- Email: [email protected]
Non-destructive identification of pull-off adhesion between concrete layers
Automation in Construction 57 (2015) 146–155
Contents lists available at ScienceDirect
Automation in Construction journal homepage: www.elsevier.com/locate/autcon
Non-destructive identification of pull-off adhesion between concrete layers Łukasz Sadowski Faculty of Civil Engineering, Wrocław University of Technology, Wybrzeże Wyspiańskiego 27, 50-370 Wrocław, Poland
a r t i c l e
i n f o
Article history: Received 13 December 2014 Received in revised form 7 May 2015 Accepted 6 June 2015 Available online 19 June 2015 Keywords: Floors Concrete layers Pull-off adhesion Non-destructive test methods Artificial neural networks Methodology
a b s t r a c t This paper presents a methodology for the non-destructive identification of the values of pull-off adhesion between concrete layers in floors on the basis of parameters determined using three non-destructive testing (NDT) methods and artificial neural networks (ANN). The methodology is based on the earlier research by the author. There are three stages in the methodology: stage 1 in which two parameters are determined on the concrete substrate layer surface using the non-destructive 3D laser scanning method and three parameters are determined on the added concrete surface using the acoustic impulse response and impact-echo methods, stage 2 in which an ANN is trained and tested, and stage 3 in which numerical analyses of the results are carried out and the values of pull-off adhesion fc,b are identified. It is shown that the methodology is practicable, as demonstrated by the provided example. © 2015 Elsevier B.V. All rights reserved.
1. Introduction The durability of concrete structures depends on the proper bond between the added concrete layer and the existing concrete substrate layer [1–5]. The measure of this bond is the value of pull-off adhesion fb. The latter, in practice determined by the semi-destructive testing (SDT) pull-off method, should amount to less than 0.5 MPa according to [6]. According to [1], in order for the tests to be reliable 1 measurement per 3 m2 of floor surface should be taken. The check measurement in this method consists in drilling a 50 mm diameter borehole in the added concrete layer of the floor. The sample created in this way is pulled off from the concrete substrate layer, as described in US standard ASTM D 7234 [7] and in European standard EN 12504-3:2006 [6]. The drawback of the pull-off method, particularly in the case of largesurface floors, is that the tested surface is locally damaged in many places and then needs to be repaired. This drawback can be eliminated through the use of non-destructive testing methods (NDT) and artificial neural networks (ANN) [8–23]. Nowadays the number of applications of ANNs in different engineering areas still increases [24–27]. Studies described in [28–32] have shown that it is possible to reliably determine the pull-off adhesion through the use of the three NDT methods: the 3D laser scanning method and the acoustic impulse response and impact-echo methods plus ANN as a tool for correlating the results obtained by means of the above methods. Whereas as shown in [33–35], it is not possible to establish reliable correlations
E-mail address: [email protected].
http://dx.doi.org/10.1016/j.autcon.2015.06.004 0926-5805/© 2015 Elsevier B.V. All rights reserved.
between pull-off adhesion fb and the individual parameters determined by different NDT methods. In [36] (similar in its subject to the present paper) a methodology for the non-destructive assessment of the bond between the concrete layers in floors was developed for two cases: 1) delamination occurs (pull-off adhesion fb amounts to 0) and 2) delamination does not occur, but without the possibility of determining intermediate pull-off adhesion values in the latter case. The methodology proposed in this paper offers such a possibility. The methodology uses parameters estimated on the concrete substrate layer surface by the latest 3D laser scanning method [37–50] and parameters estimated on the floor surface by the impulse response and impact-echo methods [51–59] and ANN. It should be explained that since paper [36] no new papers dealing with this problem have been published and therefore no typical survey of literature is presented here. 2. Synthetic description of NDT test methods used Tables 1–3 present basic information on the latest 3D laser scanning method and the acoustic NDT impulse response and impact-echo methods used in the tests which provided the basis for developing the methodology. 3. Short description of tests providing basis for methodology As presented in [28–32], the tests providing the basis for the methodology were carried out on two 2500 × 2500 mm model concrete
Ł Sadowski / Automation in Construction 57 (2015) 146–155
147
Table 1 Synthetic description of 3D laser scanning method [37–50].
Name and description of method 1 The 3D laser scanning method exploits laser triangulation. It consists in measuring the deformations of a line produced by the measuring device's laser beam. A camera driven by a stepper motor is responsible for measuring the distance between the particular points located on the tested surface. Owing to this measurements can be taken in profiles spaced at every 10 µm with a vertical accuracy of 15 µm. Each point is assigned three coordinates describing its position on the tested surface. The size of the measuring field subjected to testing by this method is 50 mm × 50 mm, which corresponds to the field subjected to testing by the pull-off method.
Schematic and view of test setup
Recorded parameters
2
3 Using this method one can determine the values of the 3D roughness parameters specified in standard [49], and analyze surface morphology by means of specialist software. In particular the following are determined: - arithmetical mean height Sa, - root mean square height Sq, - surface bearing index Sbi.
Schematic and view of test setup
Recorded parameters and exemplary test results
Table 2 Synthetic description of acoustic impulse response method [51–53].
Name and description of method 1 The impulse response method is based on the excitation of the an elastic ultrasonic wave in the tested element by means of a calibrated hammer with a rubber tip. The frequency of the excited wave is in the interval of 1–800 Hz while the excitation range around the test point amounts to about 1000 mm. This method is suitable for the fast searching of large concrete surfaces and enables, among other things, the approximate identification and surface location of delaminations and the assessment of the pull– off adhesion of concrete layers to a depth of about 1500 mm.
2
3 The values of the follow– ing parameters are re– corded: – average mobility Nav, – stiffness Kd, – mobility slope Mp/N, – average mobility times mobility slope Nav ·Mp/N, – voids index v.
Table 3 Synthetic description of acoustic impact-echo method [54–59].
Name and description of method 1 The impact echo is based on the excitation of a low frequency (1–60 kHz) wave in the tested element by striking its surface with an exciter in the form of a steel ball (with different diame– ters). Specialist software is used to record the graphic image of the elastic wave propagating in the tested element, in the amplitude–time system, and to transform this image into an amplitude–frequency spectrum by means of the fast Fourier transform. This method is suitable for testing concrete members to estimate the thickness of unilaterally accessible members, the adhesion between layers, delamination, etc.
Schematic of test setup 2
Recorded parameters and exemplary test results 3 The values of the follow– ing parameters are re– corded: – transmitting impulse amplitude A, – the frequency of ultra– sonic wave reflection from the defect ( fD), – the frequency of ultra– sonic wave reflection from the bottom ( fT.).
Ł Sadowski / Automation in Construction 57 (2015) 146–155
148
a)
b)
c)
d)
Fig. 1. Exemplary view of determining: a) surface roughness parameters of concrete substrate layer by non-destructive 3D laser scanning method, b) parameters on floor surface by nondestructive impulse response method c) parameters on floor surface by non-destructive impact-echo method, d) pull-off adhesion fb by semi-destructive pull-off method [28–32].
Table 4 Exemplary values of tested parameters for few selected measuring points [32]. No. of test point
Name of test method and symbol of tested parameter 3D laser scanning method Sa
1 2 3 4 … 460
Sq
Ssa
Impulse response method
Impact-echo method
Pull-off method
Sku
Sbi
Sci
Svi
Sv
Sp
Kd
Nav
Mp/N
v
fT
fb
mm
mm
–
–
–
–
–
mm
mm
–
m/s N
–
–
kHz
MPa
0.500 0.737 0.825 0.912 … 0.125
0.267 1.033 0.730 1.287 … 0.179
−0.478 −0.582 −0.845 −1.109 … −2.280
4.686 3.933 4.509 5.084 … 11.63
0.040 0.048 0.033 0.019 … 0.019
0.697 0.744 0.730 0.716 … 0.674
0.024 0.004 0.012 0.021 … 0.009
1.311 1.009 1.250 1.492 … 1.546
0.694 0.384 0.489 0.595 … 0.316
0.017 0.004 0.041 0.021 … 0.088
62.674 99.276 57.401 102.430 … 449.394
0.828 3.449 1.746 3.761 … 0.918
1.116 0.592 0.544 0.882 … 1.235
7.500 7.000 8.000 6.500 … 5.000
0.970 0.870 0.940 0.710 … 0.610
floor specimens consisting of a 25 mm thick layer laid on a 125 mm thick concrete substrate layer. The latter had been laid on a 100 mm thick layer of sand. The 25 mm thick layer was made of C20/25 concrete and quartz aggregate with a maximum grading of 2 mm. The concrete substrate layer was made of C30/37 concrete and crushed basalt aggregate with a maximum grading of 8 mm. Different ways of preparing the concrete substrate layer surface, ensuring pull-off adhesion fb in a relatively wide range of 0.3–1.2 MPa were adopted. As reported in [28–32], tests on the added concrete layer surface were carried out in a few hundred previously determined measuring points. In each of the points after 28 days of concrete substrate layer concrete curing the values of 9 parameters describing the roughness of the concrete substrate layer surface were determined using the 3D laser scanning method. The parameters according to [49] were: Sa—arithmetical mean height, Sq—root mean square height, Ssk—skewness, Sku—kurtosis, Sp—maximum height of summits, Sv—maximum depth of valleys, Sbi—surface bearing index, Sci—the core's liquid retention index, Svi—the pits' liquid retention index.
Stage 1 Stage 1a
Stage 1b
Stage 1c
Stage 2
Stage 3
Fig. 2. General schematic of methodology for non-destructive identification of values of pull-off adhesion between concrete layers in floors by means of ANN.
Ł Sadowski / Automation in Construction 57 (2015) 146–155
Then a added concrete layer was laid on the existing concrete substrate layer and after 28 days since concreting the following 5 parameters were determined on its surface according to [51,54] using the NDT acoustic methods of impulse response and impact-echo: Nav—average mobility, Kd—dynamic stiffness, Mp/N—mobility slope, v—void index. fT—the frequency corresponding to the ultrasonic wave reflected from the bottom. The parameters were determined in measuring points coinciding with the points in which the 3D laser scanning tests had been carried
149
out. After the NDT tests, boreholes were drilled in the added concrete layer in these places in order to determine pull-off adhesion fb by means of the semi-destructive pull-off method (Fig. 1). The results of the tests in 460 measuring points, in the form of a set of 14 parameters were subjected to statistical analyses (descriptive statistics, the Shapiro–Wilk test, Spearman's rank correlation coefficient, the C&RT algorithm), as described in detail in [29]. Some of the test results for only a few selected points are shown in Table 4. From among the 14 parameters the following most suitable 5 parameters were selected as the input variables for the training and testing of an ANN: Sa—arithmetical mean height, Sq—root mean square height,
Stage 1a – testing of existing concrete surface roughness by 3D laser scanning
Select n measuring points on Existing concrete surface n = 1, … , m
n=1
Scan 50 x 50 mm areas in measuring points 50 x 50 mm
Record isometric 3D images of tested 50 x 50 mm areas. Record analyze and process data.
YES
Record data n≥m YES
NO
n=n+1
Determine values of surface roughness parameters: - arithmetical mean height Sa, - root mean square height Sq, - surface bearing index Sbi.
Fig. 3. Methodology for non-destructive identification of pull-off adhesion between concrete layers in floors by means of ANN—stage 1a.
150
Ł Sadowski / Automation in Construction 57 (2015) 146–155
determined by the non-destructive 3D laser scanning method, Kd—dynamic stiffness, Nav—average mobility, determined by the non-destructive impulse response method, fT—the frequency for ultrasonic wave reflection from the bottom, determined by the non-destructive impact-echo method.
A unidirectional multilayer backpropagation neural network with the QUASI-NEWTON training algorithm, 10 hidden layer neurons and the tanh hidden layer activation function was selected for the nondestructive identification of the pull-off adhesion of the added concrete layer to the concrete substrate layer and subjected to training and testing [29].
Etap 1b – testing of added concrete surface by NDT acoustic impulse-response and impact echo methods
Mark n measuring points on added concrete surface in the same places as in stage 1a n = 1, … , m n=1
Carry out NDT impulse response tests. Excite elastic wave with calibrated hammer Analyze: - values of elastic force F, - graph of elastic wave velocity w, - graph of mobility N YES
Are F, w, N satisfactory? NO YES
Record results
n≥m
Repeat test
NO n=n+1
Determine values of parameters: - average mobility Nav, - stiffness Kd.
Carry out NDT impact-echo tests. Excite elastic wave with exciter and record amplitude-time spectrum Transform amplitude-time spectrum into amplitude-frequency spectrum.
Analyze amplitude-frequency spectrum
Record results
NO
n ≥m
n=n+1
YES
Determine value of parameter: frequency of ultrasonic wave reflection from bottom (ƒT) Fig. 4. Methodology for non-destructive identification of pull-off adhesion between concrete layers in floors by means of ANN—stage 1b.
Ł Sadowski / Automation in Construction 57 (2015) 146–155
As presented in [33–35], it is not possible to establish reliable correlations between pull-off adhesion fb and the individual parameters determined by different NDT methods. At first in [29,33,34] an attempt was made to determine the correlations between the Sa, Sq, Ssk, Sku, Sp, Sv, Sbi, Sci and Svi parameters and pull-off adhesion fb evaluated by the pull-off method. The highest values of determination coefficient r2 from 0.35 for ground surface and 0.65 for unprepared surface have been obtained for Sa and Sq defined by using the Eqs. (1) and (2): Sa ¼
Sq ¼
1 A
Z Z A
jZ ðx; yÞjdxdy;
ffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Z Z 1 ðZ ðx; yÞÞ2 dxdy: A A
ð1Þ
ð2Þ
where Z (x, y) is the height coordinate within the sampling area A. Sa and Sq are the amplitude parameters which are a class of surface finish parameters characterizing the distribution of heights [60]. According to [61] Sa and Sq are insensitive in differentiating peaks, valleys and the spacing of the various texture features. That is why Sa and Sq can be used to correlate with mechanical adhesion between concrete layers which is connected with the mechanical anchoring of the material of one layer in the pores and surface irregularities of another layer [28]. According to [62], mechanical adhesion has the decisive influence on the quality of the bond between concrete layers. However these values of r2 are low in case of reliable prediction of pull-off adhesion fb only based only on the values of surface roughness parameters and indicated that in this situation this approach was not good. On the other side in [35] it has been presented that the values of determination coefficient r2 between average mobility Nav obtained by impulse response method and pull-off adhesion fb evaluated by the pull-off method assumes from 0.7115 for ground surface to 0.3417 for unprepared surface. Also determination coefficient r2 for dynamic stiffness Kd assumes values from 0.40 to 0.56. This attempt indicated also that this approach was not proper and it is not possible to obtain reliable correlation between single parameters. That is why ANN has been used as a tool for correlating the results obtained by means of the 3D laser scanning method and the acoustic impulse response and impact-echo methods. In previous studies presented in [28–32] linear correlation coefficients R of 0.9775 and 0.9725 were obtained in respectively the training and testing of ANN. These high linear correlation coefficients are satisfactory in case of the relation between selected parameters and pull-off adhesion. The experimental verification of the ANN showed very good agreement between the results [28–32]. It should be explained here that in this way an adequate database was obtained. The latter can be used in practice for the nondestructive neural identification of the pull-off adhesion of the added concrete layer to the concrete substrate layer in concrete floors with a similar concrete composition and thickness of the particular layers as those used in the above experiments. On the basis of extensive studies [28–32], briefly described above, a methodology for the non-destructive identification of the values of the pull-off adhesion between concrete layers has been developed and applied in practice, as exemplified in this paper.
151
It includes three stages as shown in Fig. 2. Stage 1 covers experimental studies the results of which form a database. The stage consists of: - stage 1a, shown in Fig. 3, in which concrete substrate layer surface roughness tests should be carried out in selected measuring points by the 3D laser scanning method, - stage 1b, shown in Fig. 4, in which after laying the floor added concrete layer tests should be carried out on its surface by the acoustic methods of impulse response and impact-echo, - stage 1c, shown in Fig. 5, in which tests should be carried out in the same measuring points on the floor surface by the semi-destructive pull-off method. And so in stage 1a (Fig. 3) one should select n measuring points on the concrete substrate layer surface and then scan a 50 × 50 mm area in each of the measuring points, using the 3D laser scanning method. The acquired data (a isometric 3D image) are saved as a file. In the next step of this stage one should determine the values of the roughness parameters in the test points. Then in stage 1b (Fig. 4), having laid the added concrete layer of the floor, one should mark n test points on the surface in the same places as in stage 1a. In the next step of this stage one should carry out tests by the non-destructive impulse response method, consisting in the excitation of an elastic wave in each of the measuring points by means of the calibrated hammer. The photographs shown in Fig. 4 illustrate the tests. As a result, the values of the characteristic parameters: average mobility Nav and stiffness Kd can be acquired in each of the measuring points. Then impact-echo tests, consisting in exciting an elastic wave in each point by means of the exciter and recording the amplitude-time spectrum, as illustrated by the photographs in Fig. 4, are performed. As a result, the value of the frequency of ultrasonic wave reflection from the bottom (fT) is obtained. In stage 1c, shown in Fig. 5, seminon-destructive pull-off adhesion tests should be performed as shown in the photographs and the value of pull-off adhesion fb should be determined in the same measuring points in which the impulse response and impact-echo tests were carried. Stage 2 consists in the training, testing and experimental verification of the ANN. This stage (Fig. 5) includes the creation of a database containing the data acquired using the 3D laser scanning method, the
Etap 1c – testing by SDT pull-off method
Carry out STD pulloff tests in n points on added concrete surface
4. Description of proposed methodology On the basis of experiments and analyses involving the use of ANN [28–32] a methodology for the non-destructive identification of the values of the pull-off adhesion between concrete layers has been proposed. The methodology is based on the use of: - three NDT test methods and five parameters determined by them, - artificial neural networks.
Determine value of parameter: - pull-off adhesion ƒb Fig. 5. Methodology for non-destructive identification of values of pull-off adhesion between concrete layers in floors by means of ANN—stage 1c.
Ł Sadowski / Automation in Construction 57 (2015) 146–155
152
acoustic methods of impulse response and impact-echo and the semidestructive pull-off method. The values of the five parameters determined in n measuring points in stages 1a, 1b and 1c form the database. After a statistical analysis of the parameters an ANN (its structure, a training algorithm and the number of hidden layer neurons) is selected and trained and tested. According to [32], a unidirectional multilayer backpropagation neural network with the QUASI-NEWTON training algorithm, 10 hidden layer neurons and the tanh hidden layer activation
function, whose structure is shown in Fig. 5, guaranteeing that correlation coefficient R N 0.7 will be obtained in its training and testing, should be selected. The final step in this stage consists in the experimental verification of the ANN. Then in stage 3, after numerical analyses the prediction of the values of pull-off adhesion fc,b takes place. The obtained values of fc,b should be verified by the pull-off method in randomly selected measuring points.
Stage 2 – training, testing and experimental verification of artificial neural network
Impulse-response method
3D laser scanning method
Input layer Experimentally determined parameters
Impact-echo method Pull-off method
Hidden layer K=10
Output layer Identification parameter
Create database containing data acquired from testing by NDT 3D laser scanning and acoustic methods and SDT pull-off method
Sa
Sq Nav fc,b
Carry out statistical analysis of experimentally determined parameters. Divide database into training data, testing data and experimental verification data
Kd
fT
Select ANN: its structure, training algorithm and number of hidden layer neurons
fb
Train and test
NO
Are values of correlation coefficient R and relative error RE satisfactory? YES Experimentally verify ANN
Stage 3 – analysis of obtained results and identification of pull-off adhesion fc,b Numerical analyses of obtained results
Prediction of values of pull-off adhesion f c,b
SDT verification in selected test points through boreholes drilled with core bit
Fig. 6. Methodology for NDT identification of values of pull-off adhesion fc,b between concrete layers in floors by means of ANN—stages 2 and 3.
Ł Sadowski / Automation in Construction 57 (2015) 146–155
153
Stage 1 Stage 1a 1 4
2
5
3
Stage 1b
6 9
Stage 2
7
8
10
Fig. 7. Methodology for non-destructive identification of values of pull-off adhesion between concrete layers in floors by means of ANN in practical applications using existing database. - randomly selected places (points) in which methodology was verified in practice
5. Practical use of methodology
Fig. 9. Distribution of measuring points on tested floor.
5.1. Way of using Using the methodology presented in Sect. 4 one can estimate the pull-off adhesion of the added concrete layer to the concrete substrate layer in real floors. The methodology is fully practicable when the composition of the concrete in the particular floor layers and the thicknesses of the layers are similar to the ones used to create the database in [28–32]. Otherwise, it is necessary to create a new database for the floors tested in practice, as specified in stage 1 of the methodology described in Sect. 4. The stages in the practical use of an existing database are shown in Fig. 7 and described below. The procedure includes two stages. Stage 1 covers the testing of an inspected real concrete floor, using an existing data base created earlier. The stage consists of: stage 1a (shown in Fig. 3) in which tests of the concrete substrate layer surface roughness should be carried out in selected measuring points using the 3D laser scanning method prior to concreting the added concrete layer, stage 1b (shown in Fig. 4), in which after laying the added concrete layer tests should be carried out on the floor surface by means of the acoustic methods of impulse response and impact-echo in measuring points coinciding with the ones in which the 3D laser scanning tests were carried out.
In stage 2 numerical analyses of the results obtained from stage 1 are performed using the ANN and the values of pull-off adhesion fc,b are determined. It is proposed to use a unidirectional multilayer backpropagation neural network with the QUASI-NEWTON training algorithm, 10 hidden layer neurons and the tanh hidden layer activation function, with the structure shown in Fig. 6. The obtained fc,b values can be verified in randomly selected test points using the pull-off method. 5.2. Example of use The methodology was used to non-destructively identify the pull-off adhesion (fc,b) of the added concrete layer to the concrete substrate layer in the floor in a newly built industrial building. The 20 mm thick top layer was made of grade C20/25 concrete with a maximum quartz aggregate grading of 2 mm. The 120 mm thick base layer was made of grade C30/37 concrete with a maximum crushed basalt aggregate grading of 8 mm. The concrete substrate layer was laid on underlayment membrane and on a 100 mm thick layer of styrofoam. Considering that the tested floor was characterized by similar concrete compositions
Database created for the tested concrete floor in practice
Stage 2 – analysis of obtained results and identification of pull-off adhesion fc,b
Numerical analysis of obtained results
Identification of pull-off adhesion fc,b
Seminondestructive verification in selected test points through boreholes drilled with core bit Fig. 8. Stage 2 in practical procedure for non-destructive identification of values of pull-off adhesion fc,b between concrete layers in floors.
Ł Sadowski / Automation in Construction 57 (2015) 146–155
154
6. Conclusion
Table 5 Values of tested parameters, constituting input data for ANN. Number of measuring point
1 2 3 4 5 6 7 8 9 10
Name of NDT method and measured parameter 3D laser scanning
Impulse response
Impact-echo
Sa
Sq
Kd
Nav
fT
mm
mm
–
m/s N
kHz
0.245 0.161 0.202 0.270 0.169 0.175 0.282 0.146 0.226 0.169
0.314 0.204 0.248 0.369 0.217 0.233 0.341 0.195 0.280 0.214
3.000 5.000 4.000 0.010 3.000 0.105 5.000 1.000 6.000 0.009
83.000 82.000 100.038 80.000 68.000 73.000 66.000 78.000 78.000 98.000
12.690 13.670 15.630 13.670 13.180 13.700 15.630 13.670 13.200 11.500
and thicknesses of the particular layers as the layered elements previously used to create the database for the non-destructive neural identification of the pull-off adhesion of the added concrete layer to the concrete substrate layer in such a floor, the previously developed database was adopted [28–32]. The values of two surface roughness parameters (Sa and Sq) were determined in randomly selected 10 measuring points (denoted with digits from 1 to 10 in Fig. 9) on the concrete substrate layer surface using the 3D laser scanning method. Then three parameters (Kd, Nav and fT) were determined in the same measuring points on the added concrete layer surface using the acoustic methods. The values of the parameters are shown in Table 5. The unidirectional multilayer backpropagation neural network with the QUASI-NEWTON training algorithm, 10 hidden layer neurons and the tanh hidden layer activation function was used to non-destructively identify pull-off adhesion values. As already mentioned, this ANN had been trained and tested using the database developed in [28–32]. As a result, the ANN generated the value of pull-off adhesion fc,b in each of the 10 points. Then the values were compared with the experimental fb values determined using the semi-destructive pull-off method. The results for fc,b and fb are compared in Table 6. (See Fig. 8.) It appears from Table 6 that the fc,b and fb values obtained in the particular measuring points are similar and the relative error (RE) values are low, ranging from 0.05 to 0.17. It means that there is a very high correlation between fc,b and fb.
Table 6 Values of pull-off adhesion fc,b and fb determined by respectively ANN and semidestructive pull-off method. Number of measuring point
1 2 3 4 5 6 7 8 9 10
Pull-off adhesion identified by ANN
Pull-off adhesion experimentally determined by semi-destructive pull-off method
Absolute difference between fc,b and fb values
Relative error of fc,b and fb values
fc,b
fb
Δfc,b
RE
MPa
MPa
MPa
–
1.11 0.82 1.05 0.59 0.67 1.01 1.27 0.66 1.22 0.55
1.22 0.71 1.28 0.54 0.64 1.07 1.15 0.59 1.10 0.51
−0.11 0.11 0.23 0.05 0.03 −0.06 0.14 0.07 0.12 0.04
0.09 0.15 0.17 0.09 0.05 0.06 0.10 0.12 0.11 0.08
A methodology for the non-destructive identification of the values of concrete layers in floors on the basis of parameters determined by three NDT methods and ANN is proposed. The methodology has been developed on the basis of test and analytical results reported in [28–32] indicating that it is possible to reliably assess the bond (pull-off adhesion) between concrete layers in floors by means of ANN. In the proposed methodology the following three stages are distinguished: - stage 1 in which two parameters are tested on the concrete substrate layer surface using the non-destructive 3D laser scanning method and three parameters are tested on the added concrete layer surface using the acoustic methods of impulse response and impact-echo, - stage 2 in which the ANN is trained and tested, - stage 3 in which numerical analyses of the obtained results are carried out and pull-off adhesion fc,b values are non-destructively identified. It has been shown that the methodology is practicable, as demonstrated by the provided example. References [1] P. Berkowski, G. Dmochowski, J. Grosel, K. Schabowicz, Z. Wójcicki, Analysis of failure conditions for a dynamically loaded composite floor system of an industrial building, J. Civ. Eng. Manag. 19–4 (2013) 529–541. [2] A. Cwirzen, P. Sztermen, K. Habermehl-Cwirzen, Effect of baltic seawater and binder type on frost durability of concrete, J. Mater. Civ. Eng. 26–2 (2013) 283–287. [3] T. Błaszczyński, J. Jasiczak, B. Ksit, M. Siewczyńska, Aspects of bond layer role in concrete repairs, Arch. Civ. Mech. Eng. 6–4 (2006) 73–85. [4] N. Zoidis, E. Tatsis, C. Vlachopoulos, A. Gotzamanis, J. Clausen, D. Aggelis, T. Matikas, Inspection, evaluation and repair monitoring of cracked concrete floor using NDT methods, Constr. Build. Mater. 48 (2013) 1302–1308. [5] M. Mohamad, I. Ibrahim, R. Abdullah, A. Rahman, A. Kueh, J. Usman, Friction and cohesion coefficients of composite concrete-to-concrete bond, Cem. Concr. Compos. 56 (2015) 1–14. [6] EN 12504-3:2006, Testing Concrete in Structures, Part 3: Non-destructive tests, Determination of pull-off force (in Polish)2006. [7] ASTM D7234–05, Standard Test Method for Pull-Off Adhesion Strength of Coatings on Concrete Using Portable Pull-Off Adhesion Testers, 2005. [8] H. Adeli, Neural Networks in Civil Engineering: 1989–2000, Comput.-Aided Civ. Infrastruct. Eng. 16 (2001) 126–142. [9] M. Alhamdoosh, D. Wang, Fast decorrelated neural network ensembles with random weights, Inf. Sci. 264 (2014) 104–117. [10] L. Bal, F. Buyle-Bodin, Artificial neural network for predicting creep of concrete, Neural Comput. & Applic. 25–6 (2014) 1359–1367. [11] O. Gencel, F. Kocabas, J. del Coz Diaz, A comparative modeling study to estimate wear of concrete, Neural Comput. & Applic. 24-3-4 (2014) 649–662. [12] J. Hoła, K. Schabowicz, State-of-the-art non-destructive methods for diagnostics testing of building structures—anticipated development trends, Arch. Civ. Mech. Eng. 10–3 (2010) 5–18. [13] V. Malhotra, N. Carino, Handbook on Non-Destructive Testing of Concrete, CRC Press, 2004. [14] K. Schabowicz, Modern acoustic techniques for testing concrete structures accessible from one side only, Arch. Civ. Mech. Eng. (2015) (in press, http://dx.doi.org/ 10.1016/j.acme.2014.10.001). [15] J. Hoła, J. Bień, L. Sadowski, K. Schabowicz, Non-destructive and semi-destructive diagnostics of concrete structures in assessment of their durability, Bull. Pol. Acad. Sci. Tech. Sci. 63–1 (2015) 87–96. [16] L. Drobiec, R. Jasiński, A. Piekarczyk, Diagnostics of reinforced concrete structures, Methodology, field tests, laboratory tests of concrete and steel (in Polish), vol. 1, Państwowe Wydawnictwo Naukowe, Warsaw, 2010. [17] M. Nikoo, P. Zarfam, M. Nikoo, Determining displacement in concrete reinforcement building with using evolutionary artificial neural networks, World Appl. Sci. J. 16–12 (2012) 1699–1708. [18] M. Nikoo, P. Zarfam, H. Sayahpour, Determination of compressive strength of concrete using self organization feature map (SOFM), Engineering with Computers 31–1 (2015) 113–121. [19] E. Güneyisi, M. Gesoğlu, F. Karaboğa, K. Mermerdaş, Corrosion behavior of reinforcing steel embedded in chloride contaminated concretes with and without metakaolin, Compos. Part B 45 (2013) 1288–1295. [20] E.M. Güneyisi, K. Mermerdaş, F. Güneyisi, M. Gesoğlu, Numerical modeling of time to corrosion induced cover cracking in reinforced concrete using soft-computing based methods, Mater. Struct. (2014), http://dx.doi.org/10.1617/s11527-014-0269-8. [21] E. Güneyisi, T. Özturan, M. Gesoğlu, Effect of initial curing on chloride ingress and corrosion resistance characteristics of concretes made with plain and blended cements, Build. Environ. 42 (2007) 2676–2685.
Ł Sadowski / Automation in Construction 57 (2015) 146–155 [22] M. Korzeniowski, T. Piwowarczyk, R. Maev, Application of ultrasonic method for quality evaluation of adhesive layers, Arch. Civ. Mech. Eng. 14–4 (2014) 661–670. [23] J. Bungey, S. Millard, M. Gratham, Testing of Concrete in Structures, Taylor & Francis, London, 2006. [24] M. D'Aniello, E.M. Güneyisi, R. Landolfo, K. Mermerdaş, Predictive models of the flexural overstrength factor for steel thin-walled circular hollow section beams, ThinWalled Struct. 94 (2015) 67–78. [25] E.M. Güneyisi, M. Gesoğlu, E. Güneyisi, K. Mermerdaş, Assessment of shear capacity of adhesive anchors for structures using neural network based model, Mater. Struct. (2015), http://dx.doi.org/10.1617/s11527-015-0558-x (in press). [26] E.M. Güneyisi, M. D'Aniello, R. Landolfo, K. Mermerdaş, Prediction of the flexural overstrength factor for steel beams using artificial neural network, Steel Compos. Struct. 17 (2014) 215–236. [27] M. D'Aniello, E.M. Güneyisi, R. Landolfo, K. Mermerdaş, Analytical prediction of available rotation capacity of cold-formed rectangular and square hollow section beams, Thin-Walled Struct. 77 (2014) 141–152. [28] L. Sadowski, Non-destructive evaluation of the pull-off adhesion of concrete floor layers using RBF neural network, J. Civ. Eng. Manag. 19-4 (2013) 550–560. [29] L. Sadowski, The non-destructive assessment of the bond between concrete layers in floors by means of artificial neural networks (in Polish)(PhD thesis) Wrocław University of Technology, Wroclaw, 2012. [30] L. Sadowski, J. Hoła, Application of an artificial neural network to the evaluation of pull-off adhesion of concrete layers, in: B.H.V. Topping, P,. Iványi (Eds.), Proceedings of the Fourteenth International Conference on Civil Structural and Environmental Engineering Computing, Civil-Comp Press, Stirlingshire, UK, 2013 (Paper 179). [31] L. Sadowski, J. Hoła, Neural prediction of the pull-off adhesion of the concrete layers in floors on the basis of non-destructive tests, Procedia Eng. 57 (2013) 986–995. [32] L. Sadowski, J. Hoła, New non-destructive way of identifying the values of pull-off adhesion between concrete layers in floors, J. Civ. Eng. Manag. 20–4 (2014) 561–569. [33] L. Sadowski, Application of three-dimensional optical laser triangulation method for concrete surface morphology measurement, Indian J. Eng. Mater. Sci. 21–6 (2014) 692–700. [34] L. Sadowski, Analysis of the effect of concrete base roughness on the pull-off adhesion of the topping layer (in Polish), IAPGOŚ 1 (2013) 39–42. [35] J. Hoła, L. Sadowski, Testing interlayer pull-off adhesion in concrete floors by means of non-destructive acoustic methods, 18th World Conference on Non Destructive Testing, Durban, 2012. [36] J. Hoła, L. Sadowski, K. Schabowicz, Non-destructive identification of delaminations in concrete floor toppings with acoustic methods, Autom. Constr. 20–7 (2011) 799–807. [37] T. Mathia, P. Pawlus, M. Wieczorowski, Recent trends in surface metrology, Wear 271 (2011) 494–508. [38] M. Wieczorowski, Surface Topography Analysis, Publishing House of Poznan University of Technology, Poznan, 2009. [39] P. Santos, E. Julio, A state-of-the-art review on roughness quantification methods for concrete surfaces, Constr. Build. Mater. 38 (2013) 912–923. [40] W. Zielecki, P. Pawlus, R. Perłowski, A. Dzierwa, Surface topography effect on strength of lap adhesive joints after mechanical pre-treatment, Arch. Civ. Mech. Eng. 13–2 (2013) 175–185. [41] J. Hoła, L. Sadowski, J. Reiner, S. Stach, Usefulness of 3D surface roughness parameters for nondestructive evaluation of pull-off adhesion of concrete layers, Constr. Build. Mater. 84 (2015) 111–120.
155
[42] M. Niemczewska-Wójcik, A. Mańkowska-Snopczyńska, W. Piekoszewski, The investigation of wear tracks with the use of noncontact measurement methods, Arch. Civ. Mech. Eng. 13–2 (2013) 158–167. [43] H. Sezen, N. Fisco, Evaluation and comparison of surface macrotexture and friction measurement methods, J. Civ. Eng. Manag. 19–3 (2013) 387–399. [44] M. Niemczewska-Wójcik, T. Mathia, A. Wójcik, Measurement techniques used for analysis of the geometric structure of machined surfaces, Manag. Prod. Eng. Rev. 5–2 (2014) 27–32. [45] A. Garbacz, L. Courard, B. Bissonnette, A surface engineering approach applicable to concrete repair engineering, Bull. Pol. Acad. Sci. Tech. Sci. 61–1 (2013) 73–84. [46] K. Kubiak, M. Bigerelle, T. Mathia, A. Dubois, L. Dubar, Dynamic evolution of interface roughness during friction and wear processes, Scanning 36–1 (2014) 30–38. [47] P. Santos, E. Julio, Correlation between concrete-to-concrete bond strength and the roughness of the substrate surface, Constr. Build. Mater. 21-8 (2007). [48] R. Deltombe, K. Kubiak, M. Bigerelle, How to select the most relevant 3D roughness parameters of a surface, Scanning 36–1 (2014) 150–160. [49] ISO 25178: Geometric Product Specifications (GPS) - Surface texture: areal, 2012. [50] M. Siewczyńska, Method for determining the parameters of surface roughness by usage of a 3D scanner, Arch. Civ. Mech. Eng. 12–1 (2012) 83–89. [51] ASTM C1740–10, Standard Practice for Evaluating the Condition of Concrete Plates Using the Impulse-Response Method, 2010. [52] A. Davis, The non-destructive impulse response test in North America: 1985–2001, NDT&E Int. 36–4 (2003) 185–193. [53] A. Davis, B. Hertlein, K. Lim, K. Michols, Impact-echo and impulse response stress wave methods: advantages and limitations for the evaluation of highway pavement concrete overlays, Conference on Non-destructive Evaluation of Bridges and Highways, Scottsdale, 1996. [54] M. Sansalone, W. Streett, Impact-Echo: Non-Destructive Evaluation of Concrete and Masonry, Bullbrier Press, Ithaca, 1997. [55] K. Schabowicz, Methodology for non-destructive identification of thickness of unilaterally accessible concrete elements by means of state-of-the-art acoustic techniques, J. Civ. Eng. Manag. 19–3 (2013) 325–334. [56] H. Azari, S. Nazarian, D. Yuan, Assessing sensitivity of impact echo and ultrasonic surface waves methods for nondestructive evaluation of concrete structures, Constr. Build. Mater. 71 (2014) 384–391. [57] T. Piotrowski, The use of the impact-echo signal analysis to assess bond quality in concrete repair systems (in Polish)(PhD thesis) Warsaw University of Technology, Warsaw, 2009. [58] J. Bień, J. Hoła, L. Sadowski, Modern NDT Techniques in Diagnostics of Transport Infrastructure Concrete Structures, International Workshop on Smart Materials, Structures NDT in Canada 2013 Conference & NDT for the Energy Industry, Calgary, 2013. [59] T. Gorzelańczyk, J. Hoła, L. Sadowski, K. Schabowicz, Methodology of nondestructive identification of defective concrete zones in unilaterally accessible massive members, J. Civ. Eng. Manag. 19–6 (2013) 775–786. [60] K. Stout, T. Mathia, H. Zahouani, P. Sullivan, W. Dong, E. Mainsah, N. Lou, The Development of Methods for Characterisation of Roughness in Three Dimensions, ECSCEEC-EAEC, Brussels-Luxembourg, 1993. [61] http://www.michmet.com/3d_s_height_parameters.htm. [62] L. Czarnecki, Adhesion—a challenge for concrete repair, Proc. of the 2nd International Conference on Concrete Repair, Rehabilitation and Retrofitting, Cape Town, November, 2008.
Contents lists available at ScienceDirect
Automation in Construction journal homepage: www.elsevier.com/locate/autcon
Non-destructive identification of pull-off adhesion between concrete layers Łukasz Sadowski Faculty of Civil Engineering, Wrocław University of Technology, Wybrzeże Wyspiańskiego 27, 50-370 Wrocław, Poland
a r t i c l e
i n f o
Article history: Received 13 December 2014 Received in revised form 7 May 2015 Accepted 6 June 2015 Available online 19 June 2015 Keywords: Floors Concrete layers Pull-off adhesion Non-destructive test methods Artificial neural networks Methodology
a b s t r a c t This paper presents a methodology for the non-destructive identification of the values of pull-off adhesion between concrete layers in floors on the basis of parameters determined using three non-destructive testing (NDT) methods and artificial neural networks (ANN). The methodology is based on the earlier research by the author. There are three stages in the methodology: stage 1 in which two parameters are determined on the concrete substrate layer surface using the non-destructive 3D laser scanning method and three parameters are determined on the added concrete surface using the acoustic impulse response and impact-echo methods, stage 2 in which an ANN is trained and tested, and stage 3 in which numerical analyses of the results are carried out and the values of pull-off adhesion fc,b are identified. It is shown that the methodology is practicable, as demonstrated by the provided example. © 2015 Elsevier B.V. All rights reserved.
1. Introduction The durability of concrete structures depends on the proper bond between the added concrete layer and the existing concrete substrate layer [1–5]. The measure of this bond is the value of pull-off adhesion fb. The latter, in practice determined by the semi-destructive testing (SDT) pull-off method, should amount to less than 0.5 MPa according to [6]. According to [1], in order for the tests to be reliable 1 measurement per 3 m2 of floor surface should be taken. The check measurement in this method consists in drilling a 50 mm diameter borehole in the added concrete layer of the floor. The sample created in this way is pulled off from the concrete substrate layer, as described in US standard ASTM D 7234 [7] and in European standard EN 12504-3:2006 [6]. The drawback of the pull-off method, particularly in the case of largesurface floors, is that the tested surface is locally damaged in many places and then needs to be repaired. This drawback can be eliminated through the use of non-destructive testing methods (NDT) and artificial neural networks (ANN) [8–23]. Nowadays the number of applications of ANNs in different engineering areas still increases [24–27]. Studies described in [28–32] have shown that it is possible to reliably determine the pull-off adhesion through the use of the three NDT methods: the 3D laser scanning method and the acoustic impulse response and impact-echo methods plus ANN as a tool for correlating the results obtained by means of the above methods. Whereas as shown in [33–35], it is not possible to establish reliable correlations
E-mail address: [email protected].
http://dx.doi.org/10.1016/j.autcon.2015.06.004 0926-5805/© 2015 Elsevier B.V. All rights reserved.
between pull-off adhesion fb and the individual parameters determined by different NDT methods. In [36] (similar in its subject to the present paper) a methodology for the non-destructive assessment of the bond between the concrete layers in floors was developed for two cases: 1) delamination occurs (pull-off adhesion fb amounts to 0) and 2) delamination does not occur, but without the possibility of determining intermediate pull-off adhesion values in the latter case. The methodology proposed in this paper offers such a possibility. The methodology uses parameters estimated on the concrete substrate layer surface by the latest 3D laser scanning method [37–50] and parameters estimated on the floor surface by the impulse response and impact-echo methods [51–59] and ANN. It should be explained that since paper [36] no new papers dealing with this problem have been published and therefore no typical survey of literature is presented here. 2. Synthetic description of NDT test methods used Tables 1–3 present basic information on the latest 3D laser scanning method and the acoustic NDT impulse response and impact-echo methods used in the tests which provided the basis for developing the methodology. 3. Short description of tests providing basis for methodology As presented in [28–32], the tests providing the basis for the methodology were carried out on two 2500 × 2500 mm model concrete
Ł Sadowski / Automation in Construction 57 (2015) 146–155
147
Table 1 Synthetic description of 3D laser scanning method [37–50].
Name and description of method 1 The 3D laser scanning method exploits laser triangulation. It consists in measuring the deformations of a line produced by the measuring device's laser beam. A camera driven by a stepper motor is responsible for measuring the distance between the particular points located on the tested surface. Owing to this measurements can be taken in profiles spaced at every 10 µm with a vertical accuracy of 15 µm. Each point is assigned three coordinates describing its position on the tested surface. The size of the measuring field subjected to testing by this method is 50 mm × 50 mm, which corresponds to the field subjected to testing by the pull-off method.
Schematic and view of test setup
Recorded parameters
2
3 Using this method one can determine the values of the 3D roughness parameters specified in standard [49], and analyze surface morphology by means of specialist software. In particular the following are determined: - arithmetical mean height Sa, - root mean square height Sq, - surface bearing index Sbi.
Schematic and view of test setup
Recorded parameters and exemplary test results
Table 2 Synthetic description of acoustic impulse response method [51–53].
Name and description of method 1 The impulse response method is based on the excitation of the an elastic ultrasonic wave in the tested element by means of a calibrated hammer with a rubber tip. The frequency of the excited wave is in the interval of 1–800 Hz while the excitation range around the test point amounts to about 1000 mm. This method is suitable for the fast searching of large concrete surfaces and enables, among other things, the approximate identification and surface location of delaminations and the assessment of the pull– off adhesion of concrete layers to a depth of about 1500 mm.
2
3 The values of the follow– ing parameters are re– corded: – average mobility Nav, – stiffness Kd, – mobility slope Mp/N, – average mobility times mobility slope Nav ·Mp/N, – voids index v.
Table 3 Synthetic description of acoustic impact-echo method [54–59].
Name and description of method 1 The impact echo is based on the excitation of a low frequency (1–60 kHz) wave in the tested element by striking its surface with an exciter in the form of a steel ball (with different diame– ters). Specialist software is used to record the graphic image of the elastic wave propagating in the tested element, in the amplitude–time system, and to transform this image into an amplitude–frequency spectrum by means of the fast Fourier transform. This method is suitable for testing concrete members to estimate the thickness of unilaterally accessible members, the adhesion between layers, delamination, etc.
Schematic of test setup 2
Recorded parameters and exemplary test results 3 The values of the follow– ing parameters are re– corded: – transmitting impulse amplitude A, – the frequency of ultra– sonic wave reflection from the defect ( fD), – the frequency of ultra– sonic wave reflection from the bottom ( fT.).
Ł Sadowski / Automation in Construction 57 (2015) 146–155
148
a)
b)
c)
d)
Fig. 1. Exemplary view of determining: a) surface roughness parameters of concrete substrate layer by non-destructive 3D laser scanning method, b) parameters on floor surface by nondestructive impulse response method c) parameters on floor surface by non-destructive impact-echo method, d) pull-off adhesion fb by semi-destructive pull-off method [28–32].
Table 4 Exemplary values of tested parameters for few selected measuring points [32]. No. of test point
Name of test method and symbol of tested parameter 3D laser scanning method Sa
1 2 3 4 … 460
Sq
Ssa
Impulse response method
Impact-echo method
Pull-off method
Sku
Sbi
Sci
Svi
Sv
Sp
Kd
Nav
Mp/N
v
fT
fb
mm
mm
–
–
–
–
–
mm
mm
–
m/s N
–
–
kHz
MPa
0.500 0.737 0.825 0.912 … 0.125
0.267 1.033 0.730 1.287 … 0.179
−0.478 −0.582 −0.845 −1.109 … −2.280
4.686 3.933 4.509 5.084 … 11.63
0.040 0.048 0.033 0.019 … 0.019
0.697 0.744 0.730 0.716 … 0.674
0.024 0.004 0.012 0.021 … 0.009
1.311 1.009 1.250 1.492 … 1.546
0.694 0.384 0.489 0.595 … 0.316
0.017 0.004 0.041 0.021 … 0.088
62.674 99.276 57.401 102.430 … 449.394
0.828 3.449 1.746 3.761 … 0.918
1.116 0.592 0.544 0.882 … 1.235
7.500 7.000 8.000 6.500 … 5.000
0.970 0.870 0.940 0.710 … 0.610
floor specimens consisting of a 25 mm thick layer laid on a 125 mm thick concrete substrate layer. The latter had been laid on a 100 mm thick layer of sand. The 25 mm thick layer was made of C20/25 concrete and quartz aggregate with a maximum grading of 2 mm. The concrete substrate layer was made of C30/37 concrete and crushed basalt aggregate with a maximum grading of 8 mm. Different ways of preparing the concrete substrate layer surface, ensuring pull-off adhesion fb in a relatively wide range of 0.3–1.2 MPa were adopted. As reported in [28–32], tests on the added concrete layer surface were carried out in a few hundred previously determined measuring points. In each of the points after 28 days of concrete substrate layer concrete curing the values of 9 parameters describing the roughness of the concrete substrate layer surface were determined using the 3D laser scanning method. The parameters according to [49] were: Sa—arithmetical mean height, Sq—root mean square height, Ssk—skewness, Sku—kurtosis, Sp—maximum height of summits, Sv—maximum depth of valleys, Sbi—surface bearing index, Sci—the core's liquid retention index, Svi—the pits' liquid retention index.
Stage 1 Stage 1a
Stage 1b
Stage 1c
Stage 2
Stage 3
Fig. 2. General schematic of methodology for non-destructive identification of values of pull-off adhesion between concrete layers in floors by means of ANN.
Ł Sadowski / Automation in Construction 57 (2015) 146–155
Then a added concrete layer was laid on the existing concrete substrate layer and after 28 days since concreting the following 5 parameters were determined on its surface according to [51,54] using the NDT acoustic methods of impulse response and impact-echo: Nav—average mobility, Kd—dynamic stiffness, Mp/N—mobility slope, v—void index. fT—the frequency corresponding to the ultrasonic wave reflected from the bottom. The parameters were determined in measuring points coinciding with the points in which the 3D laser scanning tests had been carried
149
out. After the NDT tests, boreholes were drilled in the added concrete layer in these places in order to determine pull-off adhesion fb by means of the semi-destructive pull-off method (Fig. 1). The results of the tests in 460 measuring points, in the form of a set of 14 parameters were subjected to statistical analyses (descriptive statistics, the Shapiro–Wilk test, Spearman's rank correlation coefficient, the C&RT algorithm), as described in detail in [29]. Some of the test results for only a few selected points are shown in Table 4. From among the 14 parameters the following most suitable 5 parameters were selected as the input variables for the training and testing of an ANN: Sa—arithmetical mean height, Sq—root mean square height,
Stage 1a – testing of existing concrete surface roughness by 3D laser scanning
Select n measuring points on Existing concrete surface n = 1, … , m
n=1
Scan 50 x 50 mm areas in measuring points 50 x 50 mm
Record isometric 3D images of tested 50 x 50 mm areas. Record analyze and process data.
YES
Record data n≥m YES
NO
n=n+1
Determine values of surface roughness parameters: - arithmetical mean height Sa, - root mean square height Sq, - surface bearing index Sbi.
Fig. 3. Methodology for non-destructive identification of pull-off adhesion between concrete layers in floors by means of ANN—stage 1a.
150
Ł Sadowski / Automation in Construction 57 (2015) 146–155
determined by the non-destructive 3D laser scanning method, Kd—dynamic stiffness, Nav—average mobility, determined by the non-destructive impulse response method, fT—the frequency for ultrasonic wave reflection from the bottom, determined by the non-destructive impact-echo method.
A unidirectional multilayer backpropagation neural network with the QUASI-NEWTON training algorithm, 10 hidden layer neurons and the tanh hidden layer activation function was selected for the nondestructive identification of the pull-off adhesion of the added concrete layer to the concrete substrate layer and subjected to training and testing [29].
Etap 1b – testing of added concrete surface by NDT acoustic impulse-response and impact echo methods
Mark n measuring points on added concrete surface in the same places as in stage 1a n = 1, … , m n=1
Carry out NDT impulse response tests. Excite elastic wave with calibrated hammer Analyze: - values of elastic force F, - graph of elastic wave velocity w, - graph of mobility N YES
Are F, w, N satisfactory? NO YES
Record results
n≥m
Repeat test
NO n=n+1
Determine values of parameters: - average mobility Nav, - stiffness Kd.
Carry out NDT impact-echo tests. Excite elastic wave with exciter and record amplitude-time spectrum Transform amplitude-time spectrum into amplitude-frequency spectrum.
Analyze amplitude-frequency spectrum
Record results
NO
n ≥m
n=n+1
YES
Determine value of parameter: frequency of ultrasonic wave reflection from bottom (ƒT) Fig. 4. Methodology for non-destructive identification of pull-off adhesion between concrete layers in floors by means of ANN—stage 1b.
Ł Sadowski / Automation in Construction 57 (2015) 146–155
As presented in [33–35], it is not possible to establish reliable correlations between pull-off adhesion fb and the individual parameters determined by different NDT methods. At first in [29,33,34] an attempt was made to determine the correlations between the Sa, Sq, Ssk, Sku, Sp, Sv, Sbi, Sci and Svi parameters and pull-off adhesion fb evaluated by the pull-off method. The highest values of determination coefficient r2 from 0.35 for ground surface and 0.65 for unprepared surface have been obtained for Sa and Sq defined by using the Eqs. (1) and (2): Sa ¼
Sq ¼
1 A
Z Z A
jZ ðx; yÞjdxdy;
ffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Z Z 1 ðZ ðx; yÞÞ2 dxdy: A A
ð1Þ
ð2Þ
where Z (x, y) is the height coordinate within the sampling area A. Sa and Sq are the amplitude parameters which are a class of surface finish parameters characterizing the distribution of heights [60]. According to [61] Sa and Sq are insensitive in differentiating peaks, valleys and the spacing of the various texture features. That is why Sa and Sq can be used to correlate with mechanical adhesion between concrete layers which is connected with the mechanical anchoring of the material of one layer in the pores and surface irregularities of another layer [28]. According to [62], mechanical adhesion has the decisive influence on the quality of the bond between concrete layers. However these values of r2 are low in case of reliable prediction of pull-off adhesion fb only based only on the values of surface roughness parameters and indicated that in this situation this approach was not good. On the other side in [35] it has been presented that the values of determination coefficient r2 between average mobility Nav obtained by impulse response method and pull-off adhesion fb evaluated by the pull-off method assumes from 0.7115 for ground surface to 0.3417 for unprepared surface. Also determination coefficient r2 for dynamic stiffness Kd assumes values from 0.40 to 0.56. This attempt indicated also that this approach was not proper and it is not possible to obtain reliable correlation between single parameters. That is why ANN has been used as a tool for correlating the results obtained by means of the 3D laser scanning method and the acoustic impulse response and impact-echo methods. In previous studies presented in [28–32] linear correlation coefficients R of 0.9775 and 0.9725 were obtained in respectively the training and testing of ANN. These high linear correlation coefficients are satisfactory in case of the relation between selected parameters and pull-off adhesion. The experimental verification of the ANN showed very good agreement between the results [28–32]. It should be explained here that in this way an adequate database was obtained. The latter can be used in practice for the nondestructive neural identification of the pull-off adhesion of the added concrete layer to the concrete substrate layer in concrete floors with a similar concrete composition and thickness of the particular layers as those used in the above experiments. On the basis of extensive studies [28–32], briefly described above, a methodology for the non-destructive identification of the values of the pull-off adhesion between concrete layers has been developed and applied in practice, as exemplified in this paper.
151
It includes three stages as shown in Fig. 2. Stage 1 covers experimental studies the results of which form a database. The stage consists of: - stage 1a, shown in Fig. 3, in which concrete substrate layer surface roughness tests should be carried out in selected measuring points by the 3D laser scanning method, - stage 1b, shown in Fig. 4, in which after laying the floor added concrete layer tests should be carried out on its surface by the acoustic methods of impulse response and impact-echo, - stage 1c, shown in Fig. 5, in which tests should be carried out in the same measuring points on the floor surface by the semi-destructive pull-off method. And so in stage 1a (Fig. 3) one should select n measuring points on the concrete substrate layer surface and then scan a 50 × 50 mm area in each of the measuring points, using the 3D laser scanning method. The acquired data (a isometric 3D image) are saved as a file. In the next step of this stage one should determine the values of the roughness parameters in the test points. Then in stage 1b (Fig. 4), having laid the added concrete layer of the floor, one should mark n test points on the surface in the same places as in stage 1a. In the next step of this stage one should carry out tests by the non-destructive impulse response method, consisting in the excitation of an elastic wave in each of the measuring points by means of the calibrated hammer. The photographs shown in Fig. 4 illustrate the tests. As a result, the values of the characteristic parameters: average mobility Nav and stiffness Kd can be acquired in each of the measuring points. Then impact-echo tests, consisting in exciting an elastic wave in each point by means of the exciter and recording the amplitude-time spectrum, as illustrated by the photographs in Fig. 4, are performed. As a result, the value of the frequency of ultrasonic wave reflection from the bottom (fT) is obtained. In stage 1c, shown in Fig. 5, seminon-destructive pull-off adhesion tests should be performed as shown in the photographs and the value of pull-off adhesion fb should be determined in the same measuring points in which the impulse response and impact-echo tests were carried. Stage 2 consists in the training, testing and experimental verification of the ANN. This stage (Fig. 5) includes the creation of a database containing the data acquired using the 3D laser scanning method, the
Etap 1c – testing by SDT pull-off method
Carry out STD pulloff tests in n points on added concrete surface
4. Description of proposed methodology On the basis of experiments and analyses involving the use of ANN [28–32] a methodology for the non-destructive identification of the values of the pull-off adhesion between concrete layers has been proposed. The methodology is based on the use of: - three NDT test methods and five parameters determined by them, - artificial neural networks.
Determine value of parameter: - pull-off adhesion ƒb Fig. 5. Methodology for non-destructive identification of values of pull-off adhesion between concrete layers in floors by means of ANN—stage 1c.
Ł Sadowski / Automation in Construction 57 (2015) 146–155
152
acoustic methods of impulse response and impact-echo and the semidestructive pull-off method. The values of the five parameters determined in n measuring points in stages 1a, 1b and 1c form the database. After a statistical analysis of the parameters an ANN (its structure, a training algorithm and the number of hidden layer neurons) is selected and trained and tested. According to [32], a unidirectional multilayer backpropagation neural network with the QUASI-NEWTON training algorithm, 10 hidden layer neurons and the tanh hidden layer activation
function, whose structure is shown in Fig. 5, guaranteeing that correlation coefficient R N 0.7 will be obtained in its training and testing, should be selected. The final step in this stage consists in the experimental verification of the ANN. Then in stage 3, after numerical analyses the prediction of the values of pull-off adhesion fc,b takes place. The obtained values of fc,b should be verified by the pull-off method in randomly selected measuring points.
Stage 2 – training, testing and experimental verification of artificial neural network
Impulse-response method
3D laser scanning method
Input layer Experimentally determined parameters
Impact-echo method Pull-off method
Hidden layer K=10
Output layer Identification parameter
Create database containing data acquired from testing by NDT 3D laser scanning and acoustic methods and SDT pull-off method
Sa
Sq Nav fc,b
Carry out statistical analysis of experimentally determined parameters. Divide database into training data, testing data and experimental verification data
Kd
fT
Select ANN: its structure, training algorithm and number of hidden layer neurons
fb
Train and test
NO
Are values of correlation coefficient R and relative error RE satisfactory? YES Experimentally verify ANN
Stage 3 – analysis of obtained results and identification of pull-off adhesion fc,b Numerical analyses of obtained results
Prediction of values of pull-off adhesion f c,b
SDT verification in selected test points through boreholes drilled with core bit
Fig. 6. Methodology for NDT identification of values of pull-off adhesion fc,b between concrete layers in floors by means of ANN—stages 2 and 3.
Ł Sadowski / Automation in Construction 57 (2015) 146–155
153
Stage 1 Stage 1a 1 4
2
5
3
Stage 1b
6 9
Stage 2
7
8
10
Fig. 7. Methodology for non-destructive identification of values of pull-off adhesion between concrete layers in floors by means of ANN in practical applications using existing database. - randomly selected places (points) in which methodology was verified in practice
5. Practical use of methodology
Fig. 9. Distribution of measuring points on tested floor.
5.1. Way of using Using the methodology presented in Sect. 4 one can estimate the pull-off adhesion of the added concrete layer to the concrete substrate layer in real floors. The methodology is fully practicable when the composition of the concrete in the particular floor layers and the thicknesses of the layers are similar to the ones used to create the database in [28–32]. Otherwise, it is necessary to create a new database for the floors tested in practice, as specified in stage 1 of the methodology described in Sect. 4. The stages in the practical use of an existing database are shown in Fig. 7 and described below. The procedure includes two stages. Stage 1 covers the testing of an inspected real concrete floor, using an existing data base created earlier. The stage consists of: stage 1a (shown in Fig. 3) in which tests of the concrete substrate layer surface roughness should be carried out in selected measuring points using the 3D laser scanning method prior to concreting the added concrete layer, stage 1b (shown in Fig. 4), in which after laying the added concrete layer tests should be carried out on the floor surface by means of the acoustic methods of impulse response and impact-echo in measuring points coinciding with the ones in which the 3D laser scanning tests were carried out.
In stage 2 numerical analyses of the results obtained from stage 1 are performed using the ANN and the values of pull-off adhesion fc,b are determined. It is proposed to use a unidirectional multilayer backpropagation neural network with the QUASI-NEWTON training algorithm, 10 hidden layer neurons and the tanh hidden layer activation function, with the structure shown in Fig. 6. The obtained fc,b values can be verified in randomly selected test points using the pull-off method. 5.2. Example of use The methodology was used to non-destructively identify the pull-off adhesion (fc,b) of the added concrete layer to the concrete substrate layer in the floor in a newly built industrial building. The 20 mm thick top layer was made of grade C20/25 concrete with a maximum quartz aggregate grading of 2 mm. The 120 mm thick base layer was made of grade C30/37 concrete with a maximum crushed basalt aggregate grading of 8 mm. The concrete substrate layer was laid on underlayment membrane and on a 100 mm thick layer of styrofoam. Considering that the tested floor was characterized by similar concrete compositions
Database created for the tested concrete floor in practice
Stage 2 – analysis of obtained results and identification of pull-off adhesion fc,b
Numerical analysis of obtained results
Identification of pull-off adhesion fc,b
Seminondestructive verification in selected test points through boreholes drilled with core bit Fig. 8. Stage 2 in practical procedure for non-destructive identification of values of pull-off adhesion fc,b between concrete layers in floors.
Ł Sadowski / Automation in Construction 57 (2015) 146–155
154
6. Conclusion
Table 5 Values of tested parameters, constituting input data for ANN. Number of measuring point
1 2 3 4 5 6 7 8 9 10
Name of NDT method and measured parameter 3D laser scanning
Impulse response
Impact-echo
Sa
Sq
Kd
Nav
fT
mm
mm
–
m/s N
kHz
0.245 0.161 0.202 0.270 0.169 0.175 0.282 0.146 0.226 0.169
0.314 0.204 0.248 0.369 0.217 0.233 0.341 0.195 0.280 0.214
3.000 5.000 4.000 0.010 3.000 0.105 5.000 1.000 6.000 0.009
83.000 82.000 100.038 80.000 68.000 73.000 66.000 78.000 78.000 98.000
12.690 13.670 15.630 13.670 13.180 13.700 15.630 13.670 13.200 11.500
and thicknesses of the particular layers as the layered elements previously used to create the database for the non-destructive neural identification of the pull-off adhesion of the added concrete layer to the concrete substrate layer in such a floor, the previously developed database was adopted [28–32]. The values of two surface roughness parameters (Sa and Sq) were determined in randomly selected 10 measuring points (denoted with digits from 1 to 10 in Fig. 9) on the concrete substrate layer surface using the 3D laser scanning method. Then three parameters (Kd, Nav and fT) were determined in the same measuring points on the added concrete layer surface using the acoustic methods. The values of the parameters are shown in Table 5. The unidirectional multilayer backpropagation neural network with the QUASI-NEWTON training algorithm, 10 hidden layer neurons and the tanh hidden layer activation function was used to non-destructively identify pull-off adhesion values. As already mentioned, this ANN had been trained and tested using the database developed in [28–32]. As a result, the ANN generated the value of pull-off adhesion fc,b in each of the 10 points. Then the values were compared with the experimental fb values determined using the semi-destructive pull-off method. The results for fc,b and fb are compared in Table 6. (See Fig. 8.) It appears from Table 6 that the fc,b and fb values obtained in the particular measuring points are similar and the relative error (RE) values are low, ranging from 0.05 to 0.17. It means that there is a very high correlation between fc,b and fb.
Table 6 Values of pull-off adhesion fc,b and fb determined by respectively ANN and semidestructive pull-off method. Number of measuring point
1 2 3 4 5 6 7 8 9 10
Pull-off adhesion identified by ANN
Pull-off adhesion experimentally determined by semi-destructive pull-off method
Absolute difference between fc,b and fb values
Relative error of fc,b and fb values
fc,b
fb
Δfc,b
RE
MPa
MPa
MPa
–
1.11 0.82 1.05 0.59 0.67 1.01 1.27 0.66 1.22 0.55
1.22 0.71 1.28 0.54 0.64 1.07 1.15 0.59 1.10 0.51
−0.11 0.11 0.23 0.05 0.03 −0.06 0.14 0.07 0.12 0.04
0.09 0.15 0.17 0.09 0.05 0.06 0.10 0.12 0.11 0.08
A methodology for the non-destructive identification of the values of concrete layers in floors on the basis of parameters determined by three NDT methods and ANN is proposed. The methodology has been developed on the basis of test and analytical results reported in [28–32] indicating that it is possible to reliably assess the bond (pull-off adhesion) between concrete layers in floors by means of ANN. In the proposed methodology the following three stages are distinguished: - stage 1 in which two parameters are tested on the concrete substrate layer surface using the non-destructive 3D laser scanning method and three parameters are tested on the added concrete layer surface using the acoustic methods of impulse response and impact-echo, - stage 2 in which the ANN is trained and tested, - stage 3 in which numerical analyses of the obtained results are carried out and pull-off adhesion fc,b values are non-destructively identified. It has been shown that the methodology is practicable, as demonstrated by the provided example. References [1] P. Berkowski, G. Dmochowski, J. Grosel, K. Schabowicz, Z. Wójcicki, Analysis of failure conditions for a dynamically loaded composite floor system of an industrial building, J. Civ. Eng. Manag. 19–4 (2013) 529–541. [2] A. Cwirzen, P. Sztermen, K. Habermehl-Cwirzen, Effect of baltic seawater and binder type on frost durability of concrete, J. Mater. Civ. Eng. 26–2 (2013) 283–287. [3] T. Błaszczyński, J. Jasiczak, B. Ksit, M. Siewczyńska, Aspects of bond layer role in concrete repairs, Arch. Civ. Mech. Eng. 6–4 (2006) 73–85. [4] N. Zoidis, E. Tatsis, C. Vlachopoulos, A. Gotzamanis, J. Clausen, D. Aggelis, T. Matikas, Inspection, evaluation and repair monitoring of cracked concrete floor using NDT methods, Constr. Build. Mater. 48 (2013) 1302–1308. [5] M. Mohamad, I. Ibrahim, R. Abdullah, A. Rahman, A. Kueh, J. Usman, Friction and cohesion coefficients of composite concrete-to-concrete bond, Cem. Concr. Compos. 56 (2015) 1–14. [6] EN 12504-3:2006, Testing Concrete in Structures, Part 3: Non-destructive tests, Determination of pull-off force (in Polish)2006. [7] ASTM D7234–05, Standard Test Method for Pull-Off Adhesion Strength of Coatings on Concrete Using Portable Pull-Off Adhesion Testers, 2005. [8] H. Adeli, Neural Networks in Civil Engineering: 1989–2000, Comput.-Aided Civ. Infrastruct. Eng. 16 (2001) 126–142. [9] M. Alhamdoosh, D. Wang, Fast decorrelated neural network ensembles with random weights, Inf. Sci. 264 (2014) 104–117. [10] L. Bal, F. Buyle-Bodin, Artificial neural network for predicting creep of concrete, Neural Comput. & Applic. 25–6 (2014) 1359–1367. [11] O. Gencel, F. Kocabas, J. del Coz Diaz, A comparative modeling study to estimate wear of concrete, Neural Comput. & Applic. 24-3-4 (2014) 649–662. [12] J. Hoła, K. Schabowicz, State-of-the-art non-destructive methods for diagnostics testing of building structures—anticipated development trends, Arch. Civ. Mech. Eng. 10–3 (2010) 5–18. [13] V. Malhotra, N. Carino, Handbook on Non-Destructive Testing of Concrete, CRC Press, 2004. [14] K. Schabowicz, Modern acoustic techniques for testing concrete structures accessible from one side only, Arch. Civ. Mech. Eng. (2015) (in press, http://dx.doi.org/ 10.1016/j.acme.2014.10.001). [15] J. Hoła, J. Bień, L. Sadowski, K. Schabowicz, Non-destructive and semi-destructive diagnostics of concrete structures in assessment of their durability, Bull. Pol. Acad. Sci. Tech. Sci. 63–1 (2015) 87–96. [16] L. Drobiec, R. Jasiński, A. Piekarczyk, Diagnostics of reinforced concrete structures, Methodology, field tests, laboratory tests of concrete and steel (in Polish), vol. 1, Państwowe Wydawnictwo Naukowe, Warsaw, 2010. [17] M. Nikoo, P. Zarfam, M. Nikoo, Determining displacement in concrete reinforcement building with using evolutionary artificial neural networks, World Appl. Sci. J. 16–12 (2012) 1699–1708. [18] M. Nikoo, P. Zarfam, H. Sayahpour, Determination of compressive strength of concrete using self organization feature map (SOFM), Engineering with Computers 31–1 (2015) 113–121. [19] E. Güneyisi, M. Gesoğlu, F. Karaboğa, K. Mermerdaş, Corrosion behavior of reinforcing steel embedded in chloride contaminated concretes with and without metakaolin, Compos. Part B 45 (2013) 1288–1295. [20] E.M. Güneyisi, K. Mermerdaş, F. Güneyisi, M. Gesoğlu, Numerical modeling of time to corrosion induced cover cracking in reinforced concrete using soft-computing based methods, Mater. Struct. (2014), http://dx.doi.org/10.1617/s11527-014-0269-8. [21] E. Güneyisi, T. Özturan, M. Gesoğlu, Effect of initial curing on chloride ingress and corrosion resistance characteristics of concretes made with plain and blended cements, Build. Environ. 42 (2007) 2676–2685.
Ł Sadowski / Automation in Construction 57 (2015) 146–155 [22] M. Korzeniowski, T. Piwowarczyk, R. Maev, Application of ultrasonic method for quality evaluation of adhesive layers, Arch. Civ. Mech. Eng. 14–4 (2014) 661–670. [23] J. Bungey, S. Millard, M. Gratham, Testing of Concrete in Structures, Taylor & Francis, London, 2006. [24] M. D'Aniello, E.M. Güneyisi, R. Landolfo, K. Mermerdaş, Predictive models of the flexural overstrength factor for steel thin-walled circular hollow section beams, ThinWalled Struct. 94 (2015) 67–78. [25] E.M. Güneyisi, M. Gesoğlu, E. Güneyisi, K. Mermerdaş, Assessment of shear capacity of adhesive anchors for structures using neural network based model, Mater. Struct. (2015), http://dx.doi.org/10.1617/s11527-015-0558-x (in press). [26] E.M. Güneyisi, M. D'Aniello, R. Landolfo, K. Mermerdaş, Prediction of the flexural overstrength factor for steel beams using artificial neural network, Steel Compos. Struct. 17 (2014) 215–236. [27] M. D'Aniello, E.M. Güneyisi, R. Landolfo, K. Mermerdaş, Analytical prediction of available rotation capacity of cold-formed rectangular and square hollow section beams, Thin-Walled Struct. 77 (2014) 141–152. [28] L. Sadowski, Non-destructive evaluation of the pull-off adhesion of concrete floor layers using RBF neural network, J. Civ. Eng. Manag. 19-4 (2013) 550–560. [29] L. Sadowski, The non-destructive assessment of the bond between concrete layers in floors by means of artificial neural networks (in Polish)(PhD thesis) Wrocław University of Technology, Wroclaw, 2012. [30] L. Sadowski, J. Hoła, Application of an artificial neural network to the evaluation of pull-off adhesion of concrete layers, in: B.H.V. Topping, P,. Iványi (Eds.), Proceedings of the Fourteenth International Conference on Civil Structural and Environmental Engineering Computing, Civil-Comp Press, Stirlingshire, UK, 2013 (Paper 179). [31] L. Sadowski, J. Hoła, Neural prediction of the pull-off adhesion of the concrete layers in floors on the basis of non-destructive tests, Procedia Eng. 57 (2013) 986–995. [32] L. Sadowski, J. Hoła, New non-destructive way of identifying the values of pull-off adhesion between concrete layers in floors, J. Civ. Eng. Manag. 20–4 (2014) 561–569. [33] L. Sadowski, Application of three-dimensional optical laser triangulation method for concrete surface morphology measurement, Indian J. Eng. Mater. Sci. 21–6 (2014) 692–700. [34] L. Sadowski, Analysis of the effect of concrete base roughness on the pull-off adhesion of the topping layer (in Polish), IAPGOŚ 1 (2013) 39–42. [35] J. Hoła, L. Sadowski, Testing interlayer pull-off adhesion in concrete floors by means of non-destructive acoustic methods, 18th World Conference on Non Destructive Testing, Durban, 2012. [36] J. Hoła, L. Sadowski, K. Schabowicz, Non-destructive identification of delaminations in concrete floor toppings with acoustic methods, Autom. Constr. 20–7 (2011) 799–807. [37] T. Mathia, P. Pawlus, M. Wieczorowski, Recent trends in surface metrology, Wear 271 (2011) 494–508. [38] M. Wieczorowski, Surface Topography Analysis, Publishing House of Poznan University of Technology, Poznan, 2009. [39] P. Santos, E. Julio, A state-of-the-art review on roughness quantification methods for concrete surfaces, Constr. Build. Mater. 38 (2013) 912–923. [40] W. Zielecki, P. Pawlus, R. Perłowski, A. Dzierwa, Surface topography effect on strength of lap adhesive joints after mechanical pre-treatment, Arch. Civ. Mech. Eng. 13–2 (2013) 175–185. [41] J. Hoła, L. Sadowski, J. Reiner, S. Stach, Usefulness of 3D surface roughness parameters for nondestructive evaluation of pull-off adhesion of concrete layers, Constr. Build. Mater. 84 (2015) 111–120.
155
[42] M. Niemczewska-Wójcik, A. Mańkowska-Snopczyńska, W. Piekoszewski, The investigation of wear tracks with the use of noncontact measurement methods, Arch. Civ. Mech. Eng. 13–2 (2013) 158–167. [43] H. Sezen, N. Fisco, Evaluation and comparison of surface macrotexture and friction measurement methods, J. Civ. Eng. Manag. 19–3 (2013) 387–399. [44] M. Niemczewska-Wójcik, T. Mathia, A. Wójcik, Measurement techniques used for analysis of the geometric structure of machined surfaces, Manag. Prod. Eng. Rev. 5–2 (2014) 27–32. [45] A. Garbacz, L. Courard, B. Bissonnette, A surface engineering approach applicable to concrete repair engineering, Bull. Pol. Acad. Sci. Tech. Sci. 61–1 (2013) 73–84. [46] K. Kubiak, M. Bigerelle, T. Mathia, A. Dubois, L. Dubar, Dynamic evolution of interface roughness during friction and wear processes, Scanning 36–1 (2014) 30–38. [47] P. Santos, E. Julio, Correlation between concrete-to-concrete bond strength and the roughness of the substrate surface, Constr. Build. Mater. 21-8 (2007). [48] R. Deltombe, K. Kubiak, M. Bigerelle, How to select the most relevant 3D roughness parameters of a surface, Scanning 36–1 (2014) 150–160. [49] ISO 25178: Geometric Product Specifications (GPS) - Surface texture: areal, 2012. [50] M. Siewczyńska, Method for determining the parameters of surface roughness by usage of a 3D scanner, Arch. Civ. Mech. Eng. 12–1 (2012) 83–89. [51] ASTM C1740–10, Standard Practice for Evaluating the Condition of Concrete Plates Using the Impulse-Response Method, 2010. [52] A. Davis, The non-destructive impulse response test in North America: 1985–2001, NDT&E Int. 36–4 (2003) 185–193. [53] A. Davis, B. Hertlein, K. Lim, K. Michols, Impact-echo and impulse response stress wave methods: advantages and limitations for the evaluation of highway pavement concrete overlays, Conference on Non-destructive Evaluation of Bridges and Highways, Scottsdale, 1996. [54] M. Sansalone, W. Streett, Impact-Echo: Non-Destructive Evaluation of Concrete and Masonry, Bullbrier Press, Ithaca, 1997. [55] K. Schabowicz, Methodology for non-destructive identification of thickness of unilaterally accessible concrete elements by means of state-of-the-art acoustic techniques, J. Civ. Eng. Manag. 19–3 (2013) 325–334. [56] H. Azari, S. Nazarian, D. Yuan, Assessing sensitivity of impact echo and ultrasonic surface waves methods for nondestructive evaluation of concrete structures, Constr. Build. Mater. 71 (2014) 384–391. [57] T. Piotrowski, The use of the impact-echo signal analysis to assess bond quality in concrete repair systems (in Polish)(PhD thesis) Warsaw University of Technology, Warsaw, 2009. [58] J. Bień, J. Hoła, L. Sadowski, Modern NDT Techniques in Diagnostics of Transport Infrastructure Concrete Structures, International Workshop on Smart Materials, Structures NDT in Canada 2013 Conference & NDT for the Energy Industry, Calgary, 2013. [59] T. Gorzelańczyk, J. Hoła, L. Sadowski, K. Schabowicz, Methodology of nondestructive identification of defective concrete zones in unilaterally accessible massive members, J. Civ. Eng. Manag. 19–6 (2013) 775–786. [60] K. Stout, T. Mathia, H. Zahouani, P. Sullivan, W. Dong, E. Mainsah, N. Lou, The Development of Methods for Characterisation of Roughness in Three Dimensions, ECSCEEC-EAEC, Brussels-Luxembourg, 1993. [61] http://www.michmet.com/3d_s_height_parameters.htm. [62] L. Czarnecki, Adhesion—a challenge for concrete repair, Proc. of the 2nd International Conference on Concrete Repair, Rehabilitation and Retrofitting, Cape Town, November, 2008.