Special Issue Article

Advances in Mechanical Engineering

2017, Vol. 9(3) 1–6
Ó The Author(s) 2017
Damage localization for beam DOI: 10.1177/1687814017695956
journals.sagepub.com/home/ade
structure by moving load

Qiuwei Yang1, JK Liu2, BX Sun1 and CF Liang1

Abstract
A static algorithm is presented in this article for damage localization in beam structures by moving load. Central to the
damage localization approach is the combination of the moving load technique and the pure bending theory. The first
advantage of the presented algorithm is that only a few sensors are needed in the static test. The second advantage of
the presented algorithm is that only the deflection parameters of the current structure are needed in damage localiza-
tion process. The proposed method is applied to a simple supported beam and its feasibility is verified using damage
simulations. It was found that the defect location can be identified by inspecting the sudden increase in the curvature of
the deflection curve for the beam. The results showed that the presented algorithm is effective on the localization of sin-
gle damage or multiple damages in the beam structure without baseline data.

Keywords
Damage localization, moving load, static test, deflection, curvature

Date received: 7 September 2016; accepted: 6 February 2017

Academic Editor: Simon Laflamme

Introduction mode shape (measured or modeled) for beam-like struc-

tures. Kopsaftopoulos and Fassois11 proposed a
Many load-carrying structural systems have been in use vibration-based statistical time series method that is
for tens or even hundreds of years. It is unavoidable for capable of effective damage detection, precise localiza-
these structures to continuously accumulate damage in tion, and magnitude estimation within a unified sto-
their service environment. Thus, it is most desirable chastic framework. Sung et al.16 present a damage
that this damage be detected at the earliest possible detection method for beam-like structure by the nor-
stage to assure safety. Many research studies have been malized uniform load surface (NULS) curvature
conducted during the past years in the area of damage
obtained by modal flexibility. It has shown that
detection–based structural response characteristics with
changes in NULS curvature only occur at damaged ele-
different algorithms and useful databases.1–3 Generally,
ments in the structure. Xiang et al.17 used the changes
the existing damage detection approaches can be clari-
fied into two major categories, that is, the dynamic
detection methods and the static detection methods. 1
Department of Civil Engineering, Shaoxing University, Shaoxing, P.R.
Significant efforts have already been devoted to devel- China
oping dynamic detection methods using various vibra- 2
Department of Mechanics, Sun Yat-sen University, Guangzhou, P.R.
tion characteristics, such as natural frequencies,4–6 China
mode shapes,7–11 modal flexibility,12–15 curvature,11,16
Corresponding author:
modal strain energy,17–19 and moving load responses.20–22 Qiuwei Yang, Department of Civil Engineering, Shaoxing University,
Radzieński et al.10 defined new damage indicators Shaoxing 312000, P.R. China
based on the change in natural frequencies and any Email: yangqiuwei79@gmail.com; yangqiuwei@outlook.com

Creative Commons CC-BY: This article is distributed under the terms of the Creative Commons Attribution 3.0 License
(http://www.creativecommons.org/licenses/by/3.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 Advances in Mechanical Engineering

in damage-induced curvature mode shape and natural

frequency to form a hybrid damage detection approach
to detect damages on the surface of conical shell. Xu
et al.18 developed an energy damage detection strategy
by disposing strain responses for a long-span cable-
stayed bridge. Yi et al.19 proposed a multi-stage method
based on ‘‘energy-damage’’ theory using the wavelet
packet transform (WPT) and the artificial neural net-
work (ANN) for structural damage diagnosis. Using
modal strain energy, Cha and Buyukozturk20 devel-
oped hybrid multi-objective optimization algorithms to
detect damages in various three-dimensional steel struc- Figure 1. Simple supported beam and corresponding bending
tures. Khorram et al.21 compared the performances of moment diagram under two symmetrical concentrated forces.
two wavelet-based damage detection approaches to find
the location and the size of a crack in a beam subjected Theory
to a moving load. Using Hilbert–Huang transform,
Figure 1 presents a simple supported beam ‘‘A–B’’
Roveri and Carcaterra22 presented a novel method for
under two symmetrical concentrated forces F and the
damage detection of bridge structures under a traveling
corresponding bending moment diagram. Apparently,
load. The technique uses a single-point measurement
the bending moment at any cross section in the segment
and is able to identify the presence and the location of
‘‘1 2 n’’ of this beam is a constant (M = F 3 Dx), as
the damage along the beam. Cavadas et al.23 discussed
shown in Figure 1.
the application of data-driven methods on moving load
According to the theory of material mechanics, the
responses to detect the occurrence and the location of
curvature of the deflection curve for this beam can be
damage. However, these dynamic detection algorithms
calculated by
are faced with some difficulties in applications.24 One is
that the precision of dynamic data are affected by many M
factors such as stiffness, mass, and damping. Therefore, cv = ð1Þ
EI
many dynamic approaches unavoidably suppose that
there is no damping and the mass is unchanged before where cv is the curvature, M is the bending moment, E
and after damage. The next issue is that modal data for is the elastic modulus, and I is the moment of inertia of
some stiff civil structures are difficult to be measured the beam’s cross section with respect to the transversal
accurately, especially for high-order modes. In view of axis. From equation (1), one can conclude that the cur-
this, the static detection algorithms25–34 also attracted vature cv of the deflection curve is a constant if EI is
much attention in civil engineering area in recent years constant for the beam. When EI is reduced due to struc-
due to their simplicity for implementation. However, tural damage, the curvature cv of the deflection curve at
most of these methods need a detailed finite element the damaged location will increase accordingly. This
model (FEM) or the baseline parameters of the intact means that the defect location for this beam can be
system. It is sometimes impossible to obtain a detailed identified by inspecting the sudden increase in curvature
FEM or the baseline data for many existing structures. of the deflection curve.
Thus, a simple and effective localization method that On the other hand, the curvature can be calculated
does not need an FEM or the baseline data of the intact by the central difference technique of the deflection vec-
structure will be welcomed. tor d for the beam as follows
For this reason, this article presents a static method
for damage localization in beam structures by moving d i + 1  2d i + d i1
civ = ð2Þ
load. The key point of the damage location algorithm (Dx)2
lies in the combination of the moving load technique
and the pure bending theory. The first advantage of the where di is the ith coefficient of d, civ is the correspond-
presented algorithm is that only a few sensors are ing curvature, and Dx is the length of structural ele-
needed in the static test. The second advantage of the ment. The deflection vector consists of the vertical
presented algorithm is that only the deflection para- displacement data of the nodes in the beam, as shown
meters of the current structure are needed in damage in Figure 1. Equation (2) shows that damage identifica-
localization process. The proposed method is verified tion can be carried out if the deflection parameters of
by a numerical example to show its feasibility. It has the beam were measured through static testing.
been shown that the presented algorithm is effective on However, it is unrealistic and uneconomic to place
the localization of single damage or multiple damages many sensors in the static test of the beam. According
in the beam structure. to reciprocal theorem of displacement in structural
Yang et al. 3

Table 1. Damage cases in the example.

Damage case Element number Stiffness reduction (%)

Case 1 18 10
Case 2 18 20
Case 3 18 30
Case 4 18 40
Case 5 9, 18 10, 20
Case 6 9, 18 20, 30

Step 1. Measure the deflection vectors (d1 and d2)

for the two load cases as shown in Figure 3(a) and
Figure 2. Reciprocal theorem of displacement: (a) the load F is (b) using the moving load technique.
at node 1 and (a’) the load F is at node i. Step 2. Compute the deflection vector (d) of the
beam as shown in Figure 1 or 3(c) by equation (3).
Step 3. Compute the curvatures (civ ) of the deflection
curve for the beam by equation (2).
Step 4. Evaluate the possible damage locations
according to the sudden increases in the
curvatures (civ ).

Numerical example
Figure 4 presents a simple supported beam used to
demonstrate the feasibility of the moving load method
for damage localization. The beam is divided into 36
segments and the length of each segment is
Dx = 0:04 m. The physical parameters of this beam are
as follows: Young’s modulus E = 200 GPa, density
r = 7800 kg=m3 , moment of inertia I = 2:133 3
Figure 3. Superposition principle of load: (a) the load F is at
node 1, (b) the load F is at node n, and (c) the superposition of
1010 m4 , and cross-sectional area A = 1:6 3 104 m2 .
load cases (a) and (b). Six damage cases as shown in Table 1 are studied in the
example. Case 1: element 18 is damaged with a stiffness
mechanics, this difficulty can be avoided by increasing loss of 10%. Case 2: element 18 is damaged with a stiff-
the loading conditions instead of increasing the number ness loss of 20%. Case 3: element 18 is damaged with a
of measurement sensors. For example, as shown in stiffness loss of 30%. Case 4: element 18 is damaged
Figure 2, the deflection of node ‘‘i’’ (Di) in Figure 2(a) with a stiffness loss of 40%. Case 5: elements 9 and 18
equals the deflection of node ‘‘1’’ (D1) in Figure 2(a#). are damaged with stiffness losses of 10% and 20%,
This means that only a sensor is needed in measure- respectively. Case 6: elements 9 and 18 are damaged
ment of the deflection of a beam under a concentrated with stiffness losses of 20% and 30%, respectively. For
force by this moving load technique. each damage case, suppose that the deflections at all
On the other hand, according to superposition prin- nodes as shown in Figure 4 are measured by moving
ciple in linear elastic theory, the deflection of the beam in load (F = 5 N). Note that only two sensors are needed
Figure 1 under two symmetrical concentrated forces will in the static test for this example. Certainly, the reliabil-
be equal to the summation of the deflections obtained by ity of the damage localization result will increase as the
the two load cases as shown in Figure 3(a) and (b). Using number of sensors increases. From steps 1 to 3 of the
the above moving load technique, the deflection vectors proposed method, the values of curvatures civ for nodes
in Figure 3(a) and (b) can obtained as d1 and d2, respec- 1–35 in damage cases 1–4 are shown in Table 2 and
tively. And then the deflection vector of the beam as Figure 5. One can see from Table 2 and Figure 5 that
shown in Figure 1 or 3(c) can be calculated as nodes 17 and 18 have larger curvatures than the other
nodes. The sudden increases in curvatures indicate that
d = d1 + d2 ð3Þ element 18 is damaged since nodes 17 and 18 are
exactly associated with the 18th element as shown in
In the end, a summary of the overall process for this Figure 4. Figure 6 presents the damage localization
algorithm is given as follows: results for damage cases 1–4 with 1% noise on
4 Advances in Mechanical Engineering

Figure 4. The simple supported beam and sensor layout.

Table 2. Curvature values when element 18 is damaged with since their corresponding nodes 8, 9, 17, and 18 have
10%, 20%, and 40% stiffness reductions (no noise). larger curvatures than the other nodes. When 1% noise
is considered, the damage localization result for small
Node Curvatures Curvatures Curvatures Curvatures
number for case 1 for case 2 for case 3 for case 4 damage case 5 is unreliable from Figure 8. For large
damage case 6, the curvature indexes in Figure 8 can
1 0.0039 0.0039 0.0039 0.0039 still indicate the damage locations by the notable
2 0.0047 0.0047 0.0047 0.0047 increases in the curvatures. In addition, one can con-
3 0.0047 0.0047 0.0047 0.0047
4 0.0047 0.0047 0.0047 0.0047 clude that the curvatures will increase as the damage
5 0.0047 0.0047 0.0047 0.0047 extents increase from Figures 5–8. It has been shown
6 0.0047 0.0047 0.0047 0.0047 that the proposed method is effective on the localiza-
7 0.0047 0.0047 0.0047 0.0047 tion of single damage or multiple damages in the beam
8 0.0047 0.0047 0.0047 0.0047 structure.
9 0.0047 0.0047 0.0047 0.0047
10 0.0047 0.0047 0.0047 0.0047 It must be noted that the proposed method has some
11 0.0047 0.0047 0.0047 0.0047 drawbacks as follows. The first is that the proposed
12 0.0047 0.0047 0.0047 0.0047 approach can detect only damages that are between the
13 0.0047 0.0047 0.0047 0.0047 two loading locations as shown in Figure 1. It is sug-
14 0.0047 0.0047 0.0047 0.0047
gested to put the loads as close as possible to the sup-
15 0.0047 0.0047 0.0047 0.0047
16 0.0047 0.0047 0.0047 0.0047 port. This, in turn, complicates the things because
17 0.0049 0.0057 0.0053 0.0063 deflections close to the support are extremely small and
18 0.0049 0.0057 0.0053 0.0063 difficult to be measured with good accuracy. The sec-
19 0.0047 0.0047 0.0047 0.0047 ond drawback is that the proposed method can detect
20 0.0047 0.0047 0.0047 0.0047
21 0.0047 0.0047 0.0047 0.0047
only notable stiffness reduction in the structure. If the
22 0.0047 0.0047 0.0047 0.0047 beam is a reinforced concrete one and rebars are cut
23 0.0047 0.0047 0.0047 0.0047 at a certain location, the safety is dramatically
24 0.0047 0.0047 0.0047 0.0047 reduced, while the stiffness is only slightly affected.
25 0.0047 0.0047 0.0047 0.0047 This type of engineering disease cannot be detected
26 0.0047 0.0047 0.0047 0.0047
27 0.0047 0.0047 0.0047 0.0047 by the proposed method. In addition, the support
28 0.0047 0.0047 0.0047 0.0047 conditions of the beam structure have some impact
29 0.0047 0.0047 0.0047 0.0047 on the proposed method. If the support conditions
30 0.0047 0.0047 0.0047 0.0047 are non-ideal, the dynamic flexibility matrix obtained
31 0.0047 0.0047 0.0047 0.0047 by the experimental modal parameters can be used to
32 0.0047 0.0047 0.0047 0.0047
33 0.0047 0.0047 0.0047 0.0047 construct the virtual pure bending state for the beam
34 0.0047 0.0047 0.0047 0.0047 structure. This issue will be studied in the further
35 0.0039 0.0039 0.0039 0.0039 research.
The bold values denote the curvature values corresponding to the
damaged element.
Conclusion
simulated data. One can see that the curvature indexes A new deflection-based method has been proposed for
are obviously affected by the noise, and the damaged beam damage localization by moving load. This
location (element 18) cannot be clearly identified by method has two advantages: (1) only a few sensors are
inspecting the corresponding curvatures for small dam- needed in static test and (2) only the deflection para-
age cases 1 and 2. For large damage cases 3 and 4, the meters of the current structure are needed in damage
proposed method can still determine the damaged loca- localization. To confirm the feasibility of the proposed
tion even if the measurement error is considered. method, a simple supported beam was investigated for
For damage cases 5 and 6, elements 9 and 18 can be several damage scenarios. The results showed that the
determined from Figure 7 without noise to be damaged presented algorithm is effective to identify structural
Yang et al. 5

Figure 5. Curvature values when element 18 is damaged (no noise).

Figure 6. Curvature values when element 18 is damaged (1% noise).

Figure 7. Curvature values when elements 9 and 18 are Figure 8. Curvature values when elements 9 and 18 are
damaged (no noise). damaged (1% noise).

Declaration of conflicting interests

damage locations. The presented algorithm may have The author(s) declared no potential conflicts of interest with
good prospects in defect localization for beam struc- respect to the research, authorship, and/or publication of this
tures without baseline data. article.
6 Advances in Mechanical Engineering

Funding 16. Sung SH, Jung HJ and Jung HY. Damage detection for
The author(s) disclosed receipt of the following financial sup- beam-like structures using the normalized curvature of a
port for the research, authorship, and/or publication of this uniform load surface. J Sound Vib 2013; 332: 1501–1519.
article: This work was supported by National Natural Science 17. Xiang J, Matsumoto T, Wang Y, et al. Detect damages
Foundation of China (41272345, 11202138 and 41572305). in conical shells using curvature mode shape and wavelet
finite element method. Int J Mech Sci 2013; 66: 83–93.
18. Xu ZD, Liu M, Wu Z, et al. Energy damage detection
References strategy based on strain responses for long-span bridge
1. Doebling SW, Farrar CR and Prime MB. A summary structures. J Bridge Eng 2011; 16: 644–652.
review of vibration-based damage identification methods. 19. Yi TH, Li HN and Sun HM. Multi-stage structural dam-
Shock Vib Digest 1998; 30: 91–105. age diagnosis method based on ‘‘energy-damage’’ theory.
2. Carden EP and Fanning P. Vibration based condition Smart Struct Syst 2013; 12: 345–361.
monitoring: a review. Struct Health Monit 2004; 3: 20. Cha YJ and Buyukozturk O. Structural damage detection
355–377. using modal strain energy and hybrid multiobjective opti-
3. Prinaris A, Alampalli S and Ettouney M. Review of mization. Comput-Aided Civ Inf 2015; 30: 347–358.
remote sensing for condition assessment and damage 21. Khorram A, Bakhtiari-Nejad F and Rezaeian M. Com-
identification after extreme loading conditions. In: Pro- parison studies between two wavelet based crack detec-
ceedings of the 2008 structures congress, Vancouver, BC, tion methods of a beam subjected to a moving load. Int J
Canada, 24–26 April 2008. New York: ASCE. Eng Sci 2012; 51: 204–215.
4. Messina A, Williams JE and Contursi T. Structural dam- 22. Roveri N and Carcaterra A. Damage detection in struc-
age detection by a sensitivity and statistical-based tures under traveling loads by Hilbert–Huang transform.
method. J Sound Vib 1996; 216: 791–808. Mech Syst Signal Pr 2012; 28: 128–144.
5. Ashokkumar CR and Lyengar NGR. Partial eigenvalue 23. Cavadas F, Smith IFC and Figueiras J. Damage detec-
assignment for structural damage mitigation. J Sound tion using data-driven methods applied to moving-load
Vib 2011; 330: 9–16. responses. Mech Syst Signal Pr 2013; 39: 409–425.
6. Yang ZC and Wang L. Structural damage detection by 24. Wang X, Hu N, Fukunaga H, et al. Structural damage
changes in natural frequencies. J Intel Mat Syst Str 2010; identification using static test data and changes in fre-
21: 309–319. quencies. Eng Struct 2001; 23: 610–621.
7. Shi ZY, Law SS and Zhang LM. Damage localization by 25. Sanayei M and Onipede O. Assessment of structures
directly using incomplete mode shapes. J Eng Mech using static test data. AIAA J 1991; 29: 1156–1179.
2000; 126: 656–660. 26. Banan MR, Banna MR and Hjelmstad KD. Parameter
8. Ho YK and Ewins DJ. On structural damage identifica- estimation of structures from static response, I: computa-
tion with mode shapes. In: Proceeding of COST F3 con- tional aspects. J Struct Eng 1994; 120: 3243–3258.
ference on system identification and structural health 27. Banan MR, Banna MR and Hjelmstad KD. Parameter
monitoring, Madrid, Spain, June 2000, pp.677–686. estimation of structures from static response, II: numer-
9. Zhu HP, Li L and He XQ. Damage detection method ical simulation studies. J Struct Eng 1994; 120:
for shear buildings using the changes in the first mode 3259–3283.
shape slopes. Comput Struct 2011; 89: 733–743. 28. Hjelmstad KD and Shin S. Damage detection and assess-
10. Radzieński M, Krawczuk M and Palacz M. Improve- ment of structures from static response. J Eng Mech
ment of damage detection methods based on experimen- 1997; 123: 568–576.
tal modal parameters. Mech Syst Signal Pr 2011; 25: 29. Chou JH and Jamshid G. Genetic algorithm in structural
2169–2190. damage detection. Comput Struct 2001; 79: 1335–1353.
11. Kopsaftopoulos FP and Fassois SD. A functional model 30. Bakhtiari-Nejad F, Rahai A and Esfandiari A. A struc-
based statistical time series method for vibration based tural damage detection method using static noisy data.
damage detection, localization, and magnitude estima- Eng Struct 2005; 27: 1784–1793.
tion. Mech Syst Signal Pr 2013; 39: 143–161. 31. Chen XZ, Zhu HP and Chen CY. Structural damage
12. Wu D and Law SS. Model error correction from trun- identification using test static data based on grey system
cated modal flexibility sensitivity and generic parameters. theory. J Zhejiang Univ Sci 2005; 6A: 790–796.
I: simulation. Mech Syst Signal Pr 2004; 18: 1381–1399. 32. Kouchmeshky B, Aquino W, Bongard JC, et al. Co-evo-
13. Wu D and Law SS. Eigen-parameter decomposition of lutionary algorithm for structural damage identification
element matrices for structural damage detection. Eng using minimal physical testing. Int J Numer Meth Eng
Struct 2007; 29: 519–528. 2007; 69: 1085–1107.
14. Yang QW and Liu JK. Damage identification by the 33. Yang QW and Sun BX. Structural damage localization
eigenparameter decomposition of structural flexibility and quantification using static test data. Struct Health
change. Int J Numer Meth Eng 2009; 78: 444–459. Monit 2011; 10: 381–389.
15. Yang QW. A new damage identification method based 34. Abdo MAB. Parametric study of using only static
on structural flexibility disassembly. J Vib Control 2011; response in structural damage detection. Eng Struct 2012;
17: 1000–1008. 34: 124–131.