Key Engineering Materials Vol 607 (2014) pp 21-29 Online: 2014-04-09

© (2014) Trans Tech Publications, Switzerland

doi:10.4028/www.scientific.net/KEM.607.21

Damage Identification in Beams Using Experimental Data

E.U.L. Palechor1, a, R.S.Y.C. Silva1, b, L.M. Bezerra1, c, T.N. Bittencourt2, d

1
University of Brasília, Department of Civil Engineering, Campus Darcy Ribeiro SG-12 70910-900,
Brasília, Brazil
2
University of São Paulo, Department of Civil Engineering, Avenue Prof Almeida Prado, Trav.02
number 271 Butanta. 05508-070, São Paulo, Brazil.
a
erwin@aluno.unb.br, bramon@unb.br, clmbz@unb.br, dtbitten@usp.br

Keywords: Beams; Wavelet transform; Damage.

Abstract. There are several techniques for non-destructive damage detection in structures.
However, they are costly and require a very precise analysis of the extent of the structure.
Numerical methods can help in nondestructive testing of structures, showing the possible location
of the damage and thereby decreasing the area of analysis and constituting less expensive non-
destructive tests. Outstanding among the numerical methods most used to detect damage are the
finite element method and the boundary element method. This paper presents the application of
wavelet transform for damage detection to the static response of a beam (profile I) simply supported
with a point load. Damage simulation was achieved using saw cuts in the top and bottom flanges of
the beam.

Introduction
Deterioration in structures happens frequently causing the onset of damage, even though structures
are projected and built according to appropriate specifications. The damage may occur due to many
factors such as: repeated loads, environmental risks, overload, wind loads and earthquakes. In
important structures like bridges, this damage can result in the loss of human life and considerable
socio-economic impacts. In that light, an adequate inspection to identify damage is an essential
procedure.
There are forms of damage in structures that cannot be detected by visual inspection alone and
for that reason the detection, localization and quantification of damage through a static response of
the structure (beam deformation) is one of the most important damage investigation topics. The
concept of damage identification was introduced two decades ago [1, 2].
Structural damage consists of a loss of rigidity inducing variations in both static and dynamic
responses with respect to the undamaged structure [3]. In a condition assessment study of existing
engineering structures, it is typically possible to measure the structural response to both statically
and dynamically applied loads [4].
Methods that can detect damage using only information obtained from the damaged condition of
the structure are more advantageous since the structure’s condition before damage is rarely known.
In that context, the application of wavelet-based methods can be useful. These methods detect the
singularities present in the static response caused by damage and therefore do not require
information on the condition of the structure before damage [5].

Wavelet Transform
Wavelet transform analysis is based on the idea that any signal can be divided into a series of base
functions called “waves”.
The wavelet decomposition calculates a “similarity index” between the signal and the wavelet. If
the index is huge, then, the similarity is strong; otherwise, the similarity is weak.

All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of Trans
Tech Publications, www.ttp.net. (ID: 189.61.37.249-21/04/15,06:18:50)
22 Advanced Computational Engineering and Experimenting III

For a signal f x in the range a, b , the wavelet transform is given by:

∞ (1)
, ∞ , .

Where C , is the wavelet coefficient, ψ , x is a mother wavelet function that is part of a

wavelet family. The coefficients a and b are the scale and position parameters respectively [6].
Wavelet transform includes continuous wavelet transform (CWT) and discrete wavelet transform
(DWT). The difference between them is in the form of representation of the scale, as shown below:
The CWT is defined as the sum over all the time (or space) of the signal multiplied by the scale a
with translation b. The five steps to create CWT are:

1. Choose a mother wavelet and compare it to sections along the original signal;
2. Calculate the coefficient , , which represents the similarity between the wavelet and the
signal in the interval analyzed. The results depend on the shape of the wavelet chosen (see
Fig. 1(a)).
3. The next step is to translate the wavelet to the right and repeat steps 1 and 2 until all the
signal has been covered (see Fig. 1(b)).

(b)
(a)
Figure 1. Calculation of the coefficient C [7].

4. Apply the wavelet scale chosen in the analyzed section and repeat steps 1-3.
5. Repeat steps 1-4 for all scales.

The final results will have produced coefficients for different scales. For the representation of the
results, the x axis represents the position b along the signal (time or space), the y axis represents the
scale a and the z axis the magnitude of the coefficient C.
The DWT minimizes the quantity of data generated by the CWT. The DWT chooses a subset of
scales and positions. The scales and positions are chosen based on powers of two called dyadic
scales, resulting in greater efficiency [7].
For this purpose, the scale is defined as a 2 and the translation b k2 where j, k ϵ Z and Z
is the set of integers. Using these parameters the DWT is given as:
(2)
,
/
2 2 , .

The difference between the CTW and the DWT is the representation on the scale; this difference
is summarized in the Table 1:
Table 1: Difference between CWT and DWT [8].
Continuous Time Discrete Time
Continuous Time
Continuous Analysis Discrete Analysis Discrete Analysis
1 1
, , , ,
√ √
, Δ2 , Δ 2 2 , 2
, ,
 Key Engineering Materials Vol. 607 23

Experimental tests
It is well known that the presence of damage influences the static response of the loaded structure.
This section presents the experimental model of a simply supported beam (profile I), subjected to
the action of a point load. The properties of the beam are presented in Table 2.

Table 2: Geometric properties of beam.

Perfil I -Steel 102 X 11.4
h(cm) 10.16
h0(cm) 8.68
tf(cm) 0.74
t0(cm) 0.483
c(cm) 1.59
b(cm) 6.76
Area (cm2) 14.5
Ix (cm4) 252
ix (cm) 4.17
Iy (cm4) 31.7
iy(cm) 1.48
fy(Kn/cm2) 25
E (Kn/cm2) 20000
Length (m) 6

The beams used in the experimental tests were divided into 16 elements and one LVDT was
positioned in each element. Figure 2 shows the position of load and the discretization of the beam.
Load

37.5 cm
A B

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
L1 L2 L3 L4 L5 L6 L7 L8 L9 L10 L11 L12 L3 L14 L15
LVDT
Figure 2. Position of LVDT’s

The support conditions at the beam ends are presented in Fig. 3. Figure 3a shows the restrains
in the y and Z directions and Fig. 3b shows the restrains in the x, y and Z directions.

Uz=0
Uz=0

Uy=0
Uy=0 Ux=0
(a) (b)
Figure 3. Model schematic of the beam.

Figure 4 shows the beam damage locations and dimensions. The damage is illustrated as a green
line.
24 Advanced Computational Engineering and Experimenting III

(a) V2E V2E

(b) V3E V3E

Figure 4. Damage positions and dimensions.

Damage simulation. The induced damage in the steel beam represents a fatigue crack. To simulate
this crack in a real test it would be necessary to apply load cycles during several weeks, but to
simplify the damage in the experimental model, it was simulated by making cuts in the top and
bottom flanges. For this case there are two positions of damage with different lengths, as illustrated
in Fig. 5. The first is four centimeters long (Fig. 5b) and the second, in a different location, is two
centimeters long (Fig. 5a).

(a) Damage length 20 mm (b) Damage length 40 mm

Figure 5. Induced damages by fatigue.

Measuring instruments. For the measurement of displacements, LVDT’s were positioned in the
fifteen internal nodes of the beam (see Fig. 6). The data acquisition system used was the spider
team.

(a) (b)
Figure 6. Beam instrumentation.

The load was applied to the middle of the beam (node 8) and a load cell was used to accurately
determine load intensity (see Fig. 7).
 Key Engineering Materials Vol. 607 25

(a) (b)
Figure 7. Load application.
Experimental procedure. The experimental procedure is presented in flowchart form in Fig. 8.
DAMAGE IDENTIFICATION USING A WAVELET
TRANSFORM FOR A BEAM

STATIC ANALYSIS TO EXPORT THE DISPLACEMENTS

EXPERIMENTAL TESTS-
OBTAIN THE GENERATED BY THE LOAD ALONG THE
SIMPLE BEAM
DISPLACEMENTS BOTTON FLANGE IN 17 SELECTED POINTS

APPLICATION OF THE THE 17 POINTS ARE

IMPORT DISPLACEMENTS AS VECTOR
REGULARIZATION TECHINIQUE INTERPOLATED USING THE
[POSITION, DISPLACEMENT] IN THE
( THIKONOV) CUBIC-SPLINE FUNCTION OF
MATLABPROGRAM
THE MATLAB PROGRAM.

APPLICATION OF CWT AND DWT

IN THE VECTOR OF GENERATION OF
GRAPHICS WITH THE IDENTIFICATION OF THE
REGULARIZATION USING
WAVELET COEFFICIENTS DAMAGE POSITION
WAVELET TOOLBOX OF THE
MATLAB PROGRAM

Figure 8. General procedure adopted for the location of the damage.

Static response. The static tests were carried out applying an incremental load and measuring the
displacements in the 15 nodes [9]. The load increases and the variations of the displacements of the
two damaged beams and the undamaged beam are shown in Table 3 and 4.
Table 3: Beam displacement V2E
VERTICAL DISPLACEMENT Uy(cm) - BEAM V2E
LOAD CELL (N)
LVDT DISTANCE (m) 990 2040 3060 3990 4530
SUPPORT A 0 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000
LVDT01 0.375 1.3656250 2.8312500 4.3000000 5.5750000 5.7250000
LVDT02 0.75 2.8093750 5.8656250 8.9281250 11.6125000 11.9437500
LVDT03 1.125 4.1937500 8.7781250 13.4062500 17.4437500 17.9343750
LVDT04 1.5 5.5500000 11.6062500 17.7562500 23.1437500 23.7437500
LVDT05 1.875 6.5250000 13.6562500 20.8781250 27.2093750 27.9187500
LVDT06 2.25 7.2968750 15.2718750 23.2843750 30.2937500 30.4937500
LVDT07 2.625 7.8062500 16.4312500 25.0937500 32.6500000 33.5437500
LVDT08 3 7.8437500 16.5062500 25.2156250 32.7906250 33.8437500
LVDT09 3.375 7.7812500 16.3750000 25.0500000 32.6281250 33.7562500
LVDT10 3.75 7.2812500 15.3000000 23.3937500 30.4687500 31.6218750
LVDT11 4.125 6.4968750 13.6312500 20.7906250 27.0250000 28.1156250
LVDT12 4.5 5.3687500 11.3406250 17.3468750 22.5656250 23.5312500
LVDT13 4.875 4.1312500 8.7531250 13.3468750 17.3500000 18.1156250
LVDT14 5.25 2.7750000 5.8843750 8.9250000 11.5718750 12.0843750
LVDT15 5.625 1.3750000 2.9281250 4.4125000 5.6937500 5.9406250
SUPPORT B 6 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000
26 Advanced Computational Engineering and Experimenting III

Table 4: Beam displacement V3E.

VERTICAL DISPLACEMENT Uy (cm) - BEAM V3E
LOAD CELL (N)
LVDT DISTANCE(m) 960 2040 3120 4350 4620
SUPPORT A 0 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000
LVDT01 0.375 1.4687500 2.9031250 4.2937500 5.6312500 5.7156250
LVDT02 0.75 3.2375000 6.4343750 9.5218750 12.5281250 12.7093750
LVDT03 1.125 4.7781250 9.5125000 14.1062500 18.5531250 18.8250000
LVDT04 1.5 6.3312500 12.5531250 18.6156250 24.4812500 24.8093750
LVDT05 1.875 7.5312500 14.9750000 22.2312500 29.2625000 29.6593750
LVDT06 2.25 8.5218750 16.9312500 25.0250000 32.7250000 32.8656250
LVDT07 2.625 9.2187500 18.2968750 27.0343750 35.4781250 36.0062500
LVDT08 3 9.2968750 18.4218750 27.2343750 35.7781250 36.3656250
LVDT09 3.375 9.2156250 18.2687500 27.0593750 35.6281250 36.2750000
LVDT10 3.75 9.0218750 17.5281250 25.7468750 33.8375000 34.5031250
LVDT11 4.125 7.8343750 15.4500000 22.7312500 29.8812500 30.5000000
LVDT12 4.5 6.8343750 13.2218750 19.3500000 25.3687500 25.9281250
LVDT13 4.875 5.5062500 10.4187500 15.0937500 19.7281250 20.1687500
LVDT14 5.25 4.0062500 7.2937500 10.3937500 13.4750000 13.7718750
LVDT15 5.625 1.5031250 3.1843750 4.6843750 6.1437500 6.2781250
SUPPORT B 6 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000

For better localization of damage, Table 5 shows the correspondences between the node
number and distance of the damage.
Table 5: Damage Location (node-distance)
Damage Location- from the left support distance
Beam Position (m) Node (#) TDW Node (#) TCW
V2E 1.5 25 250
V3E 1.8 and 4.2 30 and 70 300 and 700

Damage identification using Discrete Wavelet Transform (DWT). To beam V2E (3.99 kN),
three different mother wavelets were applied: coiflet (coif2), Symlet (sym6) and Biorthogonal
(bior6.8). The results obtained for the beam V2E are shown in Figs. 9 to 11.

Figure 9. Damage Location-TDW Figure 10. Damage Location-TDW

(coif2)- beam V2E. (sym6)- beam V2E.
 Key Engineering Materials Vol. 607 27

Figure 11. Damage Location-TDW (bio6.8)- beam V2E.

The signal transformed by the DWT showed small discontinuities along the beam, but in the
damaged position corresponding to node 25, the wavelet coefficients achieved a large amplitude.
The DWT presents peaks at the ends that are the result of the discontinuity of the support.
The same mother wavelet applied to beam V2E was applied to beam V3E (2.04 kN). The results
obtained for beam V3E are shown in Figs. 12 to 14.

Figure 12. Damage Location-TDW Figure 13. Damage Location-TDW

(bio6.8)- beam V3E. (sym6)- beam V3E.

Figure 14. Damage Location-TDW (coif2)- beam V3E.

For the V3E beam the positions corresponding to the damage are at node 30 and node 70, the
peak representing the damage at node 30 is around the node 26, with an error of about 20
centimeters in the detection of the damage. The damage at node 70 was identified. Also it can be
seen that at node 63, the wavelet coefficients gave a false indication of damage. That indication
could be associated to measurement errors during the tests.
Damage identification using Continuous Wavelet Transform (CWT). The same procedure used
in the DWT was applied to the CWT. The mother wavelets used in the DWT were: Daubechies
(db5), Symlet (sym8) and Gauss (gaus8), for a load of 3.9 kN. The results for beam V2E are shown
in Figs. 15 to 17.

Damage Location

Damage Location

(a) 3D 2D
Figure 15. Damage Location-TCW (db5)- beam V2E.
28 Advanced Computational Engineering and Experimenting III

Damage Location

Damage Location

(a)3D (b) 2D
Figure 16. Damage Location-TCW (sym8) - beam V2E.

Damage Location

Damage Location

(a) 3D (b) 2D
Figure 17. Damage Location-TCW (gaus8) - beam V2E.
The signal transformed by the CWT showed small discontinuities along the beam V2E, but at the
damaged position corresponding to the node 250, the wavelet coefficients achieved large
amplitudes. The CWT also presented peaks at the ends that are the result of the discontinuity in the
support.
The mother wavelets used for the beam V3E were the same as for the beam V2E, but the load
used was 3.12 kN. The results for the beam V3E are shown in Figs. 18 to 20.

Damage Location

Damage Location

(a)3D (b) 2D
Figure 18. Damage Location-TCW (db5) - beam V3E.

Damage Location

Damage Location

(a) 3D (b) 2D
Figure 19. Damage Location-TCW (sym8) - beam V3E.
 Key Engineering Materials Vol. 607 29

Damage Location

Damage Location

(a) 3D (b) 2D
Figure 20. Damage Location-TCW (gaus8) - beam V3E.

For the V3E beam, the positions corresponding to the damage are at nodes 300 and 700, but
the peaks corresponding to the damage appeared around the nodes 260 and 725. The differences in
the damage detections are 24cm and 15cm respectively from the real damage position.

Conclusions
This paper has presented an alternative methodology to the traditional damage identification
methods, because Wavelet Transform only requires the damage response of the structure, while the
traditional methods need to make comparisons between the intact and damaged responses.
The Continuous and Discrete Wavelet Transform using the mother wavelet coif2, sym6,
boir6.8, db10, db9, sym3, and sym7 were very efficient in the process of identifying damage in the
beam. The wavelet transform showed that they are sensitive to the geometric discontinuities caused
by the boundary conditions; at these points the wavelet coefficients showed large amplitudes.
The beam with only one 4 cm cut presented better results than the beam with 2 points of local
damage in the form of 2cm cuts.
Structures in service require maintenance and continuous monitoring to ensure their integrity
and functionality. During recent years much research has been done and several methods for
damage identification discussed. The advances offered by using wavelets are one alternative for
achieving good maintenance, because sudden changes in the displacements and mode shapes can
indicate a possible damaged area.

References
[1] H. Li, T. Yi, M. Gu and L. Huo: Materials International Vol. 19 (2009), 461-470.
[2] I. Y. Choi: Damage Identification Techniques for Bridges Using Static Response. PhD. Thesis,
University of Hanyang, Seoul, 153p. 2002.
[3] A. Greco and A. Pau: Computers & Structures Vols. 92-93 (2012), 328-336.
[4] R. Tipireddy, H.A. Naserellah, C.S Manohar and A. Kalman: Probabilistic Engineering
Mechanics Vol. 24 (2009), 60-74.
[5] R.S.Y.C. Silva: Determinação de patologias estruturais utilizando modelagem numérica e
transformadas de wavelet. MSc. Thesis. University of Brasilia. Department of Civil
Engineering, 2011.
[6] N. Wu and Q. Wang: International Journal of Engineering Science Vol. 49 (2011), 253-261.
[7] M. Misiti: Wavelet Toolbox for use with MATLAB. 2002.
[8] A.V. Ovanesova: Application of wavelet to crack detection in frame structures. PhD Thesis.
University of Porto Rico, 235p, 2000.
[9] M.A.B. Abdo: Engineering Structures Vol. 34 (2012), 124-131.
Advanced Computational Engineering and Experimenting III
10.4028/www.scientific.net/KEM.607

Damage Identification in Beams Using Experimental Data

10.4028/www.scientific.net/KEM.607.21