Professional Documents
Culture Documents
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
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
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.
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
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).
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
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.
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