This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 143.248.122.199

This content was downloaded on 24/06/2015 at 05:25

Please note that terms and conditions apply.

Monitoring of pipelines in nuclear power plants by measuring laser-based mechanical

impedance

View the table of contents for this issue, or go to the journal homepage for more

2014 Smart Mater. Struct. 23 065008

(http://iopscience.iop.org/0964-1726/23/6/065008)

Home Search Collections Journals About Contact us My IOPscience

Monitoring of pipelines in nuclear powerplants by measuring laser-based mechanicalimpedance

Hyeonseok Lee, Hoon Sohn, Suyoung Yang and Jinyeol Yang

Department of Civil and Environmental Engineering, Korea Advanced Institute of Science andTechnology, Daejeon, 305-701, Korea

E-mail: hoonsohn@kaist.ac.kr

Received 3 December 2013, revised 10 February 2014Accepted for publication 24 March 2014Published 17 April 2014

AbstractUsing laser-based mechanical impedance (LMI) measurement, this study proposes a damagedetection technique that enables structural health monitoring of pipelines under the hightemperature and radioactive environments of nuclear power plants (NPPs). The applications ofconventional electromechanical impedance (EMI) based techniques to NPPs have been limited,mainly due to the contact nature of piezoelectric transducers, which cannot survive under thehigh temperature and high radiation environments of NPPs. The proposed LMI measurementtechnique aims to tackle the limitations of the EMI techniques by utilizing noncontact laserbeams for both ultrasound generation and sensing. An Nd:Yag pulse laser is used for ultrasoundgeneration, and a laser Doppler vibrometer is employed for the measurement of thecorresponding ultrasound responses. For the monitoring of pipes covered by insulation layers,this study utilizes optical fibers to guide the laser beams to specific target locations. Then, anoutlier analysis is adopted for autonomous damage diagnosis. Validation of the proposed LMItechnique is carried out on a carbon steel pipe elbow under varying temperatures. A corrosiondefect chemically engraved in the specimen is successfully detected.

Keywords: laser-based mechanical impedance (LMI), laser ultrasound, structural healthmonitoring (SHM), nuclear power plant (NPP), damage detection

(Some figures may appear in colour only in the online journal)

1. Introduction

Nuclear power plants (NPPs) have long been considered to beone of the safest and secure sources of electricity that cancope with the fast-growing power demands. However, it hasbeen reported that many of the current NPPs have been inoperation beyond their initial design lifespan. Furthermore,many countries are facing growing public opposition to NPPfacilities after the Fukushima nuclear disaster in 2011. Inresponse to mounting concerns over NPPs, nuclear regulatoryauthorities in many countries have tightened their main-tenance and inspection standards for operational NPPs.

Conventionally the inspection of pipelines in NPPs isbased on periodic nondestructive testing (NDT). NDT tech-niques are widely accepted by the nuclear industry, and manycommercialized NDT products are available (ASTM Com-mittee E07 2009, ISO Technical Committee ISO/TC17 2011). However, NDT often requires a periodic overhaulof the whole NPP facility, reducing its power productionefficiency. Furthermore, NDT needs to be performed bycertified engineers, so inspection can be labor-intensive andtime consuming. Another issue is that pipelines coated withinsulation materials or buried underground cannot be easilyaccessed (Adams 2007).

0964-1726/14/065008+10$33.00 © 2014 IOP Publishing Ltd Printed in the UK1

Smart Materials and Structures

Smart Mater. Struct. 23 (2014) 065008 (10pp) doi:10.1088/0964-1726/23/6/065008

For these reasons, the nuclear industry has a keen interestin developing structural health monitoring (SHM) techniquesas a vital supplement to the existing NDT techniques. Com-pared to NDT, SHM can provide the following potentialbenefits for NPP inspections: (1) automated and continuousmonitoring; (2) reduction in overhaul frequency, cost andlabor; (3) monitoring of hidden critical spots (Sohn et al 2004,Inman 2001). However, SHM techniques require the perma-nent installation of distributed sensors that can tolerate thehigh temperature and radiation inside NPPs. In addition, SHMtechniques are normally less accurate than NDT techniques inquantifying damage levels, and are vulnerable to false alarmsinduced by operational variations.

Among the various SHM techniques, ultrasonic techni-ques have gained prominence because of their long inspectionrange and responsiveness to small damage (Yang andKundu 1998, Kessler et al 2002, Giurgiutiu 2008). Piezo-electric materials are widely used for ultrasound generationand sensing, and wafer-type piezoceramic materials are par-ticularly popular for SHM applications because of theirlightweight and nonintrusive natures. However, the integrityof the piezoceramic materials and bonding layers can becompromised over time and under high temperature

conditions such as those inside NPPs. Recently, piezoelectrictransducers which can withstand high temperatures of up to350–500 °C have been developed (Zhang et al 2005,Kobayashi et al 2009, Shih et al 2010). However, piezo-electric materials can still easily be crystallized and damagedunder radiation exposure. Finally, long-range cablings oftransducers installed on distributed pipes are susceptible tohigher thermal noises, electro-magnetic interference and sig-nal attenuation (Wu and Ro 2004, Lee et al 2010).

In order to overcome the limitations of piezoelectrictransducers, laser ultrasound techniques have been explored(Saint-Pierre et al 1996, Su et al 2006). Previous applicationsof laser ultrasound have been focused on damage detection ofmetal and composite plate structures (Spicer et al 1990, Pierceet al 1997, Wu and Liu 1999, Silva et al 2003, Gaoet al 2003, Kim et al 2006, Sohn et al 2011). Severalresearchers have also applied laser ultrasound for damagedetection of pipes by measuring pipe thickness and ultrasoundvelocity (Clorennec et al 2002, Nishino et al 2004, Yashiroet al 2008, Lee and Lee 2009). A laser ultrasound systemtailored for NPPs has been developed using embeddableoptical fibers and fixing devices, and it survived well in a hightemperature environment (Lee et al 2012). This system is

Smart Mater. Struct. 23 (2014) 065008 H Lee et al

2

Figure 1. Overall schematic diagram of the proposed LMI measurement. (a) Generation of ultrasound: thermal expansion induced by a pulselaser irradiation generates structural deformation, and generates elastic waves. (b) Formation of LMI responses: the generated ultrasoundinitially produces wave propagation but eventually converges to structural vibration, which is equivalent to LMI. (c) Measurement of LMIresponses: the LMI responses are measured using a laser vibrometer.

optimized for the generation and sensing of guided waves,and the damage detection is limited to straight pipes.

The main objective of this study is to measure mechan-ical impedance (MI) by using a laser ultrasound system. AnNd:Yag pulse laser is used to generate ultrasound, and a laservibrometer is used to measure the corresponding LMI. Opticalfibers are used to transmit laser beams from the laser sourcesto target points. In addition, this study suggests an autono-mous damage detection technique based on outlier analysis sothat structural damage can be detected under changing tem-perature conditions. The feasibility of the proposed ultrasoundsystem and LMI for detecting corrosion damage is verifiedthrough lab-scale experiments. The proposed study has thefollowing uniqueness and advantages over the existing tech-niques: (a) the proposed laser ultrasound system does notrequire the attachment of contact-type transducers or adhesivematerials to a target pipe surface; (b) it can measuremechanical impedance in a fully optical way independent ofthe electrical and mechanical characteristics of conventionalpiezoelectric transducers; and (c) it can be applied under hightemperature and high radiation.

This paper is organized as follows. First, the workingprinciples of the LMI measurement and the formation of theLMI response are explained in section 2. Then, an autono-mous damage detection technique, which can compensate forsignal changes caused by temperature changes, is introducedin section 3. Corrosion diagnosis performed on an elbow pipeunder changing temperature conditions is reported in sec-tion 5. Finally, this paper concludes with a summary anddiscussions in section 6.

2. Principles of LMI measurement

Figure 1 shows the overall schematic diagram of the proposedLMI measurement. The first step of the entire process is thegeneration of ultrasound (figure 1(a)). When a solid surface ofa target structure is illuminated by a high-power pulse laser, alocalized temperature gradient is formed on an infinitesimalarea of the surface. The temperature gradient induces elasticexpansion and compression over the illuminated area actingas an ultrasound source. Then, ultrasound can be generated onthe target structure. Here, the laser duration, power level andbeam size need to be carefully tuned and maintained togenerate elastic waves without ablating the surface. Ablationtypically occurs for metals with a power density above8 · 107W cm−2 for carbon steels, but this value variesdepending on their material properties as well as surfaceconditions (Pierce et al 1998).

The principle of LMI formation is shown in figure 1(b).The generated ultrasound initially travels forward from theexcitation point and is reflected back from the structuralboundaries. Later, the superposition of these forwarding andreflected waves converges to structural vibration (i.e. LMI).The LMI response includes a structural vibration response inthe frequency domain. Since this response has relatively high-frequency components, up to hundreds of kHz as is the casein EMI, it enables one to obtain vibration characteristics andtheir changes with respect to small-size damage (Parket al 2003, Bhalla et al 2009).

For the measurement of the LMI response, this studyutilizes a laser Doppler vibrometer (figure 1(c)). The Dopplervibrometer uses a continuous wave (CW) laser as a power

Smart Mater. Struct. 23 (2014) 065008 H Lee et al

3

Figure 2. Interaction between a pulse laser beam and a target structure.

source. When the continuous laser beam is pointed at avibrating object and scattered back from the vibrating object,the phase of the reflected laser beam is shifted with respect tothe incident laser beam due to the Doppler effect. The fre-quency and phase modulation recovers the velocity infor-mation as well as the displacement from the phase shifting ofthe reflected laser beam (Kundu 2004, Staszewski et al 2004,Castellini et al 2006). Since the laser vibrometer used in thisstudy only measures the velocity component in an out-of-plane direction, the measured LMI responses contain onlyflexural vibration modes.

There has been a plethora of research work to simulatemechanical forces induced by the pulse laser beam. In pre-vious papers, modeling of laser generated ultrasound hasfocused on simulating the laser source as a combination ofmechanical stresses and boundary conditions (Scruby andDrain 1990). This equivalent stress distribution can berepresented as a point dilatation at the surface, and differentapproaches has been developed to derive more sophisticated

expressions for laser-generated ultrasound. Several modelingcases include the epicentral waveform (Telschow and Con-ant 1990), the surface acoustic waveform (Hurley et al 1998),and the directivity pattern (Davies et al 1993, Bernstein andSpicer 2000) in a thermo-elastic regime.

Among the various mechanical models used in describ-ing the laser-generated ultrasound, the analytical solution ofLMI in this study is derived using a simple flexural vibrationmodel. The laser-induced excitation force is modeled with apair of dipole forces, and the analytical solution of theresulting flexural vibration is formulated. Figure 2 shows aschematic diagram for interaction between a pulse laser beamand a 1D beam structure. When a pulse laser beam with abeam diameter of la is incident on a surface near xa, a laser-induced dipole force FL is generated on the surface. Here, FL

generates an axial force Na and an excitation moment Ma tothe target structure. Then, a frequency response function(FRF), ωH ( ), for flexural vibration at a sensing point xs

Smart Mater. Struct. 23 (2014) 065008 H Lee et al

4

Figure 3. A schematic diagram of the proposed outlier analysis.

becomes (Inman 2001, Giurgiutiu 2008):

∑

ω ωρ

ω ξω ω ω

= ˆˆ = −

×− ′ + ′ +

+ −=

⎜ ⎟⎛⎝

⎞⎠

( ) ( ) ( )

Hu

F A

h

X x X x l

iX x

( )2

2(1)

L

n

N a a a

n nn s

12 2

where u is the amplitude of the out-of-plane velocity at xs, FL

is the amplitude of the dipole force at xa, X x( )n is the nthorthonormal flexural mode shape, ρ is the material density, Ais the cross-sectional area of the beam structure, ξ is thedamping ratio and ω is the angular velocity. For a simplysupported beam with a length L, Young’s modulus E and

moment of inertia I =( )bh 123 , X x( )n and ωn can be repre-sented as follows.

π

ω πρ

=

= ⎜ ⎟⎛⎝

⎞⎠

X xn

Lx

n

L

EI

A

( ) sin ,

(2)

n

n

2

3. Outlier analysis

For the development of autonomous damage diagnosis, adamage index (DI) first needs to be identified. This studydefines a damage index based on a maximum cross-correla-tion between two LMI response signals as shown in equation(3). The motivation for using the cross-correlation value isthat the shape of the LMI signal changes with the presence ofa structural defect while temperature variation mainly pro-duces shifting of the LMI resonance peaks without shapechanges.

= − ( )DI corr X X1 max , (3)ij i j

∑ω ω ω

σ σ=

−

+ ˜ − ¯ − ¯

ωω

˜

⎧⎨⎪

⎩⎪

⎫⎬⎪

⎭⎪( )( )

( )( ) ( )

corr X X

N

X X X X

max ,

max1

1

i j

i i j j

X Xi j

where ωX ( )i and ωX ( )j are two distinctive LMI signals, Xi is

the mean of Xi, σXiis the standard deviation of Xi, N is the

total number of data points in Xi and Xj, and ω is the fre-

quency shift. The DI value will be close to zero if the shapesof the two LMI signals are similar, but its value will increaseand approach one if their shapes become different.

In field applications, temperature variations can alsoproduce changes in the LMI signals and consequentlyincrease the DI value. As mentioned previously, the proposedDI based on a correlation analysis alleviates the effect oftemperature variations, but additional processing is stillnecessary to further minimize temperature-induced false

alarms. Figure 3 shows a brief explanation of the proposedoutlier analysis to diminish the temperature effect.

(1) Collect training data from varying temperature conditionsof the intact structure.

(2) Compute DI values for all possible pairs of LMI signalsin the training data set.

(3) Fit a statistical distribution to the DI values obtained fromthe previous step, and compute a threshold correspondingto a user-specified one-sided confidence level.

(4) Measure a new LMI test signal from an unknowncondition of the structure.

(5) Compute DI values between the test signal and all thesignals in the training data set, and the minimum DIvalues among all computed DI values are assigned as thefinal DI value of the specific test signal.

(6) Determine whether the test LMI signal is from an intactor damaged state by comparing the DI value with thethreshold value.

Because the DI value obtained for a damaged state oftenfalls near the tail of the DI distribution, a general extremevalue (GEV) distribution is used to describe the distributionof the DI as follows.

μ σ ξσ

= ξ+ −( )f x t x e; , ,1

( ) (4)t x1 ( )

μσ

ξ ξ ξ

ξ

=+ − ≠

=μ σ

−

− −

⎜ ⎟

⎧

⎨⎪⎪

⎩⎪⎪

⎛⎝⎜

⎛⎝

⎞⎠

⎞⎠⎟

⎜ ⎟⎛⎝

⎞⎠

( )( )

t x

x

e

( )1

1

0

0x

where f is the density function of the GEV distribution and μ,σ and ξ are the location, scale and shape parameters estimatedfrom the data, respectively. Figure 4 shows the comparisonbetween the cumulative density function (CDF) of DI valuesfrom the actual intact-state data and the analytical curve of theGEV distribution. The result confirms that the DI values arewell subjected to the estimated GEV distribution. A good-ness-of-fit test using the Kolmogorov–Smirnov (K–S) test is

Smart Mater. Struct. 23 (2014) 065008 H Lee et al

5

Figure 4. Comparison of the cumulative density function (CDF)from the actual intact-state data and the estimated GEV distribution.

employed to test if the GEV distribution properly describesthe characteristics of the DI values obtained from the intactconditions. A threshold corresponding to a one-sided 99.7%(3σ) confidence level is obtained.

4. Experimental setup

Figure 5 shows the overall configuration of the experimentalsetup. The test is designed to simulate the structural andoperational conditions of a typical secondary coolant systemin an NPP. The target specimen (figure 6) is a carbon steel andseamless-type pipe elbow (KS D 3562) commonly used forhigh-pressure piping systems in NPPs. The outer diameter ofthe pipe is 114.3 mm, and the wall thickness is 4 mm. Thecenter-to-end distance of the pipe elbow is 152.4 mm, and thedistance between the ultrasound generation and sensing pointsis 329.2 mm. A corrosion defect with increasing depths of0.3 mm, 0.6 mm, 0.9 mm and 1.2 mm is introduced on theouter surface at the center of the pipe elbow using 50 mL of7 mol sulfuric acid solution. At each corrosion stage, the LMIsignals are obtained and the thickness at the corrosion area ismeasured using an ultrasonic thickness gauge (SC20,Picosonic).

Training LMI signals are collected from 20 °C to 300 °Cwith a 20 °C interval for the intact state, and test data from50 °C, 110 °C, 170 °C, 230 °C and 290 °C. By winding astrip-type band heater (GCTC-DP, Global Lab) around thepipe, the pipe temperature is controlled and the pipe surfacetemperature is measured by a thermocouple attached to thepipe specimen.

As mentioned in the previous chapters, the excitation ofthe structural vibration response, LMI, is a key factor in theactual detection performance. The pipe elbow specimen usedin this study is small and fairly well isolated so that excitationof the entire structure is relatively easy, resulting in globalvibration responses. In real NPPs, the pipe elbow componentwill be attached to a further pipe network with fittings andinternal flowing fluid. Therefore, further study is underway toexperimentally validate the proposed system and techniqueusing a small-scale NPP test bed.

Figure 7 shows the configuration of the optical fiberguided ultrasound generation unit. First, the Nd:Yag pulselaser (Brilliant Ultra, Quantel) emits a 532 nm laser beam at arepetition rate of 20 Hz. The excitation energy per pulse isaround 10 mJ, and the pulse duration is 8 ns. The powerdensity is set to 0.75MW cm−2 to avoid ablation on thesurface. The laser intensity profile at the laser source follows aGaussian distribution in the spatial domain, and this profile isconverted to a flat top at the entrance of the optical fiber usingMicrolens arrays (Edmund, P64-478) to avoid fiber damage.Then, the laser beam passes through a convex lens (Thorlabs,LA1805-YAG) with a focal length of 50 mm before enteringthe SMA connector at the entrance of the optical fiber. Amultimode optical fiber (Thorlabs, FT1000EMT) with a corediameter of 1000 μm is used for high-power laser transmis-sion. The surface of the optical fiber is metal coated for long-term resistance for high-temperature and radiation conditions.The other end of the optical fiber on the specimen side has ametal tube made of stainless steel and aluminum. The mainfunction of this metal tube is to minimize the heat transferfrom the specimen to the SMA connector. Finally, the dia-meter of the laser beam becomes 7–8 mm on the surface.

Figure 8 shows the configuration of the optical fiberguided ultrasound measurement unit. A phase-modulationfiber-coupled laser Doppler vibrometer (Polytec, OFV-551) isused for out-of-plane velocity measurements. This vibrometerhas a He–Ne continual laser source with 633 nm wavelength,15W maximum power, and 16 μm beam size. Since the laserbeam emitted from the laser vibrometer is less powerful thanthe pulse laser beam used for ultrasound generation, a singlemode fiber with a core diameter of 10 μm is used for the laserbeam transmission. This single mode fiber is made of apolarization maintaining (PM) fiber, which can reduce thepower loss induced by the polarization direction. At the endof the optical fiber on the specimen side, a metal tube made ofstainless steel and aluminum is installed similarly to theultrasound generation unit. Here, a focusing lens with a focallength of 15 mm is inserted into the metal tube. Additionally,a KGI filter is placed at the vibrometer side to prevent theinfrared heat radiation from entering the vibrometer sensorhead.

There are additional issues that need to be considered forNPP applications. First, many pipes inside NPPs are covered

Smart Mater. Struct. 23 (2014) 065008 H Lee et al

6

Figure 5. Overall configuration of the experimental setup.

Figure 6. Dimensions of the target pipe specimen and the corrosiondefect (all dimensions in mm).

with insulation materials to minimize the heat loss in thewater flow system. The insulation materials used in NPPs aretypically composed of two layers, and the thickness of eachlayer varies from 20 mm to 100 mm. For the ultrasoundmeasurement of covered pipes, a stainless steel strip can beused to anchor the metal tubes of the optical fiber guidedultrasound units directly to a target pipe inside the insulationlayers (Lee et al 2012). The authors have demonstrated inprevious research work that comparison of these two ultra-sound signals with and without the insulation materialsreveals only a marginal effect on the signal changes (Yanget al 2012).

Another consideration is radiation for primary coolingsystems. Radiation hardening tests are performed using a Co-60 (364 000 Ci) source on the optical fiber systems used inthis study. Figure 9 shows the gamma radiation facility atKorea Atomic Energy Research Institute (KAERI). A radia-tion dose rate of 20 kGy h−1 is used, and a total of 125 kGy,which is equivalent to a typical 10 year dose in an operatingNPP, are exposed to the fibers. The experiments verified thatthere were neither changes in the signal waveforms nordamage to the fiber and only a little (1.28 dB) signalattenuation.

5. Experimental results

Figure 10 compares wave propagation and structural vibrationresponses in the time domain. At the initial stage of laserultrasound excitation, incident wave propagations occur as

Smart Mater. Struct. 23 (2014) 065008 H Lee et al

7

Figure 7. Configuration of the optical fiber guided ultrasound generation unit (units: mm).

Figure 8. Configuration of the optical fiber guided ultrasound measurement unit.

Figure 9. Gamma radiation facility at Korea Atomic EnergyResearch Institute (KAERI).

shown in figure 9(a). Considering the material properties usedin this study, the velocities of the pressure (P) and shear (S)waves are 6311 m s−1 and 3070 m s−1, respectively. Theincident wave from the direct path between the signal gen-eration and sensing points is well identified by using thearrival times of P and S waves. With the advent of subsequentwaves from helical paths and reflections, superposition ofthese waves generates structural vibration responses (i.e.LMI) over the entire region of the structure as shown infigure 10(b). The time duration of the transient wave propa-gation is much shorter compared with that of the structuralvibration. The structural vibration response starts when thewaves from the direct path are reflected at least twice at theedge boundaries (i.e. 0.83 ms)

Figure 11 shows the variation of the LMI responses withrespect to temperature. Each signal is normalized with respectto its maximum peak value within the frequency range. TheLMI signals are measured at 100 °C, 200 °C and 300 °C fromthe intact state. For a closer investigation, the LMI signals arezoomed in for the frequency range of 58–61 kHz in figure 11.As the temperature increases, the peaks of the LMI responsesare shifted to the left with little change of the shape.

Because significant response signal changes can beobserved due to temperature even without any structuraldefects, an outlier analysis mentioned in the previous chapterneeds to be taken for reliable damage diagnosis. There havebeen several different techniques to compensate for the signalchanges induced by temperature variations. Baseline sub-traction methods (Lu and Michaels 2005, Konstantinidiset al 2006, Croxford et al 2007) have identified structuraldamage by subtracting the test signal from the baseline signalafter correcting the effect of temperature on the group velo-cities of specific wave modes. The baseline signals are mea-sured at a varying range of temperatures and the baseline withthe smallest error compared to the test signal is then selected.

Figure 12 compares LMI responses obtained from theintact and two corrosion cases (0.6 mm and 1.2 mm deepcorrosion) at 100 °C. Each signal is also normalized with

respect to its maximum peak value. Unlike for the previoustemperature experiments, the distortions of the resonancepeak shapes are observed.

For the establishment of a threshold, 15 training LMIsignals sets are obtained from the intact condition undervarying temperature conditions of 20 °C to 300 °C with a20 °C interval. Note that 10 LMI signals are obtained for eachdata set corresponding to a specific temperature. Therefore, atotal of 150 LMI signals are measured from 15 different

Smart Mater. Struct. 23 (2014) 065008 H Lee et al

8

Figure 10. Response signals in the time domain (temperature at 20 °C).

1.0

0.8

0.6

0.4

0.2

058 59 59.5 60 60.5 6158.5

Frequency (kHz)

Nor

mal

ized

Am

plitu

de100°C200°C300°C

Figure 11. LMI responses obtained from 100 °C, 200 °C and 300 °Cof the intact specimen.

1.0

0.8

0.6

0.4

0.2

059.5 60 60.5 61

Frequency (kHz)

Nor

mal

ized

Am

plitu

de

Intact0.6 mm Corrosion1.2 mm Corrosion

Figure 12. Comparison of LMI responses obtained from the intactand two corrosion cases (0.6 mm and 1.2 mm deep corrosions) at100 °C.

temperature conditions, and 11175 DI values (=150C2) arecomputed. By fitting a GEV distribution to all DI values fromthe training data, the threshold corresponding to a one-sided99.7% confidence interval is set to 0.15.

Table 1 shows a list of test data sets obtained fromvarying temperatures (50 °C, 110 °C, 170 °C, 230 °C and290 °C) and damage conditions. The first 5 data sets(♯16–♯20) are obtained from the intact state, and an additional20 data sets (♯21–♯40) from the increasing corrosion cases(the corrosion depths varied from 0.3 mm to 0.6 mm, 0.9 mmand 1.2 mm). Again, note that each data set in the tableincluded 10 LMI signals obtained from the same condition.

Figure 13 shows the damage diagnosis of the test datasets listed in table 1. The DI value for each data set is theaverage of 10 DI values obtained from 10 LMI signals in eachdata set. No false alarms are indicated for data sets ♯16–♯20.On the other hand, DI values for the corrosion cases increasefar above the threshold value, indicating the presence ofcorrosion in the target specimen. The shape changes of LMIinduced by the corrosion seem to increase the DI value morethan those induced by temperature variations. However, theDI values do not monotonically increase as the corrosionprogresses due to the complex nature of structural vibration(i.e. LMI) in a pipe elbow.

6. Conclusion

This study develops a fiber guided laser ultrasound system togenerate and measure LMI responses for damage detection inNPP pipes, which are exposed to high temperature andradiation. Optical fibers with specially designed focusinglenses are used to deliver laser beams between the lasersources (i.e. an Nd:Yag pulse laser and CW laser) and a targetstructure. An Nd:Yag pulse laser shoots a broadband laserpulse and creates wave propagation. The initial wave propa-gation converges to structural vibration, which is equivalentto LMI. The LMI responses are obtained by measuring out-of-plane velocity of the target structure using a laser vib-rometer. The automated damage diagnosis is performed usingan outlier analysis. Corrosion damage in an elbow pipe issuccessfully detected even under varying temperatureconditions.

Several follow-up studies are warranted to improve theperformance of the proposed technique. First, the proposedLMI technique is only applied to a single elbow pipe speci-men in this study. Further studies are necessary to generalizethe findings in this study to other complex structures such aswelded pipes or T-joints. Also, only temperature variation isconsidered in the present study while the operational condi-tions of NPP pipes, such as ambient vibrations, pressure andfluid velocity inside the pipes, are also subject to change.

Acknowledgements

This research was supported by the Mid-career ResearcherProgram through the National Research Foundation of Korea(NRF) (No. 2010-0017456).

References

Adams D E 2007 Health Monitoring of Structural Materials andComponents (Chichester: Wiley)

ASTM Committee E07 2009 Standard Practice for UltrasonicTesting of Metal Pipe and Tubing (West Conshohocken, PA:American Society of Mechanical Engineers)

Bernstein J R and Spicer J B 2000 Line source representation forlaser-generated ultrasound in aluminum J. Acoust. Soc. Am.107 1352–7

Bhalla S and Soh C K 2004 High frequency piezoelectric signaturesfor diagnosis of seismic/blast induced structural damages NDT& E Int. 37 23–33

Castellini P, Martarelli M and Toamasini E P 2006 Laser Dopplervibrometry: development of advanced solutions answering totechnology's needs Mech. Syst. Signal Process. 20 1265–85

Clorennec D, Royer D and Walaszek H 2002 Nondestructiveevaluation of cylindrical parts using laser ultrasonicsUltrasonics 40 783–9

Croxford A J, Wilcox P D, Drinkwater B W and Konstantinidis G2007 Strategies for guided-wave structural health monitoringProc. R. Soc. Lond. A 463 2961–81

Davies S J, Edwards C, Taylor G S and Palmer S B 1993 Laser-generated ultrasound: its properties, mechanisms andmultifarious applications J. Phys. D: Appl. Phys. 26 329–48

Smart Mater. Struct. 23 (2014) 065008 H Lee et al

9

0.7

0.6

0.5

0.4

0.3

0.2

0.1

15 20 25 30 35 40

Data Set Number

DI

Threshold Level= 0.15

Intact Corrosion 1 Corrosion 2 Corrosion 3 Corrosion 4

Figure 13.Damage diagnosis under varying temperature and damageconditions.

Table 1. Test data sets obtained from varying temperature anddamage conditions. (The depths for corrosions 1, 2, 3 and 4 are0.3 mm, 0.6 mm, 0.9 mm and 1.2 mm, respectively.)

Data♯ State

Temp.(°C)

Data♯ State

Temp.(°C)

16 Intact 50 29 Corrosion 2 23017 Intact 110 30 Corrosion 2 29018 Intact 170 30 Corrosion 2 29019 Intact 230 31 Corrosion 3 5020 Intact 290 32 Corrosion 3 11021 Corrosion 1 50 33 Corrosion 3 17022 Corrosion 1 110 34 Corrosion 3 23023 Corrosion 1 170 35 Corrosion 3 29024 Corrosion 1 230 36 Corrosion 4 5025 Corrosion 1 290 37 Corrosion 4 11026 Corrosion 2 50 38 Corrosion 4 17027 Corrosion 2 110 39 Corrosion 4 23028 Corrosion 2 170 40 Corrosion 4 29029 Corrosion 2 230

Gao W, Glorieux C and Thoen J 2003 Laser ultrasonic study ofLamb waves: determination of the thickness and velocities of athin plate Int. J. Eng. Sci. 41 219–28

Giurgiutiu V 2008 Structural Health Monitoring with PiezoelectricWafer Active Sensors (Boston, MA: Elsevier)

Grisso B L and Inman D J 2008 Autonomous hardware developmentfor impedance-based structural health monitoring Smart Struct.Syst. 4 305–18

Hurley D H, Spicer J B, Wagner J W and Murray T W 1998Investigation of the anisotropic nature of laser-generatedultrasound in zinc and unidirectional carbon epoxy compositesUltrasonics 36 355–60

Inman D J 2001 Engineering Vibration (Upper Saddle River, NJ:Pearson Education)

ISO Technical Committee ISO/TC 17 2011 Non-destructive Testingof Steel Tubes ISO 10893 (Geneva: International Organizationfor Standardization)

Kim H, Jhang K, Shin M and Kim J 2006 A noncontact NDEmethod using a laser generated focused-Lamb wave withenhanced defect-detection ability and spatial resolution NDT &E Int. 39 312–9

Kessler S S, Spearing S M and Soutis C 2002 Damage detection incomposite materials using Lamb wave methods Smart Mater.Struct. 11 269–78

Kobayashi M, Jen C K, Bussiere J F and Wu K T 2009 High-temperature integrated and flexible ultrasonic transducers fornondestructive testing NDT & E Int. 42 157–61

Konstantinidis G, Drinkwater B and Wilcox P D 2006 Thetemperature stability of guided wave structural healthmonitoring systems Smart Mater. Struct. 15 967–76

Kundu T 2004 Ultrasonic Nondestructive Evaluation: Engineeringand Biological Material Characterization (London: CRCPress)

Lee J H and Lee S J 2009 Application of laser-generated guidedwave for evaluation of corrosion in carbon steel pipe NDT & EInt. 42 222–7

Lee H, Park H J, Sohn H and Kwon I B 2010 Integrated guided wavegeneration and sensing using a single laser source Meas. Sci.Technol. 21 105207

Lee H, Yang J and Sohn H 2012 Baseline-free pipeline monitoringusing optical fiber–guided laser ultrasonics Struct. HealthMonitor. 11 684–95

Lu Y and Michaels J E 2005 A methodology for structural healthmonitoring with diffuse ultrasonic waves in the presence oftemperature variations Ultrasonics 43 717–31

Nishino H, Takemoto M and Chubachi N 2004 Estimating thediameter/thickness of a pipe using the primary wave velocityof a hollow cylindrical guided wave Appl. Phys. Lett. 851077–9

Park S, Lee C and Sohn H 2010 Reference-free crack detection usingtransfer impedance J. Sound Vib. 329 2337–48

Pierce S G, Culshaw B, Philp W R, Lecuyer F and Farlow R 1997Broadband Lamb wave measurements in aluminium andcarbon/glass fibre reinforced composite materials using non-contacting laser generation and detection Ultrasonics 35105–14

Pierce S G, Culshaw B and Shan Q 1998 Laser generation ofultrasound using a modulated continuous wave laser diodeAppl. Phys. Lett. 72 1030–2

Rens K, Wipf T and Klaiber F 1997 Review of nondestructiveevaluation techniques of civil infrastructure J. Perform.Constr. Facil. 11 152–60

Saint-Pierre N, Jayet Y, Perrissin-Fabert I and Baboux J C 1996 Theinfluence of bonding defects on the electric impedance ofpiezoelectric embedded element J. Phys. D: Appl. Phys. 292976–82

Scruby C B and Drain L E 1990 Laser Ultrasonics: Techniques andApplications (New York: Taylor & Francis)

Shih J L, Kobayashi M and Jen C K 2010 Flexible metallicultrasonic transducers for structural health monitoring of pipesat high temperatures IEEE Trans. Ultrason. Ferroelectr. Freq.Control 57 2103–2010

Silva M Z, Gouyon R and Lepoutre F 2003 Hidden corrosiondetection in aircraft aluminum structures using laser ultrasonicsand wavelet transform signal analysis Ultrasonics 41 301–5

Sohn H, Farrar C R, Francois M H, Jerry J C, Devin D S,Daniel W S and Brett R N 2004 A review of structural healthmonitoring literature: 1996-2001 Los Alamos NationalLaboratory Report LA-13976-Ms

Sohn H, Dutta D, Yang J, DeSimio M P, Olson S E andSwenson E D 2011 Automated detection of delamination anddisbond from wavefield images obtained using scanning laservibrometer Smart Mater. Struct. 20 045107

Spicer J B, McKie A D W and Wagner J W 1990 Quantitative theoryfor laser ultrasonic waves in a thin plate Appl. Phys. Lett. 571882–4

Staszewski W J, Lee B C, Mallet L and Scarpa F 2004 Structuralhealth monitoring using scanning laser vibrometry: I. Lambwave sensing Smart Mater. Struct. 13 251–60

Su Z, Ye L and Lu Y 2006 Guided Lamb waves for identification ofdamage in composite structures: a review J. Sound Vib. 295753–80

Telschow K L and Conant R J 1990 Optical and thermal parametereffects on laser-generated ultrasound J. Acoust. Soc. Am. 881494–502

Wachel J C, Morton S J and Atkins K E 1990 Piping vibrationanalysis Proc. 19th Turbomachinery Symp. (College Station,TX, Sept. 1990) pp 119–34

Wu T and Liu Y H 1999 On the measurement of anisotropic elasticconstants of fiber-reinforced composite plate using ultrasonicbulk wave and laser generated Lamb wave Ultrasonics 37405–12

Wu T and Ro P I 2004 Dynamic peak amplitude analysis andbonding layer effects of piezoelectric bimorph cantileversSmart Mater. Struct. 13 203–10

Yang J, Lee H and Sohn H 2012 An optical fiber guided ultrasonicexcitation and sensing system for online monitoring ofnuclear power plants Rev. Prog. Quant. Nondestruct. Eval. 311640–7

Yang W and Kundu T 1998 Guided waves in multilayerd plates forinternal defect detection J. Eng. Mech. 124 311–27

Yashiro S, Takatsubo J, Miyauchi H and Toyama N 2008 A noveltechnique for visualizing ultrasonic waves in general solidmedia by pulsed laser scan NDT & E Int. 41 137–44

Zhang S, Xia R, Lebrun L, Anderson D and Shrout T R 2005Piezoelectric materials for high power, high temperatureapplications Mater. Lett. 59 3471–5

Smart Mater. Struct. 23 (2014) 065008 H Lee et al

10