Available online at www.sciencedirect.

com

Physics Procedia
Physics 3 (2010)
Procedia 863–871
00 (2009) 000–000
www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia

International Congress on Ultrasonics, Universidad de Santiago de Chile, January 2009

Localization Algorithm for Acoustic Emission

V. Salinasa,*, Y. Vargasa,#, J. Ruzzanteb, L. Gaetea.
a
Universidad de Santiago de Chile, Av. Ecuador 3493, Santiago, Chile
b
Comisión Nacional de Energía Atómica CNEA, Av. General Paz 1499, San Martín Buenos Aires, Argentina

Elsevier use only: Received date here; revised date here; accepted date here

Abstract

In this paper, an iterative algorithm for localization of acoustic emission (AE) source is presented. The main advantage of the
system is that it is independent of the “ability” in the determination of signal level to triggering the signal by the researcher. The
system was tested in cylindrical samples with an AE localized in a known position; the precision in the source determination was
of about 2 mm, better than the precision obtained with classic localization algorithms (~1cm).

PACS: 43.35.Zc; 43.40.Le

Keywords: acoustic emission, iterative methods, NDT

1. Introduction

Acoustic emissions (AE) are the stress waves produced by the sudden internal stress redistribution of the
materials caused by the changes in the internal structure [1]. Internal-structure changes could be crack initiation,
crack growth, crack opening, crack closure, dislocation movement, phase transformation in monolithic materials,
fiber breakage and fiber-matrix debonding in composites. Most of the sources of AEs are damage-related; thus, the
detection and monitoring of these emissions are commonly used to predict material failure [2]. Besides the
applications of AE in research endeavors, AE has been widely used in industries, including the detection of flaws or
leakage in pressure vessels, tanks, and piping systems. AE is also used to monitoring the welding and corrosion
progress.
The difference between the AE technique and other nondestructive evaluation (NDT) is that AE detects the
activities inside the materials, so this technique can be predict a failure, while other NDT methods detect the flaw
when the failure was created. Furthermore, AE only needs the signal from one or more relatively small sensors on

* Corresponding author. Tel.: +56-2-7181248.

E-mail address: vicente.salinas@usach.cl
# Corresponding author. Tel.: +56-2-7181270.
E-mail address: yolanda.vargas@usach.cl

doi:10.1016/j.phpro.2010.01.111
864 V. Salinas et al. / Physics Procedia 3 (2010) 863–871
Vicente Salinas / Physics Procedia 00 (2010) 000–000

the surface of the structure or specimen being examined. So the structure or specimen can be examinated in-service
or in laboratory operation while the AE system continuously monitoring the progressive damage [3].

The methods of locazation are constructed through iterative and non iterative algorithms based on signals from
the source and acquired by sensors located on the surface of the sample and the position of these sensors.

Based on the arrival-time differences, observed in an array of three sensors, an exact solution for the
configuration of three sensors in a planar surface was derived. The AE source is located at the intersection points of
two hyperbolae curves Tobias [4]. Using the differences in time-of-arrival (TOA) from an AE source to three
sensors [5], the position of the source is calculated through the use of spherical coordinates. M. Asty found that, the
positions of the source, can have more than 1 value and it is needed to use fourth sensor to eliminate this problem. In
a similar approach, Barat et al. [6] developed a mathematical method for the calculation of the coordinates of an AE
source in cylindrical surfaces using the concept of geodesic, to find the path of the wave between 2 sensors on the
surface. Dong-Jin Yoon [7], generalized mathematical formalism for the path traveled by an acoustic wave in the
surface of a structure is presented, and a direct method for the calculation of the coordinates of an AE source on a
cylindrical surface by three sensors is evaluated.
Based on the cross-correlation of two sensor signals the difference of the arrival time is estimated accurately by
calculating the phase shift of the signals in the cross-correlation. Two alternative but equivalent procedures for the
best estimation of the source location, corresponding to the ellipse’s centre for three or more microphone clusters are
presented by B. V. Landau [8]. A novel localization method of the source position generating impulsive periodic
signals is proposed in [9], requiring a minimum of three microphones to identify impulsive sources in 2-D space,
while in 3-D space at least four microphones must be used.
In this paper, a method to find the coordinate of an AE source in a non dispersive material using TOA of N events
in M sensors is presented. The number of sensor must be greater than six if location in 3D is required.
The method suppose propagation wave velocity constant in all direction, this velocity was determined by TOF
before experimentation.

2. Theorical Arguments

For the source localization, only geometrical arguments were utilized. If the position of the sensors are
determined by Xi(xi,yi,zi) and the source position is X(x,y,z), the distance from the source to the sensor “ i ” can be
obtained from

Di = (x i - x) 2 +(yi - y) 2 +(z i - z) 2 (1)

The arrival time from the position of the source to the sensor “i” is given by the equation
D i = Q (t i -t 1 ) (2)
Where ti is the arrival time to sensor “i” and t is the time of event occurrence.
The position of AE source and the wave propagation velocity can be obtained solving the system composed by
equation (1) and (2). In these formulae more unknown variables than equations appear. To solve this problem we do:
D i -D 1 = Q (t i -t 1 ) (3)

Where, D1 is Di evaluated for i=1, and i=2, 3,..., m.. These equations conduct to the following relations [10]
ei (a x+bi y+ci z) (4)
di + -2 i = - 2 D1
di di

Where
 V. Salinas et al. / Physics Procedia 3 (2010) 863–871 865
Vicente Salinas / Physics Procedia 00 (2010) 000–000

ai x1  xi
bi y1  yi
ci z1  zi
di Q (ti  t1 )
ei x 21  y 21  z 21  ( x 2i  y 2i  z 2i )

The non-linear system defined by (4) can be linearized by subtracting the equation (4) evaluated for i=2 from the
equation (4) [10].

fi,1 x  fi,2 y  fi,3 z hi  gi Q 2 (5)

Where

§a a ·
fi ,1 2¨ 2  i ¸
© t2 ti ¹
§b b ·
fi ,2 2¨ 2  i ¸
© t2 ti ¹
§c c ·
fi ,3 2¨ 2  i ¸
© t 2 ti ¹
gi (t2  ti )
e2 ei
hi 
t 2 ti

With i 3,4,..., m

In the equation (5) it is assumed that the velocity v is known, in the present case, the velocity in the material will
be unknown and calculated in each one of the iteration. The matrix form of equations system is
A˜ x b (6)
Where

ª f3,1 f 3,2 f3,3 g 3wº ª xº ª h3 º

« » « y» ,b «#»
A « # # »,x « » « »
«¬ f m,1 f m ,2 f m ,3 g m w»¼ «¬ z »¼ «¬ hm »¼

Each new value of velocity calculated is compared with the input velocity (measured). This system gives the
coordinates of the source and the propagation velocity in the medium given by w .
866 V. Salinas et al. / Physics Procedia 3 (2010) 863–871
Vicente Salinas / Physics Procedia 00 (2010) 000–000

2.1. Algorithm

Fig.1 Flow diagram of development of the localization algorithm.

Figure (1) shows the flow diagram of the algorithm that begins with the selection of one AE event detected by all
the sensors. The program is feed with the coordinates of each sensor in any convenient system, according to the
geometry. For the initial configuration it is necessary to feed the first threshold to calculate a tentative position of
source in x(0). Then the program realize the calculation of the coordinates of the source and the propagation velocity
in the medium with the linear system of equations (6), these calculations are performed in computational platform.
After calculating the position and velocity, the program performs a comparison between the velocity measured and
calculated, if the difference between both is greater than the tolerance (the tolerance is given by the user, of course it
depends on the technique used in the measurement of speed, the case treated in this paper, the tolerance used was
10  m / s  ), the system changes the first threshold thus obtained new times of arrival. The decision to increase or
reduce the threshold is given by the tendency of the velocity calculated, in the first instant the program lowers the
threshold, if the value of the calculated velocity is far from the measured value i.e., if the speed difference increases,
the program increased the threshold until it converges, otherwise it continues to decline until fulfill the requirement
of tolerance. Finally, the program gives the coordinates of the source in the reference system chosen for the position
of the sensors.

3. Experimental Setup

The algorithm was tested on samples of epoxy resin (Araldite) (figure (2a)) which has in its interior a
piezoelectric ceramic (Figure (2b)), which simulate the acoustic emission. Because the cylinder is made with
semitransparent Epoxy resin is possible know the exact location of that source checking the algorithm accuracy. The
ceramic was excited with a pulse of 475 [9] for 9 [QV], emitting a pulse with shape and duration determined by its
 V. Salinas et al. / Physics Procedia 3 (2010) 863–871 867
Vicente Salinas / Physics Procedia 00 (2010) 000–000

own mechanic characteristics. In the figure (3) the impedance and phase curve for the piezoelectric ceramic it is
shown. The sample is a cylinder 5115 [PP] in diameter and 90 [PP] height.
Ceramics is 243 [PP] wide, 346 [PP] long and 290 [PP] thick. For the reception of signals, differential
transducers (16 [PP] in diameter) it was used. The resonance frequency of the sensor was measured, giving 137
[N+]].

Signals were acquired by the acquisition card with 2.5 [MHz] of sampling frequency.

(a) (b)

Fig.2 (a) Sample with BNC connector adapted to the excitation of ceramics. (b) Ceramics within the sample simulating acoustic emission.

Fig.3 (+) Phase of the ceramics used. (ź) Impedance of the ceramics used.
868 V. Salinas et al. / Physics Procedia 3 (2010) 863–871
Vicente Salinas / Physics Procedia 00 (2010) 000–000

The sensors were distributed to produce the best coverage of the volume in question. An ensemble of 12 sensors
distributed in 3 rings of 4 sensors on the surface of the sample was mounted. The signals recived by the 12 sensors
are shown in the figure (4). It can be appreciated in figure 4 that the first transducer detecting any signal is the
channel five sensor. Solid vaseline was used to couple the sensor with the sample.

Fig.4 Signals received in the 12 positions sorted by time of arrival.

 V. Salinas et al. / Physics Procedia 3 (2010) 863–871 869
Vicente Salinas / Physics Procedia 00 (2010) 000–000

4. Results

To assess the computing velocity of different methods to solve the linear system different algorithms were tested;
the results are shown in table I.

Method Accuracy (mm) Nº Iterations

Gauss-Seidel 3.350 621
Linsolve 2.033 491
Jacobi 3.562 541
Gauss 3.562 541

Analyzing the results supplied by the tested algorithms, the method obtaining greater accuracy is Linsolve. The
difference between the actual position and the calculated positions is 2.033[mm]. In figure (5) the convergence of
this method it is shown.

Fig.5 Convergence of the method according to the iterations calculated by Linsolve method.
870 V. Salinas et al. / Physics Procedia 3 (2010) 863–871
Vicente Salinas / Physics Procedia 00 (2010) 000–000

Progress Location Algorithm

0.08
Z coordinate
0.06

0.04

0.02

0
Ŧ0.02
+
Ŧ0.01
0
0.01
0.01 0.02
0.02 Ŧ0.01 0
Ŧ0.02
X coordinate Y coordinate
Fig.6 Progress of the algorithm, outlined the solution in each iteration (dots) in the vicinity of the real position (diamond).

To better understanding of the process, in figure (6) the results of position source obtained in the iterations it is
shown. The blue dots represent susesive positions calculated, where the first iteration is the first point of the left and
the next are goes to the right. The red diamonds represent the true position of the source (PZT ceramic). The green
contour represents the sample surface. In each iteration the computed solution is closer to the true position of the
source, which confirms the results shown in the convergence curve presented in figure (5).

5. Conclusion

A new approach to solve the problem of localization of acoustic emission source in 3D is presented. Localization
algorithm is evaluated in homogeneous media where the speed of propagation remains constant in all direction.
The position of the sensors is arbitrarily selected by the researcher depending of the geometry of the sample.
With this technique it is possible to find the time in which the event happened, this is easily obtained by replacing
the position found in equation (1) and (2).
The presented technique improves the location of AE source in approximately one magnitude order.

Aknowledgements

This work was supported by FONDECYT project 1085187 of the Chilean Government.

References

[1] GOST (State Standard) 27655-88: Acoustic Emission. Terms, Definitions, and Symbols, 1989.
[2] Masayasu Ohtsua,U, Y. Kaminagab, Muzo C. Munwamb Experimental and numerical crack analysis of mixed-mode
failure in concrete by acoustic emission and boundary element method Construction and Building Materials 13_1999.57]64
[3] A. Berkovits and D. Fang, Study of Fatigue Crack Characteristics by Acoustic Emission, Eng. Fracture Mech., 51 (3) (1995), pp. 401416.
[4] A. Tobias, Acoustic emission source location in two dimensions by an array of three sensors, Non-destr. Test 9I (1976) 912.
[5] M. Asty, Acoustic emission source location on a spherical or plane surface, NDT Internat. 11 (5) (1978) 223226.
[6] P. Barat, P. Kalyanasundaram, Baldev Raj, Acoustic emission source location on a cylindrical surface, NDT&E Internat. 26 (6) (1993)
295297.
[7] Dong-Jin Yoon, Young H. Kim, Oh-Yang Kwon, New algorithm for acoustic emission source location in cylindrical structure, J. Acoust.
Emiss. 9 (4)(1992) 237242.
 V. Salinas et al. / Physics Procedia 3 (2010) 863–871 871
Vicente Salinas / Physics Procedia 00 (2010) 000–000

[8] B.V. Landau, M. West, Estimation of the source location and the determination of the 50% Probability zone for an acoustic source locating
system (SLS) using multiple systems of 3 sensors, Appl. Acoust. 52 (1) (1997) 85100.
[9] Y.-C. Choi, Y.-H. Kim, Impulsive sources localisation in noisy environment using modified beam forming method, Mech. Systems Signal
Process. 20 (6) (2006) 14731481.
[10] Maochen Ge, Analysis of source location algorithms Part I: Overview and non-iterative methods, J. Acoustic Emission, 21 (2003) 14-24