Ultrasonics 54 (2014) 737–745

Contents lists available at ScienceDirect

Ultrasonics

journal homepage: www.elsevier .com/locate /ul t ras

A novel acoustic emission beamforming method with two uniform lineararrays on plate-like structures

0041-624X/$ - see front matter � 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.ultras.2013.09.020

⇑ Corresponding author. Tel.: +86 13141281907.E-mail addresses: hetian@buaa.edu.cn, ht9706@163.com (T. He).

Denghong Xiao a, Tian He a,⇑, Qiang Pan a, Xiandong Liu a, Jin Wang b, Yingchun Shan a

a School of Transportation Science and Engineering, Beihang University, Beijing 100191, PR Chinab Chongqing Key Laboratory of Computational Intelligence (Chongqing University of Posts and Telecommunications), Chongqing 400065, PR China

a r t i c l e i n f o a b s t r a c t

Article history:Received 10 June 2013Received in revised form 9 September 2013Accepted 20 September 2013Available online 2 October 2013

Keywords:LocalizationAcoustic emissionBeamformingUniform linear arrayFinite element

A novel acoustic emission (AE) source localization approach based on beamforming with two uniform lin-ear arrays is proposed, which can localize acoustic sources without accurate velocity, and is particularlysuited for plate-like structures. Two uniform line arrays are distributed in the x-axis direction and y-axisdirection. The accurate x and y coordinates of AE source are determined by the two arrays respectively. Toverify the location accuracy and effectiveness of the proposed approach, the simulation of AE wave prop-agation in a steel plate based on the finite element method and the pencil-lead-broken experiment areconducted, and the AE signals obtained from the simulations and experiments are analyzed using theproposed method. Moreover, to study the ability of the proposed method more comprehensive, a plateof carbon fiber reinforced plastics is taken for the pencil-lead-broken test, and the AE source localizationis also realized. The results indicate that the two uniform linear arrays can localize different sources accu-rately in two directions even though the localizing velocity is deviated from the real velocity, which dem-onstrates the effectiveness of the proposed method in AE source localization for plate-like structures.

� 2013 Elsevier B.V. All rights reserved.

1. Introduction

Acoustic emission (AE) is defined as the class of phenomenawhere by transient elastic wave is generated by the rapid releaseof energy from a localized source of damage [1]. Various types ofAE sources, such as fiber breakage [2], fatigue cracks [3], rubbing[4], and impact of foreign objects [5] can generate the propagationof AE waves. Since most of the damage information of structures iscontained in AE signal, AE technique (AET) is probably the mostsensitive non-destructive technique [6]. Compared with othernon-destructive testing (NDT) techniques, AET is distinctive intesting because AET can simply detect the energy released by anobject rather than supply energy to the object under examination.Moreover, AET can directly detect damage/crack extension and it isused to deal with dynamic processes in structures [7]. Due to itspotential advantages in damage monitoring and source localiza-tion, AET has been used in a variety of fields such as material [8]and manufacturing processes [9], especially in plate-like structureswhich are commonly used in civil, aerospace and otherapplications.

To accurately localize and evaluate the damage in plate-likestructures, four necessary steps are needed with increasing levelsof difficulty for AET. Step 1: qualitative indication of the occurrence

of damage. Step 2: quantitative assessment of the position of dam-age. Step 3: quantitative estimation of the severity of damage. Step4: prediction of structural safety, e.g. residual service life [10].Therefore, it is absolutely clear that the AE source localization isone of the most important pieces of information to be gained fromthe AET. The ability to localize the AE source is a step in wholedamage identification process, by which the accurate source loca-tion can indicate information about the characteristic of the dam-age and even the size of the crack with relatively few sensors onlarge and complex structures.

Currently localization of AE sources is normally performed byusing the time difference of arrival (TDOA) technique which ap-plies the propagation velocity in a material to derive the sourcelocation in one, two or three dimensions from the arrival delay be-tween sensors based on first threshold crossing. Arrival time infor-mation is a most important parameter as this arrival ispredominantly used for source location of TDOA [11]. However,when the AE wave propagates in the solid medium, the signalsmay be significantly affected by multi-mode, dispersion, energyattenuation and other factors [12], which make it difficult to accu-rately determine the arrival time. When structures and materialsare complex, it is common to be accompanied with large errors.In order to solve this problem, some techniques to improve theaccuracy of arrival time are introduced. To decrease the influenceof multi-mode and dispersion on plate wave propagation, Gorman[13] introduced plate wave theory to determine the orientation of

Fig. 1. Illustration of delay-and-sum beamforming.

738 D. Xiao et al. / Ultrasonics 54 (2014) 737–745

a source mechanism based on the amplitudes of the different wavemodes. It indicates that the wave theory can be used to improvesource location. A probabilistic approach was developed by Niriand Salamone [14] to decrease the uncertainties in the wave veloc-ity and the time of flight measurements. Ahadi and Bakhtiar [15]conducted the localization of continuous AE sources by using anintelligent locator since time delays cannot be simply estimated.But lots of repeated training must be carried out before localizationby intelligent locator. Besides, when there is more than one AEsource, the arrival time information may be confusing. This issueis also one key problem of TDOA. To make the localization simplebut effective, McLaskey et al. [16] introduced the AE beamformingmethod to localize AE source in civil structure, and it was improvedby He et al. [17] to plate-like structure, and used by Nakatani et al.[18,19] to study AE source localization on an anisotropic structure.Although beamforming has been successfully used in damagelocalization by many scholars, a difficulty is involved that thistechnique assumes constant wave speed in all propagation direc-tions and it is a velocity dependent method in direction normalto the array [16]. The propagation velocity of AE wave is not con-stant due to the multi-mode and dispersion in the solid structure.It is not only changed with propagating directions in anisotropicstructure [20], but also affected by the wave modes and frequen-cies of AE signal in isotropic structure [16,21]. Therefore, the local-ization accuracy of beamforming will be significantly influenced bythe propagation characteristics. In order to reduce the effect ofthose factors on AE beamforming, He et al. [22] revealed the mech-anism of AE propagation characteristics on the localization accu-racy of beamforming. Their results showed that the beamformingis not sensitive to the velocity along the array direction, but itchanges significantly in the direction normal to the array. Thenan approach for accurate localization velocity was presented bycombined with plate wave theory and wavelet packet transform.However, in these studies, the beamforming method is dependenton the accurate localization velocity.

This paper aims to propose a novel AE source localizationapproach based on beamforming for plate-like structures whenaccurate velocity is unavailable. The novel AE beamforming locali-zation method is established by arranging two uniform lineararrays. In order to verify this proposed method, the AE signalsobtained from both finite element (FE) simulation and pencil-lead-broken (PLB) test are analyzed for the source localization.Moreover, a plate of carbon fiber reinforced plastics (CFRP-plate)is taken as the PLB medium. Also, the obtained AE signals are usedto test the function of the localization method. The AE source local-ization results indicate that this method is effective to localize theAE source for plate-like structures.

2. Principles

In typical beamforming method, the detailed information of thesound source, especially the source localization is obtainedthrough a set of microphone array distributed in fixed positions.Delay-and-sum is the most widely used one among various beam-forming algorithms, which is simple but very effective [23].

The basic principle of near-field beamforming based on thedelay-and-sum algorithm is illustrated in Fig. 1. When the arrayis focused on a point source at limited distance, the array outputof incident waves in an isotropic plate is expressed by [17]

bð~r; tÞ ¼ 1M

XM

m¼1

wmxmðt � Dmð~rÞÞ ð1Þ

where M is the index of sensors, wm the weighting coefficient for thechannel of sensor m and xm(t) the measured signal of sensor m.Dmð~rÞ indicates the individual time delay of sensor m to the

reference point. By adjusting time delay Dmð~rÞ, the signals associ-ated with the spherical waves, emitting from sound source focus,will be aligned in time before they are summed. If the focused pointis the real source, the signals are aligned at the same wave front andthe energy output of the sensor array is maximum. However, thesignals cannot be aligned at the same wave front when the sensorarray is focused on other positions, and the energy output is notthe maximum.

As shown in Fig. 1, Dmð~rÞ can be obtained by

Dmð~rÞ ¼j~rj � j~r �~rmj

cð2Þ

where~r represents the distance of the reference to the focus point,~rm the distance between reference point and sensor m, and c thepropagation velocity of sound.

The literature [19] indicates that the resolution is high in thedirection along the array, but it is low in y-axis direction. More-over, the beamforming is not sensitive to the velocity in the direc-tion along the sensor array while it is velocity-dependent in thedirection normal to the sensor array no matter the S0 wave or A0

wave is used. Although the sensitivity of beamforming to the local-ization velocity error was discussed in references [17,22], theemphasis of these two papers is to obtain high localization accu-racy by improving accuracy of the localization velocity used inAE beamforming.

3. The presented beamforming method with two uniform lineararrays

The proposed technique aims at localizing AE source by arrang-ing two uniform linear arrays, regardless of the accurate velocity.Refer to [22], the localization accuracy of beamforming method isnot sensitive to the velocity along the array direction (x-direction),while it changes significantly in the y-direction at various veloci-ties. Although this is a drawback of beamforming, the localizationaccuracy is robust along the array direction and it is velocity-inde-pendent. Therefore, if two uniform linear arrays are distributed inthe x-axis direction and y-axis direction respectively, the accuratecoordinates of AE source can be well determined, whereby the xand y coordinate values are calculated by the sensor array alongx direction and y direction respectively. Thus, even if the measuredvelocity has a large error, the accurate AE source location can beobtained.

Fig. 2. Flowchart of presented proposal.

D. Xiao et al. / Ultrasonics 54 (2014) 737–745 739

The flowchart of the proposed approach is shown in Fig. 2. Sen-sors Sxi and Syi (i = 1, 2, 3,. . .,n) are the elements of the AE sensorarrays in x direction and y direction respectively. As previous sec-tions stated that the AE beamforming has high localization accu-racy in the direction along the array, the AE sensor array in xdirection is adopted to determine the x coordinate of the AE source.Also, the x coordinate of the AE source can be obtained through AEbeamforming with the AE sensor array in y direction. Therefore, theAE source position is identified as (x, y). And the accurate velocityis not necessary for the AE source localization work.

Fig. 3. Setup of simulation configuration.

4. AE simulation based on FEM

In order to verify the proposed AE source localization method,the simulation signals of AE are employed in the localization testsin this section. Recently, many authors have applied the finite ele-ment method (FEM) to simulate AE formation and AE signal prop-agation for plate specimens [24–27]. These researches clearlydemonstrate that it is possible to use FEM for solving AE wavepropagation problems. Modeling of AE waves can convenientlyprovide high signal-to-noise ratio modeled data that can be usedto develop useful advanced signal processing and analysistechniques. Besides, there are numerous commercial FE packagesavailable. They have some advantages such as that they are veryuser-friendly and may provide sophisticated pre-and post-process-ing functions. Moreover, compared to real AE signals, almost all ofthe most important information of the AE signal including sourcelocation, source type, source orientation, and source time historycan be exactly determined during the AE modeling process. Dueto these advantages, the AE signal is simulated in this paper byFE analysis in the general three-dimensional domain using thecommercial package ABAQUS/Explicit [28], which has been imple-mented in AE simulation by many scholars [29–31].

4.1. FE model of AE simulation

The model is shown in Fig. 3, where the rectangular plate isassumed to be an isotropic steel plate with 500 mm � 500 mm inlength and width and 5 mm thick. The numerical model consists

of a homogeneous steel plate with specifically elasticity modulus,E = 2.09 � 1011 N/m2, Poisson’s ratio t = 0.3, and densityq = 7.8 � 103 kg/m3. According to the plate wave theory, the longi-tudinal and shear wave velocities in this plate can be calculated as5940 and 3230 m/s respectively [26]. To utilize the novel AE sourcelocalization approach without accurate localization velocity basedon beamforming proposed in Section 3, two uniform linear arraysare distributed in the x direction and the y direction respectively,as shown in Fig. 3. The vertical sensor array is 50 mm away fromthe right edge of the plate, while the horizontal sensor array is also50 mm away from the bottom of the plate. For each linear arraythere are four sensors arranged with spacing of 120 mm. The posi-tive direction of x-axis points to the left side of the plate, and thepositive direction of y-axis is perpendicular to the horizontal

Fig. 4. AE waveform in time domain.

740 D. Xiao et al. / Ultrasonics 54 (2014) 737–745

sensor array direction. The cross point of the x- and y-axis isassigned as the origin of coordinates. To comprehensively studythe performance of this method, nine AE sources are excited in dif-ferent position of the plate. The locations of these AE sources arerespectively marked with 1# (340 mm, 340 mm), 2# (280 mm,280 mm), 3# (150 mm, 340 mm), 4# (210 mm, 280 mm), 5#(150 mm, 150 mm), 6# (210 mm, 210 mm), 7# (340 mm,150 mm), 8# (280 mm, 210 mm), and 9# (250 mm, 250 mm) inthe plate.

The AE waves in plates are passively generated by using amechanical PLB input, and actively generated by using a surfacebonded piezoelectric actuator [32]. To model AE waves in FE simu-lation, a transient excitation such as a delta or step function isneeded [24,25]. For the simulation work in this paper, a triangularloading function f(t) is applied to excite the AE signal. In the presentconfiguration, the total distance of the two points acting as burieddipole source is 1 mm with 1 N force in magnitude [26,33]. The goalof this model is to obtain AE signals with frequencies less than0.2 MHz, where only the zero-order symmetric mode, S0, andanti-symmetric mode, A0, are presented [22]. These two modesare selectively excited in the model by applying appropriate nodalloads. For a maximum frequency of 0.2 MHz, the minimum wave-length is for A0 and it is given by, kmin ¼ cs=fmax ¼ 15:71 mm consid-ering a theoretical phase velocity cs = 3230 m/s. In present study thevalue of kmin is assigned as 20 mm.

To model the AE propagation in a plate-like structure, ABAQUS/Explicit is adopted with an explicit integration based on the centraldifference method [34], whose stability is limited by the temporaland the spatial resolution of the analysis.

To avoid numerical instability, ABAQUS/EXPLICIT recommendsa stability limit for the integration time step Dt:

Dt ¼ lmin=cL ð3Þ

The integration time step and the element size are dependenton the maximum frequency of this AE simulation, fmax. There is asuper rule that a minimum of 20 points per cycle at the highest fre-quency, that is:

Dt ¼ 1=ð20f maxÞ ð4Þ

The size of the finite element, le, is typically derived from thesmallest wavelength to be analyzed, lmin. Based on the smallestwavelength to be analyzed, the size of the finite element, le, canbe calculated. For a good spatial resolution, 16 nodes per wave-length are normally required at least [26,34].

le ¼ kmin=16 ð5Þ

Therefore, from Eq. (5), in this simulation the element size isle = 1 mm. This AE propagation problem is solved with an integra-tion time step Dt = 2.5 ls. The sampling rate is 5 MHz, which is suf-ficient to resolve the observed signals frequency content in therange up to a maximal frequency of 0.2 MHz.

Fig. 4 shows the AE waveform of source #2 in time domain. It isobvious that the mode wave S0 reaches the sensor position prior tomode wave A0 because the velocity of S0 is as high as previous anal-ysis. It also can be found that the intensity of mode wave A0 is high-er than that of mode wave S0. As discussed by Gorman [35], whenthe pencil-lead breaks on the surface, it preferentially generatesthe A0 mode. And if the pencil-lead breaks on the plate edge, itexcites signals with dominant S0 mode components. A similarobservation was also reported by Hamstad for isotropic materials[36] and Sause for anisotropic materials [24].

4.2. Localization result

In order to illustrate the effect of the proposed localizationmethod, the wave signal obtained from one of those AE sources

is used and the #9 is selected in analysis. The real coordinates ofthis AE source is (250 mm, 250 mm). In plate-like structures, AEwave is a common form of Lamb wave propagation, which travelsin a variety of modes with different group and phase velocities[22]. According to previous analysis, in the frequency range ofinterest, only the zero-order symmetric mode, S0, and anti-sym-metric mode, A0, are presented. Therefore, in this section, the local-ization results through S0 and A0 will be presented. After that,localization results of other typical AE sources are attached.

4.2.1. Localization through S0 with two uniform linear arraysThe normalized energy of array output is shown in Fig. 5, where

the localization results with horizontal sensor array and verticalsensor array are described in Fig. 5(a) and (b) respectively. It canbe found that the side lobe in the direction along the array is smallin shape with high resolution, while the side lobes and the mainlobe in the direction normal to the array cannot be distinguishedand it has poor resolution. Fig. 6 shows the localization resultobtained by beamforming with different localization velocities inthe range from 2000 m/s to 6000 m/s for both the horizontal arrayand vertical array. The curves in Fig. 6 denote the x- and y-coordi-nates of identified source varied with different wave velocitiesrespectively. It is obvious that the beamforming with two uniformlinear arrays has fault tolerance ability of the error between thelocalization velocity and real velocities, but the fault tolerance abil-ity in x and y directions are obviously different. Fig. 6 also indicatesthat the horizontal sensor array is proper to accurately localize theAE source in x-axis direction, while the AE source in y-axis direc-tion can be accurately localized in the direction along the verticalsensor array. Moreover, for both horizontal sensor array and verti-cal sensor array, the localization result is affected by the velocityerror in the direction normal to the array. Therefore, these two sen-sor arrays can complement each other very well, where the veloc-ity effect can also be ignored.

4.2.2. Localization through A0 with two uniform linear arraysThis section focuses on the analysis of localization via A0 with

two uniform linear arrays. Fig. 7 shows the localization resultobtained by beamforming method with different localizationvelocities in the range from 2000 m/s to 6000 m/s. It also describesthe x- and y-coordinates of identified sources varied with differentwave velocities. It can be clearly seen that the beamforming meth-od with two uniform linear arrays has fault tolerance ability of theerror between the localization velocity and real velocities. But the

Fig. 5. Normalized energy of array output: (a) localization result with horizontalsensor array and (b) localization result with vertical sensor array.

Fig. 6. Localization results through S0 with two sensor arrays.

Fig. 7. Localization results through A0 with two sensor arrays.

D. Xiao et al. / Ultrasonics 54 (2014) 737–745 741

fault tolerance ability in x and y directions are obviously different.Similar to the localization via S0 with two uniform linear arrays,when the mode wave A0 is used for localization, the horizontal sen-sor array and the vertical sensor array have a strong ability to

accurately localize the AE source in x-axis and y-axis directionsrespectively. Moreover, in the direction perpendicular to the array,the localization result is affected seriously by the velocity errorwhether the horizontal sensor array or vertical sensor array isused. Therefore, the AE source localization can be well determinedunder these two uniform linear arrays and the accurate localiza-tion velocity is not necessary.

4.2.3. Localization results of other typical AE sourcesTable 1 summarizes the localization results via S0 with two uni-

form linear arrays. It is clearly seen that the localization methodproposed in this paper can localize the AE source accurately in bothx-axis and y-axis directions, where only S0 is used. Moreover, allthe locations of those AE source can be identified correctly. Thelocalization results of those AE sources near to the sensor array,2#, 4#, 6#, 8#, are more accurate than the results of the sourcesnearby the boundary of the plate, 1#, 3#, 5#, and 7#. The reasonmay be that the positions 2#, 4#, 6# and 8# are all close to the cen-ter and farther away from the sensors than the remaining positions(1#, 3#, 5# and 7#). The localization results obtained from A0 withtwo uniform linear arrays are shown in Table 2. When A0 is usedfor localization, the proposed localization method localizes all theAE sources correctly in both x-axis and y-axis directions. In addi-tion, the localization results of those AE sources, 2#, 4#, 6#, 8#,which are near to the sensor array, are more accurate than thelocalization results of the sources nearby the boundary of the plate,1#, 3#, 5#, and 7#. This is the same phenomenon as the localiza-tion results of S0 mode wave. Therefore, it can be concluded thatthe localization accuracy is better for AE source near the centerof the array (or area of interest). Compared with the localizationresults via S0, the overall localization accuracy of A0 is lower. Itmight be caused by the facts that the dispersion of A0 is more seri-ous than S0 and there maybe exist some reflection waves within A0.Similar conclusion is gained by a number of simulations conductedat other positions of the plate. Therefore, in practical engineeringapplications, it is apt to use A0 rather than S0 for AE source locali-zation because its energy is much higher than S0.

5. Test verification

In order to verify the AE source localization accuracy of the pro-posed method, firstly the PLB test on a steel plate is carried out. The

Table 1Localization results with S0 obtained by FEM simulation.

Localization velocity AE source

1# 2# 3# 4# 5# 6# 7# 8#

x y x y x y x y x y x y x y x y340 340 280 280 150 340 210 280 150 150 210 210 340 150 280 210

4000 340 360 280 280 150 350 210 280 150 140 210 210 350 150 280 2104500 340 360 280 280 150 340 210 280 150 140 210 210 340 150 280 2105000 350 350 280 280 150 350 210 280 150 140 210 210 340 150 280 2105500 350 350 280 280 150 380 210 280 150 140 210 210 330 150 280 2106000 350 340 280 280 150 370 210 280 150 120 210 210 330 150 280 210

Table 2Localization results with A0 obtained by FEM simulation.

Localization velocity AE source

1# 2# 3# 4# 5# 6# 7# 8#

x y x y x y x y x y x y x y x y340 340 280 280 150 340 210 280 150 150 210 210 340 150 280 210

2000 340 290 260 260 150 380 210 260 140 150 210 210 370 180 270 2202500 340 290 260 260 150 370 210 290 160 140 210 210 370 180 280 2203000 340 290 280 260 120 320 220 290 160 140 210 210 320 180 280 2203500 370 260 280 260 120 320 220 290 160 160 210 210 380 180 280 2204000 360 280 280 300 120 370 220 290 130 160 210 240 360 120 280 220

742 D. Xiao et al. / Ultrasonics 54 (2014) 737–745

tested AE signals are used for source localization with the proposedmethod. Moreover, to study the function of the proposed methodmore comprehensive, a plate of carbon fiber reinforced plastics(CFRP-plate) is taken and the PLB test is also carried out.

Fig. 8. The AE signal obtained by PLB test.

5.1. AE source localization of an isotropic steel plate

The steel plate specimen has the dimensions of 500 mm inlength, 500 mm in width and 5 mm in height. The array of sensorsis shown in Fig. 8(a). The AE sources are respectively marked with1# (450 mm, 450 mm), 2# (390 mm, 390 mm), 3# (210 mm,450 mm), 4# (270 mm, 390 mm), 5# (210 mm, 210 mm), 6#(270 mm, 270 mm), 7# (450 mm, 210 mm), 8# (390 mm,270 mm), 9# (330 mm, 330 mm) in tests. The sampling frequencyin tests is 3 MHz. The waveform of one signal acquired by PLB testsis shown in Fig. 8(b). It can be found that the intensity of A0 is highwhile S0 is low. Moreover, lots of waves are reflected when theyreach the boundaries.

The normalized energy of array output is shown in Fig. 9, wherethe localization results with horizontal sensor array and verticalsensor array are described in Fig. 9(a) and (b) respectively. The realcoordinate of this AE source is (330 mm, 330 mm). It can be foundthat the side lobe in the direction along the array is small, while theside lobes and the main lobe in the direction normal to the arraycannot be distinguished. Therefore, the PLB test result indicatesthat the resolution of the uniform linear array is high in the direc-tion along the array, and that is poor in the direction perpendicularto the array.

Tables 3 and 4 summarize the localization results through S0

and A0 with two uniform linear arrays respectively. Taking thelocalization results into consideration, this proposed localizationmethod can be used to localize the AE source accurately in bothx-axis and y-axis directions no matter whether S0 or A0 is adopted.Moreover, the identified locations of those AE source close to thecenter and farther away from the sensors, 2#, 4#, 6#, 8#, are moreaccurate than the remaining sources, 1#, 3#, 5#, and 7#. Besides,the overall localization accuracy of A0 is greater than the localiza-tion results obtained from S0. There are two main possible reasons:the dispersion of A0 is more serious than S0 and within A0 there

maybe exist some reflection waves. Similar conclusion is gainedby a number of PLB tests conducted at other positions of the plate.Hence, A0 can be taken for AE source localization in practical engi-neering-applications because its energy is much higher than S0.

Fig. 9. Normalized energy of array output: (a) localization result with horizontalsensor array and (b) localization result with vertical sensor array.

Table 3Localization results with S0 obtained by PLB test.

Localization velocity AE source

1# 2# 3# 4#

x y x y x y x450 450 390 390 210 450 270

2000 440 450 390 390 210 460 2602500 450 460 390 390 200 460 2603000 450 460 390 390 200 470 2603500 450 460 390 420 200 470 2604000 450 460 390 420 200 470 260

Table 4Localization results with A0 obtained by PLB test.

Localization velocity AE source

1# 2# 3# 4#

x y x y x y x450 450 390 390 210 450 270

2000 440 450 390 390 210 460 2602500 450 460 390 390 200 460 2603000 450 460 390 390 200 470 2603500 450 460 390 420 200 470 2604000 450 460 390 420 200 470 260

D. Xiao et al. / Ultrasonics 54 (2014) 737–745 743

5.2. Localization result of the CFRP plate

The CFRP plate is made of 18 equal layers with a total thicknessof 2 mm and stacking of [45 0 �45 90 0 45 0 �45 0]s. Fig. 10(A)shows the structure with dimensions 500 mm � 500 mm. EightAE sensors are attached to the surface of the CFRP plate. The coor-dinates of the AE sources are listed as following: 1# (200 mm,200 mm), 2# (200 mm, 350 mm), 3# (350 mm, 350 mm), 4#(350 mm, 200 mm), 5# (275 mm, 275 mm). Fig. 10(B) displaysthe waveforms of the AE signals tested by the array in x direction.Compared with the AE signal obtained from the steel plate, thewave shape is much more complex. It also can be found that theamplitude of A0 mode wave is large while that of S0 is small.

Compared with the AE wave in time domain propagating in thesteel plate, the waveform is much more complex in the CFRP plate.The most obvious feature is the wave structure of A0 mode wave,which is different from that of the steel plate. Besides, from thewaveforms plotted in Fig. 10(B) it is clearly seen that the waveshape of A0 mode waves tested by the four sensors are differentfrom each other. The reason may be that the A0 mode waves areinfluenced by serious dispersion and reflection waves. When thoseA0 mode waves are used for beamforming directly, the AE sourcecannot be well determined with high accuracy. In the paper, onlythe localization results of S0 mode wave is presented. Fig. 11 dis-plays the normalized energy of array output when the AE sourceis located at (275 mm, 275 mm). It can be clearly seen that thelocation of the AE source is identified and marked as the peak pointof the output. The AE source localization results are summarized inTable 5. By analyzing the localization results it concludes that theAE source location can be determined with high accuracy when S0

mode wave is used. Moreover, when the localization velocity ismuch slower the real wave velocity, the localization result willbe a little away from the real positions. It is suggested that a rela-tively large velocity can be used if the real wave propagation veloc-ity is unknown. Therefore, the proposed technique has theapplicability to localize the AE source in an anisotropic plate-likestructure. But the overall localization accuracy is not as high as that

5# 6# 7# 8#

y x y x y x y x y390 210 210 270 270 450 210 390 270

390 210 250 270 260 470 260 390 290390 210 240 270 260 470 230 390 260390 210 240 260 260 460 220 390 260390 210 250 260 260 460 220 390 260390 210 250 260 260 460 220 390 270

5# 6# 7# 8#

y x y x y x y x y390 210 210 270 270 450 210 390 270

390 210 250 270 260 470 260 390 290390 210 240 270 260 470 230 390 260390 210 240 260 260 460 220 390 260390 210 250 260 260 460 220 390 260390 210 250 260 260 460 220 390 270

Fig. 10. PLB test on the CFRP plate: (A) experiment setup. (B) AE signals tested bythe horizontal sensor array.

Fig. 11. Normalized energy of array output when the AE source is identified.

Table 5Localization results with S0 obtained by PLB test on the CFRP plate.

Localizationvelocity

AE source

1# 2# 3# 4# 5#

x y x y x y x y x y200 200 200 350 350 350 350 200 275 275

5000 190 200 190 360 310 420 300 240 260 2805500 190 200 190 360 310 420 300 210 260 2806000 190 200 180 400 310 410 370 210 260 2806500 190 200 180 300 320 320 370 210 260 2807000 210 200 180 390 320 390 370 210 260 2807500 210 200 180 380 320 370 370 210 260 2808000 210 200 180 380 320 370 360 210 260 2808500 190 200 180 370 320 360 360 210 260 2809000 190 200 180 360 320 360 360 210 260 280

744 D. Xiao et al. / Ultrasonics 54 (2014) 737–745

of the isotropic steel plate. To obtain better localization result, theAE wave propagation in anisotropic material should be consideredin future.

6. Summary and conclusions

In this paper, a novel AE source localization approach withoutmeasuring localization velocity based on beamforming is intro-duced for AE source localization of plate-like structures. The novelAE source localization method is established by two uniform lineararrays distributed in two directions. In order to verify this methodfor the source localization, the AE signals obtained from both FEsimulation and PLB are employed. Besides, a plate of carbon fiberreinforced plastics (CFRP-plate) is used to verify the applicabilityto localize the AE source in an anisotropic plate-like structure.Some conclusions are drawn based on analysis of the localizationresults as follows:

(1) Considering the directivity and velocity-dependent proper-ties of beamforming, a novel AE source localization methodwith two uniform linear arrays has been developed. If twouniform linear arrays are distributed in x direction and ydirection respectively, the accurate coordinates of AE sourcecan be well determined by the sensors array along x direc-tion and the y direction respectively. Therefore, even if thevelocity error is large, the AE source location can still belocalized accurately.

(2) Compared with sound waves propagating in air, the AE sig-nals in plate structures have some special characteristics,such as multi-mode and dispersion. In addition, some noiseand unwanted reflection waves appear in the real tested AEsignals. Therefore, in order to verify the localization perfor-mance of this proposed method, it is necessary to obtainhigh quality AE signals regardless of multi-mode, dispersion,noises and unwanted reflections. Simulation of AE wave sig-nals can conveniently provide high signal-to-noise ratiomodeled data that can be used to develop useful advancedsignal processing and analysis techniques.

(3) Compared with the localization results via S0, the overalllocalization accuracy of A0 is low. There are two main possi-ble reasons: the dispersion of A0 mode wave is more seriousthan S0 mode wave and there may be some reflection withinA0 mode waves. Therefore, when the obtained AE signals hasa high signal-to-noise ratio, S0 mode wave can be selected tolocalize the AE source. While if there is too much noise in thesignals, the A0 mode wave should be used for the localizationsince the amplitude of A0 mode wave is much larger than S0

mode wave.

D. Xiao et al. / Ultrasonics 54 (2014) 737–745 745

(4) The beamforming method with two uniform linear arrayshas fault tolerance ability of the error between the localiza-tion velocity and real velocities. Those two sensor arrays cancomplement each other very well, where the velocity effectcan also be ignored. Moreover, the identified locations ofthose AE source near to the sensor array are more accuratethan other sources nearby the boundary of the plate. Themismatch between the identified and real source locationis considered due to the dispersion of waves.

(5) The AE source localization in a CFRP plate indicates the pro-posed technique has the applicability to localize the AEsource in an anisotropic plate-like structure. However, sincethe waveforms of the tested AE signals in the CFRP plate ismuch more complex than that in the isotropic steel plate,the overall localization accuracy is declined, especially whenthe A0 mode wave is used.

Acknowledgements

This work was finically supported by the National Science Foun-dation of China (Grant No. 51105018), Chongqing Key Laboratoryof Computational Intelligence (Grant No. CQ-LCI-2013-07),Aeronautical Science Foundation of China (Grant No. 2011ZB51039)and the Innovation Foundation of BUAA for PhD Graduates.

References

[1] R. Wells, M.A. Hamstad, A.K. Mukherjee, On the origin of the first peak ofacoustic emission in 7075 aluminum alloy, J. Mater. Sci. 18 (1983) 1015–1020.

[2] M.G.R. Sause, S. Horn, Simulation of lamb wave excitation for different elasticproperties and acoustic emission source geometries, J. Acoust. Emiss. 28 (2010)142–154.

[3] L. Lin, F. Chu, HHT-based AE characteristics of natural fatigue cracks in rotatingshafts, Mech. Syst. Signal Pr. 26 (2012) 181–189.

[4] Q. Wang, F. Chu, Experimental determination of the rubbing location by meansof acoustic emission and wavelet transform, J. Sound Vibr. 248 (2001) 91–103.

[5] T. Kundu, S. Das, S.A. Martin, et al., Locating point of impact in anisotropic fiberreinforced composite plates, Ultrasonics 48 (2008) 193–201.

[6] K.M. Holford, M. Eaton, Recent advances in acoustic emission, in: Proceedingsof World Conference on Acoustic Emission, Beijing, China, 2011, pp. 58–66.

[7] J. Yu, P. Ziehl, B. Zárate, et al., Prediction of fatigue crack growth in steel bridgecomponents using acoustic emission, J. Constr. Steel Res. 67 (2011) 1254–1260.

[8] N. Choi, T. Kim, K.Y. Rhee, Kaiser effects in acoustic emission from compositesduring thermal cyclic-loading, NDT&E Int. 38 (2005) 268–274.

[9] D.E. Leea, I. Hwanga, C.M.O. Valenteb, et al., Precision manufacturing processmonitoring with acoustic emission, Int. J. Mach. Tools Manuf. 46 (2006) 176–188.

[10] K. Worden, G. Manson, D. Allman, Experimental validation of a structuralhealth monitoring methodology: Part I. Novelty detection on a laboratorystructure, J. Sound Vib. 259 (2003) 323–343.

[11] P. Sedlak, H. Yuichiro, E. Manabu, Acoustic emission localization in thin multi-layer plates using first-arrival determination, Mech. Syst. Signal Pr. 36 (2013)636–649.

[12] S. Grondela, C. Delebarrea, J. Assaada, Fatigue crack monitoring of rivetedaluminium strap joints by lamb wave analysis and acoustic emissionmeasurement techniques, NDT&E Int. 44 (2007) 3666–3684.

[13] M.R. Gorman, AE source orientation by plate wave analysis, J. Acoust. Emiss. 9(1991) 283–288.

[14] E.D. Niri, S. Salamone, A probabilistic framework for acoustic emission sourcelocalization in plate-like structures, Smart Mater. Struct. 21 (2012) 035009-1–035009-16.

[15] M. Ahadi, M.S. Bakhtiar, Leak detection in water-filled plastic pipes throughthe application of tuned wavelet transforms to acoustic emission signals, Appl.Acoust. 71 (2010) 634–639.

[16] G.C. McLaskey, S.D. Glaser, C.U. Grosse, Beamforming array techniques foracoustic emission monitoring of large concrete structures, J. Sound Vibr. 329(2010) 2384–2394.

[17] T. He, Q. Pan, X. Liu, et al., Near-field beamforming analysis for AE sourcelocalization, Ultrasonics 52 (2012) 587–592.

[18] H. Nakatani, T. Hajzargarbashi, K. Ito, et al., Impact localization on a cylindricalplate by near-field beamforming analysis, sensors and smart structurestechnologies for civil, mechanical, and aerospace systems, in: M. Tomizuka,(Ed.), SPIE’s 2012 Annual International Symposium on Smart Structures andNondestructive Evaluation, San Diego, California, March 12–15, 2012, pp.8345.

[19] H. Nakatani, T. Hajzargarbashi, K. Ito, et al., Locating point of impact on ananisotropic cylindrical surface using acoustic beamforming technique, in: 4thAsia-Pacific Workshop on Structural Health Monitoring, Melbourne, Australia,December 5–7, 2012.

[20] T. Kundu, H. Nakatani, N. Takeda, Acoustic source localization in anisotropicplates, Ultrasonics 52 (2012) 740–746.

[21] S. Santhanam, R. Demirli, Reflection of lamb waves obliquely incident on thefree edge of a plate, Ultrasonics 53 (2013) 271–282.

[22] T. He, D. Xiao, Q. Pan, et al., Analysis on accuracy improvement of rotor-statorrubbing localization based on acoustic emission beamforming method,Ultrasonics. http://dx.doi.org/10.1016/j.ultras.2013.04.017.

[23] J. Yang, W.S. Gan, K.S. Tan, M.H. Er, Acoustic beamforming of a parametricspeaker comprising ultrasonic transducers, Sens. Actuator A-Phys. 125 (2005)91–99.

[24] M.G.R. Sause, Simulation of acoustic emission in planar carbon fiber reinforcedplastic specimens, J. Nondestruct. Eval. 29 (2010) 123–142.

[25] M.A. Hamstad, A. O’Gallagher, J. Gary, A wavelet transform applied to acousticemission signals: part 1: source identification, J. Acoust. Emiss. 20 (2002) 39–61.

[26] F. Moser, L.J. Jacobs, J. Qu, Modeling elastic wave propagation in waveguideswith the finite element method, NDT&E Int. 32 (1999) 225–234.

[27] Z. Fan, M.J.S. Lowe, M. Castaings, Torsional waves propagation along awaveguide of arbitrary cross section immersed in a perfect fluid, J. AcoustSoc. Am. 124 (2008) 2002–2010.

[28] ABAQUS v6.11-1 Analysis User’s Manual, 2010.[29] B.D. Mickael, Efficient finite element modeling of ultrasound waves in elastic

media, PhD thesis, Imperial College London, 2004.[30] Z.A.B. Ahmad, U. Gabbert, Simulation of lamb wave reflections at plate edges

using the semi-analytical finite element method, Ultrasonics 52 (2012) 815–820.

[31] G. Liu, J.S. Quek, A non-reflecting boundary for analyzing wave propagationusing the finite element method, Finite Elem. Anal. Des. 39 (2003) 403–417.

[32] J.P. Berenger, Simulation of asymmetric lamb waves for sensing and actuationin plates, J. Sound Vib. 12 (2005) 243–271.

[33] A. Ghoshal, W.N. Martin, M.J. Schulz, et al., Simulation of asymmetric lambwaves for sensing and actuation in plates, Shock Vib. 12 (2005) 243–271.

[34] B. Ivan, L.S. Francesco, F. Mahmood, et al., Modeling guided wave propagationwith application to the long-range defect detection in railroad tracks, NDT&EInt. 38 (2005) 325–334.

[35] M.R. Gorman, Plate wave acoustic emission, J. Acoust. Soc. Am. 90 (1991) 358–364.

[36] M.A. Hamstad, A.O. Gallagher, J. Gary, Effects of lateral plate dimensions onacoustic emission signals from dipole sources, J. Acoust. Emiss. 19 (2001) 258–274.