IXth NDT in PROGRESS

October 9–11, 2017, Prague, Czech Republic

EXPERIMENTAL AND NUMERICAL ANALYSIS OF

A TRANSDUCER FOR THE GENERATION OF GUIDED
WAVES

Marco ZENNARO 1, Alex HAIG 2, Dan J. O’BOY 3, Stephen J. WALSH4

1
National Structural Integrity Research Center,
Granta Park, Cambridge, CB21 6AL, United Kingdom,
Phone: +44 (0)1223940396, E-mail: [email protected]
2
The Welding Institute (TWI),
Granta Park, Cambridge, CB21 6AL, United Kingdom; E-mail: [email protected]
3,4
School of Aeronautical and Automotive, Chemical and Materials Engineering (AACME), Loughborough
University, Leicestershire LE11 3TU; E-mail: [email protected], [email protected]

Abstract
Dry-coupled thickness-shear transducers represent one of the most common ways to excite guided waves for the
inspection of tubular and plate-like structures. Although already established in industry, some features of these
transducers need to be studied, i.e. the uniformity of vibration, the modes excitability and the transmission of
ultrasonic energy into the inspected structure. In particular, due to the dispersive behaviour of guided waves and
the mode coupling these transducers require a precise characterization to guarantee a uniform in-plane vibration.
A numerical and experimental characterization of the assembly has been carried out to assess the influence of the
elements of the assembly into the uniformity of vibration. The outcome of the results in terms of mode-shape,
displacement pattern and resonance frequencies is discussed to predict useful design changes to enhance the
ultrasonic performance of these transducers.
Keywords: Piezoelectric Transducers, Guided Waves, Laser Vibrometer, Finite Element Analysis, Modal Analysis

1. Introduction
The implementation of ultrasonic guided wave systems in the fields of non-destructive testing and
structural health monitoring has been present in the industry for many decades. After various
pioneering work in these fields [1-2], Alleyne and Cawley introduced dry-coupled piezoelectric
transducers to excite guided waves for inspection in tubular structures [3], where the main
advantages included the practical ability to mount them on pipelines and the capability to excite
longitudinal modes. Liu et al. used this system to excite torsional modes in pipes [4].
The dry coupling requires the use of a mounting force to guarantee the effective transmission
of the ultrasonic signal into the inspected structure. The influence of this force on the
variation of the mechanical resonances has been investigated by Engineer [5]. Through
Hertzian contact theory Engineer proved experimentally that when a force of 200 N is applied
the first thickness shear resonance of the transducer shifts from 1.2 MHz to 21 kHz.
Marques focused his research on the characterization of an array of thickness-shear
transducers for the generation of shear horizontal waves in plate-like structures. His main
objectives were the directionality of this wave mode and the suppression of unwanted modes
(Lamb waves) [6]. He proved how the capability of his array in terms of amplitude was
limited by the size of the transducers. Thus, he stated the importance of an investigation into
the possibility of miniaturizing the transducers.

- 133 -
IXth NDT in PROGRESS

The miniaturization of the transducers requires the understanding of their physical behaviour,
in order to choose which geometrical parameters can be modified to improve the ultrasonic
performance. This paper provides an insight into the identification of the mode-shapes,
displacement patterns and resonances of all the components of the assembly.
Section 2 contains a background on piezoelectric transducers, while section 3 contains the
numerical analysis, section 4 the experimental results and a discussion, section 5 the conclusions.

2. Piezoelectric transducers
2.1 Current transducer
The transducer under consideration is composed of an assembly of three elements, a thickness
shear piezoelectric device, an alumina wear-plate and a backing mass of steel. PI Ceramics
provides the first two components glued together (figure 1a).
Thickness shear transducers are defined as specimens where the application of an electric
field normal to the axis of polarization causes a shear strain into the material [7, 8]. The class
of the piezoelectric element is PZT 5 A. The geometrical dimensions are 13 x 3 x 0.5 mm.
Longitude, width and thickness throughout the study are identified as the x, y and z axes.
The necessity to protect the mechanical failure of the piezo-element requires the presence of
the alumina wear-plate in the actual applications: thus, in case of excitation of guided waves
the dry-coupling happens to be between the wear-plate and the waveguide.
The backing mass of steel (figure 1b) is added to increase the flexural stiffness of the material.
This motivation is different from conventional ultrasonics, where the backing mass is used to
damp out vibrations [3]. The electrical contact to the transducers is obtained by a positive
electrode on the upper surface and a negative electrode in the down surface wrapped around to the
upper surface. A space of 1 mm is used to separate the two electrodes on the upper face.

Positive electrode Gap Negative electrode

Figure 1: Picture of the piezoelectric element attached to the alumina wear-plate (Courtesy of TWI)

- 134 -
 IXth NDT in PROGRESS

3. Numerical modelling
3.1 Methodology
The finite element analysis was obtained through the commercial software COMSOL, which
contains a library for the piezoelectric analysis. In a previous study the finite element analysis
with COMSOL was used to calculate the natural frequencies of the piezoelectric strip, with
and without an alumina layer [9].
A further analysis has now been carried out by applying an electrical voltage as a boundary
condition to calculate the mechanical displacement and the mode shape as an output of the
model.
The study was conducted in the frequency domain: through this study it is possible to
compute the effect of a model under a harmonic excitation. This excitation has been computed
as a uniform load performed through a frequency sweep in the range 15-150 kHz. The
frequency domain study was chosen for two reasons: first of all, it is computationally less
expensive that a typical time-domain analysis; secondly it is readily seen how the eigenmodes
interact with the applied voltage.
In this first analysis the response of a thickness-shear piezoelectric transducer to an applied
voltage was computed. A voltage was applied on the positive electrode, to simulate the
behaviour of the current transducer: the applied voltage was 15 V for all the frequencies in the
sweep analysis, and the bottom surface and the wrapped around electrode were grounded. In
terms of the mechanical boundary conditions, a free-free system was implemented.
Due to the geometry relatively simple of the system a free tetrahedral method was chosen: the
total number of elements used was 15167.
As a second numerical step, the frequency response function of a piezoelectric strip attached
to an alumina layer was analysed. The methodology followed was similar to the case
presented in the previous section. However, in this case the number of elements of the mesh
rises to 30514.
As far as the contact between the two surfaces is concerned, it is assumed that the materials
are perfectly glued and the two materials are sharing the nodes of the mesh.

3.2.2 Numerical results

The numerical work indicates that in the range of applications no thickness-shear resonance
can be found. Although this result can be expected from the theory, it is still important to
quantify how the imposition of a voltage interacts with the eigenmodes of the structures. In
particular potential undesired components of motion must be identified. In addition, when a
preload is applied to the transducer, the very high frequencies will be reduced into the range
of interest of this study.
The natural frequency analysis performed in [9] indicated the presence of flexural, torsional,
shear horizontal and longitudinal modes: the experimental analysis proved that not all the
modes are actually excited, due to the uniformity of the load along the longitudinal axis.
These results are confirmed by the computational analysis here presented. Even though the
mode shapes are constrained by the geometry of the piezoelectric element, the voltage drives
the element to move along one axis. As an example the mode shape for 25 kHz is shown in
figure: it can be clearly seen that there is a superimposition of thickness-shear motion on a
flexural mode shape; moreover the presence of the wrapped around electrode changes

- 135 -
IXth NDT in PROGRESS

dramatically the motion of the upper surface. Other results will be commented later in the
sections of comparison with experimental results.

Figure 2: Picture of the numerical mode shape of piezoelement (left) and piezoelement with alumina (right) at
25 kHz. Colours indicate amplitude of displacements.

3. Experimental analysis
3.1 Methodology
The identification of natural frequencies and mode shapes obtained through numerical
analysis requires experimental validation, especially because the computation is obtained with
elements with free-free mechanical boundary conditions. One of the most widespread
techniques for validation of numerical analysis in guided wave propagation and piezoelectric
transducers is Laser Vibrometry [10]. As a matter of fact the necessity of not imposing
constraints on the transducer drives the need of a non-contact measuring technique. A Polytec
3D Scanning Laser Vibrometer PSV-400-3D-M was used in this experiment [11]. This Laser
Vibrometer is based on the emission of red laser from three heads to a specified moving
surface: the backscattered energy is then compared to a reference signal to detect the velocity
of the measured surface.
The three heads are used for the three components of velocity (see figure 3). The internal
software offers the capability of defining a grid of points on the measured specimen to be
scanned during the excitation of the element. The measurement is then repeated for every
single point.
The form of excitation used was a chirp signal between 20-150 kHz. The chirp signal was
used for two reasons: first of all it permits the frequency sweep in a specific range, secondly it
is a signal often used in guided wave excitation. Thus, it can be used later for further
comparisons when the waveguide will be introduced. As a function generator a Teletest Focus
system was used: the voltage was 250 Vpp, spanning for 500 ms.
To increase the reflective power of the backscattered signal a bright sheet was used: although
the application of the bright sheet considerably augmented the reflection, as a downsize some
of the geometrical features were not clearly distinguishable. Therefore, the surface could not
be completely measured. Moreover, as shown in figure 1, the ends of the transducers could
not be measured due to the presence of the soldered cables.

- 136 -
 IXth NDT in PROGRESS

Laser Vibrometer Piezoelectric specimen

Figure 3: Picture of the 3D Laser Vibrometer (Courtesy of TWI)

3.2 Piezoelectric element

It can be readily found from analytical calculations [8] and experimental measurements [5] of
the impedance that the first thickness-shear resonance of this analysed piezoelectric strip is
around 1.7 MHz. Thus, one would expect that for a strip driven by a voltage in a range well
below this threshold, the shape of vibration would be determined by the geometry. Flexural,
torsional and extensional modes would then be expected.

Figure 4: Average experimental spectrum of the piezo-element. Note that the resonance are labelled with letters

As an example of mode shape the magnitude of the displacement for resonance B and C are
shown in figure 5. For resonances A, B and C a z component is higher than the other two
components. Thus it is expected that those two components are actually flexural resonances.
The first resonance, A, at 20 kHz, is a flexural resonance of one complete cycle, although
there is not a complete symmetry between the two cycles: as a matter of fact there is a larger
area with an larger area with a downward deflection: the resonance C, at 64 kHz, is also a
flexural type of resonance, with two complete cycles.
Resonance B, although not showing this peak of magnitude on the left, is showing how the
left corner presents a different shape than the right corner. This area on the left corner was
identified as the area surrounding the gap between the positive and negative electrodes. Also,

- 137 -
IXth NDT in PROGRESS

resonance C on the left side presents a slight difference in the mode shape, i.e. a curvature of
the end.
It is suggested that the discontinuity of electrodes is compromising the uniformity of
excitation along the length, causing the rise of unwanted modes. Thus, almost one quarter of
the piezoelectric element is actually not moving uniformly with the positive electrode.
Therefore, this electrodes topology and its influence on the mode shape needs to be carefully
investigated. In particular, a comparison with a topology with parallel electrodes or with
different spacing of the negative electrodes needs to be investigated.
The investigation of the spacing of the electrodes would be useful for the generation of guided
waves for two reasons: first of all the elimination of the current shape should help to generate
a uniform guided wave signal. As a matter of fact an unwanted mode shape would lead to the
excitation of not uniform wave-fronts, which results in a decrease of resolution and capability
of detection of defects in plate-like and tubular structures. In addition, shortening the
electrodes would be beneficial in terms of the miniaturization of the transducers.

Figure 5 (a-b): Plot of the experimental mode shape for resonance B (a) and C (b)

3.3 Piezoelectric element with alumina layer

The addition of an alumina layer is expected to increase the natural frequencies, due to the
increase of stiffness in the material. Moreover, adding a layer of material to the piezoelectric
strip is likely to contribute to a decrease a displacement,

Figure 6: Plot of the average spectrum of the piezoelement attached to the alumina layer.

- 138 -
 IXth NDT in PROGRESS

The piezoelectric-alumina layer was measured with the same technique used for the
piezoelectric layer. The average spectrum is plotted in figure 6. In this case some important
differences can be noted: first of all, it can be seen that there is a more uniform response othe
displacement across the frequency. Secondly it should be noted that one of the resonances
(the longitudinal resonance D) is not anymore present in this case. Thirdly, it is interesting to
note that in both cases the longitudinal comparison (piezoelectric and piezoelectric plus
alumina) is in the order of 600-500 pms, except for the resonances peaks.
In figure 7 the mode shape E of material joined together is presented. The comparison with
the numerical results in figure 2 indicates a match in terms of shape, i.e. a flexural mode
shape of the first order. Also in this case as for the piezoelectric element, it can be noted how
in the left corner there is a remarkable difference in terms of displacement, due to the
presence of the gap between positive and negative electrodes.

Figure 7: Plot of the experimental mode shape E

4.4 Comparison of results

The results presented in the previous section require a comparison method to understand
physical trends and mismatches, where existing. In figure 8 the frequency response function
of the piezoelectric element is presented both numerically and experimentally. Some details
can be readily noted. First of all, the displacement pattern presents a similar response in the
two cases: the out of plane vibration decreases dramatically for frequencies higher than 90
kHz, to become almost negligible: after this threshold the higher component is the
longitudinal component. Secondly the response of the longitudinal component is flat across
the frequencies except for the resonance around 90 kHz. Therefore one of the objectives for
design changes would be to suppress the unwanted resonances in order to maintain only this
component of motion.
On the other hand, there is a difference in magnitude between numerical and experimental
results: one possible cause of explanation could be the difference of boundary conditions in
the two cases. As a matter of fact in the experiment the cables of the transducers are fixed and
there is also an interaction with the polystyrene foam: most likely these factors have
contributed to decrease the amplitude.

- 139 -
IXth NDT in PROGRESS

Figure 8: Comparison of experimental (up) and numerical results (down) for piezoelectric element

As far as the results with the alumina layer are concerned, it was expected that the presence of
this component could flatten the resonances to obtain a more uniform response across the
frequencies. In regards to the first resonance is concerned, it can be readily noted that it is
shifted, as it would be expected by the increase in stiffness. On the hand the second resonance
does not appear anymore. This fact can be explained by referring to the natural frequency
analysis: no natural frequency appears in the range 25-60 kHz. As far as the general trend is
concerned it can still be noted that the longitudinal component of motion is uniform.

Figure 9: Comparison of experimental (up) and numerical results (down) for piezoelectric element and alumina

5. Conclusions and further work

The frequency response function of a piezoelectric transducer with and without an alumina
layer has been obtained: numerical and experimental results have been described and
compared. First of all, it has been shown that there is a match of resonance frequencies in the
two cases; secondly the trend of displacement is quite similar. On the other hand, the
magnitude of displacement presents some inconsistencies: the reasons have been discussed in
the section of comparison, such as the importance of exactly replicating the boundary
condition in the experiments. Moreover, the presence of the gap of the electrodes introduces
modifications in the mode shapes that need to be investigated.
These results can now be used to define an objective function to indicate useful design
changes for the increase of the ultrasonic performance: parameters such as frequency response
across the frequency range, amplitude of displacement, ratio of components of displacement,

- 140 -
 IXth NDT in PROGRESS

geometric layout, represent core points to be analysed. However such an objective function
needs to take into account not only the response of the piezoelectric transducer, but also the
interaction with the waveguide.
Thus, a series of experiments and numerical analysis have to be carried out to test design changes
of the piezoelectric transducers and compare them with the current design. Amplitude of the
generated signal, sensitivity to defects and directionality of the modes will be assessed and
compared to provide an insight for industrial applications of the generation of guided waves.

Acknowledgements
The corresponding author would like to acknowledge the financial support of Lloyd’s
Registered Foundation, Loughborough University and TWI for this publication.

References
1. I A Viktorov, 'Rayleigh and Lamb waves: Physical theory and Applications', Plenum
Press, 1991
2. M G Silk and Bainton, 'The propagation in Metal Tubing of Ultrasonic Wave Modes
Equivalent to Lamb Waves', Ultrasonics 17:1, pp 11-19, 1979
3. D N Alleyne and Cawley P, 'The excitation of Lamb Waves in Pipes Using Dry-
Coupled Piezoelectric Transducers', J Nondestructiv Eval, 15:1, pp 11-20, 1996
4. Z Liu, C He, B Wu, X Wang, & S Yang, 'Circumferential and Longitudinal Defect
Detection Using T(0,1) Mode Excited by Thickness Shear-Mode Piezoelectric
Elements', Ultrasonics, 44, pp 1135-1138, 2006
5. B A Engineer, 'The Mechanical and Resonant Behaviour of a Dry-Coupled Thickness
Shear PZT Transducer used for Guided Wave Testing in Pipe Line' Diss. Brunel
University, 2013
6. H Rodrigues Marques, 'Omnidirectional and unidirectional SH0 mode transducer arrays
for guided wave evaluation of plate-like structures' Diss. Brunel University London, 2016.
7. W G Cady, 'Piezoelectricity', McGraw-Hill, 1946
8. N Aurelle, D Roche, C Richard & P Gonnard, 'Sample Aspect Ratio Influence on the
Shear Coefficients Measurements of a Piezoelectric Bar', Proceedings of the Ninth IEEE
International Symposium, pp 162-165, 1994
9. M Zennaro, A Haig, D J O’Boy, S J Walsh, 'Characterisation of the vibration of an
ultrasonic transducers for guided waves applications', Applied Physics, System Science
and Computers, Proceedings of the 2nd International Conference on Applied Physics,
System Science and Computers (APSAC2017), to be published
10. S E Olson, M P Desimio, Matthew, J D, E D Swenson & H Sohn, 'Computational Lamb
Wave Model Validation Using 1D and 3D Laser Vibrometer Measurement' Proc. SPIE
7650, pp. 76500M, 2010
11. H Weisbecker, H Cazzolato, S Wildy, S Marburg, J Codrington & A Kotousov, 'Surface
Strain Measurements Using a 3D Scanning Laser Vibrometer', Experimental
Mechanics, 52:7, pp 808-815, (2012)

- 141 -