Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

This article was downloaded by: [George Mason University]

On: 26 December 2014, At: 14:57


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Nondestructive Testing and Evaluation


Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/gnte20

The assessment of the compressive


strength and thickness of concrete
structures using nondestructive testing
and an artificial neural network
a a b
Young S. Cho , S.U. Hong & Moon Sung Lee
a
Department of Architectural Engineering , Hanyang University ,
Ansan-Si, Kyunggi-Do, Republic of Korea
b
BK21 Architectural Engineering Program , Hanyang University ,
Seongdong-gu, Seoul, Republic of Korea
Published online: 30 Sep 2009.

To cite this article: Young S. Cho , S.U. Hong & Moon Sung Lee (2009) The assessment of
the compressive strength and thickness of concrete structures using nondestructive testing
and an artificial neural network, Nondestructive Testing and Evaluation, 24:3, 277-288, DOI:
10.1080/10589750802375970

To link to this article: http://dx.doi.org/10.1080/10589750802375970

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Downloaded by [George Mason University] at 14:57 26 December 2014
Nondestructive Testing and Evaluation,
Vol. 24, No. 3, September 2009, 277–288

The assessment of the compressive strength and thickness of concrete


structures using nondestructive testing and an artificial neural network
Young S. Choa*, S.U. Honga and Moon Sung Leeb
a
Department of Architectural Engineering, Hanyang University, Ansan-Si, Kyunggi-Do, Republic of
Korea; bBK21 Architectural Engineering Program, Hanyang University, Seongdong-gu, Seoul,
Republic of Korea
(Received 27 April 2008; final version received 29 June 2008)
Downloaded by [George Mason University] at 14:57 26 December 2014

This study focuses on the prediction of concrete compressive strength and the unknown
thickness of concrete structures as a partial development of a concrete assessment system.
The nondestructive tests (NDTs), impact-echo method and spectral analysis of surface
wave method were applied to predict concrete compressive strength for the correlation
between NDT results and cylinder tests results. The concrete strength prediction and the
measurement of thickness were effectively achieved by using an artificial neural network
technology. As actual problems were tested in the neural network system, good agreement
between the results from the cylinder test and the results from the neural network run was
achieved. The accuracy in measuring the thickness of the specimen was successfully
achieved using the same technology. The result of this study is a basic algorithm for the
automation of predicting the compressive strength and the thickness of concrete member.
Automation of these results can contribute to predict the accurate form removal time and
unknown slab thickness in building construction practice.
Keywords: structural assessment system; neural network; nondestructive test; impact-
echo method; SASW method

1. Introduction
Nondestructive tests (NDTs) are defined as tests of concrete properties without resulting in
a significant destruction in the structure members. These methods are widely used in
quality assurance, the troubleshooting of problems within new construction, condition
evaluations of older concrete for rehabilitation purposes, and quality assurance of concrete
repairs. Because of convenience and advances in data processing, NDT methods are
evolving and new test techniques are continuing to be developed.
Recently, the impact-echo (IE) and spectral analysis of surface wave (SASW) [1] stress
wave propagation methods have been suggested to rapidly evaluate concrete or soil
conditions. These methods may allow a comprehensive, economical assessment of existing
buildings, without coring or causing other damage to the structures. Unfortunately, the
nature of stress wave propagation in layered systems is very complex. Therefore, a number
of simplifying methods and assumptions are made in interpreting the signal process
obtained from the experiments. In this paper, the concrete strength prediction and
measurement of the thickness of slab were effectively achieved by using artificial neural
network (ANN) technology [2]. Good agreement between the results from the cylinder test

*Corresponding author. Email: ycho@hanyang.ac.kr

ISSN 1058-9759 print/ISSN 1477-2671 online


q 2009 Taylor & Francis
DOI: 10.1080/10589750802375970
http://www.informaworld.com
278 Y.S. Cho et al.

and the results from the ANN was observed showing average determination coefficients of
0.9 in ANN-I and 0.95 or larger in ANN-II. The accuracy in measuring the thickness of the
specimens was successfully achieved using the neural network technique within the error
limit of 0.6%. This analysis will be used as a part of a concrete assessment system.

2. Background research
Recent research activities in NDTs such as IE method and SASW method are summarised
as follows. Erdem [3] studied transient elastic response of slender graphite rods to detect
cracks developed in the rods. Theoretical and numerical studies were conducted to detect
cracks developed in the slender graphite rods. Sansalone et al. (2006) studied sampling
techniques for evaluating large concrete structures [4,5]. The brief description of sampling
methods is given, and the procedure has been applied to a post-tensioned bridge which was
Downloaded by [George Mason University] at 14:57 26 December 2014

nondestructively tested to locate air voids within the grouted tendon ducts. Zomorodian
and Hunaidi [6] studied inversion of SASW dispersion curves based on maximum
flexibility coefficients in the wave number domain. Simplex method is used to match field
and theoretical dispersion curves automatically. Kim et al. [7] studied feasibility study of
the IE – SASW methods for nondestructive evaluation of containment building structures
in nuclear power plants. The research was to detect thickness of concrete structures, voids
inside concrete structures using IE –SASW methods, and detect reinforcement bar
location using the ground penetration radar (GPR). Hunaidi [8] studied evolution-based
genetic algorithms for analysis of nondestructive surface wave tests on pavements. Search
technique employing genetic algorithm is applied based on the principles of natural
evolution. Tsai et al. [9] studied application of neural networks to laser ultrasonic NDE of
bonded structures. The purpose of the study is to develop ANNs for the nondestructive
evaluation of a bounded structure using laser generated surface waves.
To design the neural network, 48 specimens with different concrete material properties
and strengths and 2 specimens with 2 concrete strengths and 6-slab thicknesses were
examined using IE and SASW NDT methods. The NDT methods, IE and SASW, were
applied to predict concrete compressive strength and flaw location within the concrete.
Three types of the neural network, ANN-I/II and ANN-III, were designed to predict
concrete strength and thickness using the artificial technique [10]. Another four specimens
were used to evaluate the neural network training performance.
Good agreement was achieved between the results from the test and the results from
the neural network system. For the prediction of concrete strength, an average accuracy
greater than 90% was achieved using ANN-I and ANN-II. The results from the ANN-III
showed good accuracy at predicting the concrete strength and thickness of slabs with 0.6
and 4% errors, respectively. Using the neural network obviously resulted in better
predictions of the thickness and concrete strength compared to the approximate formula,
which is the traditional method of NDT used to obtain results.

3. Stress wave propagation method for structures


Along with visual inspection, the use of an acoustic method is the oldest form of a NDT
method. Essentially, this method is based on stress wave propagation properties, which
indicate the strength and flaws in concrete. Several test methods based on stress waves
were developed from the 1930s including the ultrasonic pulse method, the IE method, and
the SASW method. Due to the limitation of the depth range that can be detected by the
ultrasonic pulse test method, low frequency stress wave methods, SASW and IE methods,
were increasingly used with a hammer or steel ball on the concrete or soil surface.
Nondestructive Testing and Evaluation 279

3.1 IE test method


The IE method as shown in Figure 1 is a NDT method which uses transient stress waves to
detect damage in the structures and measure the free boundary within the concrete structures.
When an impact is applied on a testing surface like concrete, the surface of concrete causes a
P-wave (compression wave), VP, a S-wave (shear wave), Vs, which propagates along spherical
wave fronts through the structure, and a Rayleigh (surface wave), Vr, which propagates along
circular wave fronts on the surface of the structure. The P- and S-waves are reflected from
internal flaws and external free boundaries. The accelerations caused by the arrival of these
stress waves are collected using a transducer near the location of impact and are used to
compute the depth of free boundaries inside the structures and external free boundaries. In the
IE test, the surface accelerations which were caused by compression waves form a resonance
condition and show the maximum amplitude [11].
Downloaded by [George Mason University] at 14:57 26 December 2014

The surface accelerations caused by reflecting stress waves can be collected on a time
domain basis in acceleration wave forms. The results can be more easily analysed in the
frequency domain with the amplitude spectrum computed from the time domain waveform
using a fast Fourier transform method. As the stress wave repeatedly travels from the source to
the free boundaries and back to the surface, a transient resonant condition forms, and damages
and free boundaries traditionally show up as highest amplitudes at frequencies corresponding
to their depths in the amplitude spectrum. The depth of a free boundary within a concrete
structure or an external free boundary can be determined using the following equation (1):
Vp
H¼a ; ð1Þ
2f
where H is the free boundary depth, Vp the P-wave speed within the concrete, a the correction
factor, and f the frequency due to reflection of the P-wave from the free boundaries.

3.2 SASW method


Unlike the compressive wave velocity, which is significantly affected by water content and
the degree of saturation within the material, shear wave velocity, Vs, is mainly governed

Figure 1. Schematic of IE method.


280 Y.S. Cho et al.

by properties of the material and is closely related to the shear modulus of the material.
And furthermore, according to plane elastic wave theory, the plane Rayleigh wave
velocity, Vr, is related to the shear wave velocity, Vs, of the material by Poisson’s ratio, n.
Therefore, if the Vr of an uniform layer was obtained through the Rayleigh wave’s
dispersion curve, the shear wave velocity, Vs, can be quickly obtained by the inversion of
the dispersion curve. The schematic of the SASW method is shown in Figure 2.
The SASW method consists of three steps: an in situ test, an evaluation of the
dispersion curve using a phase unwrapping method, and the determination of the shear
wave velocity profile by an inversion process. Through an iterative process, the
experimental dispersion curves were converted to the theoretical curve in order to achieve
the shear wave velocity profile.
Downloaded by [George Mason University] at 14:57 26 December 2014

4. Neural network design


To simulate the neurons in the brain, a computational model of the neural network was
recently developed with three basic elements: the weight to characterise the synapses or
connecting links, a combining function for summing the input signals, and an activation
function for limiting the amplitude of the output of a neuron, as shown in Figure 3. Lateral
processing of a large number of neurons makes an ANN efficient together with
approximation, pattern recognition, interpolation, and classification [10]. A multilayer
neural network develops a better capacity in non-linear analysis of the relationship between
inputs and desired outputs. Due to the remarkable performance of non-linear mapping
between the output vectors and input vectors, the error back propagation algorithm using
multilayer ANNs was applied to interpret the results obtained from the NDT experiments.
In this paper, three types of neural networks, ANN-I, ANN-II and ANN-III, were
designed to predict the concrete strength and thickness of slabs. ANN-I and ANN-II were
used to predict the concrete strength. ANN-I took into account the basic concrete
information (producer, designed strength, pouring time, etc.), material properties of each
component (water, cement, aggregate and adhesive), and the temperature and humidity
history to predict the concrete strength development.
The difference between ANN-I and ANN-II is that ANN II is using the NDT data in
addition to the above basic concrete information, the material properties and the
temperature and humidity history. ANN-II is designed to obtain the concrete strength
based on additional NDT data (SASW and IE methods) in addition to the ANN I data.
ANN-III, designed for the prediction of concrete strength and flaw depth in concrete,
adopted the IE test results as input data.

Figure 2. Schematic of SASW method.


Nondestructive Testing and Evaluation 281

Figure 3. Schematic of neural network model.


Downloaded by [George Mason University] at 14:57 26 December 2014

Five steps were followed to predict concrete strength as shown in Figure 4. In step 1,
based on BI, MP, and T/H, the concrete strength at 24 h after pouring was obtained by NN-
I, where BI is basic information, MP is material properties, and T/H is temperature and
humidity history from pouring day to 28 days after pouring. One of the outputs, fc, which is
the concrete compressive strength at 24 h, added to NN-II where it was considered as an
input vector. The improved concrete strength result has been added to the subsequent ANN
model. In this way, the improvement of concrete strength prediction has been added from
the front network to the next network. Finally, the increment of concrete strength between
7 and 28 days after pouring was obtained from the neural network.
The different thicknesses of testing objects show different shapes and amplitudes in the
wave spectra. The resonance frequency of the measured wave shows similar trends if it has
identical concrete strength and thickness, even if it has various magnitudes of impact [12].
Therefore, the ANN can be applied to determine the relationship between the stress wave
and the thickness or the strength of concrete.
ANN-III was also designed to determine the mapping between the data of NDT and the
thickness or strength of the concrete slab using an error back propagation algorithm.

5. Experimental studies of NDT method


5.1 Material and specimens
The 48 specimens of 800 £ 800 £ 200 mm were prepared for ANN-I and ANN-II based
on the concrete strengths of 18 –60 MPa together with the cylinder test specimens of
diameter 100 mm £ 200 mm for the test of compressive strength as shown in Figure 5.
Two specimens of 1000 £ 1000 £ 200 mm with the different thicknesses of 100 and
200 mm at different points were tested with the concrete strengths of 30 and 35 MPa for the
thickness and strength prediction in ANN-III, as shown in Figure 6.

Figure 4. Five steps of ANN-I/II.


282 Y.S. Cho et al.
Downloaded by [George Mason University] at 14:57 26 December 2014

Figure 5. Profile of specimens for ANN-I/II.

5.2 Instruments
To measure compression wave velocity, an IE test was used. Its instrumentation included a
dynamic signal analyser, a PCB 352C66 accelerometer receiver which has a frequency
range of 0.5 –10,000 Hz, a signal conditioner, and an impact source using a small steel ball
12 mm in diameter and varying diameter. The impact location was 25.4 –50.8 mm from the
receiver. Good couplings between receiver and the test surfaces were required.

5.3 Traditional concrete strength prediction


The specimens of different strengths and the cylinder specimens were tested at the
concrete ages of 16 h, 20 h, 24 h, 2 days, 3 days, 7 days, and 28 days using the concrete
cylinder tests. The IE test was conducted at locations 1 through 9 as shown in Figure 5. The
results following the NDTs are plotted in Figure 7 showing the compressive strengths, fck,
and compression wave velocities, Vp, at different ages. The horizontal axis of each figure
indicates the compressive strength, fck, measured from the cylinder tests and the vertical
axis of each graph indicates the compression wave velocities. Figures 7 and 8 represent the
correlation between the compressive strength of concrete, fck, and the compression wave
velocities, Vp, and the surface wave velocities, Vr, which can be used in the prediction of
compressive strength in early-age concrete and the existing structure using data from
beyond the concrete age of 28 days.

5.4 Traditional thickness prediction


The IE test was conducted to examine the frequency change in the concrete specimens of
30 and 35 MPa which were designed with varying thicknesses from 100 to 200 mm
increasing by 20 mm increments as shown in Table 1.

Figure 6. Profile of specimen for ANN-III.


Nondestructive Testing and Evaluation 283
Downloaded by [George Mason University] at 14:57 26 December 2014

Figure 7. Correlation between compressive strength, fck, versus compression wave velocity.

Figure 8. Correlation between compressive strength, fck, versus Rayleigh wave velocity.

Even though free boundaries existed due to the creation of steps in the concrete
specimen, the tested results agreed well with the known thickness, which can be used in
the actual structures. The error between the known thickness and the test result was less
than 3%. The test result of 30 MPa specimens showed a similar outcome to that of the
35 MPa specimens.
Using Equation (1), the thickness of concrete slabs can be derived from the resonance
frequency of the compression wave, as shown in Table 1.

6. Neural network prediction results


The training pattern consists of the experiment procedure, curing, and cylinder testing.
The concrete strength for desired output was determined at seven stages (16 h, 20 h, 24 h,

Table 1. Tested thickness versus actual thickness using IE test at 35 MPa.


Design thickness (mm) 100 120 140 160 180 200
Peak frequency (Hz) 18,261 14,965 12,288 11,392 9664 9376
Tested thickness (mm) 98.4 120.1 146.2 157.7 185.9 191.6
Error ratio (%) 1.6 0.08 4.43 1.44 3.28 4.21
284 Y.S. Cho et al.

2 days, 3 days, 7 days, and 28 days) after pouring. The input data were classified into three
categories: basic information (BI), material properties (MP), and temperature and
humidity history from pouring day to 28 days after pouring (T/H). All of these variables
were taken as input neurons of ANN and individually used for each module.

6.1 Concrete strength prediction of ANN-I/II


6.1.1 Training of ANN-I
Data for the 48 specimens were used in training with different concrete strengths. Desired
output data were obtained from cylinder testing of concrete specimens as shown in
Tables 2 –4.
After training, four specimens with different concrete strengths, 24, 27, 35, and 50 MPa,
Downloaded by [George Mason University] at 14:57 26 December 2014

were tested to evaluate the effect of the training results of NN, as shown in Figure 9. Good
agreement in predicting the concrete strength was achieved with an average determination
coefficient of 0.9.

Table 2. Training patterns of ANN-II (input data: basic information and material properties).
Input dataa
Basic information Material properties
No. a b c d e f g h i j k l m n
P1 A 180 8 11 17 157 N 215 FA 38 895 977 AE 0.63
P2 A 210 8 11 17 163 N 244 FA 43 873 953 AE 1.44

P3 A 600 8 31 16 163 N 504 FA 89 710 934 HF 8.6
Notes: a, concrete producer (A ¼ ready mixed concrete company ‘A’); b, nominal concrete strength (kgf/cm2);
c, month of pouring; d, day of pouring; e, time of pouring; f, unit weight of water (kg/m3); g, cement type
(N ¼ normal type); h, unit weight of cement (kg/m3); i, type of supplementary cementitious materials (FA ¼ fly
ash); k, unit weight of fine aggregate (kg/m3); l, unit weight of coarse aggregate (kg/m3); m, admixture type
(AE ¼ air-entraining, HF ¼ high-fowability); n, unit weight of admixture (kg/m3); 1 kgf/cm2 ¼ 0.098 MPa.
a
Excluded variables for avoiding duplication in actual training.

Table 3. Training patterns of ANN-II (input data: NDT data – IE method and SASW method).

Input data: NDT


IE method: Vp (m/s) SASW method: Vr (m/s)
No. Vp1 Vp2 Vp3 Vp4 Vp5 Vp6 Vp7 Vr1 Vr2 Vr3 Vr4 Vr5 Vr6 Vr7
P1 120 137 170 198 262 267 285 672 785 948 108 143 147 159
4 5 4 1 6 2 6 9 5 7 8
P2 145 160 209 228 293 311 322 747 897 116 124 160 168 181
1 5 1 6 2 2 6 5 8 2 9 5

P48 318 362 381 421 430 451 463 178 199 209 231 236 248 259
3 4 2 5 7 6 1 4 8 8 8 9 2 2
Notes: Vp, peak value of P-wave velocity on a curing day by IE method (m/s); Vr, peak value of R-wave velocity
on a curing day by SASW method (m/s).
Nondestructive Testing and Evaluation 285

Table 4. Training patterns of ANN-II (input data: temperature and humidity and desired output data).

Input data: temperature and


humidity Desired output dataa
0 , 28 days after pour-
ing Concrete strength (kgf/cm2)
No. aT0 aH0 , aT28 aH28 16 h 20 h 24 h 48 h 3 days 7 days 28 days
P1 29.8 57.6 , 21.3 71.9 41.2 48.2 51.1 109 112 164 190
0 6 9 2
P2 29.8 57.6 , 21.3 71.9 56.4 61.8 67.9 133 140 170 208
8 2 1 6
, ,
,
Downloaded by [George Mason University] at 14:57 26 December 2014

P48 24.2 81.5 27 70.1 110 121 137 210 292 492 608
2 2 2 2 2 1 5
Notes: aT0–aT28, average temperature on a curing day; aH0–aH28, average humidity on a curing day.
a Concrete strength on a curing day (kgf/cm2) by cylinder test, 1 kgf/cm2 ¼ 0.098 MPa.

6.1.2 Training of ANN-II


Similar to ANN-I, the additional NDT result was introduced to the input layer of the neural
network. More accurate results were obtained than those of ANN-I. An averaged
determination coefficient over 0.95 was obtained with the prediction. The results
comparing ANN-I and ANN-II are shown in Figure 9.

Figure 9. Simulation results of ANN-I/II.


286 Y.S. Cho et al.

6.2 Flaws detection with ANN-III


One hundred compressed data points in the stress wave curves were picked out and put
into the input neuron of the neural network. The concrete strength and thickness of the slab
were chosen as desired outputs. Two specimens’ data and 120 training patterns which were
composed of 2 concrete strengths (30 and 35 MPa) and 6 thicknesses (100, 120, 140, 160,
180 and 200 mm) with 10 measured wave data from experiments, were considered as
validation subsets. One specimen’s data and 12 validation patterns were used to evaluate
the performance of the neuron network after training.

6.3 Simulation results of ANN-III


The evaluation results show good agreement between the predicted thickness and concrete
Downloaded by [George Mason University] at 14:57 26 December 2014

strength with measured cylinder test results, with errors less than 0.6 and 4%, respectively.
The thickness predictions are shown in Figures 10 and 11. The neural network prediction
results were compared with traditional approximate formula results and were found to be
more accurate in thickness prediction.

Figure 10. Simulation results of ANN-III, 30 MPa.

Figure 11. Simulation results of ANN-III, 35 MPa.


Nondestructive Testing and Evaluation 287

The results of these studies including ANN-I, ANN-II, and ANN-III are basic
algorithms for the automation of predicting the compressive strength and the thickness of
concrete member. Automation of these results can contribute to predict the accurate form
removal time and unknown slab thickness in building construction practice.

7. Conclusions
The result of this study in examining the prediction of concrete compressive strength and
concrete member thickness using NDT methods (IE and SASW methods) and an ANN are
summarised as follows.
1. The correlation between the concrete compressive strength and compression wave
Downloaded by [George Mason University] at 14:57 26 December 2014

or Rayleigh wave velocity can be applied to predict the strength of concrete, which
can be adopted to evaluate the health of existing buildings.
2. The multilayer ANN, ANN-I, using an error back propagation algorithm can
describe concrete strength development by considering basic information about
concrete pouring, material properties, and temperature/humidity history.
Satisfactory results were achieved in ANN-II by considering additional NDT test
data as compared to ANN-I.
3. ANN-III can predict concrete slab thickness and strength using stress wave spectra.
Better results were obtained than those derived from IE test results. A neural
network can be used to interpret the results from the IE and SASW tests. This study
contributes to the partial development of a concrete assessment system.

Acknowledgements
This research work was supported by the SRC/ERC program of MOST (Grant#R11-2005-056-
01004-0). This research was also supported by The Korea Institute of Construction & Transportation
Technology Evaluation Planning under the title of Development of Concrete structural assessment
system based on artificial intelligence (30 December 2003– 29 December 2004). This research was
also supported by 2003 Hanyang University international journal paper competition project (HY-
2003-I-088).

References
[1] Y.S. Cho, Non-destructive testing of high-strength concrete using spectral analysis of surface
waves, NDT&E Int. (UK) 36 (2003), pp. 229– 235.
[2] D.G. Pratt and J. Lawrence, The use of a neural network in non-destructive testing, The
strategic highway research program, project C-204, 1992.
[3] I. Erdem, Transient elastic impact response of slender graphite rods, J. Nondestruct. Eval. 26
(2007), pp. 89 – 96.
[4] L.J. Sansalone and N.J. Carino, Detecting delaminations in concrete slabs with and without
overlays using the impact echo methods, ACI Mater. J. 83 (1989), pp. 175– 184.
[5] L.J. Sansalone et al., Sampling techniques for evaluating large concrete structures: Part I, ACI
Struct. J. 103, May–June (2006).
[6] S.M.A. Zomorodian and O. Hunaidi, Inversion of SASW dispersion curves based on maximum
flexibility coefficients in the wave-number domain, Soil Dyn. Earthq. Eng. 26 (2006), pp.
735– 752.
[7] D.S. Kim, H.W. Kim, K.C. Choi, and S.K. Woo, Feasibility study of the IE-SASW method for
nondestructive evaluation of containment building structures in nuclear power plants, Nucl.
Eng. Des. 219 (2003), pp. 97 –100.
288 Y.S. Cho et al.

[8] O. Hunaidi, Evolution-based genetic algorithms for analysis of non-destructive surface wave
tests on pavements, NDT&E Int. 31 (1998), pp. 273– 280.
[9] C.-D. Tsai, T.T. Wu, and Y.H. Liu, Application of neural networks to laser ultrasonic NDE of
bonded structures, NDT&E Int. 34 (2001), pp. 537– 546.
[10] B. Cho, S.-C. Lee, and Y.S. Cho, Neural network based models of a diagnosis system for
concrete structures using non-destructive test data, in Proceedings of the Eighth international
conference on the application of artificial intelligence to civil structural and environmental
engineering, Civil-Comp Press, Stirling, Scotland, 2005.
[11] J.M. Zurada, Introduction to artificial neural systems, Info Access Distribution Pte Ltd,
Singapore, 1992.
[12] D. Pratt and M. Sansalone, Impact echo signals interpretation using artificial intelligence, AIC
Mater. J. 89, March –April (1992), 178– 187.
Downloaded by [George Mason University] at 14:57 26 December 2014

You might also like