energy at the sensors
declines. In the experiment documented in Figure 9, the decline begins at 2/3 of the
ultimate load. This is due to the formation of small fractures in the concrete. Small
fractures cause a loss of energy, since some of the signal is reflected by the cracks. Figure
10 is a close up image of the signal during the destructive test for two different loads. The
first is at a load of 40 kips, while the second is at 126 kips, the last sample point before
failure occurred. The 40 kip image has clearly defined boundary while the 126 kip image
has ample noise. This is caused by the fractures in the concrete, which disrupt the signal.
Figure 10: The close-up views of the voltage sensed at the sensor are from load 40 kips and 126 kips respectively. The 126 kips
signal has visible noise caused by small fractures in the cylinder.
4. The energy at the sensor for any given test has some dependency on the placement of the
sensor in the sample. When the sensor is closer to the hydraulic load, the energy levels
are higher than in the case where the sensor is at the bottom. This is caused by a
difference in boundary conditions at the loading end and the table surface. When
14
piezocermaics are embedded in concrete, placement should be considered carefully as it
could affect the results.
Future Work
Structural monitoring is an extremely useful field. Structures can be monitored for damage
effectively maintained without waiting for visible signs of damage first. Piezosensors can be a
very effective way of monitoring structures. In future work, sensors should be picked to work
well with the charge amplifier to create an accurate measuring system for the stresses in concrete.
This research showed that wave propagation energy decreases significantly during the loading of
the first 10% of the ultimate load. To better understand the reason for the decrease in the energy
that propagates through the concrete, Scanning Electron Microscope (SEM) imaging of the
piezoceramic in its embedded state could be useful. The SEM images, taken before and after
testing, will show if any microscopic changes occur that influence the behavior of the
piezoceramic patches.
Acknowledgements
The research study described herein was sponsored by the National Science Foundation under
the Award No. EEC-0649163. The opinions expressed in this study are those of the authors and
do not necessarily reflect the views of the sponsor. This research was conducted under the
supervision of Dr. Gangbing Song of the University of Houston. Dr. Haichang Gu and Claudio
Olmi were of great help in understanding the material and conducting the research in this paper.
15
References
Fialkowski, L. (2008). (Discussion of Kistler charge amplifier ed., pp. 1). Houston.
Gassman, S. L., & Tawhed, W. F. (2004). Nondestructive Assessment of Damage in Concrete
Bridge Decks. Journal of Performance of Constructed Facilities, 18(4).
Limaye, B. R. (2002). Need for Non-Destructive Testing (NDT) of Reinforced Concrete &
Various ND Tests. Madras.
Overly, T. G. S. (2007). Development and Integration of Hardware and Software for Active-
sensors in Structural Health Monitoring. University of Cincinnati, Cincinnati.
Piezo Systems Inc. (2008). Frequently Asked Questions Retrieved 1 Aug, 2008.
Song, G., Gu, H., & Mo, y.-L. (2008). Smart aggregates: multi-functional sensors for concrete
structures - a tutorial and a review. Smart materials & structures, 17(3), 17.
16
Appendices
Appendix – 1
load calibration;
% 'calibration' is data gathered through DataDesk
% it is a structure
% X is the time
% Y(1) is the charge amp data
% Y(2) is the strain gauge data
time = calibration.X.Data;
Z = calibration.Y(1).Data;
S = calibration.Y(2).Data;
% Plot two original sets of data
subplot(2,1,1)
plot(time,Z,'-r',time,S,'-b')
xlabel('Time (s)')
ylabel('Voltage (V)')
legend('Strain Gauge','Charge Amp (Original)')
title('(a)','Fontsize',16)
% Transform charge amp data
Z = 1.577341*Z+.132023;
% Get rid off starting offset
Z = Z-(Z(1)-S(1));
subplot(2,1,2)
plot(time,Z,'-r',time,S,'-b')
legend('Strain Gauge','Charge Amp (Transformed)')
xlabel('Time (s)')
ylabel('Voltage (V)')
title('(b)','Fontsize',16)
% end
Appendix – 2
% Cylinder 2 – draw plot of energy at different loads
% SET UP SCREEN
scrsz = get(0,'ScreenSize');
figure('Position',[1 scrsz(4) scrsz(3) scrsz(4)])
axes('FontSize',16)
hold on
% axis([0 45000 10 130])
% ACTUATORS ON TOP, 3 TRIALS
lbs = [0 198 650 5400 10100 14400 18400 22900 27100 ...
31500 36100 40400];
17
E=[];
for i=1:12
x=['test2a',num2str(i)];
S=load(x);
y=detrend(S.(x).Y(2).Data);
E(i)=y*y';
end
scatter(lbs,E,'filled','CData',[1,0,0],'SizeData',100)
xlabel('Load Applied (lbs)','Fontsize',20)
ylabel('Energy','Fontsize',20)
lbs = [0 260 2800 8500 13100 17700 21900 26900 31800 ...
37600];
E=[];
for i=13:22
x=['test2a',num2str(i)];
S=load(x);
y=detrend(S.(x).Y(2).Data);
E(i-12)=y*y';
end
scatter(lbs,E,'filled','CData',[1,.3,0],'SizeData',100)
lbs = [0 120 1300 7100 11400 16300 20900 25800 30000 ...
34500 38700 0];
E=[];
for i=23:34
x=['test2a',num2str(i)];
S=load(x);
y=detrend(S.(x).Y(2).Data);
E(i-22)=y*y';
end
scatter(lbs,E,'filled','CData',[1,.6,0],'SizeData',100)
% SENSORS ON TOP, 3 TRIALS
lbs = [0 10 54 366 4800 9700 14200 18700 23100 27500 ...
31900 36200 40200];
E=[];
for i=1:13
x=['test2s',num2str(i)];
S=load(x);
y=detrend(S.(x).Y(2).Data);
E(i)=y*y';
end
scatter(lbs,E,'filled','CData',[0 0 1],'SizeData',100)
18
lbs = [0 36 435 5200 10600 17600 21900 26700 30900 ...
35100 39400];
E=[];
for i=14:24
x=['test2s',num2str(i)];
S=load(x);
y=detrend(S.(x).Y(2).Data);
E(i-13)=y*y';
end
scatter(lbs,E,'filled','CData',[0 .4 1],'SizeData',100)
lbs = [0 0 85 770 6300 10300 15600 20000 24600 29200 ...
33700 37000 0];
E=[];
for i=25:37
x=['test2s',num2str(i)];
S=load(x);
y=detrend(S.(x).Y(2).Data);
E(i-24)=y*y';
end
scatter(lbs,E,'filled','CData',[0 .8 1],'SizeData',100)
% LABEL
legend('Act-Top 1','Act-Top 2','Act-Top 3','Sen-Top 1','Sen-Top 2',...
'Sen-Top 3','Fontsize',18)
Appendix – 3