31st Conference of the

European Working Group on Acoustic Emission (EWGAE) – We.1.A.4

Finite Element Modeling of Acoustic

Emission Signal Propagation with Various
Shaped Waveguides
Andreea Manuela ZELENYAK *, Marvin A. HAMSTAD **, Markus G. R. SAUSE *
* University of Augsburg, Institute for Physics, Augsburg, Germany
** University of Denver, Daniel Felix Ritchie School of Engineering and Computer
Science, Denver, CO, USA

Abstract. In acoustic emission (AE) technology, when the examination does not
allow mounting a sensor directly because of the extreme conditions such as high
temperature, chemically environment, insulation, radiation, small or irregular
specimens, waveguides are used. Waveguides can have the form of long thin rods or
wires usually made of brass, steel or aluminum alloys. The purpose of this research
was to investigate the effects of varying waveguide shapes on acoustic emission
signal characteristics by means of finite element modeling (FEM). Different designs
of an aluminum waveguide are numerically investigated by modeling the acoustic
emission signals propagating in a 3 mm thick aluminum plate with the waveguide
attached. The signal was generated using a pencil lead break source located on the
top of the plate, acting in out-of-plane direction. The modelled signals were
evaluated at a distance of 100 mm from the source on the plate surface and on the
waveguide. The waveguide was modelled with circular cross-section and an
attached segment for sensor mounting. The latter was varied from a simple
rectangular cross-section to a conical one. Solid and hollow waveguides filled with
air or fluid were investigated. To validate the simulated acoustic emission signal
characteristics, experimental signals measured on the plate surface are compared to
simulated ones. Waveforms obtained in the simulation process of all waveguides are
analyzed. Comparison between each FEM result is presented in terms of the
detected signal on the waveguide, time delay between plate surface signal and
waveguide signal, mode conversion during waveguide transmission, attenuation in
the waveguide and shifts of the signals frequency content.

1. Introduction

Innovative non-destructive testing methods such as acoustic emissions (AE) analysis are
usually used to identify the stage of material degradation and load history [1,2]. Acoustic
emission signals are defined as microscopic displacements of a solid, originating e.g. from
crack growth, at the lower ultrasonic frequency range of 20 kHz to 500 kHz [3]. Detection
of the AE-signals involved various investigations in the past leading to the most common
type of detection systems used in ultrasonic measurements, piezoelectric materials based
sensors [4]. Behavior of the piezoelectric sensors detecting in extreme environmental
conditions such as high temperature or radiation is well studied and solutions such as using
natural crystals instead of ceramic were found. These, and careful selection of the materials

1
License: http://creativecommons.org/licenses/by/3.0/
dedicated to cable design and those used for overprotective -braid covering can lead to high
purchasing prices for the AE sensors. Piezoceramic lead-zirconate-titanate (PZT) is one
common material used for AE sensors with the big inconvenience that it cannot be directly
mounted on a structure which operates at high temperatures [5]. A solution to avoid sensor
destruction was found by the researchers like Lynnworth et al. and involves usage of thin
rods, clad rods or hollow tubes as ultrasonic waveguides (WGs) for detection [6]. Pencil
lead break AE signals detected experimentally by a WG mounted sensor were compared
with the signals obtained from a plate mounted sensor in terms of signal duplication in the
time, frequency and time/frequency domains with good results. Various short WGs with
different material properties, length, diameter, face angle were experimentally investigated
in order to show the effects upon pulsed events [7,8]. All these experimental investigations
involved the usage of more than one WG type leading to high costs and long times to
develop the right WG for one measurement purpose. With the increase of the computation
power in the past years numerical methods can be used to successfully simulate elastic
wave propagation in various materials. In AE testing of thin structures the Lamb waves
modes are the most commonly encountered type of guided wave. Often only the
fundamental symmetric mode S0 and anti-symmetric mode A0 are present in AE testing [9].
Realistic simulations of guided waves, using finite element method were presented by
different researches using software packages like Ansys, Comsol Multiphysics or Abaqus.
The purpose of the current work is to investigate the signal propagation in different
shaped thin rods as ultrasonic WGs by finite element simulation, analyzing the influence
upon the detected signal’s characteristics.

Modelling of acoustic emission in waveguides

The finite element method (FEM) was used to fulfill this investigation carrying out
simulations in Comsol Multiphysics with Structural Mechanics module to solve the
constitutive equations for linear-elastic media and piezoelectric media and the AC/DC
module for performing a P-SPICE circuit simulation. Simulation of acoustic waves, in
particular fundamental anti-symmetric A0 and symmetric S0 Lamb modes using numerical
methods was intensively investigated in recent years and results revealed signals which
were perfectly reproducing experimental ones [9].

1.1 Model setup

To investigate the effects of WG presence on the acoustic emission characteristics, two full
3-dimensional models were built. The first one involved modeling the propagation of the
acoustic emission waves in a simple plate without being influenced by the WG. The signal
obtained with this model was kept as reference signal and is called plate conical in the
following (pic. 1a). The geometry of the model considers a plate of 600 mm x 600 mm size
and 3 mm thickness. As seen in pic. 1a, symmetry boundary conditions along the xz-plane
were used to reduce the model dimensions by a factor of two. A monopole source PLB was
chosen to be acting in out-of-plane direction at (x,y,z)=(0,0,3) mm, located at the middle of
the plate. At the detection point at (x,y,z)=(100,0,3) mm a conical sensor was modeled to
avoid interference with incident reflections within the duration time of the simulation. The
sensor model was previously studied numerically by Sause et al. and all parameters used
are the same [9] (pic.1c).

2
(a)

(b) (c)

Pic.1 Model geometry for the (a) plate with conical sensor (b) plate with waveguide and conical
sensor [9] and (c) details of the conical sensor model

The second 3-dimensional model plate had the same geometry as the reference one,
but at the detection point, between the plate and the sensor, differently shaped WGs were
introduced (pic. 1b). A total number of seven types of WGs were investigated. The WGs
length was kept constant at 306.4 mm with a small diameter of 1.59 mm following the
experimental used geometry by Hamstad [7]. In the following only the upper and the lower
parts of the WGs were changed in shape.
Most of the WGs were solid rods with shape adjustments at the two ends (cf. pic
2a).
(a) (b)

306.4
mm

Pic.2 Waveguide geometries of different cases (a) and detailed view of ends (b).

A 2.5 mm sub-section at either end of the total 306.4 mm length was replaced with
various shapes for the study. From a cylindrical shape, the sub-section was changed to
conical with a large bottom radius of 3.25 mm and a semi-angle of 45° generating two
study cases. Another variation considers a conical sub-section with a very small bottom
3
radius 0.8 mm a semi-angle of 17° creating two more cases, so called sharpened WGs. In
addition, two special cases were built using hollow WGs, with the inner diameter of 0.79
mm being half of the total WG’s diameter, filled with air or water. The different geometry
variations are illustrated in pic. 2a, with a detailed view given in pic. 2b.
The material for the propagation domain and for the WGs was AlMg3 for all cases
with properties shown in table 1, with the elastic properties taken from [10].
Table1. Elastic properties of the materials used in FEM cases.
Property AlMg3
Density[kg/m3] 2660
Elastic modulus [GPa] 70
Poisson ratio 0.33

The expected waves to develop in the plate and further transmitted into the
waveguide are Lamb wave modes of fundamental order due to the small thickness of the
plate.
Due to the fact that the WGs were mounted perpendicular to the plate, the flat end
of the WG is expected to be excited only by the out-of-plane displacement as pointed out
by Hamstad in [7]. The differences can be given by the viscous couplant which is usually
used in experiments and which in this investigation was not modeled.
Convergence of the numerical solution was previously investigated through a series
of models developed specifically to create an efficient mesh and temporal resolution for
modeling pencil lead break waves in large structures. Numerical studies conducted so far in
this field by other authors showed that having a mesh resolution of 1 mm maximum
element size and signals calculated with 1×10-7 s time step will lead to a convergent
solution. Our researches concerning the convergence of the solution revealed that it is
sufficient to have 1mm mesh resolution along a line of interest using a distribution of the
mesh size for the rest of the plate within the interval [1 to 3] mm. Mesh elements are of
cubic geometric shape order. A time step of 1×10-7 s was used to calculate a total length of
100 µs of signal propagation, extended to 150 µs in WGs cases. For an accurate model of
the sensor a mesh resolution in the interval of [0.4 to 0.5] mm size was used with a
coherence level ≥0.99801 as demonstrated by [9].

2. Comparison between Simulation and Experiment

A validation of the simulation approach was done by comparing the modeled plate conical
signal to an experimental signal obtained by Hamstad in his research upon small diameter
waveguide usage in AE [7]. The sensor used is a wideband conical one, produced at the
National Institute of Standards and Technology (NIST) Colorado, which was previously
described in two publications [13,14]. The conical element has 1.5 mm aperture and 2.5
mm height, with the piezoelectric material being PZT 5A. The conical element, the backing
mass, a small part of the case and the preamplifier were modeled. This was found to be
sufficient to obtain a signal similar to the experimental one. The experimental setup also
contains a plate made from an aluminum alloy with 3.1 mm thickness, as propagation
medium and a PLB source with 0.3 mm diameter and 2H hardness to generate the AE
signals. For the modeling work we used the material properties of table 1. A cosine bell
function with the excitation time of 1×10-6 s and a maximum force of 3 N simulates the
pencil lead break behavior as acoustic emission source acting in the out-of-plane direction
on the plate surface. The geometry of this model was similar with the one presented in pic.
1a, the differences being given by presence of the sensor model at the detection point at 254
mm distance and the total plate width of 550 mm x 500 mm with 3.1 mm thickness.
4
 (a) (b)
Modeled signal
250
Experimental signal taken from [7] S0
200 Simulation vs. Experimental A0
150

100

Frequency [kHz]
Voltage [mV]

50

-50

-100

-150
Thickness 3.1 mm
-200 Propagation distance 254 mm
-250
30 40 50 60 70 80 90 100
Time [µs] Time [µs]

Pic. 3 Comparison between simulated and experimental signal (a) and wavelet transformation of the
simulated signal with superimposed dispersion curves (b)

The experimental signal was directly compared with the simulated voltage signal. The
dominating modes due to the PLB out-of-plane source applied to the surface of the plate are
the A0 mode with a weak contribution of the S0 mode in the beginning of the signal [15].
The shape of the experimental signal, as shown in pic. 3, is well reflected by the simulated
signal. There are minor differences in arrival time due to the fact that the exact properties of
the experimentally used aluminum alloy are unknown. Small discrepancies in the
magnitude can also be observed. The latter can be attributed to the attached circuitry which
models cables and preamplifier components, as demonstrated by Sause et al. in [9]. For this
simulation we choose 5 Ω resistance of the connection cables and 15 pF capacitance and for
the preamplifier 10 MΩ and 20 pF capacitance. Variation of the cable capacitance values
can lead to a better match in terms of magnitude. In pic. 3b the Wavelet transformation
(WT) of the simulated signal using the AGU Vallen software is compared to the theoretical
A0 and S0 dispersion curves calculated for aluminum for a propagation distance of 254 mm.
The WT also reveals good agreement between the dispersion curves and the simulated
signal.

3. Results and discussions

3.1 Post processing

In the following, simulation results of the cases are compared to each other and are
discussed in terms of observations from literature.
In order to discuss comparable signals spectra all simulated signals undergo a series of
post processing steps:
 The WG signal was shifted forward in time to superimpose (e.g. for the rectangular
cross section waveguide by 60.2 μs)
 All signals from the plate and WG conical sensors were terminated at a convenient zero
 All signals were extended with 0 values to a total length of 2048 points
 The Fast Fourier Transformation (FFT) were calculated with a square window function
 The resulting FFTs were smoothed by a 30 point Savitsky-Golay (5th polynomial)
method
 The FFT results were adjusted to superimpose from 0 to 100 kHz in order to fit and
compare

5
 Waveforms of simulated signals after 100 µs of propagation distance are shown in
pic. 4. The reference signal shows the arrivals of S0 and A0 modes after 20.5 µs and 32.4
µs, respectively.
(a) (b)
30 Reference signal
Waveguide 1.59 mm diameter
S0
20 A0
Amplitude [mV]

10

Frequency [kHz]
0

-10

-20
60.2 µs offset to superimpose
Propagation distance 100 mm
-30
0 20 40 60 80 100
Time [µs]
Time [µs]

60.2 µs
(c) (d)
-44
-55
Reference signal S0
-55
-66
A0
-66
-77
FFT magnitude [dB]

-77
Frequency [kHz]

-88
-88
-99
-99
-110
-110
-121
-121
Waveguide 1.59 mm diameter -132
-132
-143

0.0 0.2 0.4 0.6 0.8 1.0

Time [µs]
Frequency [MHz]

Pic. 4 Waveform signals from reference signals and WG with circular cross section (a) WT of reference
signal (b) frequency spectra of both signals (plate conical left vertical axis) (c) and WT of waveguide signal
(d).

The time domain signal of the reference case (detected directly on the plate) and
detected on the waveguide is presented in pic. 4a. The time offset deduced from the
simulation results is 60.2 µs and is in good agreement with the experimental results which
revealed 60.4 µs time delay between the signal detected on the plate and the signal detected
by the conical sensor on top of the WG [7]. The WT of the two signals are presented in
pics. 4b and 4d, respectively. Both signals exhibit a clear presence of A0 and S0 modes. The
frequency content of the signals is compared after the post processing steps mentioned
above were applied on both signals. The frequency curves presented in pic. 4c shows that
the signals are qualitatively similar also in the frequency domain after a reduction in
amplitude of the reference signal by with 7.4 dB to fit the WG signal

3.2 Waveguide shape influence

In order to see the influence of the WG shape on the frequency content of the detected
signals we compare the Fast Fourier transformation results. The frequency spectra resulted
curves are shown in pic. 5a. with the same magnitude scale. A first comparison is made in
terms of magnitude offset between the reference plate signal and the signals detected by the
WG mounted sensor.

6
 (a) Reference
(b)
Reference signal Circular
Conical element (up)
-60
15
FFT magnitude (Plate conical) [dB]

Conical element (down)

Hallow with water

FFT Magnitude offset [dB]

Hallow with air
-80 Sharp element (down)
Sharp element (up) 10

-100
5

-120
0

0.0 0.2 0.4 0.6 0.8 1.0 Circular Con_up Con_down Water Air Sh_down Sh_up
Frequency [MHz] WG shape

Pic.5 Frequency spectra a) plate conical reference case and 1.59 mm Al waveguide circular cross section b)
magnitude vs. shape

The magnitude offset of 7.4 dB corresponds to the circular cross section WG signal
and it is the lowest magnitude difference obtained amongst all geometrical implementations
investigated (cf. pic.5b). Examination of the signals detected by the different geometries
shows clear differences between the magnitudes of the WG signals in comparison with the
reference signal. The WGs with sharp tips revealed to have the highest impact on the
transmitted amplitudes, independent if the sharp sub-section was place on the lower end or
the upper end. This could be probably due to small aperture of the WG. This effect was
already pointed out by Sause et al. in the work concerning sensor modeling highlighting the
tendency to increase the amplitude of the detected signal when the sensor diameter was
increased [9]. The amplitude of the WG with sharp edge at the upper end was down with
additional 9.6 dB compared to the circular cross section WG and with 17 dB compared to
the reference signal (cf. pic.5b). A high amplitude decrease manifests also in the WG with
conical attachments in the [0.2 to 0.4] MHz frequency range. Overall the FFTs of the other
WGs signals are qualitatively similar and there were no mode conversions noticed in all
cases investigated.
The special cases in which hollow WGs were used did not show any noticeable
deviations to the duplicated signal, when compared to the reference signal. The air filled
WG showed a decrease in magnitude compared with the circular WG, 2.6 dB offset
between them, while the signal had the same arrival time and same qualitative shape. When
the air was replaced with water the magnitude offset was smaller, 1.6 dB magnitude offset,
but a delay of 3.4 µs in arrival time was noticed.

4. Conclusion

Through this numerical investigation it has been demonstrated that finite element modeling
can be used to successfully simulate signal propagation in WGs. The multiphysics coupling
provided by the Comsol environment was used in this approach to model piezoelectric
conversion and the attached electric circuit for sensor implementation. The validation of the
models was carried out using a comparison to the experimental AE signal obtained by
Hamstad [7].
Simulating different designs of the aluminum waveguides has shown that the WG
geometry can have a significant impact on transmission of the Lamb waves propagating in
the plate investigated in terms of arrival times or magnitude shifts in the frequency
spectrum. It has been shown through time domain signals and also frequency spectra that

7
WGs are capable to duplicate the out-of-plane displacement of the plate which can be
detected at the WG end by an attached wideband sensor. Such duplication using WGs was
already demonstrated experimentally but with the help of such modeling work the influence
of new geometries and WG materials can suitably be predicted.
Further numerical investigations concerning WGs usage in acoustic emission
detection can be extended to studies involving environmental changes in terms of
temperature, signal attenuation studying different evaluation distances or other parameters
influencing WG transmission like diameter, length or impedance mismatches.

References

[1] E H Saenger, G K Kocur, R Jud, M Torrilhon, ‘Application of time reverse modeling on ultrasonic non-
destructive testing of concrete’, Applied Mathematical Modelling, Vol 35, pp. 807–816, 2011.
[2] C U Grosse, M Ohtsu, ‘Acoustic Emission Testing: Basics for Research – Applications in Civil
Engineering’; With Contributions by Numerous Experts, Springer, Heidelberg, ISBN 978-3-540-69895-1,
2008.
[3] M G R Sause, ‘Identification of failure mechanisms in hybrid materials utilizing pattern recognition
techniques applied to acoustic emission signals’, PhD Dissertation, Augsburg (Germany), Augsburg
University, 2010.
[4] E S Boltz, C M Fortunko, M A Hamstad, M C Renken, ‘Absolute sensitivity of air, light and direct-
coupled wideband acoustic emission transducers’, Review of Progress in Quantitative Nondestructive
Evaluation, vol. 14, Plenum Press, New York, pp. 967–974, 1995.
[5] Ian T. Neilla, I. J. Oppenheima, D W Greveba , ‘A Wire-Guided Transducer for Acoustic Emission
Sensing’, Proceedings of the SPIE, Volume 6529, article id. 652913, 2007.
[6] L Lynnworth, Yi Liu, and J. Umina, ‘Extensional Bundle Waveguide Techniques for Measuring Flow of
Hot Fluids’, IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, Vol. 42, No. 4, 538-
544, 2005.
[7] M A Hamstad, ‘Small diameter waveguide for wideband acoustic emission’, Journal of Acoustic Emission
Vol. 24, pp. 234-247, 2006.
[8] J Sikorska and J Pan, ‘The effect of waveguide material and shape on acoustic emission transmission
characteristics Part 1: traditional features’, Journal of Acoustic Emission Vol. 22, pp. 264-273, 2004.
[9] M G R Sause, M. A. Hamstad, S. Horn, ‘Finite element modeling of conical acoustic emission sensors and
corresponding experiments’, Sensors and Actuators A 184, pp. 64– 71, 2012.
[10] M G R Sause , S. Horn, ‘Influence of internal discontinuities on ultrasonic signal propagation in carbon
fiber reinforced plastics’,30th European Conference on Acoustic Emission Testing & 7th International
Conference on Acoustic Emission University of Granada, 12-15 September 2012.
[11] M G R Sause, ‘Investigation of pencil-lead breaks as acoustic emission sources’, Journal of Acoustic
Emission 29, pp. 184-196, 2011.
[12] Y Gharaibeh1, S Soua, G. Edwards, P. Mudge, W Balachandran, ‘Modelling guided waves in complex
structures - Part 2: Wire bundles - with and without insulation’, paper presented at NDT 2009, Blackpool,
UK, 15-17 Sept. 2009.
[13] M A Hamstad, C.M. Fortunko, ‘Development of Practical Wideband High Fidelity Acoustic Emission
Sensors’, Nondestructive Evaluation of Aging Bridges and Highways, Proc. SPIE 2456, pp. 281-288, 1995.
[14] M A Hamstad, ‘Improved Signal-to-Noise Wideband Acoustic/Ultrasonic Contact Displacement Sensors
for Wood and Polymers’, Wood and Fiber Science, 29 (3), pp. 239-248, 1997.
[15] M A Hamstad, ‘On lamb modes as a function of acoustic emission source rise time, Journal of Acoustic
Emission’ 28, pp. 41–58, 2010.