VIII International Conference on Fracture Mechanics of Concrete and Concrete StructuresFraMCoS-8

J.G.M. Van Mier, G. Ruiz, C. Andrade, R.C. Yu and X.X. Zhang (Eds)

MONITORING OF CRACK PROPAGATION IN REINFORCED CONCRETEBEAMS USING EMBEDDED PIEZOELECTRIC TRANSDUCERS

CEDRIC DUMOULIN, GRIGORIS KARAISKOS AND ARNAUD DERAEMAEKER∗

∗Universite Libre de Bruxelles - BATir50 av F D Roosevelt, CP 194/02, B-1050 Brussels, Belgium

e-mail: aderaema@ulb.ac.be, web page: http://batir.ulb.ac.be/

Key words: Non Destructive Testing, Concrete Damage Monitoring, Embedded Piezoelectric Trans-ducers, Smart Aggregates

Abstract. The field of structural health monitoring and damage detection has been studied intensivelyfor the last thirty years. The main goal is to reduce inspection costs and risks of unexpected failureusing real-time on-line monitoring systems. In this paper,Non Destructive Testing is applied usingpiezoelectric transducers embedded in the structure. Embedded piezoelectric transducers (Smart Ag-gregates) have been recently developed. These transducersare ideal candidates for active sensingmethods of local damages, due to their low cost, small size and large frequency bandwidth. In thisstudy, these transducers are used to detect and follow the crack propagation in a reinforced concretebeam subjected to a three-point bending test. The beam is equipped with two pairs of embedded trans-ducers, each pair consisting in a sender and a receiver. Two different types of excitation signals areused (pulse and chirp) and the resulting waves are recorded on the receivers. Based on these signals,different damage indicators are investigated and compared. The results indicate a very high sensitivityof the method which is able to pick-up the crack initiation phase and follow the crack propagationover the whole duration of the test.

1 INTRODUCTION

A concrete structure can be subjected to sev-eral factors that may damage it during its ser-vice life. These factors are either climatic,chemical or accidental. The security require-ments are more and more specific and severe.To ensure the safety of a structure, it is impor-tant to know its state. For this purpose, dif-ferent destructive or non-destructive techniquesare used. The use of destructive methods in or-der to determine properties of a material from astructure in service should be avoided as oftenas possible. Visual inspections are costly andcan only identify macroscopic damages at ac-cessible locations.

As an alternative, the field of structuralhealth monitoring has developed during the last

thirty years. For civil engineering structures,the current trend in research is the developmentof vibration-based methods relying on ambientlow-frequency vibrations caused by the envi-ronment (traffic, wind, ...) and measured withaccelerometers or, more recently, fiber optic dy-namic strain sensors [1]. The methods are moresuitable for large scale effects than for detectinglocal damages. An alternative, extensively stud-ied in aeronautics for metallic and compositematerials, is the use of active systems in whichhigh-frequency waves are generated and mea-sured by means of piezoelectric transducers.The frequency bands used for active sensingcan be much higher which enables the detectionof local defects as cracks. During the last tenyears, a few research teams have started to ap-

1

Cedric Dumoulin,Grigoris Karaiskos and Arnaud Deraemaeker

ply these techniques to concrete structures us-ing either surface mounted or embedded trans-ducers. The latter has two obvious advantageswhich are the added flexibility in the choiceof their position, and the better integration inthe overall design of the structure. Within thisframework, the concept of Smart Aggregates(SMAGs) has been developed by researchers atthe University of Houston, Texas [2].

Several authors have used impedance curvesto assess the strength and the damage state ofconcrete. The impedance curve is measuredusing a single PZT transducer, which is veryattractive from a practical point of view. Ex-periments show that this technique is sensi-tive to damage in very local areas around thetransducers [3, 4]. Other techniques are basedon the use of at least two transducers (oneemitter and one receiver). The methods dif-fer mainly in the choice of the signal generatedat the emitter side. Harmonic signals can beused to reveal non linearities due to damage,which generate harmonics of the fundamentalfrequency [5, 6]. A second type of excitationis the chirp signal which has a much broaderfrequency content [7]. Finally, pulse excitationis traditionally used in commercial systems de-signed to estimate the quality of the concretebased on the ultrasonic pulse velocity (UPV).The systems consist of external probes whichneed to be placed on two opposite faces of theconcrete specimen, using an adequate couplingagent. In practice, for real structures, this limitsthe application to through thickness propaga-tion, or repeated wave reflections, which makethe interpretation difficult. It is also often im-practical due to limited accessibility when inservice. Coupling such systems to embeddedpiezoelectric transducers can overcome most ofthese difficulties. More complex analysis of thewave generated by pulse excitation can be car-ried out, such as backscattered waves analysis.In this method the response signal attenuationform (exponential decreasing) is used as a dam-age indicator [8]. The backscattered waves areresulting from numerous interactions such as

voids, cracks or microcracks in the concretestructure. Each of them contains informationover the state of the material.

In the present study, a three-point bendingtest is performed on a reinforced concrete (RC)beam. The aim of this work is to detect and fol-low crack growth using embedded piezoelectrictransducers (see Figure 1), recently developedat the Civil Engineering Laboratory of ULB [9]based on the concept of SMAGs [10]. Two pairsof SMAGs are cast in the beam. These pairsconsist of one receiver and one emitter. Theyare located on each side of the center of thebeam where the first crack is supposed to ap-pear. In a first step, the beam is progressivelyloaded with very small loading steps up to theappearance of the first cracks. Secondly, thebeam is subjected to larger loads reaching themaximum acceptable load. Cracks or micro-cracks in the structure appear with the load in-creasing. This modifies the internal structure ofthe beam and the paths of waves and thereforethe received signals. Two different types of ex-citation signals are used in this study. The firstexcitation signal is a high voltage pulse as usedin classical UPV and the second excitation sig-nal is a sweep sine from 0.1 to 100 kHz. Foreach of these excitation signals a specific dam-age indicator is defined. These damage indica-tors are both built by comparing the responsesignal at each damage level with a reference sig-nal recorded before any damage. The resultsindicate a very high sensitivity of the methodwhich is able to pick-up the crack initiationphase and follow the crack propagation over thewhole duration of the test.

Figure 1: Test principle.

2

Cedric Dumoulin,Grigoris Karaiskos and Arnaud Deraemaeker

2 SMART AGGREGATESSMAGs are thin PZT patches cast in small

mortar pieces. The dimensions of the PZTpatches are 12x12mm of width, and the thick-ness is about 0.2mm. Dimensions of mortarpieces are about 2cm of diameter and thickness.Such types of transducers have been recentlydeveloped in the Civil Engineering laboratoryat ULB BATir. Figure 2 shows the differentmaterial layers used during the manufacturingof the SMAGs.

Figure 2: Smart Aggregate transducer; detailsof different layers.

Figure 3: Smart Aggregate transducer manufac-tured at ULB-BATir.

To transmit the electrical signal, wires areattached to both faces of the PZT patch. Forseveral reasons the patch cannot be cast withoutprotections. At early age, concrete presents a

certain quantity of water that is relatively con-ductive. To avoid an electrical coupling be-tween SMAGs, the introduction of waterproofcoating is needed. First SMAGs produced pre-sented important interferences. Therefore, athin layer of conductive paint was placed asshielding, surrounding the waterproof layer andgrounded. The PZT patch is really thin and brit-tle. The waterproof layer is a first mechanicalprotection but not sufficient. Then, the shelteredPZT patch is cast in a small mortar cylinder.The process is summarized on Figure 3.

3 EXPERIMENTAL PROGRAM

The test consists of loading a reinforced con-crete beam and following the induced cracksgrowth. The test is a three points bending test(see Figure 4).

Figure 4: Experimental setup.

SMAGs are embedded in the concrete beam.Excitation signals are sent to the piezoelectrictransducers, which create compression wavesthat are picked up by sensors. Cracks or micro-cracks in the structure appear with the load in-creasing. This modifies the internal structure ofthe beam and the paths of waves and thereforethe received signals. The signal at each damagestate is compared to the initial state. A high-quality picture corresponds to each excitationsignal sent; it is then possible to compare modi-fications in the response signal with a visual as-sessment of the appearance of cracks.

3

Cedric Dumoulin,Grigoris Karaiskos and Arnaud Deraemaeker

3.1 SpecimenTwo couples of transducers are disposed in

the beam as exposed on Figure 5. One is lo-cated near the bottom axis and the second oneis located near the neutral axis. They are dis-posed on each side of the center of the beam ata distance of about 20cm.

A6

A18g

A9

A17g

Figure 5: Location of the SMAGs.

Due to the heterogeneity of concrete, cracksare not always exactly initiated at the centerof the beam. That is an undesirable effect forthis test. A notch permits to reduce locally thestrength of the beam. It also creates stressesconcentration. This pre-cracking should ensurethe crack initiation at the right place.

The constituents of concrete are given in Ta-ble 1 and the mechanical properties of the con-crete are given on the Table 2. Table 3 givesinformations on geometrical characteristics andthe computed strength of the beam.

3.2 Loading ProcedureThe test occurred during two different

phases. In the first phase, the beam is totallyunloaded slightly after appearance of the firstcracks. Figure 6a summarizes the loading pro-cedure of the first phase. The loading procedurefor the second phase is shown on Figure 6b; thebeam is loaded close to the maximum accept-able load. During the second phase, in addi-tion to the force and displacement sensors of thebending machine, a LVDT extensometer (Lin-ear Variable Differential Transformer) is used

to measure the deflection at center of the beamand six other LVDT extensometers (3 on eachside) are used to measure cracks width. Crackswidth has therefore only been measured on thesecond day. Extensometers are then correctlyplaced to measure the different cracks width.

Table 1: Concrete constituents of the RC beam

Water 190 kg/m3

Cement (CEM I 52.5 R HES) 350 kg/m3

Sand (0/4) 665 kg/m3

Aggregates Limestone (4/20) 1172kg/m3

Adjuvant (Viscocrete 4) 1.25 kg/m3

Table 2: Mechanical properties of the concrete

Density 2338.6 kg/m3

Compressive strengthfc 57.1 MPa

Tensile strengthft 2.86 Mpa

Young’s modulus (4/20) 33000 Gpa

Table 3: RC beam properties

L x l x h 1470x140x190 mm

Bottom reinforcement bars 2xφ12 mm

Top reinforcements bars 2xφ6 mm

Distance between stirrup 11 cm

Cracking load 9.9 kN

Maximum load 51.3 kN

4 STRUCTURAL HEALTH MONITOR-ING METHODS

As already indicated, two monitoring sys-tems are used in the current study. The UPVsystem is based on the FreshCon system, a com-mercial system designed by the University ofStuttgart for evaluating fresh parameters of con-crete.

4

Cedric Dumoulin,Grigoris Karaiskos and Arnaud Deraemaeker

0 50 100 150 200 250 300 350 400 450−1

1

3

5

7

9

11

13

1516

Time [min]

Load

[kN

]

Loading ramps − phase 1

23

1

4

5

5b

6

6b

7

8

(a) Phase 1: The beam is loaded up to the crack initiation(15kN).

0 50 100 150 200 250 300 350 400 4500

5

10

15

20

25

30

35

40

45

50

55

60Loading ramps − phase 2

Time[min]

Load

[kN

]

Re−Load

9

10

11

12

13

1416

17

18

(b) Phase 2: The beam is loaded up to the failure(50kN).

Figure 6: Loading procedure in two phases.

In this work, this system is coupled to theSMAGs. The chirp method system is developedfrom a Labview environment with a NationalInstrument (NI) data acquisition system (DAQ).Both systems are based on the same principles.An excitation signal is transmitted to a piezo-electric actuator that transforms it in a mechan-ical wave that propagates through the structure,a piezoelectric sensor measures strains createdby this wave and transmits the signal to a DAQsystem. The two systems differ in the excitationsignal and the DAQ system. But each of thesemonitoring systems are based on the same prin-ciple. On each step of load, a new measurementof these signals is recorded. A damage indexis defined by comparing each damaged signalwith the initial undamaged signal.

4.1 FreshCon SystemThe FreshCon System allows to send a high

voltage short pulse signal to a specific piezo-electric actuator and measure the time of prop-agation of theP -wave transmitted by analyzingthe sensors signal. This system is launched dur-ing the loading ramps of the beam with a fre-quency of measurements of one pulse every 10seconds. The duration of each pulse is 10µs at avoltage of 800V and the response is measured ata sampling rate of 10MHz. Previous tests haverevealed that this length of pulse gave better re-sults than other pulse time.

From the measured waves, it was found thatthe signal did not change significantly for loadsbelow 5 kN. For higher loads, the damage hasbeen found to affect two main parameters whichare the time of propagation and the amplitudeof the received signal (Figure 7). These two pa-rameters are difficult to evaluate correctly, as adecrease of the amplitude causes the signal toreach the noise level, causing potentially a largeerror in the determination of the time of prop-agation. The choice of the damage indicator isbased on the following reasoning: the first wavereceived by the sensor corresponds to the short-est wave path, which is affected only by themechanical properties of the concrete betweenthe transducers. The other waves are diffractedand refracted and arrive later to the sensor, andare therefore not considered. The first period ofthe signal mainly contains the contribution ofa direct wave between the SMAGs and the firsthalf-period corresponds to the compression partof this wave. As a consequence, we have cho-sen to define the damage indicator based on theroot main square deviation (RMSD) betweenthe healthy signal and a damaged signal com-puted in the time window corresponding to thefirst half-period of the undamaged signal. It isgiven by :

Ij =

√

√

√

√

∫ tp

tn(xj(t)− x0(t))

2dt∫ tp

tnx2

0(t)dt

(1)

Wherexj(t) corresponds to the amplitude ofthe damaged signal andx0(t) is the amplitude

5

Cedric Dumoulin,Grigoris Karaiskos and Arnaud Deraemaeker

of the healthy signal,tn is the time of receptionof x0, tp − tn corresponds to the duration of the1st half period (Figure 7).

35 40 45 50 55−40

−30

−20

−10

0

10

20

30

40

x [m

V]

Time [µs]

tpti

xj,i − x0,i

tn

7 kN

14 kN

healthy signal x0

Figure 7: Definition of the time window,tn isthe time of reception of the undamaged signal,tp − tn corresponds to the duration of the 1sthalf period of the undamaged signal.

This definition of the damage index takesinto account both the shift of the time propaga-tion and the decrease of the amplitude withoutcomputing explicitly their values. The satura-tion of the damage index means that it is con-stant and equal to one and therefore, that thedamaged signal is equal to zero in the first halfperiod.

4.2 Chirp systemThe aim of this system is to observe the ef-

fect of damage in the frequency domain. It isa high frequency vibration test. Using a sweepsine excitation has the advantage that each fre-quency is separately excited. As all the frequen-cies (in the defined bandwidth) are swept, thespectral energy is constant. This signal is there-fore particularly suitable for a frequency analy-sis. The frequency bandwidth is from 100 [Hz]to 100 [kHz] and is swept in 5 seconds. Theresponse signal is measured at a sampling rateof 204.8 kHz. In comparison with the previoustest, the length of the signal is important. Con-sidering the velocity of the sweep sine, from

5kHz the signal can assumed to be in the steadystate. Below 5kHz, the energy of the signal isvery low compared to the higher frequenciesand this part can be ignored (see Figure 8). Thechirp is generated by the NI system and am-plified by a QuickPack QPA200 voltage ampli-fier. All the sensors are plugged on the NI DAQ.This DAQ is controlled by a Labview environ-ment on a laptop. The signals are measured atthe end of each loading ramp only, as it wasnot possible to measure simultaneously with theFreshcon system on the same transducers. Asfor the FreshCon system, the RMSD value iscomputed to define the damage index. In thiscase, this value is obtained from the Amplitudeof the FRF between the actuator and the sen-sor computed for the damaged signal and thehealthy signal. The FRF is obtained by divid-ing the Discrete Fourier Transform (DFT) of theoutput signal by the DFT of the excitation sig-nal. The RMSD value is given by:

Ij =

√

∫ ωp

ωn

(Xj(ω)−X0(ω))2dω

∫ ωp

ωn

X2

0(ω)dω

(2)

Whereωn is equal to 5kHz andωp is equal to100kHz.

0 20 40 60 80 1000

0.01

0.02

0.03

0.04

0.05

0.06

Frequency[Hz]

FR

F |X

(f)|

Figure 8: FRF of a signal, the energy at thelower frequencies is very low.

6

Cedric Dumoulin,Grigoris Karaiskos and Arnaud Deraemaeker

5 EXPERIMENTAL RESULTS5.1 Visual detection of cracks and cracks

width measurementA high quality photo camera takes pictures

of the notch every 10 seconds, which cor-responds to the FreshCon measurement fre-quency. The main idea is to join the picturesto signals and later detect and follow the initia-tion and the propagation of cracks by analyzingthese pictures.

It is difficult to visually detect very smallcracks with the basic pictures as shown on Fig-ure 9a. Image processing tool as Photoshop al-lows the application of various numerical fil-ters that improve the visual detection of cracks(see Figure 9b). It is important to point out thatprocessed pictures cannot be used for evaluat-ing cracks width. Indeed, the processing toolstend to increase the width.

(a) Before numerical processing.

(b) After numerical processing.

Figure 9: Image processing for visual detectionof cracks. The crack width is approximately0.1mm.

It was observed that in the area betweenthe SMAGs, some micro cracks seem to ap-

pear around the notch corresponding approxi-matively to a load of 10kN. The crack at thecenter of the beam is totally formed under aforce of 14 kN. The width of the cracks isaround 0.07 mm, which is very small. It is im-portant to note that other cracks were observedunder smaller loads that mentioned above. Nev-ertheless they were located outside of zone ofinterest, which is the area between the SMAGs.

5.2 Damages indexFigure 10 shows the the evolution of the

damage indexes for the SMAGs that are locatedon the bottom of the beam (A18g-A17g). Un-fortunately, the other pair of SMAG was defec-tive and the results can not be exposed.

Figure 10: First 1/2 period (UPV) index andchirp index for the couple of SMAGs on the bot-tom of the beam.

The evolution of the UPV damage indexshows a good correlation with the observationsduring the experiment. This damage indicatorincreases between 5 and 15kN which are thevalues of the initiation of the observed cracks.This index saturates after the appearance ofcracks and therefore, does not evolve anymorewhen crack is growing. It saturates largely be-fore the service limit. This index is a good in-dicator for detecting crack appearance but doesno give any information about the the evolutionor the size of this damage.

7

Cedric Dumoulin,Grigoris Karaiskos and Arnaud Deraemaeker

The second damage index (chirp) has theadvantage of evolving after the appearance ofcracks. It can therefore give a qualitative ideaabout the evolution of the crack width but de-pends on many parameters that are not well un-derstood. The evolution of this damage indi-cator is difficult to correlate with the observa-tions. It increases in the first loading steps,where no cracking has been observed. Further-more, this index does not saturate even close tothe strength limit of the beam. The impossibil-ity of distinguishing the local phenomena fromthe global phenomena is one of the parametersthat probably make this index difficult to inter-pret.

6 CONCLUSIONS

The goal of this research was to develop amethod to monitor the damage of a concretestructure using embedded piezoelectric trans-ducers. Two methods have been developed forcomparison. One method is based on pulse ex-citation and post-processing of the measuredsignals in the time domain, while the other isbased on chirp excitation and post-processing ofthe measured signals in the frequency domain.

The damage index based on pulse excitationhas demonstrated a great agreement with thevisual observations of crack initiation. It candetect the appearance of cracks but does notevolve anymore when crack is growing and sat-urates largely before the service limit. The dam-age index based on the chirp excitation is lesssensitive and more difficult to interpret. It doesnot saturate even close to the failure and shouldbe used with a lot of caution.

An issue concerning the influence of thesmart aggregates on the strength of the structurecan be raised. Their size are of the same orderof magnitude as the largest aggregates used inthe concrete. Although they are made of mor-tar, they are responsible for a certain mechani-cal heterogeneity in the structure that can causea local weakness which, in turn, could inflictadditional damage.

The damage indices proposed in this studyare very simple and based on empirical obser-

vations. Further studies will focus on the de-velopment of more complex signal processingstrategies, for which a deeper understanding ofthe wave propagation will be developed throughanalytical and numerical models.

REFERENCES[1] A. Deraemaeker and K. Worden,New

trends in vibration based structural healthmonitoring. CISM Lecture Notes Vol 520,Springer, 2010.

[2] H. Gu, G. Song, H. Dhonde, Y. L. Mo,and S. Yan, “Concrete early-age strengthmonitoring using embedded piezoelectrictransducers,”Smart Materials and Struc-tures, vol. 15, no. 6, pp. 1837–1845, 2006.

[3] S. Park, S. Ahmad, C.-B. Yun, and Y. Roh,“Multiple Crack Detection of ConcreteStructures Using Impedance-based Struc-tural Health Monitoring Techniques,”Ex-perimental Mechanics, vol. 46, no. 5,pp. 609–618, 2006.

[4] R. Tawie and H. Lee, “Monitoring thestrength development in concrete byEMI sensing technique,”Constructionand Building Materials, vol. 24, no. 9,pp. 1746–1753, 2010.

[5] A. A. Shah, Y. Ribakov, and S. Hirose,“Nondestructive evaluation of damagedconcrete using nonlinear ultrasonics,”De-sign, vol. 30, pp. 775–782, 2009.

[6] A. Shah and Y. Ribakov, “Effectiveness ofnonlinear ultrasonic and acoustic emissionevaluation of concrete with distributeddamages,”Materials & Design, vol. 31,no. 8, pp. 3777–3784, 2010.

[7] G. Song, H. Gu, Y. L. Mo, T. T. C.Hsu, and H. Dhonde, “Concrete struc-tural health monitoring using embeddedpiezoceramic transducers,”Time, vol. 16,pp. 959–968, 2007.

[8] J. Chaix, “Concrete damage evolutionanalysis by backscattered ultrasonic

8

Cedric Dumoulin,Grigoris Karaiskos and Arnaud Deraemaeker

waves,”NDT & E International, vol. 36,no. 7, pp. 461–469, 2003.

[9] C. Dumoulin, G. Karaiskos, J. Carette,S. Staquet, and A. Deraemaeker, “Moni-toring of the ultrasonic p-wave velocity inearly-age concrete with embedded piezo-

electric transducers,”Smart Materials andStructures, vol. 21, 2012. 047001.

[10] G. Song, H. Gu, and Y.-L. Mo, “Smartaggregates: multi-functional sensors forconcrete structuresa tutorial and a review,”Smart Materials and Structures, vol. 17,no. 3, 2008.

9