Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Research ArticleCharged System Search Algorithm Utilized forStructural Damage Detection
Zahra Tabrizian1 Gholamreza Ghodrati Amiri2 and Morteza Hossein Ali Beigy1
1 College of Civil Engineering Babol Noshirvani University of Technology Babol Iran2 Centre of Excellence for Fundamental Studies in Structural Engineering Iran University of Science and TechnologyNarmak Tehran Iran
Correspondence should be addressed to Gholamreza Ghodrati Amiri ghodratiiustacir
Received 12 June 2013 Revised 8 April 2014 Accepted 9 April 2014 Published 20 May 2014
Academic Editor Gyuhae Park
Copyright copy 2014 Zahra Tabrizian et alThis is an open access article distributed under theCreative CommonsAttribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
This paper presents damage detection and assessment methodology based on the changes in dynamic parameters of a structuralsystemThemethod is applied at an element level using a finite elementmodel According to continuumdamagemechanics damageis represented by a reduction factor of the element stiffness A recently developed metaheuristic optimization algorithm known asthe charged system search (CSS) is utilized for locating and quantifying the damaged areas of the structure In order to demonstratethe abilities of this method three examples are included comprising of a 10-elements cantilever beam a Bowstring plane truss anda 39-element three-story three-bay plane frame The possible damage types in structures by considering several damage scenariosand using incomplete modal data are modeled Finally results are obtained from the CSS algorithm by detecting damage in thesestructures and compared to the results of the PSOPC algorithm In addition the effect of noise is shown in the results of the CSSalgorithm by suitable diagrams As is illustrated this method has acceptable results in the structural detection damage with lowcomputational time
1 Introduction
Many civil engineering structures like other structuresdesigned with older codes have been suffering damage anddeterioration in recent years which caused reduction intheir performance Due to this effect damage assessment ofthese structures is becoming increasingly essential in order todetermine their safety and reliabilityThemethodmentionedin this paper is either visual or nondestructive which leadus to the best answers Considering global damage detectionmethods applicable to complex structures methods basedon modal testing Ewins 1984 [1] and signal processing ismade to identify damage in civil engineering and mechanicsThese methods study changes in the dynamic characteristicsof the structure such as natural frequencies andmode shapeswhich were studied by Doebling et al 1998 [2] and Huet al 2001 [3] The undamaged and damaged structurecompare with the identification of the location and theseverity of damage Natural frequencies and mode shapes aredamage indicators which are used as the dynamic parameters
Methods based on the measurement of natural frequenciesare very simple since this parameter can be determinedby measuring at only one point of the structure (Salawu1997 [4] and Bicanic and Chen 1997 [5]) Some derivativesof mode shapes such as mode shape curvatures are moresensitive to small perturbations than modal displacementsand therefore can also be used to detect damage (Pandey etal 1991 [6]) However their applicability is minimal sincetheir estimation from experimental data is very difficult andtherefore they are very uncertain from a statistical point ofview Mottershead and Friswell 1993 have studied modelupdating methods In addition to the updating methodsbased on traditional optimization techniques new methodshave been developed in inverse procedures which are calledevolutionary algorithms [7]
In many structures some local damagemay occur duringtheir functional age In order to develop the efficiency of thestructures it is necessary to properly identify the damageplaces and their severity and then repair them There are
Hindawi Publishing CorporationShock and VibrationVolume 2014 Article ID 194753 13 pageshttpdxdoiorg1011552014194753
2 Shock and Vibration
many methods that have been introduced to correctly findout the place and severity of structural damage
Perera and Torres 2006 studied a nondestructive globaldamage detection and assessment methodology based onthe changes in frequencies and mode shapes of vibrationof a structural system It was shown that the proposed GAyields a suitable damage places and severity detection fromtraditional methods [8]
A two-stage method for determining place and severityof multiple structural damage by combining the adaptiveneurofuzzy inference system (ANFIS) and particle swarmoptimization (PSO) has been proposed by Fallahian andSeyedpoor [9] In a studyKoh andDyke [10] have determinedthe place and severity of multiple damage by iterativelysearching for a combination of structural responses thatmaximizes a correlation coefficient named the multipledamage location assurance criterion (MDLAC) via a geneticalgorithm (GA) Damage detection by a hybrid techniqueincluding a real-parameter genetic algorithm and grey rela-tion analysis has been presented by He and Hwang [11]They used first a grey relation analysis to exclude impossibledamage places such that the number of design variables couldbe reduced Second a real-parameter genetic algorithm wascombined with simulated annealing for finding the actualdamage The damage identification of a beam-like structurehas been formulated as an optimization problem and a GAhas been employed to find the damage place and severityby minimizing the cost function which is based on thedifference of measured and calculated natural frequencies Atwo-stage method for determining the place and severity ofmultiple structural damage by using an information fusiontechnique and a GA has been presented by Guo and Li [12]In the first stage the damage is localized by using evidencetheory and then a microsearch genetic algorithm (MSGA)has been planned to determine the damage size Yu and Chendetermined the place and severity of damage in a bridgeFirst two objective functions are defined One is defined asminimizing the sum of differences between the modal databefore and after damage in traditional wayThe other is newlydefined based on modal flexibility which is combined withanother function able to predict damage location Secondlyan improved particle swarm optimization (PSO) algorithmis developed based on the macroeconomic strategies andused to solve the multiple-objective optimization problemon bridge damage identification The results show that theprocedure is very promising for locating and quantifyingdamaged elements of bridge structures and considerablyimproves predictions based on the modal flexibility as wellas the PSO method [13] An approach for detecting damagein structural members based on continuum damage modelusing an algorithmnamedBig Bang-Big Crunch (BB-BC) haspresented suitable accuracy in results [14]
In recent years the use of metaheuristic methods hasincreased and great efforts to increase the power of algo-rithms (Efficiency) and reduce optimization time (moreconvergence speed) have been done Recently a new androbust algorithm was presented by Kaveh and Talatahari socalled charged system search [15] This algorithm is based onsome laws from electrostatics and the Newtonian mechanics
The charged system search (CSS) utilizes a number of chargedparticles (CP) which affect each other based on their fitnessvalues and separation distances consideringCoulombGaussandNewtonian lawsThe resultant forces and themotion lawsdetermine the new location of the particles The harmonysearch-based handling approach is utilized for controlling thevariable constraint
In this paper the CSS algorithm is utilized to predict thedamage place and severity for different types of structuresThe structure is modeled with the finite element method andthe damage identification is performed at element level withincomplete modal data
In order to show the power of the CSS algorithm in struc-tural damage detection the obtained results are comparedto the results of the PSOPC algorithm [16] PSOPC algo-rithm is a particle swarm optimization method with passivecongregationrsquos capacity [17] High abilities of this algorithmwith acceptable results in structural damage detection areillustrated [16]
The results show that the proposed method is capableof detecting the location and severity of diagnosis evenin large-scale structures with a large number of damagedelements it achieved acceptable resultsThe effect of the noiseis considered by assigning the noise in natural frequencieson the results of the CSS algorithm The graphs indicate thatthe noise in input data may reduce the accuracy of damagedetection
2 Theory
Existing Structural damage detection techniques can beclassified into two major categories the dynamic and staticmethods [18] Both methods are based on the finite elementutilizing experimental test data requiring dynamic and statictest data correspondingly In addition dynamic methodshave shown their advantages in comparison with static onesThe natural frequencies of a structure can be considered tobe valuable with the dynamic data Determining the levelof correlation between the calculated and predicted naturalfrequencies can provide a simple tool for finding the place andseverity of structural damage [19] Structural damage detec-tion is calculated with changes in structural characteristicsand it is possible with global evaluating Doebling et al havediscussed vibration-based techniques in a literature review[20] In this section the construction of dynamics of damagedstructures is discussed
The parameter vector used for evaluating the correlationcoefficients (the ratio of the changes first 119899119891 natural fre-quency Δ119865 due to structural damage) is
Δ119865 =
119865119873 minus 119865119864
119865119873
(1)
where 119865119873 and 119865119864 denote the natural frequency vectors of theundamaged and damaged structure Similarly the parametervector predicted from an analytical model can be definedcorrespondingly as
120575119865 (119883) =
119865119873 minus 119865 (119883)
119865119873
(2)
Shock and Vibration 3
where 119865(119883) is a natural frequency vector that can be pre-dicted from an analytic model and 119883 = [119909
1 1199092 119909
119899]
represents a damage variable vector containing the damageseverity of all 119899 structural elements Given a pair of param-eter vectors one can estimate the level of correlation inseveral ways An efficient way is to evaluate a correlation-based index termed themultiples damage location assurancecriterion (MDLAC) and covered in the following form [19]
MDLAC (119883) =10038161003816100381610038161003816Δ119865119879sdot 120575119865 (119883)
10038161003816100381610038161003816
2
(Δ119865119879sdot Δ119865) (120575119865
119879(119883) sdot 120575119865 (119883))
0 lt MDLAC lt 1
(3)
Two frequency change vectors are compared with MDLACone calculated from the structural tests and the other froma structural model analysis When the vector of analyticalfrequencies becomes the same as the frequency vector ofthe damaged structure MDLAC will be maximal That is119865(119883) = 119865 so considering this theory can be used to finda set of damage variables maximizing the MDLAC using anoptimization algorithm
Find 119883 = [119909119894 1199092 119909119899]
Maximize 119908 (119883) = MDLAC (119883) (4)
The damage severity can take values only from the set thatis given from [ 0 1 ] a set of continuous values Moreoverthe objective function that should be maximized is 119908 Asmentioned the damage occurrence in a structural elementdecreases the element stiffness Thus one of the methods forthe damage identification problem is simulation damage bydecreasing one of the stiffness parameters of the element suchas the modulus of elasticity (119864) cross-sectional area (119860) andinertia moment (119868) In this study the damage variables aredefined via a relative reduction of the elasticity modulus ofan element as
119870119889119894 = (1 minus 119909119894)119870
ℎ119894
(5)
119870119889119894 is the stiffness matrix of damaged element 119894 and119870ℎ119894 is the
stiffness matrix of healthy element 119894119864 is the primary modulus of elasticity and 119864119894 is the final
modulus of elasticity of the 119894th element The MDLAC as anobjective function for the optimization algorithm is moresensitive to damaged elements than undamaged elementsIt means that this method can find the true place of thedamaged elements but it may find an undamaged element asa damaged one Therefore in this study a new function andnew optimization algorithm is discussed the new function ispresented as [21]
obj (119883) = 1
119899119891
119899119891
sum
119894=1
min (119891119909119894 119891119864119894)max (119891119909119894 119891119864119894)
(6)
where119891119909119894119891119864119894 are the 119894th components of vectors 119865(119883) and 119865119864correspondingly
The obj(119883) function can rapidly find the locations ofhealthy elements when compared to theMDLAC however it
is very probable that it finds a damaged element as a healthyone Therefore in this study a combinational function of(3) and (6) called here the efficient correlation-based index(ECBI) is used as [21]
ECBI (119883) = 1
2
(MDLAC (119883) + obj (119883)) (7)
3 Charged System Search Algorithm
A new type of metaheuristic algorithms is introduced byKaveh and Talatahari [15] which is called charge systemsearch The charged system search (CSS) is based onCoulomb and Gauss laws from electrical physics and thegoverning laws of motion from the Newtonian mechanicsIn this algorithm each agent is a charged particle (CP) EachCP is considered as a charged sphere which exerts an electricforce on other CPs according to Coulomb and Gauss lawsThe resultant forces and the motion laws determine the newlocation of the CPs [22] The new positions of the chargedparticles in the first iteration are determined randomly andfor next iterations are obtained as follows
119883119895new = rand1198951 sdot 119896119886 sdot119865119895
119898119895
sdot Δ1199052
+ rand1198952 sdot 119896V sdot 119881119895old sdot Δ119905 + 119883119895old
119881119895new =119883119895new minus 119883119895old
Δ119905
(8)
where 119870119886 and 119870V are the acceleration and the velocitycoefficients respectively rand1198951 and rand1198952 are two randomuniformly distributed in the range (0 1) and the resultantforces vector for 119895th CP 119865119895 is calculated as
119865119895 = 119902119894 sum
119894119894 = 119895
(
119902119894
1198863119903119894119895 sdot 1198941 +
119902119894
1199032119894119895
sdot 1198942)119901119894119895 (119883119894 minus 119883119895)
119895 = 1 2 119873
1198941 = 1 1198942 = 0 lArrrArr 119903119894119895 lt 119886
1198941 = 0 1198942 = 1 lArrrArr 119903119894119895 ge 119886
(9)
where the magnitude of charge for each CP 119902119894 is defined as
119902119894 =fit (119894) minus fitworstfitbest minus fitworst
119894 = 1 2 119873 (10)
where fitbest and fitworst are the best and the worst fitness ofall the CPs fit(119894) is the fitness of the agent 119894 and119873 is the totalnumber of charged particlesThe separation distance betweentwo CPs 119903119894119895 is obtained as follows
119903119894119895 =
10038171003817100381710038171003817119883119894 minus 119883119895
10038171003817100381710038171003817
10038171003817100381710038171003817(119883119894 + 119883119895) 2 minus 119883best
10038171003817100381710038171003817+ 120576
(11)
where119883119894 and119883119895 are the positions of the 119894th and 119895th CPs119883bestis the position of the best current CP and 120576 is a small positive
4 Shock and Vibration
Search
No
Yes
No
Yes
Terminationcriterion
Step 1 Step 2
Step 3
Step 4
Step 1 Step 2
Step 3
Initialize specification of problem and algorithm
parameters and determinethe initial charged particles
Find the values of the objective function for the
CPs and rank of CPs in an increasing order
Store some of the best CPs in
the charged memory (CM)
Determine the probability of
moving and force vector for each CP
Determine the new positions and
velocities of the CPs
Correct the CP positions using an algorithm based on HS if the CP does not satisfy
the side limits
Evaluate the objective function and rank the
CPs according to their quality
Include the better vectors in the CM and exclude the worst ones
from the CM
Stop
Are the new CPs better than the stored ones
in CM
Initialization
Figure 1 Flowchart of the CSS algorithm
1 2 96 8 1043 5 7
1007mW12 times 65
Figure 2 A cantilever beam geometry
0 500 1000 1500 2000 2500 30000
005
01
015
02
025
Number of analyses
Obj
ectiv
e fun
ctio
n
CSSPSOPC
(a) Scenario 1
CSSPSOPC
0 500 1000 1500 2000 2500 30000
005
01
015
02
025
Number of analyses
Obj
ectiv
e fun
ctio
n
(b) Scenario 2
Figure 3 The convergence history of the 10-element beam for the CSS and PSOPC algorithms
Shock and Vibration 5
0 1 2 3 4 5 6 7 8 9 100
5
10
15
Element number
Dam
age (
)
Incomplete noisy dataComplete noisy dataActual damage
Figure 4 Damage distribution of cantilever beam using the com-plete and incomplete noisy data in Scenario 1
numberThe probability of moving each CP toward the otherCPs is determined using the following function
119901119894119895 =
1
fit (119894) minus fitbestfit (119895) minus fit (119894)
gt rand or fit (119895) gt fit (119894)
0 otherwise(12)
After production of the new position of the CPs if anycomponent of the solution vector swerves off the allowablebounds correct its position using the harmony search-basedhandling approach as described in [23]
The charged memory (CM) is used to save a number ofthe best solutions up to the iterationThe better new solutionsare included in the CM and the worst ones are excluded fromthe CM The flowchart of the CSS algorithm is illustrated inFigure 1
4 Objective Function
The CSS algorithm attempts to minimize an objective assess-ment function for the best solution to a given problem Thisfunction is used to provide a measure of how individualshave performed in the problem domain In the case of aminimization problem the fittest individuals will have thelowest numerical value of the associated objective functionIn this study the statement for the objective function is givenas
119865 = 119891 (119889) (13)
where 119889 = 1198891 1198892 119889119873 are damage parameters at the 119873elements
To create the objective function 119865 it is essential touse some kind of output variables of the structure thatare sufficiently sensitive to the damage parameter beingidentified in order to avoid ill conditioning problems Themode shapes and the natural frequencies can be obtained by
Incomplete noisy dataComplete noisy dataActual damage
0 1 2 3 4 5 6 7 8 9 100
5
10
15
20
25
30
35
40
45
Element number
Dam
age (
)
Figure 5 Damage distribution of cantilever beam using the com-plete and incomplete noisy data in Scenario 2
9 128 11 117 10
20
20
1221
21
19
19
17
17
15
15
13
13
18
18
16
16
14
14
22 9
75
5
3 3
1
1
10
86
6
44
22
25
24
23
160
4060
Figure 6 A 25-bar Bowstring truss
modal analysis methods Natural frequencies are relativelyeasy to measure and have been used by many researchers
To identify localized damage because of the greaterexperience variations in the locality of the affected areamodeshapes offer a better option (Salane and Baldwin 1990 [24]Salawu and Williams 1995 [25] Ndambi et al 2002 [26]) Itis necessary to note that the success of this process depends onthe quality and place selection (for test) of the measurementswhich are able to reflect the damage
Because of this assuming that only a few natural fre-quencies and mode shapes of the lower modes are availablefrequency objective functions from incomplete data areconsidered in this paper to detect damage
5 Numerical Simulation Study
In this section two examples consisting of a ten-element can-tilever beam a Bowstring plane truss and a three-story three-bay unbraced frame structures are presented to examine thecharged system search algorithm The final results are thencompared to the solutions of particle swarm optimization
6 Shock and Vibration
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
005
01
015
02
025
03
035
04
Number of analyses
Obj
ectiv
e fun
ctio
n
(a) Scenario 1
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
005
01
015
02
025
03
035
Number of analyses
Obj
ectiv
e fun
ctio
n
(b) Scenario 2
Figure 7 The convergence history of the Bowstring truss for the CSS and PSOPC algorithms
Incomplete noisy dataComplete noisy dataActual damage
0 2 4 6 8 10 12 14 16 18 20 22 240
5
10
15
20
25
30
Element number
Dam
age (
)
Figure 8 Damage distribution of Bowstring truss using the com-plete and incomplete noisy data in Scenario 1
with passive congregation algorithm to demonstrate theperformance of this work
In CSS algorithm the effect of the pervious velocity andthe resultant force affecting a CP can decrease or increasebased on the values of the 119896V and 119896119886 defined as [22]
119896V = 119888(1 minusiter
itermax)
119896119886 = 119888(1 +iter
itermax)
(14)
where iter is the iteration number itermaxis the maximumnumber of the iterations and 119888 is set to 1
For the CSS algorithm a population of 16 CPs and forPSOPC algorithm a population of 50 particles are used for all
Incomplete noisy dataComplete noisy dataActual damage
Element number0 2 4 6 8 10 12 14 16 18 20 22 24
0
5
10
15
20
25
30
35
40
45D
amag
e (
)
Figure 9 Damage distribution of Bowstring truss using the com-plete and incomplete noisy data in Scenario 2
Table 1 Simulated damage scenarios in cantilever beam
Element number Scenario 1 Scenario 21 1035 206 107 359
the examples the stop criterion is considered as maximumnumber of 3000 and 5000 analyses in first and two otherexamples respectively
Shock and Vibration 7
1
2
13
23 24
14 16 181715
22 25 26 27
19 2120
28 3029
32 3331 34 35 36 37 3938
3 6 9 12
5 8 11
4 7 104m
35m
8m 7m 9m
35m
W12 times 65
W14 times 132
Figure 10 Geometry of unbraced plane frame
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
01
02
03
04
05
Number of analyses
Obj
ectiv
e fun
ctio
n
(a) Scenario 1
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000Number of analyses
0
005
01
015
02
025
03
035
04
Obj
ectiv
e fun
ctio
n
(b) Scenario 2
Figure 11 The convergence history of the 39-element frame for the CSS and PSOPC algorithms
The algorithm has been implemented in the commercialMATLAB software and because of the stochastic nature of thealgorithms each example is independently solved ten times
Damage values have been limited to a region between 0and 1
Example 1 A cantilever steel beam is shown in Figure 2 Forthe reason of modal analyzing the beam was divided into 10two-dimensional beam elements with 20 degrees of freedom
The section of the beam is W12 times 65 with mechanicalproperties of
119860 = 00123 m2 (191 in2)
119868 = 2218 times 10minus4 m4 (533 in4)
119864 = 207 times 109Nm2
120588 = 7860 kgm3
The measured modal responses of the beam before and afterdamage were created using the proposed forward solver(for each scenario) Instead of experimental measurementsnumerically generated measurements were used to estimatethe proposed inverse procedure In this example the first tennatural frequencies are used for damage detection Two dif-ferent simulated damage scenarios are considered (Table 1)
These damage scenarios are considered to represent theeffect of severity of damage (stiffness reduction) numberof damaged elements and contribution of damage elementson the results The results of damage detecting of CSS andPSOPC algorithms are given in Table 2 In order to evaluatethe performance of the methods used the summation ofestimated error of damage detection was calculated (Table 2)that is presented as follows
Error = sum(119889119894119860 minus 119889119894119878) (15)
8 Shock and Vibration
Table 2 Results of damage detection of the 10-element beam using CSS and PSOPC algorithms
Elementnumber
Scenario 1 Scenario 2
CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0000 0 0000 0101 01 0105
2 0000 0 0000 0003 0 0000
3 0000 0 0000 0001 0 0000
4 0000 0 0000 0000 0 0000
5 0000 0 0000 0185 02 0169
6 0100 01 0100 0016 0 0037
7 0000 0 0000 0349 035 0348
8 0000 0 0000 0002 0 0000
9 0000 0 0000 0001 0 0000
10 0000 0 0000 0000 0 0000
Error 00002 00000 00385 00752
Table 3 The values of objective function of the 10-element beam for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 34 times 10
minus0652 times 10
minus0628 times 10
minus0627 times 10
minus0683 times 10
minus0665 times 10
minus0641 times 10
minus0638 times 10
minus0619 times 10
minus0644 times 10
minus06 100PSOPC 25 times 10
minus0713 times 10
minus0799 times 10
minus0814 times 10
minus0721 times 10
minus0746 times 10
minus0726 times 10
minus0716 times 10
minus0688 times 10
minus0811 times 10
minus07 100Scenario 2
CSS 00008 00001 00009 00004 00006 00002 00005 00008 00009 00006 80PSOPC 00003 00013 00056 00004 00074 00074 00080 00011 00006 00085 40
Table 4 Cross-sectional areas of truss elements
Member Area (cm2)1ndash6 187ndash12 1513ndash17 1018ndash25 12
Table 5 Simulated damage scenarios in Bowstring truss
Element number Scenario 1 Scenario 235 108 251011 4024 20
where 119889119894119860 and 119889119894119878 are the actual and estimated damage of the
119894th element using the presented method respectivelyIn Table 3 the values of the objective function for CSS
and PSOPC algorithms in independent 10 runs are given theconvergence history for the 10-element beam in scenarios 1and 2 is shown in Figure 3
It is seen in Table 2 that in the first scenario where thenumber of damaged elements is low the damage identifi-cation is well by both algorithms in the second scenariothe algorithm PSOPC has some wrong in damage detectionso that there is difference between the obtained damage viathis algorithm and the actual intensity of damage in thefifth element and the undamaged sixth element mistakenlyhas been detected to be damaged Meanwhile the locationand amount of damage in structure are obtained by the CSSalgorithm with acceptable accuracy In order to investigatethe noise effects on the results of the CSSmethod 015 noisein measurement is considered for the natural frequencies
Diagrams of the CSS algorithmrsquos damage detection withcomplete and incomplete dynamic noisy data are plotted inFigures 4 and 5
As the results show this method is able to detect thelocation andmagnitude of damaged elements in all scenariosusing complete and incomplete noisy data
Example 2 A Bowstring plane truss with 25 elements isshown in Figure 6 The properties of members of this struc-ture are 119864 = 207 times 0
9Nm2 and 120588 = 7860 kgm3 and cross-sectional areas of elements are given in Table 4
Three scenarios for this truss are considered according toTable 5
Shock and Vibration 9
Table 6 Results of damage detection of the 25-bar Bowstring truss using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0009 0 0000 0000 0 00002 0000 0 0000 0000 0 00003 0000 0 0000 0000 0 00004 0000 0 0000 0000 0 00005 0000 0 0000 0098 01 00956 0000 0 0000 0000 0 00007 0000 0 0000 0000 0 00008 0246 025 0250 0000 0 00009 0001 0 0000 0000 0 000010 0000 0 0000 0000 0 000011 0000 0 0000 0402 04 042612 0000 0 0000 0000 0 000013 0000 0 0000 0000 0 000014 0001 0 0000 0000 0 000015 0000 0 0000 0000 0 000016 0000 0 0000 0000 0 000017 0000 0 0000 0000 0 000018 0000 0 0000 0005 0 000019 0000 0 0000 0004 0 007520 0000 0 0000 0001 0 000021 0000 0 0000 0000 0 000022 0000 0 0000 0000 0 000023 0000 0 0000 0000 0 000024 0000 0 0000 0185 02 000025 0000 0 0000 0014 0 0265Error 00085 00000 00119 01609
In Table 6 the results of two algorithms in all scenarios arecompared Only the first ten natural frequencies are used fordamage detection
The comparison of the objective function for CSS andPSOPC algorithms is made in Table 7 and the convergencehistory for the Bowstring truss in scenarios 1 and 2 is shownin Figure 7
In Figures 8 and 9 the results of CSS algorithm in damagedetection of this truss that is affected by noise are shown
Table 7 and Figures 7 8 and 9 show better results indamage detection by CSS algorithm rather than PSOPCalgorithm and also by applying noise in structures
Example 3 A three-story three-bay frame as shown inFigure 10 is used to verify the damage detection methodexplained in this paper The number of elements and nodesis 39 and 34 respectively
For unbraced plane frame problem all columns areW14 times 132 (119860 = 0025m2 (388 in2) and 119868 = 6386 times
10minus4m4 (1530 in4)) and all beams are W12 times 65 Youngrsquos
modulus is 119864 = 207GPa (30 000 ksi) Poissonrsquos ratio is ] =
03 and the mass density is 120588 = 7780 kgm3 (0000728 lb minuss2in4)
In this example two scenarios from the aspect of elementnumber and its level of damage were assumed that are givenin Table 8 Between ninety natural frequencies only the firsttwenty natural frequencies are utilized for damage detectionin the first scenario
The severity of damage detection by PSOPC and CSSalgorithms in different members of the frame and in eachscenario is shown separately in Table 9 In Table 10 theobtained objective function for both algorithms in each runis given also the convergence history is shown in Figure 11
Also the results of the CSS algorithm damage detectionwith noisy data are plotted in the diagrams of Figures 12 and13
The results of damage detection with two algorithmsin this large frame show that in case of only one memberdamaged there is more accuracy with PSOPC algorithm inidentifying of desired member and obtaining the amount ofdamage than the CSS algorithm It is observed that increasingthe number of injured elements in the structures reducesthe result precision with algorithms From results errors itcan be concluded that the PSOPC algorithm has problemin detecting the location of some of the damaged structuralelements so that in the third scenario the tenth and twenty-fourth members which are injured are diagnosed as healthy
10 Shock and Vibration
Table 7 The values of objective function of the Bowstring truss for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 32 times 10
minus0543 times 10
minus0549 times 10
minus0569 times 10
minus0571 times 10
minus0533 times 10
minus0526 times 10
minus0575 times 10
minus0534 times 10
minus0526 times 10
minus05 100PSOPC 26 times 10
minus0749 times 10
minus0218 times 10
minus0713 times 10
minus0112 times 10
minus0124 times 10
minus0850 times 10
minus0819 times 10
minus0826 times 10
minus0217 times 10
minus02 60Scenario 2
CSS 00002 00003 00003 00001 00039 00038 00002 00003 00015 00051 60PSOPC 00005 00134 00672 00629 00598 00050 00049 00065 00123 00251 10
Incomplete noisy dataComplete noisy dataActual damage
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 390
5
10
15
20
25
30
35
40
Element number
Dam
age (
)
Figure 12 Damage distribution of plane frame using the completeand incomplete noisy data in Scenario 1
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 400
10
20
30
40
50
60
Element number
Dam
age (
)
Predicted damageActual damage
Figure 13 Damage distribution of plane frame using the completenoisy data in Scenario 2
Table 8 Simulated damage scenarios in unbraced plane frame
Element number Scenario 1 Scenario 21 4045 309 1010 1013 2014 4019 2524 2032 50
and a considerable amount of damage in the fifteenth andtwenty-third elements that have no damage has been gainedHowever the CSS algorithm has well identified places ofall damaged elements in the structure and has achieved theamount of their damage with high accuracy and low errorFrom Figures 12 and 13 it can be observed that even in thatcase a lot of damage to the frame is considered the effectof noise on the results obtained by the CSS algorithm islow so that in case of incomplete data accuracy of resultshas reduced slightly and the performance of this algorithmdespite the noise was considered acceptable
In all three examples the convergences diagrams areshown that the process of convergence is gradual So it causesa more comprehensive search algorithm for finding optimalsolutions
The result increases the success rate of this type ofalgorithm performance in comparison with the PSOPCalgorithm
6 Conclusions
An approach for detecting damage based on continuumdamage model using charged system search algorithm ispresented The algorithm evaluates the location and severityof damage in three structures a cantilever beam a Bowstringplane truss and a three-story three-bay unbraced plane frameby minimizing an objective function by measuring completeand incomplete noisy modal data with different damagescenarios
Shock and Vibration 11
Table 9 Results of damage detection of the 39-element frame using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0399 0400 0393 0005 0 00002 0001 0 0000 0001 0 00003 0001 0 0000 0000 0 00004 0001 0 0027 0037 0 00005 0003 0 0000 0254 0300 03566 0000 0 0000 0016 0 00007 0000 0 0000 0013 0 00008 0005 0 0000 0001 0 00009 0056 0100 0000 0001 0 000010 0002 0 0000 0095 0100 000011 0005 0 0000 0002 0 000012 0008 0 0000 0001 0 000013 0193 0200 0195 0058 0 000014 0001 0 0000 0374 0400 034915 0004 0 0000 0005 0 011516 0000 0 0000 0001 0 000017 0000 0 0000 0001 0 000018 0000 0 0000 0004 0 000019 0000 0 0000 0212 0250 025620 0000 0 0000 0004 0 000021 0000 0 0000 0029 0 000022 0000 0 0002 0015 0 000023 0000 0 0000 0001 0 009224 0000 0 0000 0174 0200 000025 0000 0 0000 0001 0 000026 0000 0 0000 0002 0 000027 0002 0 0000 0005 0 000028 0000 0 0000 0001 0 004029 0000 0 0001 0007 0 000030 0000 0 0000 0000 0 000031 0001 0 0003 0002 0 000032 0000 0 0000 0497 0500 049533 0000 0 0000 0000 0 000034 0006 0 0000 0000 0 000035 0005 0 0025 0001 0 000036 0006 0 0000 0001 0 000037 0002 0 0000 0001 0 001538 0007 0 0000 0001 0 000039 0003 0 0000 0002 0 0000Error 01162 01700 03600 06808
In order to demonstrate the power of this algorithm in thediagnosis of damage a comparison has been made betweenthe results of this algorithm and the PSOPC algorithm
By comparing the damage detection results of the twovarious methods some interesting points have been con-cluded In scenarios where the number of damaged elementsis considered low both algorithms have acceptable accuracyin the results considering the more damaged members
of structures the PSOPC algorithm has been mistaken toidentify the damaged elements of the structureThis problemin larger-scale structures and the increasing number ofmembers has beenmore visible CSS algorithm has identifiedthe exact location of damage as well as achieving the amountof damage with acceptable accuracy It can be concluded thatthe CSS algorithm has great potential in global and localsearch of damage
12 Shock and Vibration
Table 10 The values of objective function of the 39-element frame for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 44 times 10
minus0346 times 10
minus0387 times 10
minus0462 times 10
minus0411 times 10
minus0366 times 10
minus0437 times 10
minus0480 times 10
minus0429 times 10
minus0398 times 10
minus04 50PSOPC 12 times 10
minus0213 times 10
minus0244 times 10
minus0315 times 10
minus0243 times 10
minus0213 times 10
minus0339 times 10
minus0214 times 10
minus0239 times 10
minus0315 times 10
minus02 10Scenario 2
CSS 19 times 10minus03
13 times 10minus03
17 times 10minus03
12 times 10minus03
23 times 10minus03
28 times 10minus03
19 times 10minus03
11 times 10minus03
21 times 10minus03
16 times 10minus03 30
PSOPC 29 times 10minus02
39 times 10minus02
11 times 10minus02
18 times 10minus02
11 times 10minus02
59 times 10minus03
69 times 10minus02
36 times 10minus02
53 times 10minus02
13 times 10minus02 0
Due to noise the real value of natural frequencies ofstructure will change it can be seen from the diagramsthat in these cases accuracy reduction of the results of theCSS algorithm is very low This indicates the high powerof this algorithm in damage detection considering noise instructure
Charged system search by considering small populationand low iteration cycle can detect the severity and location ofdamage with acceptable accuracy which shows high powerand speed of convergence of this algorithm The proposedmethod by the charged system search algorithm producedbetter results compared with particle swarm optimizationmethod
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] D J Ewins Modal Testing Theory and Practice John Wiley ampSons New York NY USA 1984
[2] S W Doebling C R Farrar and M B Prime ldquoA summaryreview of vibration-based damage identification methodsrdquoShock and Vibration Digest vol 30 no 2 pp 91ndash105 1998
[3] N Hu XWang H Fukunaga Z H Yao H X Zhang and Z SWu ldquoDamage assessment of structures using modal test datardquoInternational Journal of Solids and Structures vol 38 no 18 pp3111ndash3126 2001
[4] O S Salawu ldquoDetection of structural damage through changesin frequency a reviewrdquo Engineering Structures vol 19 no 9 pp718ndash723 1997
[5] N Bicanic and H-P Chen ldquoDamage identification in framedstructures using natural frequenciesrdquo International Journal forNumerical Methods in Engineering vol 40 no 23 pp 4451ndash4468 1997
[6] A K Pandey M Biswas and M M Samman ldquoDamagedetection from changes in curvature mode shapesrdquo Journal ofSound and Vibration vol 145 no 2 pp 321ndash332 1991
[7] J E Mottershead and M I Friswell ldquoModel updating instructural dynamics a surveyrdquo Journal of Sound and Vibrationvol 167 no 2 pp 347ndash375 1993
[8] R Perera andR Torres ldquoStructural damage detection viamodaldata with genetic algorithmsrdquo Journal of Structural Engineeringvol 132 no 9 pp 1491ndash1501 2006
[9] S Fallahian and S M Seyedpoor ldquoA two stage method forstructural damage identification using an adaptive neuro-fuzzyinference system and particle Swarm optimizationrdquo AsianJournal of Civil Engineering (Building and Housing) vol 11 pp797ndash810 2010
[10] B H Koh and S J Dyke ldquoStructural health monitoring forflexible bridge structures using correlation and sensitivity ofmodal datardquo Computers and Structures vol 85 no 3-4 pp 117ndash130 2007
[11] R-S He and S-F Hwang ldquoDamage detection by a hybrid real-parameter genetic algorithm under the assistance of grey rela-tion analysisrdquo Engineering Applications of Artificial Intelligencevol 20 no 7 pp 980ndash992 2007
[12] H Y Guo and Z L Li ldquoA two-stage method to identifystructural damage sites and extents by using evidence theoryand micro-search genetic algorithmrdquo Mechanical Systems andSignal Processing vol 23 no 3 pp 769ndash782 2009
[13] L Yu and X Chen ldquoBridge damage identification by combiningmodal flexibility and PSO methodsrdquo in Proceedings of thePrognostics and System Health Management Conference (PHMrsquo10) Macau China January 2010
[14] Z Tabrizian E Afshari G Ghodrati Amiri M H A Beigyand SM PourhoseiniNejad ldquoAnewdamage detectionmethodBig Bang-Big Crunch (BB-BC) algorithmrdquo Journal of Shock andVibration vol 20 no 4 pp 633ndash648 2013
[15] A Kaveh and S Talatahari ldquoA novel heuristic optimizationmethod charged system searchrdquo Acta Mechanica vol 213 no3-4 pp 267ndash289 2010
[16] S M Pourhoseini Nejad G Ghodrati Amiri A Asadi EAfshari and Z Tabrizian ldquoDamage detection of skeletal struc-tures using particle swarm optimizer with passive congregation(PSOPC) algorithm via incomplete modal datardquo Journal ofComputational Methods in Civil Engineering vol 3 no 1 pp1ndash13 2012
[17] S He Q H Wu J Y Wen J R Saunders and R CPaton ldquoA particle swarm optimizer with passive congregationrdquoBioSystems vol 78 no 1ndash3 pp 135ndash147 2004
[18] X Wang N Hu H Fukunaga and Z H Yao ldquoStructuraldamage identification using static test data and changes infrequenciesrdquo Engineering Structures vol 23 no 6 pp 610ndash6212001
[19] A Messina E J Williams and T Contursi ldquoStructural damagedetection by a sensitivity and statistical-based methodrdquo Journalof Sound and Vibration vol 216 no 5 pp 791ndash808 1998
[20] S W Doebling C R Farrar M B Prime and D W ShevitzldquoDamage identification and healthmonitoring of structural and
Shock and Vibration 13
mechanical systems from changes in their vibration character-istics a literature reviewrdquo Research Report LA-13070-MS ESA-EA Los Alamos National Laboratory Los Alamos NM USA1996
[21] M Nobahari and S M Seyedpoor ldquoStructural damage detec-tion using an efficient correlation-based index and a modifiedgenetic algorithmrdquoMathematical and Computer Modelling vol53 no 9-10 pp 1798ndash1809 2011
[22] A Kaveh and S Talatahari ldquoOptimal design of skeletal struc-tures via the charged system search algorithmrdquo Structural andMultidisciplinary Optimization vol 41 no 6 pp 893ndash911 2010
[23] A Kaveh and S Talatahari ldquoParticle swarm optimizer antcolony strategy and harmony search scheme hybridized foroptimization of truss structuresrdquoComputers and Structures vol87 no 5-6 pp 267ndash283 2009
[24] H J Salane and J W Baldwin Jr ldquoIdentification of modalproperties of bridgesrdquo Journal of Structural Engineering vol 116no 7 pp 2008ndash2021 1990
[25] O S Salawu and CWilliams ldquoBridge assessment using forced-vibration testingrdquo Journal of Structural Engineering vol 121 no2 pp 161ndash173 1995
[26] J-M Ndambi J Vantomme and K Harri ldquoDamage assessmentin reinforced concrete beams using eigenfrequencies and modeshape derivativesrdquo Engineering Structures vol 24 no 4 pp 501ndash515 2002
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
2 Shock and Vibration
many methods that have been introduced to correctly findout the place and severity of structural damage
Perera and Torres 2006 studied a nondestructive globaldamage detection and assessment methodology based onthe changes in frequencies and mode shapes of vibrationof a structural system It was shown that the proposed GAyields a suitable damage places and severity detection fromtraditional methods [8]
A two-stage method for determining place and severityof multiple structural damage by combining the adaptiveneurofuzzy inference system (ANFIS) and particle swarmoptimization (PSO) has been proposed by Fallahian andSeyedpoor [9] In a studyKoh andDyke [10] have determinedthe place and severity of multiple damage by iterativelysearching for a combination of structural responses thatmaximizes a correlation coefficient named the multipledamage location assurance criterion (MDLAC) via a geneticalgorithm (GA) Damage detection by a hybrid techniqueincluding a real-parameter genetic algorithm and grey rela-tion analysis has been presented by He and Hwang [11]They used first a grey relation analysis to exclude impossibledamage places such that the number of design variables couldbe reduced Second a real-parameter genetic algorithm wascombined with simulated annealing for finding the actualdamage The damage identification of a beam-like structurehas been formulated as an optimization problem and a GAhas been employed to find the damage place and severityby minimizing the cost function which is based on thedifference of measured and calculated natural frequencies Atwo-stage method for determining the place and severity ofmultiple structural damage by using an information fusiontechnique and a GA has been presented by Guo and Li [12]In the first stage the damage is localized by using evidencetheory and then a microsearch genetic algorithm (MSGA)has been planned to determine the damage size Yu and Chendetermined the place and severity of damage in a bridgeFirst two objective functions are defined One is defined asminimizing the sum of differences between the modal databefore and after damage in traditional wayThe other is newlydefined based on modal flexibility which is combined withanother function able to predict damage location Secondlyan improved particle swarm optimization (PSO) algorithmis developed based on the macroeconomic strategies andused to solve the multiple-objective optimization problemon bridge damage identification The results show that theprocedure is very promising for locating and quantifyingdamaged elements of bridge structures and considerablyimproves predictions based on the modal flexibility as wellas the PSO method [13] An approach for detecting damagein structural members based on continuum damage modelusing an algorithmnamedBig Bang-Big Crunch (BB-BC) haspresented suitable accuracy in results [14]
In recent years the use of metaheuristic methods hasincreased and great efforts to increase the power of algo-rithms (Efficiency) and reduce optimization time (moreconvergence speed) have been done Recently a new androbust algorithm was presented by Kaveh and Talatahari socalled charged system search [15] This algorithm is based onsome laws from electrostatics and the Newtonian mechanics
The charged system search (CSS) utilizes a number of chargedparticles (CP) which affect each other based on their fitnessvalues and separation distances consideringCoulombGaussandNewtonian lawsThe resultant forces and themotion lawsdetermine the new location of the particles The harmonysearch-based handling approach is utilized for controlling thevariable constraint
In this paper the CSS algorithm is utilized to predict thedamage place and severity for different types of structuresThe structure is modeled with the finite element method andthe damage identification is performed at element level withincomplete modal data
In order to show the power of the CSS algorithm in struc-tural damage detection the obtained results are comparedto the results of the PSOPC algorithm [16] PSOPC algo-rithm is a particle swarm optimization method with passivecongregationrsquos capacity [17] High abilities of this algorithmwith acceptable results in structural damage detection areillustrated [16]
The results show that the proposed method is capableof detecting the location and severity of diagnosis evenin large-scale structures with a large number of damagedelements it achieved acceptable resultsThe effect of the noiseis considered by assigning the noise in natural frequencieson the results of the CSS algorithm The graphs indicate thatthe noise in input data may reduce the accuracy of damagedetection
2 Theory
Existing Structural damage detection techniques can beclassified into two major categories the dynamic and staticmethods [18] Both methods are based on the finite elementutilizing experimental test data requiring dynamic and statictest data correspondingly In addition dynamic methodshave shown their advantages in comparison with static onesThe natural frequencies of a structure can be considered tobe valuable with the dynamic data Determining the levelof correlation between the calculated and predicted naturalfrequencies can provide a simple tool for finding the place andseverity of structural damage [19] Structural damage detec-tion is calculated with changes in structural characteristicsand it is possible with global evaluating Doebling et al havediscussed vibration-based techniques in a literature review[20] In this section the construction of dynamics of damagedstructures is discussed
The parameter vector used for evaluating the correlationcoefficients (the ratio of the changes first 119899119891 natural fre-quency Δ119865 due to structural damage) is
Δ119865 =
119865119873 minus 119865119864
119865119873
(1)
where 119865119873 and 119865119864 denote the natural frequency vectors of theundamaged and damaged structure Similarly the parametervector predicted from an analytical model can be definedcorrespondingly as
120575119865 (119883) =
119865119873 minus 119865 (119883)
119865119873
(2)
Shock and Vibration 3
where 119865(119883) is a natural frequency vector that can be pre-dicted from an analytic model and 119883 = [119909
1 1199092 119909
119899]
represents a damage variable vector containing the damageseverity of all 119899 structural elements Given a pair of param-eter vectors one can estimate the level of correlation inseveral ways An efficient way is to evaluate a correlation-based index termed themultiples damage location assurancecriterion (MDLAC) and covered in the following form [19]
MDLAC (119883) =10038161003816100381610038161003816Δ119865119879sdot 120575119865 (119883)
10038161003816100381610038161003816
2
(Δ119865119879sdot Δ119865) (120575119865
119879(119883) sdot 120575119865 (119883))
0 lt MDLAC lt 1
(3)
Two frequency change vectors are compared with MDLACone calculated from the structural tests and the other froma structural model analysis When the vector of analyticalfrequencies becomes the same as the frequency vector ofthe damaged structure MDLAC will be maximal That is119865(119883) = 119865 so considering this theory can be used to finda set of damage variables maximizing the MDLAC using anoptimization algorithm
Find 119883 = [119909119894 1199092 119909119899]
Maximize 119908 (119883) = MDLAC (119883) (4)
The damage severity can take values only from the set thatis given from [ 0 1 ] a set of continuous values Moreoverthe objective function that should be maximized is 119908 Asmentioned the damage occurrence in a structural elementdecreases the element stiffness Thus one of the methods forthe damage identification problem is simulation damage bydecreasing one of the stiffness parameters of the element suchas the modulus of elasticity (119864) cross-sectional area (119860) andinertia moment (119868) In this study the damage variables aredefined via a relative reduction of the elasticity modulus ofan element as
119870119889119894 = (1 minus 119909119894)119870
ℎ119894
(5)
119870119889119894 is the stiffness matrix of damaged element 119894 and119870ℎ119894 is the
stiffness matrix of healthy element 119894119864 is the primary modulus of elasticity and 119864119894 is the final
modulus of elasticity of the 119894th element The MDLAC as anobjective function for the optimization algorithm is moresensitive to damaged elements than undamaged elementsIt means that this method can find the true place of thedamaged elements but it may find an undamaged element asa damaged one Therefore in this study a new function andnew optimization algorithm is discussed the new function ispresented as [21]
obj (119883) = 1
119899119891
119899119891
sum
119894=1
min (119891119909119894 119891119864119894)max (119891119909119894 119891119864119894)
(6)
where119891119909119894119891119864119894 are the 119894th components of vectors 119865(119883) and 119865119864correspondingly
The obj(119883) function can rapidly find the locations ofhealthy elements when compared to theMDLAC however it
is very probable that it finds a damaged element as a healthyone Therefore in this study a combinational function of(3) and (6) called here the efficient correlation-based index(ECBI) is used as [21]
ECBI (119883) = 1
2
(MDLAC (119883) + obj (119883)) (7)
3 Charged System Search Algorithm
A new type of metaheuristic algorithms is introduced byKaveh and Talatahari [15] which is called charge systemsearch The charged system search (CSS) is based onCoulomb and Gauss laws from electrical physics and thegoverning laws of motion from the Newtonian mechanicsIn this algorithm each agent is a charged particle (CP) EachCP is considered as a charged sphere which exerts an electricforce on other CPs according to Coulomb and Gauss lawsThe resultant forces and the motion laws determine the newlocation of the CPs [22] The new positions of the chargedparticles in the first iteration are determined randomly andfor next iterations are obtained as follows
119883119895new = rand1198951 sdot 119896119886 sdot119865119895
119898119895
sdot Δ1199052
+ rand1198952 sdot 119896V sdot 119881119895old sdot Δ119905 + 119883119895old
119881119895new =119883119895new minus 119883119895old
Δ119905
(8)
where 119870119886 and 119870V are the acceleration and the velocitycoefficients respectively rand1198951 and rand1198952 are two randomuniformly distributed in the range (0 1) and the resultantforces vector for 119895th CP 119865119895 is calculated as
119865119895 = 119902119894 sum
119894119894 = 119895
(
119902119894
1198863119903119894119895 sdot 1198941 +
119902119894
1199032119894119895
sdot 1198942)119901119894119895 (119883119894 minus 119883119895)
119895 = 1 2 119873
1198941 = 1 1198942 = 0 lArrrArr 119903119894119895 lt 119886
1198941 = 0 1198942 = 1 lArrrArr 119903119894119895 ge 119886
(9)
where the magnitude of charge for each CP 119902119894 is defined as
119902119894 =fit (119894) minus fitworstfitbest minus fitworst
119894 = 1 2 119873 (10)
where fitbest and fitworst are the best and the worst fitness ofall the CPs fit(119894) is the fitness of the agent 119894 and119873 is the totalnumber of charged particlesThe separation distance betweentwo CPs 119903119894119895 is obtained as follows
119903119894119895 =
10038171003817100381710038171003817119883119894 minus 119883119895
10038171003817100381710038171003817
10038171003817100381710038171003817(119883119894 + 119883119895) 2 minus 119883best
10038171003817100381710038171003817+ 120576
(11)
where119883119894 and119883119895 are the positions of the 119894th and 119895th CPs119883bestis the position of the best current CP and 120576 is a small positive
4 Shock and Vibration
Search
No
Yes
No
Yes
Terminationcriterion
Step 1 Step 2
Step 3
Step 4
Step 1 Step 2
Step 3
Initialize specification of problem and algorithm
parameters and determinethe initial charged particles
Find the values of the objective function for the
CPs and rank of CPs in an increasing order
Store some of the best CPs in
the charged memory (CM)
Determine the probability of
moving and force vector for each CP
Determine the new positions and
velocities of the CPs
Correct the CP positions using an algorithm based on HS if the CP does not satisfy
the side limits
Evaluate the objective function and rank the
CPs according to their quality
Include the better vectors in the CM and exclude the worst ones
from the CM
Stop
Are the new CPs better than the stored ones
in CM
Initialization
Figure 1 Flowchart of the CSS algorithm
1 2 96 8 1043 5 7
1007mW12 times 65
Figure 2 A cantilever beam geometry
0 500 1000 1500 2000 2500 30000
005
01
015
02
025
Number of analyses
Obj
ectiv
e fun
ctio
n
CSSPSOPC
(a) Scenario 1
CSSPSOPC
0 500 1000 1500 2000 2500 30000
005
01
015
02
025
Number of analyses
Obj
ectiv
e fun
ctio
n
(b) Scenario 2
Figure 3 The convergence history of the 10-element beam for the CSS and PSOPC algorithms
Shock and Vibration 5
0 1 2 3 4 5 6 7 8 9 100
5
10
15
Element number
Dam
age (
)
Incomplete noisy dataComplete noisy dataActual damage
Figure 4 Damage distribution of cantilever beam using the com-plete and incomplete noisy data in Scenario 1
numberThe probability of moving each CP toward the otherCPs is determined using the following function
119901119894119895 =
1
fit (119894) minus fitbestfit (119895) minus fit (119894)
gt rand or fit (119895) gt fit (119894)
0 otherwise(12)
After production of the new position of the CPs if anycomponent of the solution vector swerves off the allowablebounds correct its position using the harmony search-basedhandling approach as described in [23]
The charged memory (CM) is used to save a number ofthe best solutions up to the iterationThe better new solutionsare included in the CM and the worst ones are excluded fromthe CM The flowchart of the CSS algorithm is illustrated inFigure 1
4 Objective Function
The CSS algorithm attempts to minimize an objective assess-ment function for the best solution to a given problem Thisfunction is used to provide a measure of how individualshave performed in the problem domain In the case of aminimization problem the fittest individuals will have thelowest numerical value of the associated objective functionIn this study the statement for the objective function is givenas
119865 = 119891 (119889) (13)
where 119889 = 1198891 1198892 119889119873 are damage parameters at the 119873elements
To create the objective function 119865 it is essential touse some kind of output variables of the structure thatare sufficiently sensitive to the damage parameter beingidentified in order to avoid ill conditioning problems Themode shapes and the natural frequencies can be obtained by
Incomplete noisy dataComplete noisy dataActual damage
0 1 2 3 4 5 6 7 8 9 100
5
10
15
20
25
30
35
40
45
Element number
Dam
age (
)
Figure 5 Damage distribution of cantilever beam using the com-plete and incomplete noisy data in Scenario 2
9 128 11 117 10
20
20
1221
21
19
19
17
17
15
15
13
13
18
18
16
16
14
14
22 9
75
5
3 3
1
1
10
86
6
44
22
25
24
23
160
4060
Figure 6 A 25-bar Bowstring truss
modal analysis methods Natural frequencies are relativelyeasy to measure and have been used by many researchers
To identify localized damage because of the greaterexperience variations in the locality of the affected areamodeshapes offer a better option (Salane and Baldwin 1990 [24]Salawu and Williams 1995 [25] Ndambi et al 2002 [26]) Itis necessary to note that the success of this process depends onthe quality and place selection (for test) of the measurementswhich are able to reflect the damage
Because of this assuming that only a few natural fre-quencies and mode shapes of the lower modes are availablefrequency objective functions from incomplete data areconsidered in this paper to detect damage
5 Numerical Simulation Study
In this section two examples consisting of a ten-element can-tilever beam a Bowstring plane truss and a three-story three-bay unbraced frame structures are presented to examine thecharged system search algorithm The final results are thencompared to the solutions of particle swarm optimization
6 Shock and Vibration
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
005
01
015
02
025
03
035
04
Number of analyses
Obj
ectiv
e fun
ctio
n
(a) Scenario 1
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
005
01
015
02
025
03
035
Number of analyses
Obj
ectiv
e fun
ctio
n
(b) Scenario 2
Figure 7 The convergence history of the Bowstring truss for the CSS and PSOPC algorithms
Incomplete noisy dataComplete noisy dataActual damage
0 2 4 6 8 10 12 14 16 18 20 22 240
5
10
15
20
25
30
Element number
Dam
age (
)
Figure 8 Damage distribution of Bowstring truss using the com-plete and incomplete noisy data in Scenario 1
with passive congregation algorithm to demonstrate theperformance of this work
In CSS algorithm the effect of the pervious velocity andthe resultant force affecting a CP can decrease or increasebased on the values of the 119896V and 119896119886 defined as [22]
119896V = 119888(1 minusiter
itermax)
119896119886 = 119888(1 +iter
itermax)
(14)
where iter is the iteration number itermaxis the maximumnumber of the iterations and 119888 is set to 1
For the CSS algorithm a population of 16 CPs and forPSOPC algorithm a population of 50 particles are used for all
Incomplete noisy dataComplete noisy dataActual damage
Element number0 2 4 6 8 10 12 14 16 18 20 22 24
0
5
10
15
20
25
30
35
40
45D
amag
e (
)
Figure 9 Damage distribution of Bowstring truss using the com-plete and incomplete noisy data in Scenario 2
Table 1 Simulated damage scenarios in cantilever beam
Element number Scenario 1 Scenario 21 1035 206 107 359
the examples the stop criterion is considered as maximumnumber of 3000 and 5000 analyses in first and two otherexamples respectively
Shock and Vibration 7
1
2
13
23 24
14 16 181715
22 25 26 27
19 2120
28 3029
32 3331 34 35 36 37 3938
3 6 9 12
5 8 11
4 7 104m
35m
8m 7m 9m
35m
W12 times 65
W14 times 132
Figure 10 Geometry of unbraced plane frame
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
01
02
03
04
05
Number of analyses
Obj
ectiv
e fun
ctio
n
(a) Scenario 1
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000Number of analyses
0
005
01
015
02
025
03
035
04
Obj
ectiv
e fun
ctio
n
(b) Scenario 2
Figure 11 The convergence history of the 39-element frame for the CSS and PSOPC algorithms
The algorithm has been implemented in the commercialMATLAB software and because of the stochastic nature of thealgorithms each example is independently solved ten times
Damage values have been limited to a region between 0and 1
Example 1 A cantilever steel beam is shown in Figure 2 Forthe reason of modal analyzing the beam was divided into 10two-dimensional beam elements with 20 degrees of freedom
The section of the beam is W12 times 65 with mechanicalproperties of
119860 = 00123 m2 (191 in2)
119868 = 2218 times 10minus4 m4 (533 in4)
119864 = 207 times 109Nm2
120588 = 7860 kgm3
The measured modal responses of the beam before and afterdamage were created using the proposed forward solver(for each scenario) Instead of experimental measurementsnumerically generated measurements were used to estimatethe proposed inverse procedure In this example the first tennatural frequencies are used for damage detection Two dif-ferent simulated damage scenarios are considered (Table 1)
These damage scenarios are considered to represent theeffect of severity of damage (stiffness reduction) numberof damaged elements and contribution of damage elementson the results The results of damage detecting of CSS andPSOPC algorithms are given in Table 2 In order to evaluatethe performance of the methods used the summation ofestimated error of damage detection was calculated (Table 2)that is presented as follows
Error = sum(119889119894119860 minus 119889119894119878) (15)
8 Shock and Vibration
Table 2 Results of damage detection of the 10-element beam using CSS and PSOPC algorithms
Elementnumber
Scenario 1 Scenario 2
CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0000 0 0000 0101 01 0105
2 0000 0 0000 0003 0 0000
3 0000 0 0000 0001 0 0000
4 0000 0 0000 0000 0 0000
5 0000 0 0000 0185 02 0169
6 0100 01 0100 0016 0 0037
7 0000 0 0000 0349 035 0348
8 0000 0 0000 0002 0 0000
9 0000 0 0000 0001 0 0000
10 0000 0 0000 0000 0 0000
Error 00002 00000 00385 00752
Table 3 The values of objective function of the 10-element beam for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 34 times 10
minus0652 times 10
minus0628 times 10
minus0627 times 10
minus0683 times 10
minus0665 times 10
minus0641 times 10
minus0638 times 10
minus0619 times 10
minus0644 times 10
minus06 100PSOPC 25 times 10
minus0713 times 10
minus0799 times 10
minus0814 times 10
minus0721 times 10
minus0746 times 10
minus0726 times 10
minus0716 times 10
minus0688 times 10
minus0811 times 10
minus07 100Scenario 2
CSS 00008 00001 00009 00004 00006 00002 00005 00008 00009 00006 80PSOPC 00003 00013 00056 00004 00074 00074 00080 00011 00006 00085 40
Table 4 Cross-sectional areas of truss elements
Member Area (cm2)1ndash6 187ndash12 1513ndash17 1018ndash25 12
Table 5 Simulated damage scenarios in Bowstring truss
Element number Scenario 1 Scenario 235 108 251011 4024 20
where 119889119894119860 and 119889119894119878 are the actual and estimated damage of the
119894th element using the presented method respectivelyIn Table 3 the values of the objective function for CSS
and PSOPC algorithms in independent 10 runs are given theconvergence history for the 10-element beam in scenarios 1and 2 is shown in Figure 3
It is seen in Table 2 that in the first scenario where thenumber of damaged elements is low the damage identifi-cation is well by both algorithms in the second scenariothe algorithm PSOPC has some wrong in damage detectionso that there is difference between the obtained damage viathis algorithm and the actual intensity of damage in thefifth element and the undamaged sixth element mistakenlyhas been detected to be damaged Meanwhile the locationand amount of damage in structure are obtained by the CSSalgorithm with acceptable accuracy In order to investigatethe noise effects on the results of the CSSmethod 015 noisein measurement is considered for the natural frequencies
Diagrams of the CSS algorithmrsquos damage detection withcomplete and incomplete dynamic noisy data are plotted inFigures 4 and 5
As the results show this method is able to detect thelocation andmagnitude of damaged elements in all scenariosusing complete and incomplete noisy data
Example 2 A Bowstring plane truss with 25 elements isshown in Figure 6 The properties of members of this struc-ture are 119864 = 207 times 0
9Nm2 and 120588 = 7860 kgm3 and cross-sectional areas of elements are given in Table 4
Three scenarios for this truss are considered according toTable 5
Shock and Vibration 9
Table 6 Results of damage detection of the 25-bar Bowstring truss using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0009 0 0000 0000 0 00002 0000 0 0000 0000 0 00003 0000 0 0000 0000 0 00004 0000 0 0000 0000 0 00005 0000 0 0000 0098 01 00956 0000 0 0000 0000 0 00007 0000 0 0000 0000 0 00008 0246 025 0250 0000 0 00009 0001 0 0000 0000 0 000010 0000 0 0000 0000 0 000011 0000 0 0000 0402 04 042612 0000 0 0000 0000 0 000013 0000 0 0000 0000 0 000014 0001 0 0000 0000 0 000015 0000 0 0000 0000 0 000016 0000 0 0000 0000 0 000017 0000 0 0000 0000 0 000018 0000 0 0000 0005 0 000019 0000 0 0000 0004 0 007520 0000 0 0000 0001 0 000021 0000 0 0000 0000 0 000022 0000 0 0000 0000 0 000023 0000 0 0000 0000 0 000024 0000 0 0000 0185 02 000025 0000 0 0000 0014 0 0265Error 00085 00000 00119 01609
In Table 6 the results of two algorithms in all scenarios arecompared Only the first ten natural frequencies are used fordamage detection
The comparison of the objective function for CSS andPSOPC algorithms is made in Table 7 and the convergencehistory for the Bowstring truss in scenarios 1 and 2 is shownin Figure 7
In Figures 8 and 9 the results of CSS algorithm in damagedetection of this truss that is affected by noise are shown
Table 7 and Figures 7 8 and 9 show better results indamage detection by CSS algorithm rather than PSOPCalgorithm and also by applying noise in structures
Example 3 A three-story three-bay frame as shown inFigure 10 is used to verify the damage detection methodexplained in this paper The number of elements and nodesis 39 and 34 respectively
For unbraced plane frame problem all columns areW14 times 132 (119860 = 0025m2 (388 in2) and 119868 = 6386 times
10minus4m4 (1530 in4)) and all beams are W12 times 65 Youngrsquos
modulus is 119864 = 207GPa (30 000 ksi) Poissonrsquos ratio is ] =
03 and the mass density is 120588 = 7780 kgm3 (0000728 lb minuss2in4)
In this example two scenarios from the aspect of elementnumber and its level of damage were assumed that are givenin Table 8 Between ninety natural frequencies only the firsttwenty natural frequencies are utilized for damage detectionin the first scenario
The severity of damage detection by PSOPC and CSSalgorithms in different members of the frame and in eachscenario is shown separately in Table 9 In Table 10 theobtained objective function for both algorithms in each runis given also the convergence history is shown in Figure 11
Also the results of the CSS algorithm damage detectionwith noisy data are plotted in the diagrams of Figures 12 and13
The results of damage detection with two algorithmsin this large frame show that in case of only one memberdamaged there is more accuracy with PSOPC algorithm inidentifying of desired member and obtaining the amount ofdamage than the CSS algorithm It is observed that increasingthe number of injured elements in the structures reducesthe result precision with algorithms From results errors itcan be concluded that the PSOPC algorithm has problemin detecting the location of some of the damaged structuralelements so that in the third scenario the tenth and twenty-fourth members which are injured are diagnosed as healthy
10 Shock and Vibration
Table 7 The values of objective function of the Bowstring truss for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 32 times 10
minus0543 times 10
minus0549 times 10
minus0569 times 10
minus0571 times 10
minus0533 times 10
minus0526 times 10
minus0575 times 10
minus0534 times 10
minus0526 times 10
minus05 100PSOPC 26 times 10
minus0749 times 10
minus0218 times 10
minus0713 times 10
minus0112 times 10
minus0124 times 10
minus0850 times 10
minus0819 times 10
minus0826 times 10
minus0217 times 10
minus02 60Scenario 2
CSS 00002 00003 00003 00001 00039 00038 00002 00003 00015 00051 60PSOPC 00005 00134 00672 00629 00598 00050 00049 00065 00123 00251 10
Incomplete noisy dataComplete noisy dataActual damage
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 390
5
10
15
20
25
30
35
40
Element number
Dam
age (
)
Figure 12 Damage distribution of plane frame using the completeand incomplete noisy data in Scenario 1
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 400
10
20
30
40
50
60
Element number
Dam
age (
)
Predicted damageActual damage
Figure 13 Damage distribution of plane frame using the completenoisy data in Scenario 2
Table 8 Simulated damage scenarios in unbraced plane frame
Element number Scenario 1 Scenario 21 4045 309 1010 1013 2014 4019 2524 2032 50
and a considerable amount of damage in the fifteenth andtwenty-third elements that have no damage has been gainedHowever the CSS algorithm has well identified places ofall damaged elements in the structure and has achieved theamount of their damage with high accuracy and low errorFrom Figures 12 and 13 it can be observed that even in thatcase a lot of damage to the frame is considered the effectof noise on the results obtained by the CSS algorithm islow so that in case of incomplete data accuracy of resultshas reduced slightly and the performance of this algorithmdespite the noise was considered acceptable
In all three examples the convergences diagrams areshown that the process of convergence is gradual So it causesa more comprehensive search algorithm for finding optimalsolutions
The result increases the success rate of this type ofalgorithm performance in comparison with the PSOPCalgorithm
6 Conclusions
An approach for detecting damage based on continuumdamage model using charged system search algorithm ispresented The algorithm evaluates the location and severityof damage in three structures a cantilever beam a Bowstringplane truss and a three-story three-bay unbraced plane frameby minimizing an objective function by measuring completeand incomplete noisy modal data with different damagescenarios
Shock and Vibration 11
Table 9 Results of damage detection of the 39-element frame using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0399 0400 0393 0005 0 00002 0001 0 0000 0001 0 00003 0001 0 0000 0000 0 00004 0001 0 0027 0037 0 00005 0003 0 0000 0254 0300 03566 0000 0 0000 0016 0 00007 0000 0 0000 0013 0 00008 0005 0 0000 0001 0 00009 0056 0100 0000 0001 0 000010 0002 0 0000 0095 0100 000011 0005 0 0000 0002 0 000012 0008 0 0000 0001 0 000013 0193 0200 0195 0058 0 000014 0001 0 0000 0374 0400 034915 0004 0 0000 0005 0 011516 0000 0 0000 0001 0 000017 0000 0 0000 0001 0 000018 0000 0 0000 0004 0 000019 0000 0 0000 0212 0250 025620 0000 0 0000 0004 0 000021 0000 0 0000 0029 0 000022 0000 0 0002 0015 0 000023 0000 0 0000 0001 0 009224 0000 0 0000 0174 0200 000025 0000 0 0000 0001 0 000026 0000 0 0000 0002 0 000027 0002 0 0000 0005 0 000028 0000 0 0000 0001 0 004029 0000 0 0001 0007 0 000030 0000 0 0000 0000 0 000031 0001 0 0003 0002 0 000032 0000 0 0000 0497 0500 049533 0000 0 0000 0000 0 000034 0006 0 0000 0000 0 000035 0005 0 0025 0001 0 000036 0006 0 0000 0001 0 000037 0002 0 0000 0001 0 001538 0007 0 0000 0001 0 000039 0003 0 0000 0002 0 0000Error 01162 01700 03600 06808
In order to demonstrate the power of this algorithm in thediagnosis of damage a comparison has been made betweenthe results of this algorithm and the PSOPC algorithm
By comparing the damage detection results of the twovarious methods some interesting points have been con-cluded In scenarios where the number of damaged elementsis considered low both algorithms have acceptable accuracyin the results considering the more damaged members
of structures the PSOPC algorithm has been mistaken toidentify the damaged elements of the structureThis problemin larger-scale structures and the increasing number ofmembers has beenmore visible CSS algorithm has identifiedthe exact location of damage as well as achieving the amountof damage with acceptable accuracy It can be concluded thatthe CSS algorithm has great potential in global and localsearch of damage
12 Shock and Vibration
Table 10 The values of objective function of the 39-element frame for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 44 times 10
minus0346 times 10
minus0387 times 10
minus0462 times 10
minus0411 times 10
minus0366 times 10
minus0437 times 10
minus0480 times 10
minus0429 times 10
minus0398 times 10
minus04 50PSOPC 12 times 10
minus0213 times 10
minus0244 times 10
minus0315 times 10
minus0243 times 10
minus0213 times 10
minus0339 times 10
minus0214 times 10
minus0239 times 10
minus0315 times 10
minus02 10Scenario 2
CSS 19 times 10minus03
13 times 10minus03
17 times 10minus03
12 times 10minus03
23 times 10minus03
28 times 10minus03
19 times 10minus03
11 times 10minus03
21 times 10minus03
16 times 10minus03 30
PSOPC 29 times 10minus02
39 times 10minus02
11 times 10minus02
18 times 10minus02
11 times 10minus02
59 times 10minus03
69 times 10minus02
36 times 10minus02
53 times 10minus02
13 times 10minus02 0
Due to noise the real value of natural frequencies ofstructure will change it can be seen from the diagramsthat in these cases accuracy reduction of the results of theCSS algorithm is very low This indicates the high powerof this algorithm in damage detection considering noise instructure
Charged system search by considering small populationand low iteration cycle can detect the severity and location ofdamage with acceptable accuracy which shows high powerand speed of convergence of this algorithm The proposedmethod by the charged system search algorithm producedbetter results compared with particle swarm optimizationmethod
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] D J Ewins Modal Testing Theory and Practice John Wiley ampSons New York NY USA 1984
[2] S W Doebling C R Farrar and M B Prime ldquoA summaryreview of vibration-based damage identification methodsrdquoShock and Vibration Digest vol 30 no 2 pp 91ndash105 1998
[3] N Hu XWang H Fukunaga Z H Yao H X Zhang and Z SWu ldquoDamage assessment of structures using modal test datardquoInternational Journal of Solids and Structures vol 38 no 18 pp3111ndash3126 2001
[4] O S Salawu ldquoDetection of structural damage through changesin frequency a reviewrdquo Engineering Structures vol 19 no 9 pp718ndash723 1997
[5] N Bicanic and H-P Chen ldquoDamage identification in framedstructures using natural frequenciesrdquo International Journal forNumerical Methods in Engineering vol 40 no 23 pp 4451ndash4468 1997
[6] A K Pandey M Biswas and M M Samman ldquoDamagedetection from changes in curvature mode shapesrdquo Journal ofSound and Vibration vol 145 no 2 pp 321ndash332 1991
[7] J E Mottershead and M I Friswell ldquoModel updating instructural dynamics a surveyrdquo Journal of Sound and Vibrationvol 167 no 2 pp 347ndash375 1993
[8] R Perera andR Torres ldquoStructural damage detection viamodaldata with genetic algorithmsrdquo Journal of Structural Engineeringvol 132 no 9 pp 1491ndash1501 2006
[9] S Fallahian and S M Seyedpoor ldquoA two stage method forstructural damage identification using an adaptive neuro-fuzzyinference system and particle Swarm optimizationrdquo AsianJournal of Civil Engineering (Building and Housing) vol 11 pp797ndash810 2010
[10] B H Koh and S J Dyke ldquoStructural health monitoring forflexible bridge structures using correlation and sensitivity ofmodal datardquo Computers and Structures vol 85 no 3-4 pp 117ndash130 2007
[11] R-S He and S-F Hwang ldquoDamage detection by a hybrid real-parameter genetic algorithm under the assistance of grey rela-tion analysisrdquo Engineering Applications of Artificial Intelligencevol 20 no 7 pp 980ndash992 2007
[12] H Y Guo and Z L Li ldquoA two-stage method to identifystructural damage sites and extents by using evidence theoryand micro-search genetic algorithmrdquo Mechanical Systems andSignal Processing vol 23 no 3 pp 769ndash782 2009
[13] L Yu and X Chen ldquoBridge damage identification by combiningmodal flexibility and PSO methodsrdquo in Proceedings of thePrognostics and System Health Management Conference (PHMrsquo10) Macau China January 2010
[14] Z Tabrizian E Afshari G Ghodrati Amiri M H A Beigyand SM PourhoseiniNejad ldquoAnewdamage detectionmethodBig Bang-Big Crunch (BB-BC) algorithmrdquo Journal of Shock andVibration vol 20 no 4 pp 633ndash648 2013
[15] A Kaveh and S Talatahari ldquoA novel heuristic optimizationmethod charged system searchrdquo Acta Mechanica vol 213 no3-4 pp 267ndash289 2010
[16] S M Pourhoseini Nejad G Ghodrati Amiri A Asadi EAfshari and Z Tabrizian ldquoDamage detection of skeletal struc-tures using particle swarm optimizer with passive congregation(PSOPC) algorithm via incomplete modal datardquo Journal ofComputational Methods in Civil Engineering vol 3 no 1 pp1ndash13 2012
[17] S He Q H Wu J Y Wen J R Saunders and R CPaton ldquoA particle swarm optimizer with passive congregationrdquoBioSystems vol 78 no 1ndash3 pp 135ndash147 2004
[18] X Wang N Hu H Fukunaga and Z H Yao ldquoStructuraldamage identification using static test data and changes infrequenciesrdquo Engineering Structures vol 23 no 6 pp 610ndash6212001
[19] A Messina E J Williams and T Contursi ldquoStructural damagedetection by a sensitivity and statistical-based methodrdquo Journalof Sound and Vibration vol 216 no 5 pp 791ndash808 1998
[20] S W Doebling C R Farrar M B Prime and D W ShevitzldquoDamage identification and healthmonitoring of structural and
Shock and Vibration 13
mechanical systems from changes in their vibration character-istics a literature reviewrdquo Research Report LA-13070-MS ESA-EA Los Alamos National Laboratory Los Alamos NM USA1996
[21] M Nobahari and S M Seyedpoor ldquoStructural damage detec-tion using an efficient correlation-based index and a modifiedgenetic algorithmrdquoMathematical and Computer Modelling vol53 no 9-10 pp 1798ndash1809 2011
[22] A Kaveh and S Talatahari ldquoOptimal design of skeletal struc-tures via the charged system search algorithmrdquo Structural andMultidisciplinary Optimization vol 41 no 6 pp 893ndash911 2010
[23] A Kaveh and S Talatahari ldquoParticle swarm optimizer antcolony strategy and harmony search scheme hybridized foroptimization of truss structuresrdquoComputers and Structures vol87 no 5-6 pp 267ndash283 2009
[24] H J Salane and J W Baldwin Jr ldquoIdentification of modalproperties of bridgesrdquo Journal of Structural Engineering vol 116no 7 pp 2008ndash2021 1990
[25] O S Salawu and CWilliams ldquoBridge assessment using forced-vibration testingrdquo Journal of Structural Engineering vol 121 no2 pp 161ndash173 1995
[26] J-M Ndambi J Vantomme and K Harri ldquoDamage assessmentin reinforced concrete beams using eigenfrequencies and modeshape derivativesrdquo Engineering Structures vol 24 no 4 pp 501ndash515 2002
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
Shock and Vibration 3
where 119865(119883) is a natural frequency vector that can be pre-dicted from an analytic model and 119883 = [119909
1 1199092 119909
119899]
represents a damage variable vector containing the damageseverity of all 119899 structural elements Given a pair of param-eter vectors one can estimate the level of correlation inseveral ways An efficient way is to evaluate a correlation-based index termed themultiples damage location assurancecriterion (MDLAC) and covered in the following form [19]
MDLAC (119883) =10038161003816100381610038161003816Δ119865119879sdot 120575119865 (119883)
10038161003816100381610038161003816
2
(Δ119865119879sdot Δ119865) (120575119865
119879(119883) sdot 120575119865 (119883))
0 lt MDLAC lt 1
(3)
Two frequency change vectors are compared with MDLACone calculated from the structural tests and the other froma structural model analysis When the vector of analyticalfrequencies becomes the same as the frequency vector ofthe damaged structure MDLAC will be maximal That is119865(119883) = 119865 so considering this theory can be used to finda set of damage variables maximizing the MDLAC using anoptimization algorithm
Find 119883 = [119909119894 1199092 119909119899]
Maximize 119908 (119883) = MDLAC (119883) (4)
The damage severity can take values only from the set thatis given from [ 0 1 ] a set of continuous values Moreoverthe objective function that should be maximized is 119908 Asmentioned the damage occurrence in a structural elementdecreases the element stiffness Thus one of the methods forthe damage identification problem is simulation damage bydecreasing one of the stiffness parameters of the element suchas the modulus of elasticity (119864) cross-sectional area (119860) andinertia moment (119868) In this study the damage variables aredefined via a relative reduction of the elasticity modulus ofan element as
119870119889119894 = (1 minus 119909119894)119870
ℎ119894
(5)
119870119889119894 is the stiffness matrix of damaged element 119894 and119870ℎ119894 is the
stiffness matrix of healthy element 119894119864 is the primary modulus of elasticity and 119864119894 is the final
modulus of elasticity of the 119894th element The MDLAC as anobjective function for the optimization algorithm is moresensitive to damaged elements than undamaged elementsIt means that this method can find the true place of thedamaged elements but it may find an undamaged element asa damaged one Therefore in this study a new function andnew optimization algorithm is discussed the new function ispresented as [21]
obj (119883) = 1
119899119891
119899119891
sum
119894=1
min (119891119909119894 119891119864119894)max (119891119909119894 119891119864119894)
(6)
where119891119909119894119891119864119894 are the 119894th components of vectors 119865(119883) and 119865119864correspondingly
The obj(119883) function can rapidly find the locations ofhealthy elements when compared to theMDLAC however it
is very probable that it finds a damaged element as a healthyone Therefore in this study a combinational function of(3) and (6) called here the efficient correlation-based index(ECBI) is used as [21]
ECBI (119883) = 1
2
(MDLAC (119883) + obj (119883)) (7)
3 Charged System Search Algorithm
A new type of metaheuristic algorithms is introduced byKaveh and Talatahari [15] which is called charge systemsearch The charged system search (CSS) is based onCoulomb and Gauss laws from electrical physics and thegoverning laws of motion from the Newtonian mechanicsIn this algorithm each agent is a charged particle (CP) EachCP is considered as a charged sphere which exerts an electricforce on other CPs according to Coulomb and Gauss lawsThe resultant forces and the motion laws determine the newlocation of the CPs [22] The new positions of the chargedparticles in the first iteration are determined randomly andfor next iterations are obtained as follows
119883119895new = rand1198951 sdot 119896119886 sdot119865119895
119898119895
sdot Δ1199052
+ rand1198952 sdot 119896V sdot 119881119895old sdot Δ119905 + 119883119895old
119881119895new =119883119895new minus 119883119895old
Δ119905
(8)
where 119870119886 and 119870V are the acceleration and the velocitycoefficients respectively rand1198951 and rand1198952 are two randomuniformly distributed in the range (0 1) and the resultantforces vector for 119895th CP 119865119895 is calculated as
119865119895 = 119902119894 sum
119894119894 = 119895
(
119902119894
1198863119903119894119895 sdot 1198941 +
119902119894
1199032119894119895
sdot 1198942)119901119894119895 (119883119894 minus 119883119895)
119895 = 1 2 119873
1198941 = 1 1198942 = 0 lArrrArr 119903119894119895 lt 119886
1198941 = 0 1198942 = 1 lArrrArr 119903119894119895 ge 119886
(9)
where the magnitude of charge for each CP 119902119894 is defined as
119902119894 =fit (119894) minus fitworstfitbest minus fitworst
119894 = 1 2 119873 (10)
where fitbest and fitworst are the best and the worst fitness ofall the CPs fit(119894) is the fitness of the agent 119894 and119873 is the totalnumber of charged particlesThe separation distance betweentwo CPs 119903119894119895 is obtained as follows
119903119894119895 =
10038171003817100381710038171003817119883119894 minus 119883119895
10038171003817100381710038171003817
10038171003817100381710038171003817(119883119894 + 119883119895) 2 minus 119883best
10038171003817100381710038171003817+ 120576
(11)
where119883119894 and119883119895 are the positions of the 119894th and 119895th CPs119883bestis the position of the best current CP and 120576 is a small positive
4 Shock and Vibration
Search
No
Yes
No
Yes
Terminationcriterion
Step 1 Step 2
Step 3
Step 4
Step 1 Step 2
Step 3
Initialize specification of problem and algorithm
parameters and determinethe initial charged particles
Find the values of the objective function for the
CPs and rank of CPs in an increasing order
Store some of the best CPs in
the charged memory (CM)
Determine the probability of
moving and force vector for each CP
Determine the new positions and
velocities of the CPs
Correct the CP positions using an algorithm based on HS if the CP does not satisfy
the side limits
Evaluate the objective function and rank the
CPs according to their quality
Include the better vectors in the CM and exclude the worst ones
from the CM
Stop
Are the new CPs better than the stored ones
in CM
Initialization
Figure 1 Flowchart of the CSS algorithm
1 2 96 8 1043 5 7
1007mW12 times 65
Figure 2 A cantilever beam geometry
0 500 1000 1500 2000 2500 30000
005
01
015
02
025
Number of analyses
Obj
ectiv
e fun
ctio
n
CSSPSOPC
(a) Scenario 1
CSSPSOPC
0 500 1000 1500 2000 2500 30000
005
01
015
02
025
Number of analyses
Obj
ectiv
e fun
ctio
n
(b) Scenario 2
Figure 3 The convergence history of the 10-element beam for the CSS and PSOPC algorithms
Shock and Vibration 5
0 1 2 3 4 5 6 7 8 9 100
5
10
15
Element number
Dam
age (
)
Incomplete noisy dataComplete noisy dataActual damage
Figure 4 Damage distribution of cantilever beam using the com-plete and incomplete noisy data in Scenario 1
numberThe probability of moving each CP toward the otherCPs is determined using the following function
119901119894119895 =
1
fit (119894) minus fitbestfit (119895) minus fit (119894)
gt rand or fit (119895) gt fit (119894)
0 otherwise(12)
After production of the new position of the CPs if anycomponent of the solution vector swerves off the allowablebounds correct its position using the harmony search-basedhandling approach as described in [23]
The charged memory (CM) is used to save a number ofthe best solutions up to the iterationThe better new solutionsare included in the CM and the worst ones are excluded fromthe CM The flowchart of the CSS algorithm is illustrated inFigure 1
4 Objective Function
The CSS algorithm attempts to minimize an objective assess-ment function for the best solution to a given problem Thisfunction is used to provide a measure of how individualshave performed in the problem domain In the case of aminimization problem the fittest individuals will have thelowest numerical value of the associated objective functionIn this study the statement for the objective function is givenas
119865 = 119891 (119889) (13)
where 119889 = 1198891 1198892 119889119873 are damage parameters at the 119873elements
To create the objective function 119865 it is essential touse some kind of output variables of the structure thatare sufficiently sensitive to the damage parameter beingidentified in order to avoid ill conditioning problems Themode shapes and the natural frequencies can be obtained by
Incomplete noisy dataComplete noisy dataActual damage
0 1 2 3 4 5 6 7 8 9 100
5
10
15
20
25
30
35
40
45
Element number
Dam
age (
)
Figure 5 Damage distribution of cantilever beam using the com-plete and incomplete noisy data in Scenario 2
9 128 11 117 10
20
20
1221
21
19
19
17
17
15
15
13
13
18
18
16
16
14
14
22 9
75
5
3 3
1
1
10
86
6
44
22
25
24
23
160
4060
Figure 6 A 25-bar Bowstring truss
modal analysis methods Natural frequencies are relativelyeasy to measure and have been used by many researchers
To identify localized damage because of the greaterexperience variations in the locality of the affected areamodeshapes offer a better option (Salane and Baldwin 1990 [24]Salawu and Williams 1995 [25] Ndambi et al 2002 [26]) Itis necessary to note that the success of this process depends onthe quality and place selection (for test) of the measurementswhich are able to reflect the damage
Because of this assuming that only a few natural fre-quencies and mode shapes of the lower modes are availablefrequency objective functions from incomplete data areconsidered in this paper to detect damage
5 Numerical Simulation Study
In this section two examples consisting of a ten-element can-tilever beam a Bowstring plane truss and a three-story three-bay unbraced frame structures are presented to examine thecharged system search algorithm The final results are thencompared to the solutions of particle swarm optimization
6 Shock and Vibration
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
005
01
015
02
025
03
035
04
Number of analyses
Obj
ectiv
e fun
ctio
n
(a) Scenario 1
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
005
01
015
02
025
03
035
Number of analyses
Obj
ectiv
e fun
ctio
n
(b) Scenario 2
Figure 7 The convergence history of the Bowstring truss for the CSS and PSOPC algorithms
Incomplete noisy dataComplete noisy dataActual damage
0 2 4 6 8 10 12 14 16 18 20 22 240
5
10
15
20
25
30
Element number
Dam
age (
)
Figure 8 Damage distribution of Bowstring truss using the com-plete and incomplete noisy data in Scenario 1
with passive congregation algorithm to demonstrate theperformance of this work
In CSS algorithm the effect of the pervious velocity andthe resultant force affecting a CP can decrease or increasebased on the values of the 119896V and 119896119886 defined as [22]
119896V = 119888(1 minusiter
itermax)
119896119886 = 119888(1 +iter
itermax)
(14)
where iter is the iteration number itermaxis the maximumnumber of the iterations and 119888 is set to 1
For the CSS algorithm a population of 16 CPs and forPSOPC algorithm a population of 50 particles are used for all
Incomplete noisy dataComplete noisy dataActual damage
Element number0 2 4 6 8 10 12 14 16 18 20 22 24
0
5
10
15
20
25
30
35
40
45D
amag
e (
)
Figure 9 Damage distribution of Bowstring truss using the com-plete and incomplete noisy data in Scenario 2
Table 1 Simulated damage scenarios in cantilever beam
Element number Scenario 1 Scenario 21 1035 206 107 359
the examples the stop criterion is considered as maximumnumber of 3000 and 5000 analyses in first and two otherexamples respectively
Shock and Vibration 7
1
2
13
23 24
14 16 181715
22 25 26 27
19 2120
28 3029
32 3331 34 35 36 37 3938
3 6 9 12
5 8 11
4 7 104m
35m
8m 7m 9m
35m
W12 times 65
W14 times 132
Figure 10 Geometry of unbraced plane frame
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
01
02
03
04
05
Number of analyses
Obj
ectiv
e fun
ctio
n
(a) Scenario 1
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000Number of analyses
0
005
01
015
02
025
03
035
04
Obj
ectiv
e fun
ctio
n
(b) Scenario 2
Figure 11 The convergence history of the 39-element frame for the CSS and PSOPC algorithms
The algorithm has been implemented in the commercialMATLAB software and because of the stochastic nature of thealgorithms each example is independently solved ten times
Damage values have been limited to a region between 0and 1
Example 1 A cantilever steel beam is shown in Figure 2 Forthe reason of modal analyzing the beam was divided into 10two-dimensional beam elements with 20 degrees of freedom
The section of the beam is W12 times 65 with mechanicalproperties of
119860 = 00123 m2 (191 in2)
119868 = 2218 times 10minus4 m4 (533 in4)
119864 = 207 times 109Nm2
120588 = 7860 kgm3
The measured modal responses of the beam before and afterdamage were created using the proposed forward solver(for each scenario) Instead of experimental measurementsnumerically generated measurements were used to estimatethe proposed inverse procedure In this example the first tennatural frequencies are used for damage detection Two dif-ferent simulated damage scenarios are considered (Table 1)
These damage scenarios are considered to represent theeffect of severity of damage (stiffness reduction) numberof damaged elements and contribution of damage elementson the results The results of damage detecting of CSS andPSOPC algorithms are given in Table 2 In order to evaluatethe performance of the methods used the summation ofestimated error of damage detection was calculated (Table 2)that is presented as follows
Error = sum(119889119894119860 minus 119889119894119878) (15)
8 Shock and Vibration
Table 2 Results of damage detection of the 10-element beam using CSS and PSOPC algorithms
Elementnumber
Scenario 1 Scenario 2
CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0000 0 0000 0101 01 0105
2 0000 0 0000 0003 0 0000
3 0000 0 0000 0001 0 0000
4 0000 0 0000 0000 0 0000
5 0000 0 0000 0185 02 0169
6 0100 01 0100 0016 0 0037
7 0000 0 0000 0349 035 0348
8 0000 0 0000 0002 0 0000
9 0000 0 0000 0001 0 0000
10 0000 0 0000 0000 0 0000
Error 00002 00000 00385 00752
Table 3 The values of objective function of the 10-element beam for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 34 times 10
minus0652 times 10
minus0628 times 10
minus0627 times 10
minus0683 times 10
minus0665 times 10
minus0641 times 10
minus0638 times 10
minus0619 times 10
minus0644 times 10
minus06 100PSOPC 25 times 10
minus0713 times 10
minus0799 times 10
minus0814 times 10
minus0721 times 10
minus0746 times 10
minus0726 times 10
minus0716 times 10
minus0688 times 10
minus0811 times 10
minus07 100Scenario 2
CSS 00008 00001 00009 00004 00006 00002 00005 00008 00009 00006 80PSOPC 00003 00013 00056 00004 00074 00074 00080 00011 00006 00085 40
Table 4 Cross-sectional areas of truss elements
Member Area (cm2)1ndash6 187ndash12 1513ndash17 1018ndash25 12
Table 5 Simulated damage scenarios in Bowstring truss
Element number Scenario 1 Scenario 235 108 251011 4024 20
where 119889119894119860 and 119889119894119878 are the actual and estimated damage of the
119894th element using the presented method respectivelyIn Table 3 the values of the objective function for CSS
and PSOPC algorithms in independent 10 runs are given theconvergence history for the 10-element beam in scenarios 1and 2 is shown in Figure 3
It is seen in Table 2 that in the first scenario where thenumber of damaged elements is low the damage identifi-cation is well by both algorithms in the second scenariothe algorithm PSOPC has some wrong in damage detectionso that there is difference between the obtained damage viathis algorithm and the actual intensity of damage in thefifth element and the undamaged sixth element mistakenlyhas been detected to be damaged Meanwhile the locationand amount of damage in structure are obtained by the CSSalgorithm with acceptable accuracy In order to investigatethe noise effects on the results of the CSSmethod 015 noisein measurement is considered for the natural frequencies
Diagrams of the CSS algorithmrsquos damage detection withcomplete and incomplete dynamic noisy data are plotted inFigures 4 and 5
As the results show this method is able to detect thelocation andmagnitude of damaged elements in all scenariosusing complete and incomplete noisy data
Example 2 A Bowstring plane truss with 25 elements isshown in Figure 6 The properties of members of this struc-ture are 119864 = 207 times 0
9Nm2 and 120588 = 7860 kgm3 and cross-sectional areas of elements are given in Table 4
Three scenarios for this truss are considered according toTable 5
Shock and Vibration 9
Table 6 Results of damage detection of the 25-bar Bowstring truss using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0009 0 0000 0000 0 00002 0000 0 0000 0000 0 00003 0000 0 0000 0000 0 00004 0000 0 0000 0000 0 00005 0000 0 0000 0098 01 00956 0000 0 0000 0000 0 00007 0000 0 0000 0000 0 00008 0246 025 0250 0000 0 00009 0001 0 0000 0000 0 000010 0000 0 0000 0000 0 000011 0000 0 0000 0402 04 042612 0000 0 0000 0000 0 000013 0000 0 0000 0000 0 000014 0001 0 0000 0000 0 000015 0000 0 0000 0000 0 000016 0000 0 0000 0000 0 000017 0000 0 0000 0000 0 000018 0000 0 0000 0005 0 000019 0000 0 0000 0004 0 007520 0000 0 0000 0001 0 000021 0000 0 0000 0000 0 000022 0000 0 0000 0000 0 000023 0000 0 0000 0000 0 000024 0000 0 0000 0185 02 000025 0000 0 0000 0014 0 0265Error 00085 00000 00119 01609
In Table 6 the results of two algorithms in all scenarios arecompared Only the first ten natural frequencies are used fordamage detection
The comparison of the objective function for CSS andPSOPC algorithms is made in Table 7 and the convergencehistory for the Bowstring truss in scenarios 1 and 2 is shownin Figure 7
In Figures 8 and 9 the results of CSS algorithm in damagedetection of this truss that is affected by noise are shown
Table 7 and Figures 7 8 and 9 show better results indamage detection by CSS algorithm rather than PSOPCalgorithm and also by applying noise in structures
Example 3 A three-story three-bay frame as shown inFigure 10 is used to verify the damage detection methodexplained in this paper The number of elements and nodesis 39 and 34 respectively
For unbraced plane frame problem all columns areW14 times 132 (119860 = 0025m2 (388 in2) and 119868 = 6386 times
10minus4m4 (1530 in4)) and all beams are W12 times 65 Youngrsquos
modulus is 119864 = 207GPa (30 000 ksi) Poissonrsquos ratio is ] =
03 and the mass density is 120588 = 7780 kgm3 (0000728 lb minuss2in4)
In this example two scenarios from the aspect of elementnumber and its level of damage were assumed that are givenin Table 8 Between ninety natural frequencies only the firsttwenty natural frequencies are utilized for damage detectionin the first scenario
The severity of damage detection by PSOPC and CSSalgorithms in different members of the frame and in eachscenario is shown separately in Table 9 In Table 10 theobtained objective function for both algorithms in each runis given also the convergence history is shown in Figure 11
Also the results of the CSS algorithm damage detectionwith noisy data are plotted in the diagrams of Figures 12 and13
The results of damage detection with two algorithmsin this large frame show that in case of only one memberdamaged there is more accuracy with PSOPC algorithm inidentifying of desired member and obtaining the amount ofdamage than the CSS algorithm It is observed that increasingthe number of injured elements in the structures reducesthe result precision with algorithms From results errors itcan be concluded that the PSOPC algorithm has problemin detecting the location of some of the damaged structuralelements so that in the third scenario the tenth and twenty-fourth members which are injured are diagnosed as healthy
10 Shock and Vibration
Table 7 The values of objective function of the Bowstring truss for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 32 times 10
minus0543 times 10
minus0549 times 10
minus0569 times 10
minus0571 times 10
minus0533 times 10
minus0526 times 10
minus0575 times 10
minus0534 times 10
minus0526 times 10
minus05 100PSOPC 26 times 10
minus0749 times 10
minus0218 times 10
minus0713 times 10
minus0112 times 10
minus0124 times 10
minus0850 times 10
minus0819 times 10
minus0826 times 10
minus0217 times 10
minus02 60Scenario 2
CSS 00002 00003 00003 00001 00039 00038 00002 00003 00015 00051 60PSOPC 00005 00134 00672 00629 00598 00050 00049 00065 00123 00251 10
Incomplete noisy dataComplete noisy dataActual damage
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 390
5
10
15
20
25
30
35
40
Element number
Dam
age (
)
Figure 12 Damage distribution of plane frame using the completeand incomplete noisy data in Scenario 1
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 400
10
20
30
40
50
60
Element number
Dam
age (
)
Predicted damageActual damage
Figure 13 Damage distribution of plane frame using the completenoisy data in Scenario 2
Table 8 Simulated damage scenarios in unbraced plane frame
Element number Scenario 1 Scenario 21 4045 309 1010 1013 2014 4019 2524 2032 50
and a considerable amount of damage in the fifteenth andtwenty-third elements that have no damage has been gainedHowever the CSS algorithm has well identified places ofall damaged elements in the structure and has achieved theamount of their damage with high accuracy and low errorFrom Figures 12 and 13 it can be observed that even in thatcase a lot of damage to the frame is considered the effectof noise on the results obtained by the CSS algorithm islow so that in case of incomplete data accuracy of resultshas reduced slightly and the performance of this algorithmdespite the noise was considered acceptable
In all three examples the convergences diagrams areshown that the process of convergence is gradual So it causesa more comprehensive search algorithm for finding optimalsolutions
The result increases the success rate of this type ofalgorithm performance in comparison with the PSOPCalgorithm
6 Conclusions
An approach for detecting damage based on continuumdamage model using charged system search algorithm ispresented The algorithm evaluates the location and severityof damage in three structures a cantilever beam a Bowstringplane truss and a three-story three-bay unbraced plane frameby minimizing an objective function by measuring completeand incomplete noisy modal data with different damagescenarios
Shock and Vibration 11
Table 9 Results of damage detection of the 39-element frame using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0399 0400 0393 0005 0 00002 0001 0 0000 0001 0 00003 0001 0 0000 0000 0 00004 0001 0 0027 0037 0 00005 0003 0 0000 0254 0300 03566 0000 0 0000 0016 0 00007 0000 0 0000 0013 0 00008 0005 0 0000 0001 0 00009 0056 0100 0000 0001 0 000010 0002 0 0000 0095 0100 000011 0005 0 0000 0002 0 000012 0008 0 0000 0001 0 000013 0193 0200 0195 0058 0 000014 0001 0 0000 0374 0400 034915 0004 0 0000 0005 0 011516 0000 0 0000 0001 0 000017 0000 0 0000 0001 0 000018 0000 0 0000 0004 0 000019 0000 0 0000 0212 0250 025620 0000 0 0000 0004 0 000021 0000 0 0000 0029 0 000022 0000 0 0002 0015 0 000023 0000 0 0000 0001 0 009224 0000 0 0000 0174 0200 000025 0000 0 0000 0001 0 000026 0000 0 0000 0002 0 000027 0002 0 0000 0005 0 000028 0000 0 0000 0001 0 004029 0000 0 0001 0007 0 000030 0000 0 0000 0000 0 000031 0001 0 0003 0002 0 000032 0000 0 0000 0497 0500 049533 0000 0 0000 0000 0 000034 0006 0 0000 0000 0 000035 0005 0 0025 0001 0 000036 0006 0 0000 0001 0 000037 0002 0 0000 0001 0 001538 0007 0 0000 0001 0 000039 0003 0 0000 0002 0 0000Error 01162 01700 03600 06808
In order to demonstrate the power of this algorithm in thediagnosis of damage a comparison has been made betweenthe results of this algorithm and the PSOPC algorithm
By comparing the damage detection results of the twovarious methods some interesting points have been con-cluded In scenarios where the number of damaged elementsis considered low both algorithms have acceptable accuracyin the results considering the more damaged members
of structures the PSOPC algorithm has been mistaken toidentify the damaged elements of the structureThis problemin larger-scale structures and the increasing number ofmembers has beenmore visible CSS algorithm has identifiedthe exact location of damage as well as achieving the amountof damage with acceptable accuracy It can be concluded thatthe CSS algorithm has great potential in global and localsearch of damage
12 Shock and Vibration
Table 10 The values of objective function of the 39-element frame for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 44 times 10
minus0346 times 10
minus0387 times 10
minus0462 times 10
minus0411 times 10
minus0366 times 10
minus0437 times 10
minus0480 times 10
minus0429 times 10
minus0398 times 10
minus04 50PSOPC 12 times 10
minus0213 times 10
minus0244 times 10
minus0315 times 10
minus0243 times 10
minus0213 times 10
minus0339 times 10
minus0214 times 10
minus0239 times 10
minus0315 times 10
minus02 10Scenario 2
CSS 19 times 10minus03
13 times 10minus03
17 times 10minus03
12 times 10minus03
23 times 10minus03
28 times 10minus03
19 times 10minus03
11 times 10minus03
21 times 10minus03
16 times 10minus03 30
PSOPC 29 times 10minus02
39 times 10minus02
11 times 10minus02
18 times 10minus02
11 times 10minus02
59 times 10minus03
69 times 10minus02
36 times 10minus02
53 times 10minus02
13 times 10minus02 0
Due to noise the real value of natural frequencies ofstructure will change it can be seen from the diagramsthat in these cases accuracy reduction of the results of theCSS algorithm is very low This indicates the high powerof this algorithm in damage detection considering noise instructure
Charged system search by considering small populationand low iteration cycle can detect the severity and location ofdamage with acceptable accuracy which shows high powerand speed of convergence of this algorithm The proposedmethod by the charged system search algorithm producedbetter results compared with particle swarm optimizationmethod
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] D J Ewins Modal Testing Theory and Practice John Wiley ampSons New York NY USA 1984
[2] S W Doebling C R Farrar and M B Prime ldquoA summaryreview of vibration-based damage identification methodsrdquoShock and Vibration Digest vol 30 no 2 pp 91ndash105 1998
[3] N Hu XWang H Fukunaga Z H Yao H X Zhang and Z SWu ldquoDamage assessment of structures using modal test datardquoInternational Journal of Solids and Structures vol 38 no 18 pp3111ndash3126 2001
[4] O S Salawu ldquoDetection of structural damage through changesin frequency a reviewrdquo Engineering Structures vol 19 no 9 pp718ndash723 1997
[5] N Bicanic and H-P Chen ldquoDamage identification in framedstructures using natural frequenciesrdquo International Journal forNumerical Methods in Engineering vol 40 no 23 pp 4451ndash4468 1997
[6] A K Pandey M Biswas and M M Samman ldquoDamagedetection from changes in curvature mode shapesrdquo Journal ofSound and Vibration vol 145 no 2 pp 321ndash332 1991
[7] J E Mottershead and M I Friswell ldquoModel updating instructural dynamics a surveyrdquo Journal of Sound and Vibrationvol 167 no 2 pp 347ndash375 1993
[8] R Perera andR Torres ldquoStructural damage detection viamodaldata with genetic algorithmsrdquo Journal of Structural Engineeringvol 132 no 9 pp 1491ndash1501 2006
[9] S Fallahian and S M Seyedpoor ldquoA two stage method forstructural damage identification using an adaptive neuro-fuzzyinference system and particle Swarm optimizationrdquo AsianJournal of Civil Engineering (Building and Housing) vol 11 pp797ndash810 2010
[10] B H Koh and S J Dyke ldquoStructural health monitoring forflexible bridge structures using correlation and sensitivity ofmodal datardquo Computers and Structures vol 85 no 3-4 pp 117ndash130 2007
[11] R-S He and S-F Hwang ldquoDamage detection by a hybrid real-parameter genetic algorithm under the assistance of grey rela-tion analysisrdquo Engineering Applications of Artificial Intelligencevol 20 no 7 pp 980ndash992 2007
[12] H Y Guo and Z L Li ldquoA two-stage method to identifystructural damage sites and extents by using evidence theoryand micro-search genetic algorithmrdquo Mechanical Systems andSignal Processing vol 23 no 3 pp 769ndash782 2009
[13] L Yu and X Chen ldquoBridge damage identification by combiningmodal flexibility and PSO methodsrdquo in Proceedings of thePrognostics and System Health Management Conference (PHMrsquo10) Macau China January 2010
[14] Z Tabrizian E Afshari G Ghodrati Amiri M H A Beigyand SM PourhoseiniNejad ldquoAnewdamage detectionmethodBig Bang-Big Crunch (BB-BC) algorithmrdquo Journal of Shock andVibration vol 20 no 4 pp 633ndash648 2013
[15] A Kaveh and S Talatahari ldquoA novel heuristic optimizationmethod charged system searchrdquo Acta Mechanica vol 213 no3-4 pp 267ndash289 2010
[16] S M Pourhoseini Nejad G Ghodrati Amiri A Asadi EAfshari and Z Tabrizian ldquoDamage detection of skeletal struc-tures using particle swarm optimizer with passive congregation(PSOPC) algorithm via incomplete modal datardquo Journal ofComputational Methods in Civil Engineering vol 3 no 1 pp1ndash13 2012
[17] S He Q H Wu J Y Wen J R Saunders and R CPaton ldquoA particle swarm optimizer with passive congregationrdquoBioSystems vol 78 no 1ndash3 pp 135ndash147 2004
[18] X Wang N Hu H Fukunaga and Z H Yao ldquoStructuraldamage identification using static test data and changes infrequenciesrdquo Engineering Structures vol 23 no 6 pp 610ndash6212001
[19] A Messina E J Williams and T Contursi ldquoStructural damagedetection by a sensitivity and statistical-based methodrdquo Journalof Sound and Vibration vol 216 no 5 pp 791ndash808 1998
[20] S W Doebling C R Farrar M B Prime and D W ShevitzldquoDamage identification and healthmonitoring of structural and
Shock and Vibration 13
mechanical systems from changes in their vibration character-istics a literature reviewrdquo Research Report LA-13070-MS ESA-EA Los Alamos National Laboratory Los Alamos NM USA1996
[21] M Nobahari and S M Seyedpoor ldquoStructural damage detec-tion using an efficient correlation-based index and a modifiedgenetic algorithmrdquoMathematical and Computer Modelling vol53 no 9-10 pp 1798ndash1809 2011
[22] A Kaveh and S Talatahari ldquoOptimal design of skeletal struc-tures via the charged system search algorithmrdquo Structural andMultidisciplinary Optimization vol 41 no 6 pp 893ndash911 2010
[23] A Kaveh and S Talatahari ldquoParticle swarm optimizer antcolony strategy and harmony search scheme hybridized foroptimization of truss structuresrdquoComputers and Structures vol87 no 5-6 pp 267ndash283 2009
[24] H J Salane and J W Baldwin Jr ldquoIdentification of modalproperties of bridgesrdquo Journal of Structural Engineering vol 116no 7 pp 2008ndash2021 1990
[25] O S Salawu and CWilliams ldquoBridge assessment using forced-vibration testingrdquo Journal of Structural Engineering vol 121 no2 pp 161ndash173 1995
[26] J-M Ndambi J Vantomme and K Harri ldquoDamage assessmentin reinforced concrete beams using eigenfrequencies and modeshape derivativesrdquo Engineering Structures vol 24 no 4 pp 501ndash515 2002
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
4 Shock and Vibration
Search
No
Yes
No
Yes
Terminationcriterion
Step 1 Step 2
Step 3
Step 4
Step 1 Step 2
Step 3
Initialize specification of problem and algorithm
parameters and determinethe initial charged particles
Find the values of the objective function for the
CPs and rank of CPs in an increasing order
Store some of the best CPs in
the charged memory (CM)
Determine the probability of
moving and force vector for each CP
Determine the new positions and
velocities of the CPs
Correct the CP positions using an algorithm based on HS if the CP does not satisfy
the side limits
Evaluate the objective function and rank the
CPs according to their quality
Include the better vectors in the CM and exclude the worst ones
from the CM
Stop
Are the new CPs better than the stored ones
in CM
Initialization
Figure 1 Flowchart of the CSS algorithm
1 2 96 8 1043 5 7
1007mW12 times 65
Figure 2 A cantilever beam geometry
0 500 1000 1500 2000 2500 30000
005
01
015
02
025
Number of analyses
Obj
ectiv
e fun
ctio
n
CSSPSOPC
(a) Scenario 1
CSSPSOPC
0 500 1000 1500 2000 2500 30000
005
01
015
02
025
Number of analyses
Obj
ectiv
e fun
ctio
n
(b) Scenario 2
Figure 3 The convergence history of the 10-element beam for the CSS and PSOPC algorithms
Shock and Vibration 5
0 1 2 3 4 5 6 7 8 9 100
5
10
15
Element number
Dam
age (
)
Incomplete noisy dataComplete noisy dataActual damage
Figure 4 Damage distribution of cantilever beam using the com-plete and incomplete noisy data in Scenario 1
numberThe probability of moving each CP toward the otherCPs is determined using the following function
119901119894119895 =
1
fit (119894) minus fitbestfit (119895) minus fit (119894)
gt rand or fit (119895) gt fit (119894)
0 otherwise(12)
After production of the new position of the CPs if anycomponent of the solution vector swerves off the allowablebounds correct its position using the harmony search-basedhandling approach as described in [23]
The charged memory (CM) is used to save a number ofthe best solutions up to the iterationThe better new solutionsare included in the CM and the worst ones are excluded fromthe CM The flowchart of the CSS algorithm is illustrated inFigure 1
4 Objective Function
The CSS algorithm attempts to minimize an objective assess-ment function for the best solution to a given problem Thisfunction is used to provide a measure of how individualshave performed in the problem domain In the case of aminimization problem the fittest individuals will have thelowest numerical value of the associated objective functionIn this study the statement for the objective function is givenas
119865 = 119891 (119889) (13)
where 119889 = 1198891 1198892 119889119873 are damage parameters at the 119873elements
To create the objective function 119865 it is essential touse some kind of output variables of the structure thatare sufficiently sensitive to the damage parameter beingidentified in order to avoid ill conditioning problems Themode shapes and the natural frequencies can be obtained by
Incomplete noisy dataComplete noisy dataActual damage
0 1 2 3 4 5 6 7 8 9 100
5
10
15
20
25
30
35
40
45
Element number
Dam
age (
)
Figure 5 Damage distribution of cantilever beam using the com-plete and incomplete noisy data in Scenario 2
9 128 11 117 10
20
20
1221
21
19
19
17
17
15
15
13
13
18
18
16
16
14
14
22 9
75
5
3 3
1
1
10
86
6
44
22
25
24
23
160
4060
Figure 6 A 25-bar Bowstring truss
modal analysis methods Natural frequencies are relativelyeasy to measure and have been used by many researchers
To identify localized damage because of the greaterexperience variations in the locality of the affected areamodeshapes offer a better option (Salane and Baldwin 1990 [24]Salawu and Williams 1995 [25] Ndambi et al 2002 [26]) Itis necessary to note that the success of this process depends onthe quality and place selection (for test) of the measurementswhich are able to reflect the damage
Because of this assuming that only a few natural fre-quencies and mode shapes of the lower modes are availablefrequency objective functions from incomplete data areconsidered in this paper to detect damage
5 Numerical Simulation Study
In this section two examples consisting of a ten-element can-tilever beam a Bowstring plane truss and a three-story three-bay unbraced frame structures are presented to examine thecharged system search algorithm The final results are thencompared to the solutions of particle swarm optimization
6 Shock and Vibration
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
005
01
015
02
025
03
035
04
Number of analyses
Obj
ectiv
e fun
ctio
n
(a) Scenario 1
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
005
01
015
02
025
03
035
Number of analyses
Obj
ectiv
e fun
ctio
n
(b) Scenario 2
Figure 7 The convergence history of the Bowstring truss for the CSS and PSOPC algorithms
Incomplete noisy dataComplete noisy dataActual damage
0 2 4 6 8 10 12 14 16 18 20 22 240
5
10
15
20
25
30
Element number
Dam
age (
)
Figure 8 Damage distribution of Bowstring truss using the com-plete and incomplete noisy data in Scenario 1
with passive congregation algorithm to demonstrate theperformance of this work
In CSS algorithm the effect of the pervious velocity andthe resultant force affecting a CP can decrease or increasebased on the values of the 119896V and 119896119886 defined as [22]
119896V = 119888(1 minusiter
itermax)
119896119886 = 119888(1 +iter
itermax)
(14)
where iter is the iteration number itermaxis the maximumnumber of the iterations and 119888 is set to 1
For the CSS algorithm a population of 16 CPs and forPSOPC algorithm a population of 50 particles are used for all
Incomplete noisy dataComplete noisy dataActual damage
Element number0 2 4 6 8 10 12 14 16 18 20 22 24
0
5
10
15
20
25
30
35
40
45D
amag
e (
)
Figure 9 Damage distribution of Bowstring truss using the com-plete and incomplete noisy data in Scenario 2
Table 1 Simulated damage scenarios in cantilever beam
Element number Scenario 1 Scenario 21 1035 206 107 359
the examples the stop criterion is considered as maximumnumber of 3000 and 5000 analyses in first and two otherexamples respectively
Shock and Vibration 7
1
2
13
23 24
14 16 181715
22 25 26 27
19 2120
28 3029
32 3331 34 35 36 37 3938
3 6 9 12
5 8 11
4 7 104m
35m
8m 7m 9m
35m
W12 times 65
W14 times 132
Figure 10 Geometry of unbraced plane frame
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
01
02
03
04
05
Number of analyses
Obj
ectiv
e fun
ctio
n
(a) Scenario 1
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000Number of analyses
0
005
01
015
02
025
03
035
04
Obj
ectiv
e fun
ctio
n
(b) Scenario 2
Figure 11 The convergence history of the 39-element frame for the CSS and PSOPC algorithms
The algorithm has been implemented in the commercialMATLAB software and because of the stochastic nature of thealgorithms each example is independently solved ten times
Damage values have been limited to a region between 0and 1
Example 1 A cantilever steel beam is shown in Figure 2 Forthe reason of modal analyzing the beam was divided into 10two-dimensional beam elements with 20 degrees of freedom
The section of the beam is W12 times 65 with mechanicalproperties of
119860 = 00123 m2 (191 in2)
119868 = 2218 times 10minus4 m4 (533 in4)
119864 = 207 times 109Nm2
120588 = 7860 kgm3
The measured modal responses of the beam before and afterdamage were created using the proposed forward solver(for each scenario) Instead of experimental measurementsnumerically generated measurements were used to estimatethe proposed inverse procedure In this example the first tennatural frequencies are used for damage detection Two dif-ferent simulated damage scenarios are considered (Table 1)
These damage scenarios are considered to represent theeffect of severity of damage (stiffness reduction) numberof damaged elements and contribution of damage elementson the results The results of damage detecting of CSS andPSOPC algorithms are given in Table 2 In order to evaluatethe performance of the methods used the summation ofestimated error of damage detection was calculated (Table 2)that is presented as follows
Error = sum(119889119894119860 minus 119889119894119878) (15)
8 Shock and Vibration
Table 2 Results of damage detection of the 10-element beam using CSS and PSOPC algorithms
Elementnumber
Scenario 1 Scenario 2
CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0000 0 0000 0101 01 0105
2 0000 0 0000 0003 0 0000
3 0000 0 0000 0001 0 0000
4 0000 0 0000 0000 0 0000
5 0000 0 0000 0185 02 0169
6 0100 01 0100 0016 0 0037
7 0000 0 0000 0349 035 0348
8 0000 0 0000 0002 0 0000
9 0000 0 0000 0001 0 0000
10 0000 0 0000 0000 0 0000
Error 00002 00000 00385 00752
Table 3 The values of objective function of the 10-element beam for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 34 times 10
minus0652 times 10
minus0628 times 10
minus0627 times 10
minus0683 times 10
minus0665 times 10
minus0641 times 10
minus0638 times 10
minus0619 times 10
minus0644 times 10
minus06 100PSOPC 25 times 10
minus0713 times 10
minus0799 times 10
minus0814 times 10
minus0721 times 10
minus0746 times 10
minus0726 times 10
minus0716 times 10
minus0688 times 10
minus0811 times 10
minus07 100Scenario 2
CSS 00008 00001 00009 00004 00006 00002 00005 00008 00009 00006 80PSOPC 00003 00013 00056 00004 00074 00074 00080 00011 00006 00085 40
Table 4 Cross-sectional areas of truss elements
Member Area (cm2)1ndash6 187ndash12 1513ndash17 1018ndash25 12
Table 5 Simulated damage scenarios in Bowstring truss
Element number Scenario 1 Scenario 235 108 251011 4024 20
where 119889119894119860 and 119889119894119878 are the actual and estimated damage of the
119894th element using the presented method respectivelyIn Table 3 the values of the objective function for CSS
and PSOPC algorithms in independent 10 runs are given theconvergence history for the 10-element beam in scenarios 1and 2 is shown in Figure 3
It is seen in Table 2 that in the first scenario where thenumber of damaged elements is low the damage identifi-cation is well by both algorithms in the second scenariothe algorithm PSOPC has some wrong in damage detectionso that there is difference between the obtained damage viathis algorithm and the actual intensity of damage in thefifth element and the undamaged sixth element mistakenlyhas been detected to be damaged Meanwhile the locationand amount of damage in structure are obtained by the CSSalgorithm with acceptable accuracy In order to investigatethe noise effects on the results of the CSSmethod 015 noisein measurement is considered for the natural frequencies
Diagrams of the CSS algorithmrsquos damage detection withcomplete and incomplete dynamic noisy data are plotted inFigures 4 and 5
As the results show this method is able to detect thelocation andmagnitude of damaged elements in all scenariosusing complete and incomplete noisy data
Example 2 A Bowstring plane truss with 25 elements isshown in Figure 6 The properties of members of this struc-ture are 119864 = 207 times 0
9Nm2 and 120588 = 7860 kgm3 and cross-sectional areas of elements are given in Table 4
Three scenarios for this truss are considered according toTable 5
Shock and Vibration 9
Table 6 Results of damage detection of the 25-bar Bowstring truss using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0009 0 0000 0000 0 00002 0000 0 0000 0000 0 00003 0000 0 0000 0000 0 00004 0000 0 0000 0000 0 00005 0000 0 0000 0098 01 00956 0000 0 0000 0000 0 00007 0000 0 0000 0000 0 00008 0246 025 0250 0000 0 00009 0001 0 0000 0000 0 000010 0000 0 0000 0000 0 000011 0000 0 0000 0402 04 042612 0000 0 0000 0000 0 000013 0000 0 0000 0000 0 000014 0001 0 0000 0000 0 000015 0000 0 0000 0000 0 000016 0000 0 0000 0000 0 000017 0000 0 0000 0000 0 000018 0000 0 0000 0005 0 000019 0000 0 0000 0004 0 007520 0000 0 0000 0001 0 000021 0000 0 0000 0000 0 000022 0000 0 0000 0000 0 000023 0000 0 0000 0000 0 000024 0000 0 0000 0185 02 000025 0000 0 0000 0014 0 0265Error 00085 00000 00119 01609
In Table 6 the results of two algorithms in all scenarios arecompared Only the first ten natural frequencies are used fordamage detection
The comparison of the objective function for CSS andPSOPC algorithms is made in Table 7 and the convergencehistory for the Bowstring truss in scenarios 1 and 2 is shownin Figure 7
In Figures 8 and 9 the results of CSS algorithm in damagedetection of this truss that is affected by noise are shown
Table 7 and Figures 7 8 and 9 show better results indamage detection by CSS algorithm rather than PSOPCalgorithm and also by applying noise in structures
Example 3 A three-story three-bay frame as shown inFigure 10 is used to verify the damage detection methodexplained in this paper The number of elements and nodesis 39 and 34 respectively
For unbraced plane frame problem all columns areW14 times 132 (119860 = 0025m2 (388 in2) and 119868 = 6386 times
10minus4m4 (1530 in4)) and all beams are W12 times 65 Youngrsquos
modulus is 119864 = 207GPa (30 000 ksi) Poissonrsquos ratio is ] =
03 and the mass density is 120588 = 7780 kgm3 (0000728 lb minuss2in4)
In this example two scenarios from the aspect of elementnumber and its level of damage were assumed that are givenin Table 8 Between ninety natural frequencies only the firsttwenty natural frequencies are utilized for damage detectionin the first scenario
The severity of damage detection by PSOPC and CSSalgorithms in different members of the frame and in eachscenario is shown separately in Table 9 In Table 10 theobtained objective function for both algorithms in each runis given also the convergence history is shown in Figure 11
Also the results of the CSS algorithm damage detectionwith noisy data are plotted in the diagrams of Figures 12 and13
The results of damage detection with two algorithmsin this large frame show that in case of only one memberdamaged there is more accuracy with PSOPC algorithm inidentifying of desired member and obtaining the amount ofdamage than the CSS algorithm It is observed that increasingthe number of injured elements in the structures reducesthe result precision with algorithms From results errors itcan be concluded that the PSOPC algorithm has problemin detecting the location of some of the damaged structuralelements so that in the third scenario the tenth and twenty-fourth members which are injured are diagnosed as healthy
10 Shock and Vibration
Table 7 The values of objective function of the Bowstring truss for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 32 times 10
minus0543 times 10
minus0549 times 10
minus0569 times 10
minus0571 times 10
minus0533 times 10
minus0526 times 10
minus0575 times 10
minus0534 times 10
minus0526 times 10
minus05 100PSOPC 26 times 10
minus0749 times 10
minus0218 times 10
minus0713 times 10
minus0112 times 10
minus0124 times 10
minus0850 times 10
minus0819 times 10
minus0826 times 10
minus0217 times 10
minus02 60Scenario 2
CSS 00002 00003 00003 00001 00039 00038 00002 00003 00015 00051 60PSOPC 00005 00134 00672 00629 00598 00050 00049 00065 00123 00251 10
Incomplete noisy dataComplete noisy dataActual damage
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 390
5
10
15
20
25
30
35
40
Element number
Dam
age (
)
Figure 12 Damage distribution of plane frame using the completeand incomplete noisy data in Scenario 1
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 400
10
20
30
40
50
60
Element number
Dam
age (
)
Predicted damageActual damage
Figure 13 Damage distribution of plane frame using the completenoisy data in Scenario 2
Table 8 Simulated damage scenarios in unbraced plane frame
Element number Scenario 1 Scenario 21 4045 309 1010 1013 2014 4019 2524 2032 50
and a considerable amount of damage in the fifteenth andtwenty-third elements that have no damage has been gainedHowever the CSS algorithm has well identified places ofall damaged elements in the structure and has achieved theamount of their damage with high accuracy and low errorFrom Figures 12 and 13 it can be observed that even in thatcase a lot of damage to the frame is considered the effectof noise on the results obtained by the CSS algorithm islow so that in case of incomplete data accuracy of resultshas reduced slightly and the performance of this algorithmdespite the noise was considered acceptable
In all three examples the convergences diagrams areshown that the process of convergence is gradual So it causesa more comprehensive search algorithm for finding optimalsolutions
The result increases the success rate of this type ofalgorithm performance in comparison with the PSOPCalgorithm
6 Conclusions
An approach for detecting damage based on continuumdamage model using charged system search algorithm ispresented The algorithm evaluates the location and severityof damage in three structures a cantilever beam a Bowstringplane truss and a three-story three-bay unbraced plane frameby minimizing an objective function by measuring completeand incomplete noisy modal data with different damagescenarios
Shock and Vibration 11
Table 9 Results of damage detection of the 39-element frame using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0399 0400 0393 0005 0 00002 0001 0 0000 0001 0 00003 0001 0 0000 0000 0 00004 0001 0 0027 0037 0 00005 0003 0 0000 0254 0300 03566 0000 0 0000 0016 0 00007 0000 0 0000 0013 0 00008 0005 0 0000 0001 0 00009 0056 0100 0000 0001 0 000010 0002 0 0000 0095 0100 000011 0005 0 0000 0002 0 000012 0008 0 0000 0001 0 000013 0193 0200 0195 0058 0 000014 0001 0 0000 0374 0400 034915 0004 0 0000 0005 0 011516 0000 0 0000 0001 0 000017 0000 0 0000 0001 0 000018 0000 0 0000 0004 0 000019 0000 0 0000 0212 0250 025620 0000 0 0000 0004 0 000021 0000 0 0000 0029 0 000022 0000 0 0002 0015 0 000023 0000 0 0000 0001 0 009224 0000 0 0000 0174 0200 000025 0000 0 0000 0001 0 000026 0000 0 0000 0002 0 000027 0002 0 0000 0005 0 000028 0000 0 0000 0001 0 004029 0000 0 0001 0007 0 000030 0000 0 0000 0000 0 000031 0001 0 0003 0002 0 000032 0000 0 0000 0497 0500 049533 0000 0 0000 0000 0 000034 0006 0 0000 0000 0 000035 0005 0 0025 0001 0 000036 0006 0 0000 0001 0 000037 0002 0 0000 0001 0 001538 0007 0 0000 0001 0 000039 0003 0 0000 0002 0 0000Error 01162 01700 03600 06808
In order to demonstrate the power of this algorithm in thediagnosis of damage a comparison has been made betweenthe results of this algorithm and the PSOPC algorithm
By comparing the damage detection results of the twovarious methods some interesting points have been con-cluded In scenarios where the number of damaged elementsis considered low both algorithms have acceptable accuracyin the results considering the more damaged members
of structures the PSOPC algorithm has been mistaken toidentify the damaged elements of the structureThis problemin larger-scale structures and the increasing number ofmembers has beenmore visible CSS algorithm has identifiedthe exact location of damage as well as achieving the amountof damage with acceptable accuracy It can be concluded thatthe CSS algorithm has great potential in global and localsearch of damage
12 Shock and Vibration
Table 10 The values of objective function of the 39-element frame for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 44 times 10
minus0346 times 10
minus0387 times 10
minus0462 times 10
minus0411 times 10
minus0366 times 10
minus0437 times 10
minus0480 times 10
minus0429 times 10
minus0398 times 10
minus04 50PSOPC 12 times 10
minus0213 times 10
minus0244 times 10
minus0315 times 10
minus0243 times 10
minus0213 times 10
minus0339 times 10
minus0214 times 10
minus0239 times 10
minus0315 times 10
minus02 10Scenario 2
CSS 19 times 10minus03
13 times 10minus03
17 times 10minus03
12 times 10minus03
23 times 10minus03
28 times 10minus03
19 times 10minus03
11 times 10minus03
21 times 10minus03
16 times 10minus03 30
PSOPC 29 times 10minus02
39 times 10minus02
11 times 10minus02
18 times 10minus02
11 times 10minus02
59 times 10minus03
69 times 10minus02
36 times 10minus02
53 times 10minus02
13 times 10minus02 0
Due to noise the real value of natural frequencies ofstructure will change it can be seen from the diagramsthat in these cases accuracy reduction of the results of theCSS algorithm is very low This indicates the high powerof this algorithm in damage detection considering noise instructure
Charged system search by considering small populationand low iteration cycle can detect the severity and location ofdamage with acceptable accuracy which shows high powerand speed of convergence of this algorithm The proposedmethod by the charged system search algorithm producedbetter results compared with particle swarm optimizationmethod
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] D J Ewins Modal Testing Theory and Practice John Wiley ampSons New York NY USA 1984
[2] S W Doebling C R Farrar and M B Prime ldquoA summaryreview of vibration-based damage identification methodsrdquoShock and Vibration Digest vol 30 no 2 pp 91ndash105 1998
[3] N Hu XWang H Fukunaga Z H Yao H X Zhang and Z SWu ldquoDamage assessment of structures using modal test datardquoInternational Journal of Solids and Structures vol 38 no 18 pp3111ndash3126 2001
[4] O S Salawu ldquoDetection of structural damage through changesin frequency a reviewrdquo Engineering Structures vol 19 no 9 pp718ndash723 1997
[5] N Bicanic and H-P Chen ldquoDamage identification in framedstructures using natural frequenciesrdquo International Journal forNumerical Methods in Engineering vol 40 no 23 pp 4451ndash4468 1997
[6] A K Pandey M Biswas and M M Samman ldquoDamagedetection from changes in curvature mode shapesrdquo Journal ofSound and Vibration vol 145 no 2 pp 321ndash332 1991
[7] J E Mottershead and M I Friswell ldquoModel updating instructural dynamics a surveyrdquo Journal of Sound and Vibrationvol 167 no 2 pp 347ndash375 1993
[8] R Perera andR Torres ldquoStructural damage detection viamodaldata with genetic algorithmsrdquo Journal of Structural Engineeringvol 132 no 9 pp 1491ndash1501 2006
[9] S Fallahian and S M Seyedpoor ldquoA two stage method forstructural damage identification using an adaptive neuro-fuzzyinference system and particle Swarm optimizationrdquo AsianJournal of Civil Engineering (Building and Housing) vol 11 pp797ndash810 2010
[10] B H Koh and S J Dyke ldquoStructural health monitoring forflexible bridge structures using correlation and sensitivity ofmodal datardquo Computers and Structures vol 85 no 3-4 pp 117ndash130 2007
[11] R-S He and S-F Hwang ldquoDamage detection by a hybrid real-parameter genetic algorithm under the assistance of grey rela-tion analysisrdquo Engineering Applications of Artificial Intelligencevol 20 no 7 pp 980ndash992 2007
[12] H Y Guo and Z L Li ldquoA two-stage method to identifystructural damage sites and extents by using evidence theoryand micro-search genetic algorithmrdquo Mechanical Systems andSignal Processing vol 23 no 3 pp 769ndash782 2009
[13] L Yu and X Chen ldquoBridge damage identification by combiningmodal flexibility and PSO methodsrdquo in Proceedings of thePrognostics and System Health Management Conference (PHMrsquo10) Macau China January 2010
[14] Z Tabrizian E Afshari G Ghodrati Amiri M H A Beigyand SM PourhoseiniNejad ldquoAnewdamage detectionmethodBig Bang-Big Crunch (BB-BC) algorithmrdquo Journal of Shock andVibration vol 20 no 4 pp 633ndash648 2013
[15] A Kaveh and S Talatahari ldquoA novel heuristic optimizationmethod charged system searchrdquo Acta Mechanica vol 213 no3-4 pp 267ndash289 2010
[16] S M Pourhoseini Nejad G Ghodrati Amiri A Asadi EAfshari and Z Tabrizian ldquoDamage detection of skeletal struc-tures using particle swarm optimizer with passive congregation(PSOPC) algorithm via incomplete modal datardquo Journal ofComputational Methods in Civil Engineering vol 3 no 1 pp1ndash13 2012
[17] S He Q H Wu J Y Wen J R Saunders and R CPaton ldquoA particle swarm optimizer with passive congregationrdquoBioSystems vol 78 no 1ndash3 pp 135ndash147 2004
[18] X Wang N Hu H Fukunaga and Z H Yao ldquoStructuraldamage identification using static test data and changes infrequenciesrdquo Engineering Structures vol 23 no 6 pp 610ndash6212001
[19] A Messina E J Williams and T Contursi ldquoStructural damagedetection by a sensitivity and statistical-based methodrdquo Journalof Sound and Vibration vol 216 no 5 pp 791ndash808 1998
[20] S W Doebling C R Farrar M B Prime and D W ShevitzldquoDamage identification and healthmonitoring of structural and
Shock and Vibration 13
mechanical systems from changes in their vibration character-istics a literature reviewrdquo Research Report LA-13070-MS ESA-EA Los Alamos National Laboratory Los Alamos NM USA1996
[21] M Nobahari and S M Seyedpoor ldquoStructural damage detec-tion using an efficient correlation-based index and a modifiedgenetic algorithmrdquoMathematical and Computer Modelling vol53 no 9-10 pp 1798ndash1809 2011
[22] A Kaveh and S Talatahari ldquoOptimal design of skeletal struc-tures via the charged system search algorithmrdquo Structural andMultidisciplinary Optimization vol 41 no 6 pp 893ndash911 2010
[23] A Kaveh and S Talatahari ldquoParticle swarm optimizer antcolony strategy and harmony search scheme hybridized foroptimization of truss structuresrdquoComputers and Structures vol87 no 5-6 pp 267ndash283 2009
[24] H J Salane and J W Baldwin Jr ldquoIdentification of modalproperties of bridgesrdquo Journal of Structural Engineering vol 116no 7 pp 2008ndash2021 1990
[25] O S Salawu and CWilliams ldquoBridge assessment using forced-vibration testingrdquo Journal of Structural Engineering vol 121 no2 pp 161ndash173 1995
[26] J-M Ndambi J Vantomme and K Harri ldquoDamage assessmentin reinforced concrete beams using eigenfrequencies and modeshape derivativesrdquo Engineering Structures vol 24 no 4 pp 501ndash515 2002
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
Shock and Vibration 5
0 1 2 3 4 5 6 7 8 9 100
5
10
15
Element number
Dam
age (
)
Incomplete noisy dataComplete noisy dataActual damage
Figure 4 Damage distribution of cantilever beam using the com-plete and incomplete noisy data in Scenario 1
numberThe probability of moving each CP toward the otherCPs is determined using the following function
119901119894119895 =
1
fit (119894) minus fitbestfit (119895) minus fit (119894)
gt rand or fit (119895) gt fit (119894)
0 otherwise(12)
After production of the new position of the CPs if anycomponent of the solution vector swerves off the allowablebounds correct its position using the harmony search-basedhandling approach as described in [23]
The charged memory (CM) is used to save a number ofthe best solutions up to the iterationThe better new solutionsare included in the CM and the worst ones are excluded fromthe CM The flowchart of the CSS algorithm is illustrated inFigure 1
4 Objective Function
The CSS algorithm attempts to minimize an objective assess-ment function for the best solution to a given problem Thisfunction is used to provide a measure of how individualshave performed in the problem domain In the case of aminimization problem the fittest individuals will have thelowest numerical value of the associated objective functionIn this study the statement for the objective function is givenas
119865 = 119891 (119889) (13)
where 119889 = 1198891 1198892 119889119873 are damage parameters at the 119873elements
To create the objective function 119865 it is essential touse some kind of output variables of the structure thatare sufficiently sensitive to the damage parameter beingidentified in order to avoid ill conditioning problems Themode shapes and the natural frequencies can be obtained by
Incomplete noisy dataComplete noisy dataActual damage
0 1 2 3 4 5 6 7 8 9 100
5
10
15
20
25
30
35
40
45
Element number
Dam
age (
)
Figure 5 Damage distribution of cantilever beam using the com-plete and incomplete noisy data in Scenario 2
9 128 11 117 10
20
20
1221
21
19
19
17
17
15
15
13
13
18
18
16
16
14
14
22 9
75
5
3 3
1
1
10
86
6
44
22
25
24
23
160
4060
Figure 6 A 25-bar Bowstring truss
modal analysis methods Natural frequencies are relativelyeasy to measure and have been used by many researchers
To identify localized damage because of the greaterexperience variations in the locality of the affected areamodeshapes offer a better option (Salane and Baldwin 1990 [24]Salawu and Williams 1995 [25] Ndambi et al 2002 [26]) Itis necessary to note that the success of this process depends onthe quality and place selection (for test) of the measurementswhich are able to reflect the damage
Because of this assuming that only a few natural fre-quencies and mode shapes of the lower modes are availablefrequency objective functions from incomplete data areconsidered in this paper to detect damage
5 Numerical Simulation Study
In this section two examples consisting of a ten-element can-tilever beam a Bowstring plane truss and a three-story three-bay unbraced frame structures are presented to examine thecharged system search algorithm The final results are thencompared to the solutions of particle swarm optimization
6 Shock and Vibration
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
005
01
015
02
025
03
035
04
Number of analyses
Obj
ectiv
e fun
ctio
n
(a) Scenario 1
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
005
01
015
02
025
03
035
Number of analyses
Obj
ectiv
e fun
ctio
n
(b) Scenario 2
Figure 7 The convergence history of the Bowstring truss for the CSS and PSOPC algorithms
Incomplete noisy dataComplete noisy dataActual damage
0 2 4 6 8 10 12 14 16 18 20 22 240
5
10
15
20
25
30
Element number
Dam
age (
)
Figure 8 Damage distribution of Bowstring truss using the com-plete and incomplete noisy data in Scenario 1
with passive congregation algorithm to demonstrate theperformance of this work
In CSS algorithm the effect of the pervious velocity andthe resultant force affecting a CP can decrease or increasebased on the values of the 119896V and 119896119886 defined as [22]
119896V = 119888(1 minusiter
itermax)
119896119886 = 119888(1 +iter
itermax)
(14)
where iter is the iteration number itermaxis the maximumnumber of the iterations and 119888 is set to 1
For the CSS algorithm a population of 16 CPs and forPSOPC algorithm a population of 50 particles are used for all
Incomplete noisy dataComplete noisy dataActual damage
Element number0 2 4 6 8 10 12 14 16 18 20 22 24
0
5
10
15
20
25
30
35
40
45D
amag
e (
)
Figure 9 Damage distribution of Bowstring truss using the com-plete and incomplete noisy data in Scenario 2
Table 1 Simulated damage scenarios in cantilever beam
Element number Scenario 1 Scenario 21 1035 206 107 359
the examples the stop criterion is considered as maximumnumber of 3000 and 5000 analyses in first and two otherexamples respectively
Shock and Vibration 7
1
2
13
23 24
14 16 181715
22 25 26 27
19 2120
28 3029
32 3331 34 35 36 37 3938
3 6 9 12
5 8 11
4 7 104m
35m
8m 7m 9m
35m
W12 times 65
W14 times 132
Figure 10 Geometry of unbraced plane frame
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
01
02
03
04
05
Number of analyses
Obj
ectiv
e fun
ctio
n
(a) Scenario 1
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000Number of analyses
0
005
01
015
02
025
03
035
04
Obj
ectiv
e fun
ctio
n
(b) Scenario 2
Figure 11 The convergence history of the 39-element frame for the CSS and PSOPC algorithms
The algorithm has been implemented in the commercialMATLAB software and because of the stochastic nature of thealgorithms each example is independently solved ten times
Damage values have been limited to a region between 0and 1
Example 1 A cantilever steel beam is shown in Figure 2 Forthe reason of modal analyzing the beam was divided into 10two-dimensional beam elements with 20 degrees of freedom
The section of the beam is W12 times 65 with mechanicalproperties of
119860 = 00123 m2 (191 in2)
119868 = 2218 times 10minus4 m4 (533 in4)
119864 = 207 times 109Nm2
120588 = 7860 kgm3
The measured modal responses of the beam before and afterdamage were created using the proposed forward solver(for each scenario) Instead of experimental measurementsnumerically generated measurements were used to estimatethe proposed inverse procedure In this example the first tennatural frequencies are used for damage detection Two dif-ferent simulated damage scenarios are considered (Table 1)
These damage scenarios are considered to represent theeffect of severity of damage (stiffness reduction) numberof damaged elements and contribution of damage elementson the results The results of damage detecting of CSS andPSOPC algorithms are given in Table 2 In order to evaluatethe performance of the methods used the summation ofestimated error of damage detection was calculated (Table 2)that is presented as follows
Error = sum(119889119894119860 minus 119889119894119878) (15)
8 Shock and Vibration
Table 2 Results of damage detection of the 10-element beam using CSS and PSOPC algorithms
Elementnumber
Scenario 1 Scenario 2
CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0000 0 0000 0101 01 0105
2 0000 0 0000 0003 0 0000
3 0000 0 0000 0001 0 0000
4 0000 0 0000 0000 0 0000
5 0000 0 0000 0185 02 0169
6 0100 01 0100 0016 0 0037
7 0000 0 0000 0349 035 0348
8 0000 0 0000 0002 0 0000
9 0000 0 0000 0001 0 0000
10 0000 0 0000 0000 0 0000
Error 00002 00000 00385 00752
Table 3 The values of objective function of the 10-element beam for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 34 times 10
minus0652 times 10
minus0628 times 10
minus0627 times 10
minus0683 times 10
minus0665 times 10
minus0641 times 10
minus0638 times 10
minus0619 times 10
minus0644 times 10
minus06 100PSOPC 25 times 10
minus0713 times 10
minus0799 times 10
minus0814 times 10
minus0721 times 10
minus0746 times 10
minus0726 times 10
minus0716 times 10
minus0688 times 10
minus0811 times 10
minus07 100Scenario 2
CSS 00008 00001 00009 00004 00006 00002 00005 00008 00009 00006 80PSOPC 00003 00013 00056 00004 00074 00074 00080 00011 00006 00085 40
Table 4 Cross-sectional areas of truss elements
Member Area (cm2)1ndash6 187ndash12 1513ndash17 1018ndash25 12
Table 5 Simulated damage scenarios in Bowstring truss
Element number Scenario 1 Scenario 235 108 251011 4024 20
where 119889119894119860 and 119889119894119878 are the actual and estimated damage of the
119894th element using the presented method respectivelyIn Table 3 the values of the objective function for CSS
and PSOPC algorithms in independent 10 runs are given theconvergence history for the 10-element beam in scenarios 1and 2 is shown in Figure 3
It is seen in Table 2 that in the first scenario where thenumber of damaged elements is low the damage identifi-cation is well by both algorithms in the second scenariothe algorithm PSOPC has some wrong in damage detectionso that there is difference between the obtained damage viathis algorithm and the actual intensity of damage in thefifth element and the undamaged sixth element mistakenlyhas been detected to be damaged Meanwhile the locationand amount of damage in structure are obtained by the CSSalgorithm with acceptable accuracy In order to investigatethe noise effects on the results of the CSSmethod 015 noisein measurement is considered for the natural frequencies
Diagrams of the CSS algorithmrsquos damage detection withcomplete and incomplete dynamic noisy data are plotted inFigures 4 and 5
As the results show this method is able to detect thelocation andmagnitude of damaged elements in all scenariosusing complete and incomplete noisy data
Example 2 A Bowstring plane truss with 25 elements isshown in Figure 6 The properties of members of this struc-ture are 119864 = 207 times 0
9Nm2 and 120588 = 7860 kgm3 and cross-sectional areas of elements are given in Table 4
Three scenarios for this truss are considered according toTable 5
Shock and Vibration 9
Table 6 Results of damage detection of the 25-bar Bowstring truss using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0009 0 0000 0000 0 00002 0000 0 0000 0000 0 00003 0000 0 0000 0000 0 00004 0000 0 0000 0000 0 00005 0000 0 0000 0098 01 00956 0000 0 0000 0000 0 00007 0000 0 0000 0000 0 00008 0246 025 0250 0000 0 00009 0001 0 0000 0000 0 000010 0000 0 0000 0000 0 000011 0000 0 0000 0402 04 042612 0000 0 0000 0000 0 000013 0000 0 0000 0000 0 000014 0001 0 0000 0000 0 000015 0000 0 0000 0000 0 000016 0000 0 0000 0000 0 000017 0000 0 0000 0000 0 000018 0000 0 0000 0005 0 000019 0000 0 0000 0004 0 007520 0000 0 0000 0001 0 000021 0000 0 0000 0000 0 000022 0000 0 0000 0000 0 000023 0000 0 0000 0000 0 000024 0000 0 0000 0185 02 000025 0000 0 0000 0014 0 0265Error 00085 00000 00119 01609
In Table 6 the results of two algorithms in all scenarios arecompared Only the first ten natural frequencies are used fordamage detection
The comparison of the objective function for CSS andPSOPC algorithms is made in Table 7 and the convergencehistory for the Bowstring truss in scenarios 1 and 2 is shownin Figure 7
In Figures 8 and 9 the results of CSS algorithm in damagedetection of this truss that is affected by noise are shown
Table 7 and Figures 7 8 and 9 show better results indamage detection by CSS algorithm rather than PSOPCalgorithm and also by applying noise in structures
Example 3 A three-story three-bay frame as shown inFigure 10 is used to verify the damage detection methodexplained in this paper The number of elements and nodesis 39 and 34 respectively
For unbraced plane frame problem all columns areW14 times 132 (119860 = 0025m2 (388 in2) and 119868 = 6386 times
10minus4m4 (1530 in4)) and all beams are W12 times 65 Youngrsquos
modulus is 119864 = 207GPa (30 000 ksi) Poissonrsquos ratio is ] =
03 and the mass density is 120588 = 7780 kgm3 (0000728 lb minuss2in4)
In this example two scenarios from the aspect of elementnumber and its level of damage were assumed that are givenin Table 8 Between ninety natural frequencies only the firsttwenty natural frequencies are utilized for damage detectionin the first scenario
The severity of damage detection by PSOPC and CSSalgorithms in different members of the frame and in eachscenario is shown separately in Table 9 In Table 10 theobtained objective function for both algorithms in each runis given also the convergence history is shown in Figure 11
Also the results of the CSS algorithm damage detectionwith noisy data are plotted in the diagrams of Figures 12 and13
The results of damage detection with two algorithmsin this large frame show that in case of only one memberdamaged there is more accuracy with PSOPC algorithm inidentifying of desired member and obtaining the amount ofdamage than the CSS algorithm It is observed that increasingthe number of injured elements in the structures reducesthe result precision with algorithms From results errors itcan be concluded that the PSOPC algorithm has problemin detecting the location of some of the damaged structuralelements so that in the third scenario the tenth and twenty-fourth members which are injured are diagnosed as healthy
10 Shock and Vibration
Table 7 The values of objective function of the Bowstring truss for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 32 times 10
minus0543 times 10
minus0549 times 10
minus0569 times 10
minus0571 times 10
minus0533 times 10
minus0526 times 10
minus0575 times 10
minus0534 times 10
minus0526 times 10
minus05 100PSOPC 26 times 10
minus0749 times 10
minus0218 times 10
minus0713 times 10
minus0112 times 10
minus0124 times 10
minus0850 times 10
minus0819 times 10
minus0826 times 10
minus0217 times 10
minus02 60Scenario 2
CSS 00002 00003 00003 00001 00039 00038 00002 00003 00015 00051 60PSOPC 00005 00134 00672 00629 00598 00050 00049 00065 00123 00251 10
Incomplete noisy dataComplete noisy dataActual damage
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 390
5
10
15
20
25
30
35
40
Element number
Dam
age (
)
Figure 12 Damage distribution of plane frame using the completeand incomplete noisy data in Scenario 1
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 400
10
20
30
40
50
60
Element number
Dam
age (
)
Predicted damageActual damage
Figure 13 Damage distribution of plane frame using the completenoisy data in Scenario 2
Table 8 Simulated damage scenarios in unbraced plane frame
Element number Scenario 1 Scenario 21 4045 309 1010 1013 2014 4019 2524 2032 50
and a considerable amount of damage in the fifteenth andtwenty-third elements that have no damage has been gainedHowever the CSS algorithm has well identified places ofall damaged elements in the structure and has achieved theamount of their damage with high accuracy and low errorFrom Figures 12 and 13 it can be observed that even in thatcase a lot of damage to the frame is considered the effectof noise on the results obtained by the CSS algorithm islow so that in case of incomplete data accuracy of resultshas reduced slightly and the performance of this algorithmdespite the noise was considered acceptable
In all three examples the convergences diagrams areshown that the process of convergence is gradual So it causesa more comprehensive search algorithm for finding optimalsolutions
The result increases the success rate of this type ofalgorithm performance in comparison with the PSOPCalgorithm
6 Conclusions
An approach for detecting damage based on continuumdamage model using charged system search algorithm ispresented The algorithm evaluates the location and severityof damage in three structures a cantilever beam a Bowstringplane truss and a three-story three-bay unbraced plane frameby minimizing an objective function by measuring completeand incomplete noisy modal data with different damagescenarios
Shock and Vibration 11
Table 9 Results of damage detection of the 39-element frame using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0399 0400 0393 0005 0 00002 0001 0 0000 0001 0 00003 0001 0 0000 0000 0 00004 0001 0 0027 0037 0 00005 0003 0 0000 0254 0300 03566 0000 0 0000 0016 0 00007 0000 0 0000 0013 0 00008 0005 0 0000 0001 0 00009 0056 0100 0000 0001 0 000010 0002 0 0000 0095 0100 000011 0005 0 0000 0002 0 000012 0008 0 0000 0001 0 000013 0193 0200 0195 0058 0 000014 0001 0 0000 0374 0400 034915 0004 0 0000 0005 0 011516 0000 0 0000 0001 0 000017 0000 0 0000 0001 0 000018 0000 0 0000 0004 0 000019 0000 0 0000 0212 0250 025620 0000 0 0000 0004 0 000021 0000 0 0000 0029 0 000022 0000 0 0002 0015 0 000023 0000 0 0000 0001 0 009224 0000 0 0000 0174 0200 000025 0000 0 0000 0001 0 000026 0000 0 0000 0002 0 000027 0002 0 0000 0005 0 000028 0000 0 0000 0001 0 004029 0000 0 0001 0007 0 000030 0000 0 0000 0000 0 000031 0001 0 0003 0002 0 000032 0000 0 0000 0497 0500 049533 0000 0 0000 0000 0 000034 0006 0 0000 0000 0 000035 0005 0 0025 0001 0 000036 0006 0 0000 0001 0 000037 0002 0 0000 0001 0 001538 0007 0 0000 0001 0 000039 0003 0 0000 0002 0 0000Error 01162 01700 03600 06808
In order to demonstrate the power of this algorithm in thediagnosis of damage a comparison has been made betweenthe results of this algorithm and the PSOPC algorithm
By comparing the damage detection results of the twovarious methods some interesting points have been con-cluded In scenarios where the number of damaged elementsis considered low both algorithms have acceptable accuracyin the results considering the more damaged members
of structures the PSOPC algorithm has been mistaken toidentify the damaged elements of the structureThis problemin larger-scale structures and the increasing number ofmembers has beenmore visible CSS algorithm has identifiedthe exact location of damage as well as achieving the amountof damage with acceptable accuracy It can be concluded thatthe CSS algorithm has great potential in global and localsearch of damage
12 Shock and Vibration
Table 10 The values of objective function of the 39-element frame for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 44 times 10
minus0346 times 10
minus0387 times 10
minus0462 times 10
minus0411 times 10
minus0366 times 10
minus0437 times 10
minus0480 times 10
minus0429 times 10
minus0398 times 10
minus04 50PSOPC 12 times 10
minus0213 times 10
minus0244 times 10
minus0315 times 10
minus0243 times 10
minus0213 times 10
minus0339 times 10
minus0214 times 10
minus0239 times 10
minus0315 times 10
minus02 10Scenario 2
CSS 19 times 10minus03
13 times 10minus03
17 times 10minus03
12 times 10minus03
23 times 10minus03
28 times 10minus03
19 times 10minus03
11 times 10minus03
21 times 10minus03
16 times 10minus03 30
PSOPC 29 times 10minus02
39 times 10minus02
11 times 10minus02
18 times 10minus02
11 times 10minus02
59 times 10minus03
69 times 10minus02
36 times 10minus02
53 times 10minus02
13 times 10minus02 0
Due to noise the real value of natural frequencies ofstructure will change it can be seen from the diagramsthat in these cases accuracy reduction of the results of theCSS algorithm is very low This indicates the high powerof this algorithm in damage detection considering noise instructure
Charged system search by considering small populationand low iteration cycle can detect the severity and location ofdamage with acceptable accuracy which shows high powerand speed of convergence of this algorithm The proposedmethod by the charged system search algorithm producedbetter results compared with particle swarm optimizationmethod
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] D J Ewins Modal Testing Theory and Practice John Wiley ampSons New York NY USA 1984
[2] S W Doebling C R Farrar and M B Prime ldquoA summaryreview of vibration-based damage identification methodsrdquoShock and Vibration Digest vol 30 no 2 pp 91ndash105 1998
[3] N Hu XWang H Fukunaga Z H Yao H X Zhang and Z SWu ldquoDamage assessment of structures using modal test datardquoInternational Journal of Solids and Structures vol 38 no 18 pp3111ndash3126 2001
[4] O S Salawu ldquoDetection of structural damage through changesin frequency a reviewrdquo Engineering Structures vol 19 no 9 pp718ndash723 1997
[5] N Bicanic and H-P Chen ldquoDamage identification in framedstructures using natural frequenciesrdquo International Journal forNumerical Methods in Engineering vol 40 no 23 pp 4451ndash4468 1997
[6] A K Pandey M Biswas and M M Samman ldquoDamagedetection from changes in curvature mode shapesrdquo Journal ofSound and Vibration vol 145 no 2 pp 321ndash332 1991
[7] J E Mottershead and M I Friswell ldquoModel updating instructural dynamics a surveyrdquo Journal of Sound and Vibrationvol 167 no 2 pp 347ndash375 1993
[8] R Perera andR Torres ldquoStructural damage detection viamodaldata with genetic algorithmsrdquo Journal of Structural Engineeringvol 132 no 9 pp 1491ndash1501 2006
[9] S Fallahian and S M Seyedpoor ldquoA two stage method forstructural damage identification using an adaptive neuro-fuzzyinference system and particle Swarm optimizationrdquo AsianJournal of Civil Engineering (Building and Housing) vol 11 pp797ndash810 2010
[10] B H Koh and S J Dyke ldquoStructural health monitoring forflexible bridge structures using correlation and sensitivity ofmodal datardquo Computers and Structures vol 85 no 3-4 pp 117ndash130 2007
[11] R-S He and S-F Hwang ldquoDamage detection by a hybrid real-parameter genetic algorithm under the assistance of grey rela-tion analysisrdquo Engineering Applications of Artificial Intelligencevol 20 no 7 pp 980ndash992 2007
[12] H Y Guo and Z L Li ldquoA two-stage method to identifystructural damage sites and extents by using evidence theoryand micro-search genetic algorithmrdquo Mechanical Systems andSignal Processing vol 23 no 3 pp 769ndash782 2009
[13] L Yu and X Chen ldquoBridge damage identification by combiningmodal flexibility and PSO methodsrdquo in Proceedings of thePrognostics and System Health Management Conference (PHMrsquo10) Macau China January 2010
[14] Z Tabrizian E Afshari G Ghodrati Amiri M H A Beigyand SM PourhoseiniNejad ldquoAnewdamage detectionmethodBig Bang-Big Crunch (BB-BC) algorithmrdquo Journal of Shock andVibration vol 20 no 4 pp 633ndash648 2013
[15] A Kaveh and S Talatahari ldquoA novel heuristic optimizationmethod charged system searchrdquo Acta Mechanica vol 213 no3-4 pp 267ndash289 2010
[16] S M Pourhoseini Nejad G Ghodrati Amiri A Asadi EAfshari and Z Tabrizian ldquoDamage detection of skeletal struc-tures using particle swarm optimizer with passive congregation(PSOPC) algorithm via incomplete modal datardquo Journal ofComputational Methods in Civil Engineering vol 3 no 1 pp1ndash13 2012
[17] S He Q H Wu J Y Wen J R Saunders and R CPaton ldquoA particle swarm optimizer with passive congregationrdquoBioSystems vol 78 no 1ndash3 pp 135ndash147 2004
[18] X Wang N Hu H Fukunaga and Z H Yao ldquoStructuraldamage identification using static test data and changes infrequenciesrdquo Engineering Structures vol 23 no 6 pp 610ndash6212001
[19] A Messina E J Williams and T Contursi ldquoStructural damagedetection by a sensitivity and statistical-based methodrdquo Journalof Sound and Vibration vol 216 no 5 pp 791ndash808 1998
[20] S W Doebling C R Farrar M B Prime and D W ShevitzldquoDamage identification and healthmonitoring of structural and
Shock and Vibration 13
mechanical systems from changes in their vibration character-istics a literature reviewrdquo Research Report LA-13070-MS ESA-EA Los Alamos National Laboratory Los Alamos NM USA1996
[21] M Nobahari and S M Seyedpoor ldquoStructural damage detec-tion using an efficient correlation-based index and a modifiedgenetic algorithmrdquoMathematical and Computer Modelling vol53 no 9-10 pp 1798ndash1809 2011
[22] A Kaveh and S Talatahari ldquoOptimal design of skeletal struc-tures via the charged system search algorithmrdquo Structural andMultidisciplinary Optimization vol 41 no 6 pp 893ndash911 2010
[23] A Kaveh and S Talatahari ldquoParticle swarm optimizer antcolony strategy and harmony search scheme hybridized foroptimization of truss structuresrdquoComputers and Structures vol87 no 5-6 pp 267ndash283 2009
[24] H J Salane and J W Baldwin Jr ldquoIdentification of modalproperties of bridgesrdquo Journal of Structural Engineering vol 116no 7 pp 2008ndash2021 1990
[25] O S Salawu and CWilliams ldquoBridge assessment using forced-vibration testingrdquo Journal of Structural Engineering vol 121 no2 pp 161ndash173 1995
[26] J-M Ndambi J Vantomme and K Harri ldquoDamage assessmentin reinforced concrete beams using eigenfrequencies and modeshape derivativesrdquo Engineering Structures vol 24 no 4 pp 501ndash515 2002
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
6 Shock and Vibration
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
005
01
015
02
025
03
035
04
Number of analyses
Obj
ectiv
e fun
ctio
n
(a) Scenario 1
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
005
01
015
02
025
03
035
Number of analyses
Obj
ectiv
e fun
ctio
n
(b) Scenario 2
Figure 7 The convergence history of the Bowstring truss for the CSS and PSOPC algorithms
Incomplete noisy dataComplete noisy dataActual damage
0 2 4 6 8 10 12 14 16 18 20 22 240
5
10
15
20
25
30
Element number
Dam
age (
)
Figure 8 Damage distribution of Bowstring truss using the com-plete and incomplete noisy data in Scenario 1
with passive congregation algorithm to demonstrate theperformance of this work
In CSS algorithm the effect of the pervious velocity andthe resultant force affecting a CP can decrease or increasebased on the values of the 119896V and 119896119886 defined as [22]
119896V = 119888(1 minusiter
itermax)
119896119886 = 119888(1 +iter
itermax)
(14)
where iter is the iteration number itermaxis the maximumnumber of the iterations and 119888 is set to 1
For the CSS algorithm a population of 16 CPs and forPSOPC algorithm a population of 50 particles are used for all
Incomplete noisy dataComplete noisy dataActual damage
Element number0 2 4 6 8 10 12 14 16 18 20 22 24
0
5
10
15
20
25
30
35
40
45D
amag
e (
)
Figure 9 Damage distribution of Bowstring truss using the com-plete and incomplete noisy data in Scenario 2
Table 1 Simulated damage scenarios in cantilever beam
Element number Scenario 1 Scenario 21 1035 206 107 359
the examples the stop criterion is considered as maximumnumber of 3000 and 5000 analyses in first and two otherexamples respectively
Shock and Vibration 7
1
2
13
23 24
14 16 181715
22 25 26 27
19 2120
28 3029
32 3331 34 35 36 37 3938
3 6 9 12
5 8 11
4 7 104m
35m
8m 7m 9m
35m
W12 times 65
W14 times 132
Figure 10 Geometry of unbraced plane frame
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
01
02
03
04
05
Number of analyses
Obj
ectiv
e fun
ctio
n
(a) Scenario 1
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000Number of analyses
0
005
01
015
02
025
03
035
04
Obj
ectiv
e fun
ctio
n
(b) Scenario 2
Figure 11 The convergence history of the 39-element frame for the CSS and PSOPC algorithms
The algorithm has been implemented in the commercialMATLAB software and because of the stochastic nature of thealgorithms each example is independently solved ten times
Damage values have been limited to a region between 0and 1
Example 1 A cantilever steel beam is shown in Figure 2 Forthe reason of modal analyzing the beam was divided into 10two-dimensional beam elements with 20 degrees of freedom
The section of the beam is W12 times 65 with mechanicalproperties of
119860 = 00123 m2 (191 in2)
119868 = 2218 times 10minus4 m4 (533 in4)
119864 = 207 times 109Nm2
120588 = 7860 kgm3
The measured modal responses of the beam before and afterdamage were created using the proposed forward solver(for each scenario) Instead of experimental measurementsnumerically generated measurements were used to estimatethe proposed inverse procedure In this example the first tennatural frequencies are used for damage detection Two dif-ferent simulated damage scenarios are considered (Table 1)
These damage scenarios are considered to represent theeffect of severity of damage (stiffness reduction) numberof damaged elements and contribution of damage elementson the results The results of damage detecting of CSS andPSOPC algorithms are given in Table 2 In order to evaluatethe performance of the methods used the summation ofestimated error of damage detection was calculated (Table 2)that is presented as follows
Error = sum(119889119894119860 minus 119889119894119878) (15)
8 Shock and Vibration
Table 2 Results of damage detection of the 10-element beam using CSS and PSOPC algorithms
Elementnumber
Scenario 1 Scenario 2
CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0000 0 0000 0101 01 0105
2 0000 0 0000 0003 0 0000
3 0000 0 0000 0001 0 0000
4 0000 0 0000 0000 0 0000
5 0000 0 0000 0185 02 0169
6 0100 01 0100 0016 0 0037
7 0000 0 0000 0349 035 0348
8 0000 0 0000 0002 0 0000
9 0000 0 0000 0001 0 0000
10 0000 0 0000 0000 0 0000
Error 00002 00000 00385 00752
Table 3 The values of objective function of the 10-element beam for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 34 times 10
minus0652 times 10
minus0628 times 10
minus0627 times 10
minus0683 times 10
minus0665 times 10
minus0641 times 10
minus0638 times 10
minus0619 times 10
minus0644 times 10
minus06 100PSOPC 25 times 10
minus0713 times 10
minus0799 times 10
minus0814 times 10
minus0721 times 10
minus0746 times 10
minus0726 times 10
minus0716 times 10
minus0688 times 10
minus0811 times 10
minus07 100Scenario 2
CSS 00008 00001 00009 00004 00006 00002 00005 00008 00009 00006 80PSOPC 00003 00013 00056 00004 00074 00074 00080 00011 00006 00085 40
Table 4 Cross-sectional areas of truss elements
Member Area (cm2)1ndash6 187ndash12 1513ndash17 1018ndash25 12
Table 5 Simulated damage scenarios in Bowstring truss
Element number Scenario 1 Scenario 235 108 251011 4024 20
where 119889119894119860 and 119889119894119878 are the actual and estimated damage of the
119894th element using the presented method respectivelyIn Table 3 the values of the objective function for CSS
and PSOPC algorithms in independent 10 runs are given theconvergence history for the 10-element beam in scenarios 1and 2 is shown in Figure 3
It is seen in Table 2 that in the first scenario where thenumber of damaged elements is low the damage identifi-cation is well by both algorithms in the second scenariothe algorithm PSOPC has some wrong in damage detectionso that there is difference between the obtained damage viathis algorithm and the actual intensity of damage in thefifth element and the undamaged sixth element mistakenlyhas been detected to be damaged Meanwhile the locationand amount of damage in structure are obtained by the CSSalgorithm with acceptable accuracy In order to investigatethe noise effects on the results of the CSSmethod 015 noisein measurement is considered for the natural frequencies
Diagrams of the CSS algorithmrsquos damage detection withcomplete and incomplete dynamic noisy data are plotted inFigures 4 and 5
As the results show this method is able to detect thelocation andmagnitude of damaged elements in all scenariosusing complete and incomplete noisy data
Example 2 A Bowstring plane truss with 25 elements isshown in Figure 6 The properties of members of this struc-ture are 119864 = 207 times 0
9Nm2 and 120588 = 7860 kgm3 and cross-sectional areas of elements are given in Table 4
Three scenarios for this truss are considered according toTable 5
Shock and Vibration 9
Table 6 Results of damage detection of the 25-bar Bowstring truss using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0009 0 0000 0000 0 00002 0000 0 0000 0000 0 00003 0000 0 0000 0000 0 00004 0000 0 0000 0000 0 00005 0000 0 0000 0098 01 00956 0000 0 0000 0000 0 00007 0000 0 0000 0000 0 00008 0246 025 0250 0000 0 00009 0001 0 0000 0000 0 000010 0000 0 0000 0000 0 000011 0000 0 0000 0402 04 042612 0000 0 0000 0000 0 000013 0000 0 0000 0000 0 000014 0001 0 0000 0000 0 000015 0000 0 0000 0000 0 000016 0000 0 0000 0000 0 000017 0000 0 0000 0000 0 000018 0000 0 0000 0005 0 000019 0000 0 0000 0004 0 007520 0000 0 0000 0001 0 000021 0000 0 0000 0000 0 000022 0000 0 0000 0000 0 000023 0000 0 0000 0000 0 000024 0000 0 0000 0185 02 000025 0000 0 0000 0014 0 0265Error 00085 00000 00119 01609
In Table 6 the results of two algorithms in all scenarios arecompared Only the first ten natural frequencies are used fordamage detection
The comparison of the objective function for CSS andPSOPC algorithms is made in Table 7 and the convergencehistory for the Bowstring truss in scenarios 1 and 2 is shownin Figure 7
In Figures 8 and 9 the results of CSS algorithm in damagedetection of this truss that is affected by noise are shown
Table 7 and Figures 7 8 and 9 show better results indamage detection by CSS algorithm rather than PSOPCalgorithm and also by applying noise in structures
Example 3 A three-story three-bay frame as shown inFigure 10 is used to verify the damage detection methodexplained in this paper The number of elements and nodesis 39 and 34 respectively
For unbraced plane frame problem all columns areW14 times 132 (119860 = 0025m2 (388 in2) and 119868 = 6386 times
10minus4m4 (1530 in4)) and all beams are W12 times 65 Youngrsquos
modulus is 119864 = 207GPa (30 000 ksi) Poissonrsquos ratio is ] =
03 and the mass density is 120588 = 7780 kgm3 (0000728 lb minuss2in4)
In this example two scenarios from the aspect of elementnumber and its level of damage were assumed that are givenin Table 8 Between ninety natural frequencies only the firsttwenty natural frequencies are utilized for damage detectionin the first scenario
The severity of damage detection by PSOPC and CSSalgorithms in different members of the frame and in eachscenario is shown separately in Table 9 In Table 10 theobtained objective function for both algorithms in each runis given also the convergence history is shown in Figure 11
Also the results of the CSS algorithm damage detectionwith noisy data are plotted in the diagrams of Figures 12 and13
The results of damage detection with two algorithmsin this large frame show that in case of only one memberdamaged there is more accuracy with PSOPC algorithm inidentifying of desired member and obtaining the amount ofdamage than the CSS algorithm It is observed that increasingthe number of injured elements in the structures reducesthe result precision with algorithms From results errors itcan be concluded that the PSOPC algorithm has problemin detecting the location of some of the damaged structuralelements so that in the third scenario the tenth and twenty-fourth members which are injured are diagnosed as healthy
10 Shock and Vibration
Table 7 The values of objective function of the Bowstring truss for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 32 times 10
minus0543 times 10
minus0549 times 10
minus0569 times 10
minus0571 times 10
minus0533 times 10
minus0526 times 10
minus0575 times 10
minus0534 times 10
minus0526 times 10
minus05 100PSOPC 26 times 10
minus0749 times 10
minus0218 times 10
minus0713 times 10
minus0112 times 10
minus0124 times 10
minus0850 times 10
minus0819 times 10
minus0826 times 10
minus0217 times 10
minus02 60Scenario 2
CSS 00002 00003 00003 00001 00039 00038 00002 00003 00015 00051 60PSOPC 00005 00134 00672 00629 00598 00050 00049 00065 00123 00251 10
Incomplete noisy dataComplete noisy dataActual damage
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 390
5
10
15
20
25
30
35
40
Element number
Dam
age (
)
Figure 12 Damage distribution of plane frame using the completeand incomplete noisy data in Scenario 1
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 400
10
20
30
40
50
60
Element number
Dam
age (
)
Predicted damageActual damage
Figure 13 Damage distribution of plane frame using the completenoisy data in Scenario 2
Table 8 Simulated damage scenarios in unbraced plane frame
Element number Scenario 1 Scenario 21 4045 309 1010 1013 2014 4019 2524 2032 50
and a considerable amount of damage in the fifteenth andtwenty-third elements that have no damage has been gainedHowever the CSS algorithm has well identified places ofall damaged elements in the structure and has achieved theamount of their damage with high accuracy and low errorFrom Figures 12 and 13 it can be observed that even in thatcase a lot of damage to the frame is considered the effectof noise on the results obtained by the CSS algorithm islow so that in case of incomplete data accuracy of resultshas reduced slightly and the performance of this algorithmdespite the noise was considered acceptable
In all three examples the convergences diagrams areshown that the process of convergence is gradual So it causesa more comprehensive search algorithm for finding optimalsolutions
The result increases the success rate of this type ofalgorithm performance in comparison with the PSOPCalgorithm
6 Conclusions
An approach for detecting damage based on continuumdamage model using charged system search algorithm ispresented The algorithm evaluates the location and severityof damage in three structures a cantilever beam a Bowstringplane truss and a three-story three-bay unbraced plane frameby minimizing an objective function by measuring completeand incomplete noisy modal data with different damagescenarios
Shock and Vibration 11
Table 9 Results of damage detection of the 39-element frame using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0399 0400 0393 0005 0 00002 0001 0 0000 0001 0 00003 0001 0 0000 0000 0 00004 0001 0 0027 0037 0 00005 0003 0 0000 0254 0300 03566 0000 0 0000 0016 0 00007 0000 0 0000 0013 0 00008 0005 0 0000 0001 0 00009 0056 0100 0000 0001 0 000010 0002 0 0000 0095 0100 000011 0005 0 0000 0002 0 000012 0008 0 0000 0001 0 000013 0193 0200 0195 0058 0 000014 0001 0 0000 0374 0400 034915 0004 0 0000 0005 0 011516 0000 0 0000 0001 0 000017 0000 0 0000 0001 0 000018 0000 0 0000 0004 0 000019 0000 0 0000 0212 0250 025620 0000 0 0000 0004 0 000021 0000 0 0000 0029 0 000022 0000 0 0002 0015 0 000023 0000 0 0000 0001 0 009224 0000 0 0000 0174 0200 000025 0000 0 0000 0001 0 000026 0000 0 0000 0002 0 000027 0002 0 0000 0005 0 000028 0000 0 0000 0001 0 004029 0000 0 0001 0007 0 000030 0000 0 0000 0000 0 000031 0001 0 0003 0002 0 000032 0000 0 0000 0497 0500 049533 0000 0 0000 0000 0 000034 0006 0 0000 0000 0 000035 0005 0 0025 0001 0 000036 0006 0 0000 0001 0 000037 0002 0 0000 0001 0 001538 0007 0 0000 0001 0 000039 0003 0 0000 0002 0 0000Error 01162 01700 03600 06808
In order to demonstrate the power of this algorithm in thediagnosis of damage a comparison has been made betweenthe results of this algorithm and the PSOPC algorithm
By comparing the damage detection results of the twovarious methods some interesting points have been con-cluded In scenarios where the number of damaged elementsis considered low both algorithms have acceptable accuracyin the results considering the more damaged members
of structures the PSOPC algorithm has been mistaken toidentify the damaged elements of the structureThis problemin larger-scale structures and the increasing number ofmembers has beenmore visible CSS algorithm has identifiedthe exact location of damage as well as achieving the amountof damage with acceptable accuracy It can be concluded thatthe CSS algorithm has great potential in global and localsearch of damage
12 Shock and Vibration
Table 10 The values of objective function of the 39-element frame for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 44 times 10
minus0346 times 10
minus0387 times 10
minus0462 times 10
minus0411 times 10
minus0366 times 10
minus0437 times 10
minus0480 times 10
minus0429 times 10
minus0398 times 10
minus04 50PSOPC 12 times 10
minus0213 times 10
minus0244 times 10
minus0315 times 10
minus0243 times 10
minus0213 times 10
minus0339 times 10
minus0214 times 10
minus0239 times 10
minus0315 times 10
minus02 10Scenario 2
CSS 19 times 10minus03
13 times 10minus03
17 times 10minus03
12 times 10minus03
23 times 10minus03
28 times 10minus03
19 times 10minus03
11 times 10minus03
21 times 10minus03
16 times 10minus03 30
PSOPC 29 times 10minus02
39 times 10minus02
11 times 10minus02
18 times 10minus02
11 times 10minus02
59 times 10minus03
69 times 10minus02
36 times 10minus02
53 times 10minus02
13 times 10minus02 0
Due to noise the real value of natural frequencies ofstructure will change it can be seen from the diagramsthat in these cases accuracy reduction of the results of theCSS algorithm is very low This indicates the high powerof this algorithm in damage detection considering noise instructure
Charged system search by considering small populationand low iteration cycle can detect the severity and location ofdamage with acceptable accuracy which shows high powerand speed of convergence of this algorithm The proposedmethod by the charged system search algorithm producedbetter results compared with particle swarm optimizationmethod
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] D J Ewins Modal Testing Theory and Practice John Wiley ampSons New York NY USA 1984
[2] S W Doebling C R Farrar and M B Prime ldquoA summaryreview of vibration-based damage identification methodsrdquoShock and Vibration Digest vol 30 no 2 pp 91ndash105 1998
[3] N Hu XWang H Fukunaga Z H Yao H X Zhang and Z SWu ldquoDamage assessment of structures using modal test datardquoInternational Journal of Solids and Structures vol 38 no 18 pp3111ndash3126 2001
[4] O S Salawu ldquoDetection of structural damage through changesin frequency a reviewrdquo Engineering Structures vol 19 no 9 pp718ndash723 1997
[5] N Bicanic and H-P Chen ldquoDamage identification in framedstructures using natural frequenciesrdquo International Journal forNumerical Methods in Engineering vol 40 no 23 pp 4451ndash4468 1997
[6] A K Pandey M Biswas and M M Samman ldquoDamagedetection from changes in curvature mode shapesrdquo Journal ofSound and Vibration vol 145 no 2 pp 321ndash332 1991
[7] J E Mottershead and M I Friswell ldquoModel updating instructural dynamics a surveyrdquo Journal of Sound and Vibrationvol 167 no 2 pp 347ndash375 1993
[8] R Perera andR Torres ldquoStructural damage detection viamodaldata with genetic algorithmsrdquo Journal of Structural Engineeringvol 132 no 9 pp 1491ndash1501 2006
[9] S Fallahian and S M Seyedpoor ldquoA two stage method forstructural damage identification using an adaptive neuro-fuzzyinference system and particle Swarm optimizationrdquo AsianJournal of Civil Engineering (Building and Housing) vol 11 pp797ndash810 2010
[10] B H Koh and S J Dyke ldquoStructural health monitoring forflexible bridge structures using correlation and sensitivity ofmodal datardquo Computers and Structures vol 85 no 3-4 pp 117ndash130 2007
[11] R-S He and S-F Hwang ldquoDamage detection by a hybrid real-parameter genetic algorithm under the assistance of grey rela-tion analysisrdquo Engineering Applications of Artificial Intelligencevol 20 no 7 pp 980ndash992 2007
[12] H Y Guo and Z L Li ldquoA two-stage method to identifystructural damage sites and extents by using evidence theoryand micro-search genetic algorithmrdquo Mechanical Systems andSignal Processing vol 23 no 3 pp 769ndash782 2009
[13] L Yu and X Chen ldquoBridge damage identification by combiningmodal flexibility and PSO methodsrdquo in Proceedings of thePrognostics and System Health Management Conference (PHMrsquo10) Macau China January 2010
[14] Z Tabrizian E Afshari G Ghodrati Amiri M H A Beigyand SM PourhoseiniNejad ldquoAnewdamage detectionmethodBig Bang-Big Crunch (BB-BC) algorithmrdquo Journal of Shock andVibration vol 20 no 4 pp 633ndash648 2013
[15] A Kaveh and S Talatahari ldquoA novel heuristic optimizationmethod charged system searchrdquo Acta Mechanica vol 213 no3-4 pp 267ndash289 2010
[16] S M Pourhoseini Nejad G Ghodrati Amiri A Asadi EAfshari and Z Tabrizian ldquoDamage detection of skeletal struc-tures using particle swarm optimizer with passive congregation(PSOPC) algorithm via incomplete modal datardquo Journal ofComputational Methods in Civil Engineering vol 3 no 1 pp1ndash13 2012
[17] S He Q H Wu J Y Wen J R Saunders and R CPaton ldquoA particle swarm optimizer with passive congregationrdquoBioSystems vol 78 no 1ndash3 pp 135ndash147 2004
[18] X Wang N Hu H Fukunaga and Z H Yao ldquoStructuraldamage identification using static test data and changes infrequenciesrdquo Engineering Structures vol 23 no 6 pp 610ndash6212001
[19] A Messina E J Williams and T Contursi ldquoStructural damagedetection by a sensitivity and statistical-based methodrdquo Journalof Sound and Vibration vol 216 no 5 pp 791ndash808 1998
[20] S W Doebling C R Farrar M B Prime and D W ShevitzldquoDamage identification and healthmonitoring of structural and
Shock and Vibration 13
mechanical systems from changes in their vibration character-istics a literature reviewrdquo Research Report LA-13070-MS ESA-EA Los Alamos National Laboratory Los Alamos NM USA1996
[21] M Nobahari and S M Seyedpoor ldquoStructural damage detec-tion using an efficient correlation-based index and a modifiedgenetic algorithmrdquoMathematical and Computer Modelling vol53 no 9-10 pp 1798ndash1809 2011
[22] A Kaveh and S Talatahari ldquoOptimal design of skeletal struc-tures via the charged system search algorithmrdquo Structural andMultidisciplinary Optimization vol 41 no 6 pp 893ndash911 2010
[23] A Kaveh and S Talatahari ldquoParticle swarm optimizer antcolony strategy and harmony search scheme hybridized foroptimization of truss structuresrdquoComputers and Structures vol87 no 5-6 pp 267ndash283 2009
[24] H J Salane and J W Baldwin Jr ldquoIdentification of modalproperties of bridgesrdquo Journal of Structural Engineering vol 116no 7 pp 2008ndash2021 1990
[25] O S Salawu and CWilliams ldquoBridge assessment using forced-vibration testingrdquo Journal of Structural Engineering vol 121 no2 pp 161ndash173 1995
[26] J-M Ndambi J Vantomme and K Harri ldquoDamage assessmentin reinforced concrete beams using eigenfrequencies and modeshape derivativesrdquo Engineering Structures vol 24 no 4 pp 501ndash515 2002
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
Shock and Vibration 7
1
2
13
23 24
14 16 181715
22 25 26 27
19 2120
28 3029
32 3331 34 35 36 37 3938
3 6 9 12
5 8 11
4 7 104m
35m
8m 7m 9m
35m
W12 times 65
W14 times 132
Figure 10 Geometry of unbraced plane frame
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000
01
02
03
04
05
Number of analyses
Obj
ectiv
e fun
ctio
n
(a) Scenario 1
CSSPSOPC
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000Number of analyses
0
005
01
015
02
025
03
035
04
Obj
ectiv
e fun
ctio
n
(b) Scenario 2
Figure 11 The convergence history of the 39-element frame for the CSS and PSOPC algorithms
The algorithm has been implemented in the commercialMATLAB software and because of the stochastic nature of thealgorithms each example is independently solved ten times
Damage values have been limited to a region between 0and 1
Example 1 A cantilever steel beam is shown in Figure 2 Forthe reason of modal analyzing the beam was divided into 10two-dimensional beam elements with 20 degrees of freedom
The section of the beam is W12 times 65 with mechanicalproperties of
119860 = 00123 m2 (191 in2)
119868 = 2218 times 10minus4 m4 (533 in4)
119864 = 207 times 109Nm2
120588 = 7860 kgm3
The measured modal responses of the beam before and afterdamage were created using the proposed forward solver(for each scenario) Instead of experimental measurementsnumerically generated measurements were used to estimatethe proposed inverse procedure In this example the first tennatural frequencies are used for damage detection Two dif-ferent simulated damage scenarios are considered (Table 1)
These damage scenarios are considered to represent theeffect of severity of damage (stiffness reduction) numberof damaged elements and contribution of damage elementson the results The results of damage detecting of CSS andPSOPC algorithms are given in Table 2 In order to evaluatethe performance of the methods used the summation ofestimated error of damage detection was calculated (Table 2)that is presented as follows
Error = sum(119889119894119860 minus 119889119894119878) (15)
8 Shock and Vibration
Table 2 Results of damage detection of the 10-element beam using CSS and PSOPC algorithms
Elementnumber
Scenario 1 Scenario 2
CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0000 0 0000 0101 01 0105
2 0000 0 0000 0003 0 0000
3 0000 0 0000 0001 0 0000
4 0000 0 0000 0000 0 0000
5 0000 0 0000 0185 02 0169
6 0100 01 0100 0016 0 0037
7 0000 0 0000 0349 035 0348
8 0000 0 0000 0002 0 0000
9 0000 0 0000 0001 0 0000
10 0000 0 0000 0000 0 0000
Error 00002 00000 00385 00752
Table 3 The values of objective function of the 10-element beam for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 34 times 10
minus0652 times 10
minus0628 times 10
minus0627 times 10
minus0683 times 10
minus0665 times 10
minus0641 times 10
minus0638 times 10
minus0619 times 10
minus0644 times 10
minus06 100PSOPC 25 times 10
minus0713 times 10
minus0799 times 10
minus0814 times 10
minus0721 times 10
minus0746 times 10
minus0726 times 10
minus0716 times 10
minus0688 times 10
minus0811 times 10
minus07 100Scenario 2
CSS 00008 00001 00009 00004 00006 00002 00005 00008 00009 00006 80PSOPC 00003 00013 00056 00004 00074 00074 00080 00011 00006 00085 40
Table 4 Cross-sectional areas of truss elements
Member Area (cm2)1ndash6 187ndash12 1513ndash17 1018ndash25 12
Table 5 Simulated damage scenarios in Bowstring truss
Element number Scenario 1 Scenario 235 108 251011 4024 20
where 119889119894119860 and 119889119894119878 are the actual and estimated damage of the
119894th element using the presented method respectivelyIn Table 3 the values of the objective function for CSS
and PSOPC algorithms in independent 10 runs are given theconvergence history for the 10-element beam in scenarios 1and 2 is shown in Figure 3
It is seen in Table 2 that in the first scenario where thenumber of damaged elements is low the damage identifi-cation is well by both algorithms in the second scenariothe algorithm PSOPC has some wrong in damage detectionso that there is difference between the obtained damage viathis algorithm and the actual intensity of damage in thefifth element and the undamaged sixth element mistakenlyhas been detected to be damaged Meanwhile the locationand amount of damage in structure are obtained by the CSSalgorithm with acceptable accuracy In order to investigatethe noise effects on the results of the CSSmethod 015 noisein measurement is considered for the natural frequencies
Diagrams of the CSS algorithmrsquos damage detection withcomplete and incomplete dynamic noisy data are plotted inFigures 4 and 5
As the results show this method is able to detect thelocation andmagnitude of damaged elements in all scenariosusing complete and incomplete noisy data
Example 2 A Bowstring plane truss with 25 elements isshown in Figure 6 The properties of members of this struc-ture are 119864 = 207 times 0
9Nm2 and 120588 = 7860 kgm3 and cross-sectional areas of elements are given in Table 4
Three scenarios for this truss are considered according toTable 5
Shock and Vibration 9
Table 6 Results of damage detection of the 25-bar Bowstring truss using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0009 0 0000 0000 0 00002 0000 0 0000 0000 0 00003 0000 0 0000 0000 0 00004 0000 0 0000 0000 0 00005 0000 0 0000 0098 01 00956 0000 0 0000 0000 0 00007 0000 0 0000 0000 0 00008 0246 025 0250 0000 0 00009 0001 0 0000 0000 0 000010 0000 0 0000 0000 0 000011 0000 0 0000 0402 04 042612 0000 0 0000 0000 0 000013 0000 0 0000 0000 0 000014 0001 0 0000 0000 0 000015 0000 0 0000 0000 0 000016 0000 0 0000 0000 0 000017 0000 0 0000 0000 0 000018 0000 0 0000 0005 0 000019 0000 0 0000 0004 0 007520 0000 0 0000 0001 0 000021 0000 0 0000 0000 0 000022 0000 0 0000 0000 0 000023 0000 0 0000 0000 0 000024 0000 0 0000 0185 02 000025 0000 0 0000 0014 0 0265Error 00085 00000 00119 01609
In Table 6 the results of two algorithms in all scenarios arecompared Only the first ten natural frequencies are used fordamage detection
The comparison of the objective function for CSS andPSOPC algorithms is made in Table 7 and the convergencehistory for the Bowstring truss in scenarios 1 and 2 is shownin Figure 7
In Figures 8 and 9 the results of CSS algorithm in damagedetection of this truss that is affected by noise are shown
Table 7 and Figures 7 8 and 9 show better results indamage detection by CSS algorithm rather than PSOPCalgorithm and also by applying noise in structures
Example 3 A three-story three-bay frame as shown inFigure 10 is used to verify the damage detection methodexplained in this paper The number of elements and nodesis 39 and 34 respectively
For unbraced plane frame problem all columns areW14 times 132 (119860 = 0025m2 (388 in2) and 119868 = 6386 times
10minus4m4 (1530 in4)) and all beams are W12 times 65 Youngrsquos
modulus is 119864 = 207GPa (30 000 ksi) Poissonrsquos ratio is ] =
03 and the mass density is 120588 = 7780 kgm3 (0000728 lb minuss2in4)
In this example two scenarios from the aspect of elementnumber and its level of damage were assumed that are givenin Table 8 Between ninety natural frequencies only the firsttwenty natural frequencies are utilized for damage detectionin the first scenario
The severity of damage detection by PSOPC and CSSalgorithms in different members of the frame and in eachscenario is shown separately in Table 9 In Table 10 theobtained objective function for both algorithms in each runis given also the convergence history is shown in Figure 11
Also the results of the CSS algorithm damage detectionwith noisy data are plotted in the diagrams of Figures 12 and13
The results of damage detection with two algorithmsin this large frame show that in case of only one memberdamaged there is more accuracy with PSOPC algorithm inidentifying of desired member and obtaining the amount ofdamage than the CSS algorithm It is observed that increasingthe number of injured elements in the structures reducesthe result precision with algorithms From results errors itcan be concluded that the PSOPC algorithm has problemin detecting the location of some of the damaged structuralelements so that in the third scenario the tenth and twenty-fourth members which are injured are diagnosed as healthy
10 Shock and Vibration
Table 7 The values of objective function of the Bowstring truss for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 32 times 10
minus0543 times 10
minus0549 times 10
minus0569 times 10
minus0571 times 10
minus0533 times 10
minus0526 times 10
minus0575 times 10
minus0534 times 10
minus0526 times 10
minus05 100PSOPC 26 times 10
minus0749 times 10
minus0218 times 10
minus0713 times 10
minus0112 times 10
minus0124 times 10
minus0850 times 10
minus0819 times 10
minus0826 times 10
minus0217 times 10
minus02 60Scenario 2
CSS 00002 00003 00003 00001 00039 00038 00002 00003 00015 00051 60PSOPC 00005 00134 00672 00629 00598 00050 00049 00065 00123 00251 10
Incomplete noisy dataComplete noisy dataActual damage
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 390
5
10
15
20
25
30
35
40
Element number
Dam
age (
)
Figure 12 Damage distribution of plane frame using the completeand incomplete noisy data in Scenario 1
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 400
10
20
30
40
50
60
Element number
Dam
age (
)
Predicted damageActual damage
Figure 13 Damage distribution of plane frame using the completenoisy data in Scenario 2
Table 8 Simulated damage scenarios in unbraced plane frame
Element number Scenario 1 Scenario 21 4045 309 1010 1013 2014 4019 2524 2032 50
and a considerable amount of damage in the fifteenth andtwenty-third elements that have no damage has been gainedHowever the CSS algorithm has well identified places ofall damaged elements in the structure and has achieved theamount of their damage with high accuracy and low errorFrom Figures 12 and 13 it can be observed that even in thatcase a lot of damage to the frame is considered the effectof noise on the results obtained by the CSS algorithm islow so that in case of incomplete data accuracy of resultshas reduced slightly and the performance of this algorithmdespite the noise was considered acceptable
In all three examples the convergences diagrams areshown that the process of convergence is gradual So it causesa more comprehensive search algorithm for finding optimalsolutions
The result increases the success rate of this type ofalgorithm performance in comparison with the PSOPCalgorithm
6 Conclusions
An approach for detecting damage based on continuumdamage model using charged system search algorithm ispresented The algorithm evaluates the location and severityof damage in three structures a cantilever beam a Bowstringplane truss and a three-story three-bay unbraced plane frameby minimizing an objective function by measuring completeand incomplete noisy modal data with different damagescenarios
Shock and Vibration 11
Table 9 Results of damage detection of the 39-element frame using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0399 0400 0393 0005 0 00002 0001 0 0000 0001 0 00003 0001 0 0000 0000 0 00004 0001 0 0027 0037 0 00005 0003 0 0000 0254 0300 03566 0000 0 0000 0016 0 00007 0000 0 0000 0013 0 00008 0005 0 0000 0001 0 00009 0056 0100 0000 0001 0 000010 0002 0 0000 0095 0100 000011 0005 0 0000 0002 0 000012 0008 0 0000 0001 0 000013 0193 0200 0195 0058 0 000014 0001 0 0000 0374 0400 034915 0004 0 0000 0005 0 011516 0000 0 0000 0001 0 000017 0000 0 0000 0001 0 000018 0000 0 0000 0004 0 000019 0000 0 0000 0212 0250 025620 0000 0 0000 0004 0 000021 0000 0 0000 0029 0 000022 0000 0 0002 0015 0 000023 0000 0 0000 0001 0 009224 0000 0 0000 0174 0200 000025 0000 0 0000 0001 0 000026 0000 0 0000 0002 0 000027 0002 0 0000 0005 0 000028 0000 0 0000 0001 0 004029 0000 0 0001 0007 0 000030 0000 0 0000 0000 0 000031 0001 0 0003 0002 0 000032 0000 0 0000 0497 0500 049533 0000 0 0000 0000 0 000034 0006 0 0000 0000 0 000035 0005 0 0025 0001 0 000036 0006 0 0000 0001 0 000037 0002 0 0000 0001 0 001538 0007 0 0000 0001 0 000039 0003 0 0000 0002 0 0000Error 01162 01700 03600 06808
In order to demonstrate the power of this algorithm in thediagnosis of damage a comparison has been made betweenthe results of this algorithm and the PSOPC algorithm
By comparing the damage detection results of the twovarious methods some interesting points have been con-cluded In scenarios where the number of damaged elementsis considered low both algorithms have acceptable accuracyin the results considering the more damaged members
of structures the PSOPC algorithm has been mistaken toidentify the damaged elements of the structureThis problemin larger-scale structures and the increasing number ofmembers has beenmore visible CSS algorithm has identifiedthe exact location of damage as well as achieving the amountof damage with acceptable accuracy It can be concluded thatthe CSS algorithm has great potential in global and localsearch of damage
12 Shock and Vibration
Table 10 The values of objective function of the 39-element frame for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 44 times 10
minus0346 times 10
minus0387 times 10
minus0462 times 10
minus0411 times 10
minus0366 times 10
minus0437 times 10
minus0480 times 10
minus0429 times 10
minus0398 times 10
minus04 50PSOPC 12 times 10
minus0213 times 10
minus0244 times 10
minus0315 times 10
minus0243 times 10
minus0213 times 10
minus0339 times 10
minus0214 times 10
minus0239 times 10
minus0315 times 10
minus02 10Scenario 2
CSS 19 times 10minus03
13 times 10minus03
17 times 10minus03
12 times 10minus03
23 times 10minus03
28 times 10minus03
19 times 10minus03
11 times 10minus03
21 times 10minus03
16 times 10minus03 30
PSOPC 29 times 10minus02
39 times 10minus02
11 times 10minus02
18 times 10minus02
11 times 10minus02
59 times 10minus03
69 times 10minus02
36 times 10minus02
53 times 10minus02
13 times 10minus02 0
Due to noise the real value of natural frequencies ofstructure will change it can be seen from the diagramsthat in these cases accuracy reduction of the results of theCSS algorithm is very low This indicates the high powerof this algorithm in damage detection considering noise instructure
Charged system search by considering small populationand low iteration cycle can detect the severity and location ofdamage with acceptable accuracy which shows high powerand speed of convergence of this algorithm The proposedmethod by the charged system search algorithm producedbetter results compared with particle swarm optimizationmethod
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] D J Ewins Modal Testing Theory and Practice John Wiley ampSons New York NY USA 1984
[2] S W Doebling C R Farrar and M B Prime ldquoA summaryreview of vibration-based damage identification methodsrdquoShock and Vibration Digest vol 30 no 2 pp 91ndash105 1998
[3] N Hu XWang H Fukunaga Z H Yao H X Zhang and Z SWu ldquoDamage assessment of structures using modal test datardquoInternational Journal of Solids and Structures vol 38 no 18 pp3111ndash3126 2001
[4] O S Salawu ldquoDetection of structural damage through changesin frequency a reviewrdquo Engineering Structures vol 19 no 9 pp718ndash723 1997
[5] N Bicanic and H-P Chen ldquoDamage identification in framedstructures using natural frequenciesrdquo International Journal forNumerical Methods in Engineering vol 40 no 23 pp 4451ndash4468 1997
[6] A K Pandey M Biswas and M M Samman ldquoDamagedetection from changes in curvature mode shapesrdquo Journal ofSound and Vibration vol 145 no 2 pp 321ndash332 1991
[7] J E Mottershead and M I Friswell ldquoModel updating instructural dynamics a surveyrdquo Journal of Sound and Vibrationvol 167 no 2 pp 347ndash375 1993
[8] R Perera andR Torres ldquoStructural damage detection viamodaldata with genetic algorithmsrdquo Journal of Structural Engineeringvol 132 no 9 pp 1491ndash1501 2006
[9] S Fallahian and S M Seyedpoor ldquoA two stage method forstructural damage identification using an adaptive neuro-fuzzyinference system and particle Swarm optimizationrdquo AsianJournal of Civil Engineering (Building and Housing) vol 11 pp797ndash810 2010
[10] B H Koh and S J Dyke ldquoStructural health monitoring forflexible bridge structures using correlation and sensitivity ofmodal datardquo Computers and Structures vol 85 no 3-4 pp 117ndash130 2007
[11] R-S He and S-F Hwang ldquoDamage detection by a hybrid real-parameter genetic algorithm under the assistance of grey rela-tion analysisrdquo Engineering Applications of Artificial Intelligencevol 20 no 7 pp 980ndash992 2007
[12] H Y Guo and Z L Li ldquoA two-stage method to identifystructural damage sites and extents by using evidence theoryand micro-search genetic algorithmrdquo Mechanical Systems andSignal Processing vol 23 no 3 pp 769ndash782 2009
[13] L Yu and X Chen ldquoBridge damage identification by combiningmodal flexibility and PSO methodsrdquo in Proceedings of thePrognostics and System Health Management Conference (PHMrsquo10) Macau China January 2010
[14] Z Tabrizian E Afshari G Ghodrati Amiri M H A Beigyand SM PourhoseiniNejad ldquoAnewdamage detectionmethodBig Bang-Big Crunch (BB-BC) algorithmrdquo Journal of Shock andVibration vol 20 no 4 pp 633ndash648 2013
[15] A Kaveh and S Talatahari ldquoA novel heuristic optimizationmethod charged system searchrdquo Acta Mechanica vol 213 no3-4 pp 267ndash289 2010
[16] S M Pourhoseini Nejad G Ghodrati Amiri A Asadi EAfshari and Z Tabrizian ldquoDamage detection of skeletal struc-tures using particle swarm optimizer with passive congregation(PSOPC) algorithm via incomplete modal datardquo Journal ofComputational Methods in Civil Engineering vol 3 no 1 pp1ndash13 2012
[17] S He Q H Wu J Y Wen J R Saunders and R CPaton ldquoA particle swarm optimizer with passive congregationrdquoBioSystems vol 78 no 1ndash3 pp 135ndash147 2004
[18] X Wang N Hu H Fukunaga and Z H Yao ldquoStructuraldamage identification using static test data and changes infrequenciesrdquo Engineering Structures vol 23 no 6 pp 610ndash6212001
[19] A Messina E J Williams and T Contursi ldquoStructural damagedetection by a sensitivity and statistical-based methodrdquo Journalof Sound and Vibration vol 216 no 5 pp 791ndash808 1998
[20] S W Doebling C R Farrar M B Prime and D W ShevitzldquoDamage identification and healthmonitoring of structural and
Shock and Vibration 13
mechanical systems from changes in their vibration character-istics a literature reviewrdquo Research Report LA-13070-MS ESA-EA Los Alamos National Laboratory Los Alamos NM USA1996
[21] M Nobahari and S M Seyedpoor ldquoStructural damage detec-tion using an efficient correlation-based index and a modifiedgenetic algorithmrdquoMathematical and Computer Modelling vol53 no 9-10 pp 1798ndash1809 2011
[22] A Kaveh and S Talatahari ldquoOptimal design of skeletal struc-tures via the charged system search algorithmrdquo Structural andMultidisciplinary Optimization vol 41 no 6 pp 893ndash911 2010
[23] A Kaveh and S Talatahari ldquoParticle swarm optimizer antcolony strategy and harmony search scheme hybridized foroptimization of truss structuresrdquoComputers and Structures vol87 no 5-6 pp 267ndash283 2009
[24] H J Salane and J W Baldwin Jr ldquoIdentification of modalproperties of bridgesrdquo Journal of Structural Engineering vol 116no 7 pp 2008ndash2021 1990
[25] O S Salawu and CWilliams ldquoBridge assessment using forced-vibration testingrdquo Journal of Structural Engineering vol 121 no2 pp 161ndash173 1995
[26] J-M Ndambi J Vantomme and K Harri ldquoDamage assessmentin reinforced concrete beams using eigenfrequencies and modeshape derivativesrdquo Engineering Structures vol 24 no 4 pp 501ndash515 2002
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
8 Shock and Vibration
Table 2 Results of damage detection of the 10-element beam using CSS and PSOPC algorithms
Elementnumber
Scenario 1 Scenario 2
CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0000 0 0000 0101 01 0105
2 0000 0 0000 0003 0 0000
3 0000 0 0000 0001 0 0000
4 0000 0 0000 0000 0 0000
5 0000 0 0000 0185 02 0169
6 0100 01 0100 0016 0 0037
7 0000 0 0000 0349 035 0348
8 0000 0 0000 0002 0 0000
9 0000 0 0000 0001 0 0000
10 0000 0 0000 0000 0 0000
Error 00002 00000 00385 00752
Table 3 The values of objective function of the 10-element beam for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 34 times 10
minus0652 times 10
minus0628 times 10
minus0627 times 10
minus0683 times 10
minus0665 times 10
minus0641 times 10
minus0638 times 10
minus0619 times 10
minus0644 times 10
minus06 100PSOPC 25 times 10
minus0713 times 10
minus0799 times 10
minus0814 times 10
minus0721 times 10
minus0746 times 10
minus0726 times 10
minus0716 times 10
minus0688 times 10
minus0811 times 10
minus07 100Scenario 2
CSS 00008 00001 00009 00004 00006 00002 00005 00008 00009 00006 80PSOPC 00003 00013 00056 00004 00074 00074 00080 00011 00006 00085 40
Table 4 Cross-sectional areas of truss elements
Member Area (cm2)1ndash6 187ndash12 1513ndash17 1018ndash25 12
Table 5 Simulated damage scenarios in Bowstring truss
Element number Scenario 1 Scenario 235 108 251011 4024 20
where 119889119894119860 and 119889119894119878 are the actual and estimated damage of the
119894th element using the presented method respectivelyIn Table 3 the values of the objective function for CSS
and PSOPC algorithms in independent 10 runs are given theconvergence history for the 10-element beam in scenarios 1and 2 is shown in Figure 3
It is seen in Table 2 that in the first scenario where thenumber of damaged elements is low the damage identifi-cation is well by both algorithms in the second scenariothe algorithm PSOPC has some wrong in damage detectionso that there is difference between the obtained damage viathis algorithm and the actual intensity of damage in thefifth element and the undamaged sixth element mistakenlyhas been detected to be damaged Meanwhile the locationand amount of damage in structure are obtained by the CSSalgorithm with acceptable accuracy In order to investigatethe noise effects on the results of the CSSmethod 015 noisein measurement is considered for the natural frequencies
Diagrams of the CSS algorithmrsquos damage detection withcomplete and incomplete dynamic noisy data are plotted inFigures 4 and 5
As the results show this method is able to detect thelocation andmagnitude of damaged elements in all scenariosusing complete and incomplete noisy data
Example 2 A Bowstring plane truss with 25 elements isshown in Figure 6 The properties of members of this struc-ture are 119864 = 207 times 0
9Nm2 and 120588 = 7860 kgm3 and cross-sectional areas of elements are given in Table 4
Three scenarios for this truss are considered according toTable 5
Shock and Vibration 9
Table 6 Results of damage detection of the 25-bar Bowstring truss using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0009 0 0000 0000 0 00002 0000 0 0000 0000 0 00003 0000 0 0000 0000 0 00004 0000 0 0000 0000 0 00005 0000 0 0000 0098 01 00956 0000 0 0000 0000 0 00007 0000 0 0000 0000 0 00008 0246 025 0250 0000 0 00009 0001 0 0000 0000 0 000010 0000 0 0000 0000 0 000011 0000 0 0000 0402 04 042612 0000 0 0000 0000 0 000013 0000 0 0000 0000 0 000014 0001 0 0000 0000 0 000015 0000 0 0000 0000 0 000016 0000 0 0000 0000 0 000017 0000 0 0000 0000 0 000018 0000 0 0000 0005 0 000019 0000 0 0000 0004 0 007520 0000 0 0000 0001 0 000021 0000 0 0000 0000 0 000022 0000 0 0000 0000 0 000023 0000 0 0000 0000 0 000024 0000 0 0000 0185 02 000025 0000 0 0000 0014 0 0265Error 00085 00000 00119 01609
In Table 6 the results of two algorithms in all scenarios arecompared Only the first ten natural frequencies are used fordamage detection
The comparison of the objective function for CSS andPSOPC algorithms is made in Table 7 and the convergencehistory for the Bowstring truss in scenarios 1 and 2 is shownin Figure 7
In Figures 8 and 9 the results of CSS algorithm in damagedetection of this truss that is affected by noise are shown
Table 7 and Figures 7 8 and 9 show better results indamage detection by CSS algorithm rather than PSOPCalgorithm and also by applying noise in structures
Example 3 A three-story three-bay frame as shown inFigure 10 is used to verify the damage detection methodexplained in this paper The number of elements and nodesis 39 and 34 respectively
For unbraced plane frame problem all columns areW14 times 132 (119860 = 0025m2 (388 in2) and 119868 = 6386 times
10minus4m4 (1530 in4)) and all beams are W12 times 65 Youngrsquos
modulus is 119864 = 207GPa (30 000 ksi) Poissonrsquos ratio is ] =
03 and the mass density is 120588 = 7780 kgm3 (0000728 lb minuss2in4)
In this example two scenarios from the aspect of elementnumber and its level of damage were assumed that are givenin Table 8 Between ninety natural frequencies only the firsttwenty natural frequencies are utilized for damage detectionin the first scenario
The severity of damage detection by PSOPC and CSSalgorithms in different members of the frame and in eachscenario is shown separately in Table 9 In Table 10 theobtained objective function for both algorithms in each runis given also the convergence history is shown in Figure 11
Also the results of the CSS algorithm damage detectionwith noisy data are plotted in the diagrams of Figures 12 and13
The results of damage detection with two algorithmsin this large frame show that in case of only one memberdamaged there is more accuracy with PSOPC algorithm inidentifying of desired member and obtaining the amount ofdamage than the CSS algorithm It is observed that increasingthe number of injured elements in the structures reducesthe result precision with algorithms From results errors itcan be concluded that the PSOPC algorithm has problemin detecting the location of some of the damaged structuralelements so that in the third scenario the tenth and twenty-fourth members which are injured are diagnosed as healthy
10 Shock and Vibration
Table 7 The values of objective function of the Bowstring truss for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 32 times 10
minus0543 times 10
minus0549 times 10
minus0569 times 10
minus0571 times 10
minus0533 times 10
minus0526 times 10
minus0575 times 10
minus0534 times 10
minus0526 times 10
minus05 100PSOPC 26 times 10
minus0749 times 10
minus0218 times 10
minus0713 times 10
minus0112 times 10
minus0124 times 10
minus0850 times 10
minus0819 times 10
minus0826 times 10
minus0217 times 10
minus02 60Scenario 2
CSS 00002 00003 00003 00001 00039 00038 00002 00003 00015 00051 60PSOPC 00005 00134 00672 00629 00598 00050 00049 00065 00123 00251 10
Incomplete noisy dataComplete noisy dataActual damage
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 390
5
10
15
20
25
30
35
40
Element number
Dam
age (
)
Figure 12 Damage distribution of plane frame using the completeand incomplete noisy data in Scenario 1
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 400
10
20
30
40
50
60
Element number
Dam
age (
)
Predicted damageActual damage
Figure 13 Damage distribution of plane frame using the completenoisy data in Scenario 2
Table 8 Simulated damage scenarios in unbraced plane frame
Element number Scenario 1 Scenario 21 4045 309 1010 1013 2014 4019 2524 2032 50
and a considerable amount of damage in the fifteenth andtwenty-third elements that have no damage has been gainedHowever the CSS algorithm has well identified places ofall damaged elements in the structure and has achieved theamount of their damage with high accuracy and low errorFrom Figures 12 and 13 it can be observed that even in thatcase a lot of damage to the frame is considered the effectof noise on the results obtained by the CSS algorithm islow so that in case of incomplete data accuracy of resultshas reduced slightly and the performance of this algorithmdespite the noise was considered acceptable
In all three examples the convergences diagrams areshown that the process of convergence is gradual So it causesa more comprehensive search algorithm for finding optimalsolutions
The result increases the success rate of this type ofalgorithm performance in comparison with the PSOPCalgorithm
6 Conclusions
An approach for detecting damage based on continuumdamage model using charged system search algorithm ispresented The algorithm evaluates the location and severityof damage in three structures a cantilever beam a Bowstringplane truss and a three-story three-bay unbraced plane frameby minimizing an objective function by measuring completeand incomplete noisy modal data with different damagescenarios
Shock and Vibration 11
Table 9 Results of damage detection of the 39-element frame using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0399 0400 0393 0005 0 00002 0001 0 0000 0001 0 00003 0001 0 0000 0000 0 00004 0001 0 0027 0037 0 00005 0003 0 0000 0254 0300 03566 0000 0 0000 0016 0 00007 0000 0 0000 0013 0 00008 0005 0 0000 0001 0 00009 0056 0100 0000 0001 0 000010 0002 0 0000 0095 0100 000011 0005 0 0000 0002 0 000012 0008 0 0000 0001 0 000013 0193 0200 0195 0058 0 000014 0001 0 0000 0374 0400 034915 0004 0 0000 0005 0 011516 0000 0 0000 0001 0 000017 0000 0 0000 0001 0 000018 0000 0 0000 0004 0 000019 0000 0 0000 0212 0250 025620 0000 0 0000 0004 0 000021 0000 0 0000 0029 0 000022 0000 0 0002 0015 0 000023 0000 0 0000 0001 0 009224 0000 0 0000 0174 0200 000025 0000 0 0000 0001 0 000026 0000 0 0000 0002 0 000027 0002 0 0000 0005 0 000028 0000 0 0000 0001 0 004029 0000 0 0001 0007 0 000030 0000 0 0000 0000 0 000031 0001 0 0003 0002 0 000032 0000 0 0000 0497 0500 049533 0000 0 0000 0000 0 000034 0006 0 0000 0000 0 000035 0005 0 0025 0001 0 000036 0006 0 0000 0001 0 000037 0002 0 0000 0001 0 001538 0007 0 0000 0001 0 000039 0003 0 0000 0002 0 0000Error 01162 01700 03600 06808
In order to demonstrate the power of this algorithm in thediagnosis of damage a comparison has been made betweenthe results of this algorithm and the PSOPC algorithm
By comparing the damage detection results of the twovarious methods some interesting points have been con-cluded In scenarios where the number of damaged elementsis considered low both algorithms have acceptable accuracyin the results considering the more damaged members
of structures the PSOPC algorithm has been mistaken toidentify the damaged elements of the structureThis problemin larger-scale structures and the increasing number ofmembers has beenmore visible CSS algorithm has identifiedthe exact location of damage as well as achieving the amountof damage with acceptable accuracy It can be concluded thatthe CSS algorithm has great potential in global and localsearch of damage
12 Shock and Vibration
Table 10 The values of objective function of the 39-element frame for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 44 times 10
minus0346 times 10
minus0387 times 10
minus0462 times 10
minus0411 times 10
minus0366 times 10
minus0437 times 10
minus0480 times 10
minus0429 times 10
minus0398 times 10
minus04 50PSOPC 12 times 10
minus0213 times 10
minus0244 times 10
minus0315 times 10
minus0243 times 10
minus0213 times 10
minus0339 times 10
minus0214 times 10
minus0239 times 10
minus0315 times 10
minus02 10Scenario 2
CSS 19 times 10minus03
13 times 10minus03
17 times 10minus03
12 times 10minus03
23 times 10minus03
28 times 10minus03
19 times 10minus03
11 times 10minus03
21 times 10minus03
16 times 10minus03 30
PSOPC 29 times 10minus02
39 times 10minus02
11 times 10minus02
18 times 10minus02
11 times 10minus02
59 times 10minus03
69 times 10minus02
36 times 10minus02
53 times 10minus02
13 times 10minus02 0
Due to noise the real value of natural frequencies ofstructure will change it can be seen from the diagramsthat in these cases accuracy reduction of the results of theCSS algorithm is very low This indicates the high powerof this algorithm in damage detection considering noise instructure
Charged system search by considering small populationand low iteration cycle can detect the severity and location ofdamage with acceptable accuracy which shows high powerand speed of convergence of this algorithm The proposedmethod by the charged system search algorithm producedbetter results compared with particle swarm optimizationmethod
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] D J Ewins Modal Testing Theory and Practice John Wiley ampSons New York NY USA 1984
[2] S W Doebling C R Farrar and M B Prime ldquoA summaryreview of vibration-based damage identification methodsrdquoShock and Vibration Digest vol 30 no 2 pp 91ndash105 1998
[3] N Hu XWang H Fukunaga Z H Yao H X Zhang and Z SWu ldquoDamage assessment of structures using modal test datardquoInternational Journal of Solids and Structures vol 38 no 18 pp3111ndash3126 2001
[4] O S Salawu ldquoDetection of structural damage through changesin frequency a reviewrdquo Engineering Structures vol 19 no 9 pp718ndash723 1997
[5] N Bicanic and H-P Chen ldquoDamage identification in framedstructures using natural frequenciesrdquo International Journal forNumerical Methods in Engineering vol 40 no 23 pp 4451ndash4468 1997
[6] A K Pandey M Biswas and M M Samman ldquoDamagedetection from changes in curvature mode shapesrdquo Journal ofSound and Vibration vol 145 no 2 pp 321ndash332 1991
[7] J E Mottershead and M I Friswell ldquoModel updating instructural dynamics a surveyrdquo Journal of Sound and Vibrationvol 167 no 2 pp 347ndash375 1993
[8] R Perera andR Torres ldquoStructural damage detection viamodaldata with genetic algorithmsrdquo Journal of Structural Engineeringvol 132 no 9 pp 1491ndash1501 2006
[9] S Fallahian and S M Seyedpoor ldquoA two stage method forstructural damage identification using an adaptive neuro-fuzzyinference system and particle Swarm optimizationrdquo AsianJournal of Civil Engineering (Building and Housing) vol 11 pp797ndash810 2010
[10] B H Koh and S J Dyke ldquoStructural health monitoring forflexible bridge structures using correlation and sensitivity ofmodal datardquo Computers and Structures vol 85 no 3-4 pp 117ndash130 2007
[11] R-S He and S-F Hwang ldquoDamage detection by a hybrid real-parameter genetic algorithm under the assistance of grey rela-tion analysisrdquo Engineering Applications of Artificial Intelligencevol 20 no 7 pp 980ndash992 2007
[12] H Y Guo and Z L Li ldquoA two-stage method to identifystructural damage sites and extents by using evidence theoryand micro-search genetic algorithmrdquo Mechanical Systems andSignal Processing vol 23 no 3 pp 769ndash782 2009
[13] L Yu and X Chen ldquoBridge damage identification by combiningmodal flexibility and PSO methodsrdquo in Proceedings of thePrognostics and System Health Management Conference (PHMrsquo10) Macau China January 2010
[14] Z Tabrizian E Afshari G Ghodrati Amiri M H A Beigyand SM PourhoseiniNejad ldquoAnewdamage detectionmethodBig Bang-Big Crunch (BB-BC) algorithmrdquo Journal of Shock andVibration vol 20 no 4 pp 633ndash648 2013
[15] A Kaveh and S Talatahari ldquoA novel heuristic optimizationmethod charged system searchrdquo Acta Mechanica vol 213 no3-4 pp 267ndash289 2010
[16] S M Pourhoseini Nejad G Ghodrati Amiri A Asadi EAfshari and Z Tabrizian ldquoDamage detection of skeletal struc-tures using particle swarm optimizer with passive congregation(PSOPC) algorithm via incomplete modal datardquo Journal ofComputational Methods in Civil Engineering vol 3 no 1 pp1ndash13 2012
[17] S He Q H Wu J Y Wen J R Saunders and R CPaton ldquoA particle swarm optimizer with passive congregationrdquoBioSystems vol 78 no 1ndash3 pp 135ndash147 2004
[18] X Wang N Hu H Fukunaga and Z H Yao ldquoStructuraldamage identification using static test data and changes infrequenciesrdquo Engineering Structures vol 23 no 6 pp 610ndash6212001
[19] A Messina E J Williams and T Contursi ldquoStructural damagedetection by a sensitivity and statistical-based methodrdquo Journalof Sound and Vibration vol 216 no 5 pp 791ndash808 1998
[20] S W Doebling C R Farrar M B Prime and D W ShevitzldquoDamage identification and healthmonitoring of structural and
Shock and Vibration 13
mechanical systems from changes in their vibration character-istics a literature reviewrdquo Research Report LA-13070-MS ESA-EA Los Alamos National Laboratory Los Alamos NM USA1996
[21] M Nobahari and S M Seyedpoor ldquoStructural damage detec-tion using an efficient correlation-based index and a modifiedgenetic algorithmrdquoMathematical and Computer Modelling vol53 no 9-10 pp 1798ndash1809 2011
[22] A Kaveh and S Talatahari ldquoOptimal design of skeletal struc-tures via the charged system search algorithmrdquo Structural andMultidisciplinary Optimization vol 41 no 6 pp 893ndash911 2010
[23] A Kaveh and S Talatahari ldquoParticle swarm optimizer antcolony strategy and harmony search scheme hybridized foroptimization of truss structuresrdquoComputers and Structures vol87 no 5-6 pp 267ndash283 2009
[24] H J Salane and J W Baldwin Jr ldquoIdentification of modalproperties of bridgesrdquo Journal of Structural Engineering vol 116no 7 pp 2008ndash2021 1990
[25] O S Salawu and CWilliams ldquoBridge assessment using forced-vibration testingrdquo Journal of Structural Engineering vol 121 no2 pp 161ndash173 1995
[26] J-M Ndambi J Vantomme and K Harri ldquoDamage assessmentin reinforced concrete beams using eigenfrequencies and modeshape derivativesrdquo Engineering Structures vol 24 no 4 pp 501ndash515 2002
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
Shock and Vibration 9
Table 6 Results of damage detection of the 25-bar Bowstring truss using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0009 0 0000 0000 0 00002 0000 0 0000 0000 0 00003 0000 0 0000 0000 0 00004 0000 0 0000 0000 0 00005 0000 0 0000 0098 01 00956 0000 0 0000 0000 0 00007 0000 0 0000 0000 0 00008 0246 025 0250 0000 0 00009 0001 0 0000 0000 0 000010 0000 0 0000 0000 0 000011 0000 0 0000 0402 04 042612 0000 0 0000 0000 0 000013 0000 0 0000 0000 0 000014 0001 0 0000 0000 0 000015 0000 0 0000 0000 0 000016 0000 0 0000 0000 0 000017 0000 0 0000 0000 0 000018 0000 0 0000 0005 0 000019 0000 0 0000 0004 0 007520 0000 0 0000 0001 0 000021 0000 0 0000 0000 0 000022 0000 0 0000 0000 0 000023 0000 0 0000 0000 0 000024 0000 0 0000 0185 02 000025 0000 0 0000 0014 0 0265Error 00085 00000 00119 01609
In Table 6 the results of two algorithms in all scenarios arecompared Only the first ten natural frequencies are used fordamage detection
The comparison of the objective function for CSS andPSOPC algorithms is made in Table 7 and the convergencehistory for the Bowstring truss in scenarios 1 and 2 is shownin Figure 7
In Figures 8 and 9 the results of CSS algorithm in damagedetection of this truss that is affected by noise are shown
Table 7 and Figures 7 8 and 9 show better results indamage detection by CSS algorithm rather than PSOPCalgorithm and also by applying noise in structures
Example 3 A three-story three-bay frame as shown inFigure 10 is used to verify the damage detection methodexplained in this paper The number of elements and nodesis 39 and 34 respectively
For unbraced plane frame problem all columns areW14 times 132 (119860 = 0025m2 (388 in2) and 119868 = 6386 times
10minus4m4 (1530 in4)) and all beams are W12 times 65 Youngrsquos
modulus is 119864 = 207GPa (30 000 ksi) Poissonrsquos ratio is ] =
03 and the mass density is 120588 = 7780 kgm3 (0000728 lb minuss2in4)
In this example two scenarios from the aspect of elementnumber and its level of damage were assumed that are givenin Table 8 Between ninety natural frequencies only the firsttwenty natural frequencies are utilized for damage detectionin the first scenario
The severity of damage detection by PSOPC and CSSalgorithms in different members of the frame and in eachscenario is shown separately in Table 9 In Table 10 theobtained objective function for both algorithms in each runis given also the convergence history is shown in Figure 11
Also the results of the CSS algorithm damage detectionwith noisy data are plotted in the diagrams of Figures 12 and13
The results of damage detection with two algorithmsin this large frame show that in case of only one memberdamaged there is more accuracy with PSOPC algorithm inidentifying of desired member and obtaining the amount ofdamage than the CSS algorithm It is observed that increasingthe number of injured elements in the structures reducesthe result precision with algorithms From results errors itcan be concluded that the PSOPC algorithm has problemin detecting the location of some of the damaged structuralelements so that in the third scenario the tenth and twenty-fourth members which are injured are diagnosed as healthy
10 Shock and Vibration
Table 7 The values of objective function of the Bowstring truss for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 32 times 10
minus0543 times 10
minus0549 times 10
minus0569 times 10
minus0571 times 10
minus0533 times 10
minus0526 times 10
minus0575 times 10
minus0534 times 10
minus0526 times 10
minus05 100PSOPC 26 times 10
minus0749 times 10
minus0218 times 10
minus0713 times 10
minus0112 times 10
minus0124 times 10
minus0850 times 10
minus0819 times 10
minus0826 times 10
minus0217 times 10
minus02 60Scenario 2
CSS 00002 00003 00003 00001 00039 00038 00002 00003 00015 00051 60PSOPC 00005 00134 00672 00629 00598 00050 00049 00065 00123 00251 10
Incomplete noisy dataComplete noisy dataActual damage
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 390
5
10
15
20
25
30
35
40
Element number
Dam
age (
)
Figure 12 Damage distribution of plane frame using the completeand incomplete noisy data in Scenario 1
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 400
10
20
30
40
50
60
Element number
Dam
age (
)
Predicted damageActual damage
Figure 13 Damage distribution of plane frame using the completenoisy data in Scenario 2
Table 8 Simulated damage scenarios in unbraced plane frame
Element number Scenario 1 Scenario 21 4045 309 1010 1013 2014 4019 2524 2032 50
and a considerable amount of damage in the fifteenth andtwenty-third elements that have no damage has been gainedHowever the CSS algorithm has well identified places ofall damaged elements in the structure and has achieved theamount of their damage with high accuracy and low errorFrom Figures 12 and 13 it can be observed that even in thatcase a lot of damage to the frame is considered the effectof noise on the results obtained by the CSS algorithm islow so that in case of incomplete data accuracy of resultshas reduced slightly and the performance of this algorithmdespite the noise was considered acceptable
In all three examples the convergences diagrams areshown that the process of convergence is gradual So it causesa more comprehensive search algorithm for finding optimalsolutions
The result increases the success rate of this type ofalgorithm performance in comparison with the PSOPCalgorithm
6 Conclusions
An approach for detecting damage based on continuumdamage model using charged system search algorithm ispresented The algorithm evaluates the location and severityof damage in three structures a cantilever beam a Bowstringplane truss and a three-story three-bay unbraced plane frameby minimizing an objective function by measuring completeand incomplete noisy modal data with different damagescenarios
Shock and Vibration 11
Table 9 Results of damage detection of the 39-element frame using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0399 0400 0393 0005 0 00002 0001 0 0000 0001 0 00003 0001 0 0000 0000 0 00004 0001 0 0027 0037 0 00005 0003 0 0000 0254 0300 03566 0000 0 0000 0016 0 00007 0000 0 0000 0013 0 00008 0005 0 0000 0001 0 00009 0056 0100 0000 0001 0 000010 0002 0 0000 0095 0100 000011 0005 0 0000 0002 0 000012 0008 0 0000 0001 0 000013 0193 0200 0195 0058 0 000014 0001 0 0000 0374 0400 034915 0004 0 0000 0005 0 011516 0000 0 0000 0001 0 000017 0000 0 0000 0001 0 000018 0000 0 0000 0004 0 000019 0000 0 0000 0212 0250 025620 0000 0 0000 0004 0 000021 0000 0 0000 0029 0 000022 0000 0 0002 0015 0 000023 0000 0 0000 0001 0 009224 0000 0 0000 0174 0200 000025 0000 0 0000 0001 0 000026 0000 0 0000 0002 0 000027 0002 0 0000 0005 0 000028 0000 0 0000 0001 0 004029 0000 0 0001 0007 0 000030 0000 0 0000 0000 0 000031 0001 0 0003 0002 0 000032 0000 0 0000 0497 0500 049533 0000 0 0000 0000 0 000034 0006 0 0000 0000 0 000035 0005 0 0025 0001 0 000036 0006 0 0000 0001 0 000037 0002 0 0000 0001 0 001538 0007 0 0000 0001 0 000039 0003 0 0000 0002 0 0000Error 01162 01700 03600 06808
In order to demonstrate the power of this algorithm in thediagnosis of damage a comparison has been made betweenthe results of this algorithm and the PSOPC algorithm
By comparing the damage detection results of the twovarious methods some interesting points have been con-cluded In scenarios where the number of damaged elementsis considered low both algorithms have acceptable accuracyin the results considering the more damaged members
of structures the PSOPC algorithm has been mistaken toidentify the damaged elements of the structureThis problemin larger-scale structures and the increasing number ofmembers has beenmore visible CSS algorithm has identifiedthe exact location of damage as well as achieving the amountof damage with acceptable accuracy It can be concluded thatthe CSS algorithm has great potential in global and localsearch of damage
12 Shock and Vibration
Table 10 The values of objective function of the 39-element frame for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 44 times 10
minus0346 times 10
minus0387 times 10
minus0462 times 10
minus0411 times 10
minus0366 times 10
minus0437 times 10
minus0480 times 10
minus0429 times 10
minus0398 times 10
minus04 50PSOPC 12 times 10
minus0213 times 10
minus0244 times 10
minus0315 times 10
minus0243 times 10
minus0213 times 10
minus0339 times 10
minus0214 times 10
minus0239 times 10
minus0315 times 10
minus02 10Scenario 2
CSS 19 times 10minus03
13 times 10minus03
17 times 10minus03
12 times 10minus03
23 times 10minus03
28 times 10minus03
19 times 10minus03
11 times 10minus03
21 times 10minus03
16 times 10minus03 30
PSOPC 29 times 10minus02
39 times 10minus02
11 times 10minus02
18 times 10minus02
11 times 10minus02
59 times 10minus03
69 times 10minus02
36 times 10minus02
53 times 10minus02
13 times 10minus02 0
Due to noise the real value of natural frequencies ofstructure will change it can be seen from the diagramsthat in these cases accuracy reduction of the results of theCSS algorithm is very low This indicates the high powerof this algorithm in damage detection considering noise instructure
Charged system search by considering small populationand low iteration cycle can detect the severity and location ofdamage with acceptable accuracy which shows high powerand speed of convergence of this algorithm The proposedmethod by the charged system search algorithm producedbetter results compared with particle swarm optimizationmethod
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] D J Ewins Modal Testing Theory and Practice John Wiley ampSons New York NY USA 1984
[2] S W Doebling C R Farrar and M B Prime ldquoA summaryreview of vibration-based damage identification methodsrdquoShock and Vibration Digest vol 30 no 2 pp 91ndash105 1998
[3] N Hu XWang H Fukunaga Z H Yao H X Zhang and Z SWu ldquoDamage assessment of structures using modal test datardquoInternational Journal of Solids and Structures vol 38 no 18 pp3111ndash3126 2001
[4] O S Salawu ldquoDetection of structural damage through changesin frequency a reviewrdquo Engineering Structures vol 19 no 9 pp718ndash723 1997
[5] N Bicanic and H-P Chen ldquoDamage identification in framedstructures using natural frequenciesrdquo International Journal forNumerical Methods in Engineering vol 40 no 23 pp 4451ndash4468 1997
[6] A K Pandey M Biswas and M M Samman ldquoDamagedetection from changes in curvature mode shapesrdquo Journal ofSound and Vibration vol 145 no 2 pp 321ndash332 1991
[7] J E Mottershead and M I Friswell ldquoModel updating instructural dynamics a surveyrdquo Journal of Sound and Vibrationvol 167 no 2 pp 347ndash375 1993
[8] R Perera andR Torres ldquoStructural damage detection viamodaldata with genetic algorithmsrdquo Journal of Structural Engineeringvol 132 no 9 pp 1491ndash1501 2006
[9] S Fallahian and S M Seyedpoor ldquoA two stage method forstructural damage identification using an adaptive neuro-fuzzyinference system and particle Swarm optimizationrdquo AsianJournal of Civil Engineering (Building and Housing) vol 11 pp797ndash810 2010
[10] B H Koh and S J Dyke ldquoStructural health monitoring forflexible bridge structures using correlation and sensitivity ofmodal datardquo Computers and Structures vol 85 no 3-4 pp 117ndash130 2007
[11] R-S He and S-F Hwang ldquoDamage detection by a hybrid real-parameter genetic algorithm under the assistance of grey rela-tion analysisrdquo Engineering Applications of Artificial Intelligencevol 20 no 7 pp 980ndash992 2007
[12] H Y Guo and Z L Li ldquoA two-stage method to identifystructural damage sites and extents by using evidence theoryand micro-search genetic algorithmrdquo Mechanical Systems andSignal Processing vol 23 no 3 pp 769ndash782 2009
[13] L Yu and X Chen ldquoBridge damage identification by combiningmodal flexibility and PSO methodsrdquo in Proceedings of thePrognostics and System Health Management Conference (PHMrsquo10) Macau China January 2010
[14] Z Tabrizian E Afshari G Ghodrati Amiri M H A Beigyand SM PourhoseiniNejad ldquoAnewdamage detectionmethodBig Bang-Big Crunch (BB-BC) algorithmrdquo Journal of Shock andVibration vol 20 no 4 pp 633ndash648 2013
[15] A Kaveh and S Talatahari ldquoA novel heuristic optimizationmethod charged system searchrdquo Acta Mechanica vol 213 no3-4 pp 267ndash289 2010
[16] S M Pourhoseini Nejad G Ghodrati Amiri A Asadi EAfshari and Z Tabrizian ldquoDamage detection of skeletal struc-tures using particle swarm optimizer with passive congregation(PSOPC) algorithm via incomplete modal datardquo Journal ofComputational Methods in Civil Engineering vol 3 no 1 pp1ndash13 2012
[17] S He Q H Wu J Y Wen J R Saunders and R CPaton ldquoA particle swarm optimizer with passive congregationrdquoBioSystems vol 78 no 1ndash3 pp 135ndash147 2004
[18] X Wang N Hu H Fukunaga and Z H Yao ldquoStructuraldamage identification using static test data and changes infrequenciesrdquo Engineering Structures vol 23 no 6 pp 610ndash6212001
[19] A Messina E J Williams and T Contursi ldquoStructural damagedetection by a sensitivity and statistical-based methodrdquo Journalof Sound and Vibration vol 216 no 5 pp 791ndash808 1998
[20] S W Doebling C R Farrar M B Prime and D W ShevitzldquoDamage identification and healthmonitoring of structural and
Shock and Vibration 13
mechanical systems from changes in their vibration character-istics a literature reviewrdquo Research Report LA-13070-MS ESA-EA Los Alamos National Laboratory Los Alamos NM USA1996
[21] M Nobahari and S M Seyedpoor ldquoStructural damage detec-tion using an efficient correlation-based index and a modifiedgenetic algorithmrdquoMathematical and Computer Modelling vol53 no 9-10 pp 1798ndash1809 2011
[22] A Kaveh and S Talatahari ldquoOptimal design of skeletal struc-tures via the charged system search algorithmrdquo Structural andMultidisciplinary Optimization vol 41 no 6 pp 893ndash911 2010
[23] A Kaveh and S Talatahari ldquoParticle swarm optimizer antcolony strategy and harmony search scheme hybridized foroptimization of truss structuresrdquoComputers and Structures vol87 no 5-6 pp 267ndash283 2009
[24] H J Salane and J W Baldwin Jr ldquoIdentification of modalproperties of bridgesrdquo Journal of Structural Engineering vol 116no 7 pp 2008ndash2021 1990
[25] O S Salawu and CWilliams ldquoBridge assessment using forced-vibration testingrdquo Journal of Structural Engineering vol 121 no2 pp 161ndash173 1995
[26] J-M Ndambi J Vantomme and K Harri ldquoDamage assessmentin reinforced concrete beams using eigenfrequencies and modeshape derivativesrdquo Engineering Structures vol 24 no 4 pp 501ndash515 2002
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
10 Shock and Vibration
Table 7 The values of objective function of the Bowstring truss for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 32 times 10
minus0543 times 10
minus0549 times 10
minus0569 times 10
minus0571 times 10
minus0533 times 10
minus0526 times 10
minus0575 times 10
minus0534 times 10
minus0526 times 10
minus05 100PSOPC 26 times 10
minus0749 times 10
minus0218 times 10
minus0713 times 10
minus0112 times 10
minus0124 times 10
minus0850 times 10
minus0819 times 10
minus0826 times 10
minus0217 times 10
minus02 60Scenario 2
CSS 00002 00003 00003 00001 00039 00038 00002 00003 00015 00051 60PSOPC 00005 00134 00672 00629 00598 00050 00049 00065 00123 00251 10
Incomplete noisy dataComplete noisy dataActual damage
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 390
5
10
15
20
25
30
35
40
Element number
Dam
age (
)
Figure 12 Damage distribution of plane frame using the completeand incomplete noisy data in Scenario 1
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 400
10
20
30
40
50
60
Element number
Dam
age (
)
Predicted damageActual damage
Figure 13 Damage distribution of plane frame using the completenoisy data in Scenario 2
Table 8 Simulated damage scenarios in unbraced plane frame
Element number Scenario 1 Scenario 21 4045 309 1010 1013 2014 4019 2524 2032 50
and a considerable amount of damage in the fifteenth andtwenty-third elements that have no damage has been gainedHowever the CSS algorithm has well identified places ofall damaged elements in the structure and has achieved theamount of their damage with high accuracy and low errorFrom Figures 12 and 13 it can be observed that even in thatcase a lot of damage to the frame is considered the effectof noise on the results obtained by the CSS algorithm islow so that in case of incomplete data accuracy of resultshas reduced slightly and the performance of this algorithmdespite the noise was considered acceptable
In all three examples the convergences diagrams areshown that the process of convergence is gradual So it causesa more comprehensive search algorithm for finding optimalsolutions
The result increases the success rate of this type ofalgorithm performance in comparison with the PSOPCalgorithm
6 Conclusions
An approach for detecting damage based on continuumdamage model using charged system search algorithm ispresented The algorithm evaluates the location and severityof damage in three structures a cantilever beam a Bowstringplane truss and a three-story three-bay unbraced plane frameby minimizing an objective function by measuring completeand incomplete noisy modal data with different damagescenarios
Shock and Vibration 11
Table 9 Results of damage detection of the 39-element frame using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0399 0400 0393 0005 0 00002 0001 0 0000 0001 0 00003 0001 0 0000 0000 0 00004 0001 0 0027 0037 0 00005 0003 0 0000 0254 0300 03566 0000 0 0000 0016 0 00007 0000 0 0000 0013 0 00008 0005 0 0000 0001 0 00009 0056 0100 0000 0001 0 000010 0002 0 0000 0095 0100 000011 0005 0 0000 0002 0 000012 0008 0 0000 0001 0 000013 0193 0200 0195 0058 0 000014 0001 0 0000 0374 0400 034915 0004 0 0000 0005 0 011516 0000 0 0000 0001 0 000017 0000 0 0000 0001 0 000018 0000 0 0000 0004 0 000019 0000 0 0000 0212 0250 025620 0000 0 0000 0004 0 000021 0000 0 0000 0029 0 000022 0000 0 0002 0015 0 000023 0000 0 0000 0001 0 009224 0000 0 0000 0174 0200 000025 0000 0 0000 0001 0 000026 0000 0 0000 0002 0 000027 0002 0 0000 0005 0 000028 0000 0 0000 0001 0 004029 0000 0 0001 0007 0 000030 0000 0 0000 0000 0 000031 0001 0 0003 0002 0 000032 0000 0 0000 0497 0500 049533 0000 0 0000 0000 0 000034 0006 0 0000 0000 0 000035 0005 0 0025 0001 0 000036 0006 0 0000 0001 0 000037 0002 0 0000 0001 0 001538 0007 0 0000 0001 0 000039 0003 0 0000 0002 0 0000Error 01162 01700 03600 06808
In order to demonstrate the power of this algorithm in thediagnosis of damage a comparison has been made betweenthe results of this algorithm and the PSOPC algorithm
By comparing the damage detection results of the twovarious methods some interesting points have been con-cluded In scenarios where the number of damaged elementsis considered low both algorithms have acceptable accuracyin the results considering the more damaged members
of structures the PSOPC algorithm has been mistaken toidentify the damaged elements of the structureThis problemin larger-scale structures and the increasing number ofmembers has beenmore visible CSS algorithm has identifiedthe exact location of damage as well as achieving the amountof damage with acceptable accuracy It can be concluded thatthe CSS algorithm has great potential in global and localsearch of damage
12 Shock and Vibration
Table 10 The values of objective function of the 39-element frame for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 44 times 10
minus0346 times 10
minus0387 times 10
minus0462 times 10
minus0411 times 10
minus0366 times 10
minus0437 times 10
minus0480 times 10
minus0429 times 10
minus0398 times 10
minus04 50PSOPC 12 times 10
minus0213 times 10
minus0244 times 10
minus0315 times 10
minus0243 times 10
minus0213 times 10
minus0339 times 10
minus0214 times 10
minus0239 times 10
minus0315 times 10
minus02 10Scenario 2
CSS 19 times 10minus03
13 times 10minus03
17 times 10minus03
12 times 10minus03
23 times 10minus03
28 times 10minus03
19 times 10minus03
11 times 10minus03
21 times 10minus03
16 times 10minus03 30
PSOPC 29 times 10minus02
39 times 10minus02
11 times 10minus02
18 times 10minus02
11 times 10minus02
59 times 10minus03
69 times 10minus02
36 times 10minus02
53 times 10minus02
13 times 10minus02 0
Due to noise the real value of natural frequencies ofstructure will change it can be seen from the diagramsthat in these cases accuracy reduction of the results of theCSS algorithm is very low This indicates the high powerof this algorithm in damage detection considering noise instructure
Charged system search by considering small populationand low iteration cycle can detect the severity and location ofdamage with acceptable accuracy which shows high powerand speed of convergence of this algorithm The proposedmethod by the charged system search algorithm producedbetter results compared with particle swarm optimizationmethod
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] D J Ewins Modal Testing Theory and Practice John Wiley ampSons New York NY USA 1984
[2] S W Doebling C R Farrar and M B Prime ldquoA summaryreview of vibration-based damage identification methodsrdquoShock and Vibration Digest vol 30 no 2 pp 91ndash105 1998
[3] N Hu XWang H Fukunaga Z H Yao H X Zhang and Z SWu ldquoDamage assessment of structures using modal test datardquoInternational Journal of Solids and Structures vol 38 no 18 pp3111ndash3126 2001
[4] O S Salawu ldquoDetection of structural damage through changesin frequency a reviewrdquo Engineering Structures vol 19 no 9 pp718ndash723 1997
[5] N Bicanic and H-P Chen ldquoDamage identification in framedstructures using natural frequenciesrdquo International Journal forNumerical Methods in Engineering vol 40 no 23 pp 4451ndash4468 1997
[6] A K Pandey M Biswas and M M Samman ldquoDamagedetection from changes in curvature mode shapesrdquo Journal ofSound and Vibration vol 145 no 2 pp 321ndash332 1991
[7] J E Mottershead and M I Friswell ldquoModel updating instructural dynamics a surveyrdquo Journal of Sound and Vibrationvol 167 no 2 pp 347ndash375 1993
[8] R Perera andR Torres ldquoStructural damage detection viamodaldata with genetic algorithmsrdquo Journal of Structural Engineeringvol 132 no 9 pp 1491ndash1501 2006
[9] S Fallahian and S M Seyedpoor ldquoA two stage method forstructural damage identification using an adaptive neuro-fuzzyinference system and particle Swarm optimizationrdquo AsianJournal of Civil Engineering (Building and Housing) vol 11 pp797ndash810 2010
[10] B H Koh and S J Dyke ldquoStructural health monitoring forflexible bridge structures using correlation and sensitivity ofmodal datardquo Computers and Structures vol 85 no 3-4 pp 117ndash130 2007
[11] R-S He and S-F Hwang ldquoDamage detection by a hybrid real-parameter genetic algorithm under the assistance of grey rela-tion analysisrdquo Engineering Applications of Artificial Intelligencevol 20 no 7 pp 980ndash992 2007
[12] H Y Guo and Z L Li ldquoA two-stage method to identifystructural damage sites and extents by using evidence theoryand micro-search genetic algorithmrdquo Mechanical Systems andSignal Processing vol 23 no 3 pp 769ndash782 2009
[13] L Yu and X Chen ldquoBridge damage identification by combiningmodal flexibility and PSO methodsrdquo in Proceedings of thePrognostics and System Health Management Conference (PHMrsquo10) Macau China January 2010
[14] Z Tabrizian E Afshari G Ghodrati Amiri M H A Beigyand SM PourhoseiniNejad ldquoAnewdamage detectionmethodBig Bang-Big Crunch (BB-BC) algorithmrdquo Journal of Shock andVibration vol 20 no 4 pp 633ndash648 2013
[15] A Kaveh and S Talatahari ldquoA novel heuristic optimizationmethod charged system searchrdquo Acta Mechanica vol 213 no3-4 pp 267ndash289 2010
[16] S M Pourhoseini Nejad G Ghodrati Amiri A Asadi EAfshari and Z Tabrizian ldquoDamage detection of skeletal struc-tures using particle swarm optimizer with passive congregation(PSOPC) algorithm via incomplete modal datardquo Journal ofComputational Methods in Civil Engineering vol 3 no 1 pp1ndash13 2012
[17] S He Q H Wu J Y Wen J R Saunders and R CPaton ldquoA particle swarm optimizer with passive congregationrdquoBioSystems vol 78 no 1ndash3 pp 135ndash147 2004
[18] X Wang N Hu H Fukunaga and Z H Yao ldquoStructuraldamage identification using static test data and changes infrequenciesrdquo Engineering Structures vol 23 no 6 pp 610ndash6212001
[19] A Messina E J Williams and T Contursi ldquoStructural damagedetection by a sensitivity and statistical-based methodrdquo Journalof Sound and Vibration vol 216 no 5 pp 791ndash808 1998
[20] S W Doebling C R Farrar M B Prime and D W ShevitzldquoDamage identification and healthmonitoring of structural and
Shock and Vibration 13
mechanical systems from changes in their vibration character-istics a literature reviewrdquo Research Report LA-13070-MS ESA-EA Los Alamos National Laboratory Los Alamos NM USA1996
[21] M Nobahari and S M Seyedpoor ldquoStructural damage detec-tion using an efficient correlation-based index and a modifiedgenetic algorithmrdquoMathematical and Computer Modelling vol53 no 9-10 pp 1798ndash1809 2011
[22] A Kaveh and S Talatahari ldquoOptimal design of skeletal struc-tures via the charged system search algorithmrdquo Structural andMultidisciplinary Optimization vol 41 no 6 pp 893ndash911 2010
[23] A Kaveh and S Talatahari ldquoParticle swarm optimizer antcolony strategy and harmony search scheme hybridized foroptimization of truss structuresrdquoComputers and Structures vol87 no 5-6 pp 267ndash283 2009
[24] H J Salane and J W Baldwin Jr ldquoIdentification of modalproperties of bridgesrdquo Journal of Structural Engineering vol 116no 7 pp 2008ndash2021 1990
[25] O S Salawu and CWilliams ldquoBridge assessment using forced-vibration testingrdquo Journal of Structural Engineering vol 121 no2 pp 161ndash173 1995
[26] J-M Ndambi J Vantomme and K Harri ldquoDamage assessmentin reinforced concrete beams using eigenfrequencies and modeshape derivativesrdquo Engineering Structures vol 24 no 4 pp 501ndash515 2002
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
Shock and Vibration 11
Table 9 Results of damage detection of the 39-element frame using CSS and PSOPC algorithms
Element number Scenario 1 Scenario 2CSS Actual damage PSOPC CSS Actual damage PSOPC
1 0399 0400 0393 0005 0 00002 0001 0 0000 0001 0 00003 0001 0 0000 0000 0 00004 0001 0 0027 0037 0 00005 0003 0 0000 0254 0300 03566 0000 0 0000 0016 0 00007 0000 0 0000 0013 0 00008 0005 0 0000 0001 0 00009 0056 0100 0000 0001 0 000010 0002 0 0000 0095 0100 000011 0005 0 0000 0002 0 000012 0008 0 0000 0001 0 000013 0193 0200 0195 0058 0 000014 0001 0 0000 0374 0400 034915 0004 0 0000 0005 0 011516 0000 0 0000 0001 0 000017 0000 0 0000 0001 0 000018 0000 0 0000 0004 0 000019 0000 0 0000 0212 0250 025620 0000 0 0000 0004 0 000021 0000 0 0000 0029 0 000022 0000 0 0002 0015 0 000023 0000 0 0000 0001 0 009224 0000 0 0000 0174 0200 000025 0000 0 0000 0001 0 000026 0000 0 0000 0002 0 000027 0002 0 0000 0005 0 000028 0000 0 0000 0001 0 004029 0000 0 0001 0007 0 000030 0000 0 0000 0000 0 000031 0001 0 0003 0002 0 000032 0000 0 0000 0497 0500 049533 0000 0 0000 0000 0 000034 0006 0 0000 0000 0 000035 0005 0 0025 0001 0 000036 0006 0 0000 0001 0 000037 0002 0 0000 0001 0 001538 0007 0 0000 0001 0 000039 0003 0 0000 0002 0 0000Error 01162 01700 03600 06808
In order to demonstrate the power of this algorithm in thediagnosis of damage a comparison has been made betweenthe results of this algorithm and the PSOPC algorithm
By comparing the damage detection results of the twovarious methods some interesting points have been con-cluded In scenarios where the number of damaged elementsis considered low both algorithms have acceptable accuracyin the results considering the more damaged members
of structures the PSOPC algorithm has been mistaken toidentify the damaged elements of the structureThis problemin larger-scale structures and the increasing number ofmembers has beenmore visible CSS algorithm has identifiedthe exact location of damage as well as achieving the amountof damage with acceptable accuracy It can be concluded thatthe CSS algorithm has great potential in global and localsearch of damage
12 Shock and Vibration
Table 10 The values of objective function of the 39-element frame for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 44 times 10
minus0346 times 10
minus0387 times 10
minus0462 times 10
minus0411 times 10
minus0366 times 10
minus0437 times 10
minus0480 times 10
minus0429 times 10
minus0398 times 10
minus04 50PSOPC 12 times 10
minus0213 times 10
minus0244 times 10
minus0315 times 10
minus0243 times 10
minus0213 times 10
minus0339 times 10
minus0214 times 10
minus0239 times 10
minus0315 times 10
minus02 10Scenario 2
CSS 19 times 10minus03
13 times 10minus03
17 times 10minus03
12 times 10minus03
23 times 10minus03
28 times 10minus03
19 times 10minus03
11 times 10minus03
21 times 10minus03
16 times 10minus03 30
PSOPC 29 times 10minus02
39 times 10minus02
11 times 10minus02
18 times 10minus02
11 times 10minus02
59 times 10minus03
69 times 10minus02
36 times 10minus02
53 times 10minus02
13 times 10minus02 0
Due to noise the real value of natural frequencies ofstructure will change it can be seen from the diagramsthat in these cases accuracy reduction of the results of theCSS algorithm is very low This indicates the high powerof this algorithm in damage detection considering noise instructure
Charged system search by considering small populationand low iteration cycle can detect the severity and location ofdamage with acceptable accuracy which shows high powerand speed of convergence of this algorithm The proposedmethod by the charged system search algorithm producedbetter results compared with particle swarm optimizationmethod
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] D J Ewins Modal Testing Theory and Practice John Wiley ampSons New York NY USA 1984
[2] S W Doebling C R Farrar and M B Prime ldquoA summaryreview of vibration-based damage identification methodsrdquoShock and Vibration Digest vol 30 no 2 pp 91ndash105 1998
[3] N Hu XWang H Fukunaga Z H Yao H X Zhang and Z SWu ldquoDamage assessment of structures using modal test datardquoInternational Journal of Solids and Structures vol 38 no 18 pp3111ndash3126 2001
[4] O S Salawu ldquoDetection of structural damage through changesin frequency a reviewrdquo Engineering Structures vol 19 no 9 pp718ndash723 1997
[5] N Bicanic and H-P Chen ldquoDamage identification in framedstructures using natural frequenciesrdquo International Journal forNumerical Methods in Engineering vol 40 no 23 pp 4451ndash4468 1997
[6] A K Pandey M Biswas and M M Samman ldquoDamagedetection from changes in curvature mode shapesrdquo Journal ofSound and Vibration vol 145 no 2 pp 321ndash332 1991
[7] J E Mottershead and M I Friswell ldquoModel updating instructural dynamics a surveyrdquo Journal of Sound and Vibrationvol 167 no 2 pp 347ndash375 1993
[8] R Perera andR Torres ldquoStructural damage detection viamodaldata with genetic algorithmsrdquo Journal of Structural Engineeringvol 132 no 9 pp 1491ndash1501 2006
[9] S Fallahian and S M Seyedpoor ldquoA two stage method forstructural damage identification using an adaptive neuro-fuzzyinference system and particle Swarm optimizationrdquo AsianJournal of Civil Engineering (Building and Housing) vol 11 pp797ndash810 2010
[10] B H Koh and S J Dyke ldquoStructural health monitoring forflexible bridge structures using correlation and sensitivity ofmodal datardquo Computers and Structures vol 85 no 3-4 pp 117ndash130 2007
[11] R-S He and S-F Hwang ldquoDamage detection by a hybrid real-parameter genetic algorithm under the assistance of grey rela-tion analysisrdquo Engineering Applications of Artificial Intelligencevol 20 no 7 pp 980ndash992 2007
[12] H Y Guo and Z L Li ldquoA two-stage method to identifystructural damage sites and extents by using evidence theoryand micro-search genetic algorithmrdquo Mechanical Systems andSignal Processing vol 23 no 3 pp 769ndash782 2009
[13] L Yu and X Chen ldquoBridge damage identification by combiningmodal flexibility and PSO methodsrdquo in Proceedings of thePrognostics and System Health Management Conference (PHMrsquo10) Macau China January 2010
[14] Z Tabrizian E Afshari G Ghodrati Amiri M H A Beigyand SM PourhoseiniNejad ldquoAnewdamage detectionmethodBig Bang-Big Crunch (BB-BC) algorithmrdquo Journal of Shock andVibration vol 20 no 4 pp 633ndash648 2013
[15] A Kaveh and S Talatahari ldquoA novel heuristic optimizationmethod charged system searchrdquo Acta Mechanica vol 213 no3-4 pp 267ndash289 2010
[16] S M Pourhoseini Nejad G Ghodrati Amiri A Asadi EAfshari and Z Tabrizian ldquoDamage detection of skeletal struc-tures using particle swarm optimizer with passive congregation(PSOPC) algorithm via incomplete modal datardquo Journal ofComputational Methods in Civil Engineering vol 3 no 1 pp1ndash13 2012
[17] S He Q H Wu J Y Wen J R Saunders and R CPaton ldquoA particle swarm optimizer with passive congregationrdquoBioSystems vol 78 no 1ndash3 pp 135ndash147 2004
[18] X Wang N Hu H Fukunaga and Z H Yao ldquoStructuraldamage identification using static test data and changes infrequenciesrdquo Engineering Structures vol 23 no 6 pp 610ndash6212001
[19] A Messina E J Williams and T Contursi ldquoStructural damagedetection by a sensitivity and statistical-based methodrdquo Journalof Sound and Vibration vol 216 no 5 pp 791ndash808 1998
[20] S W Doebling C R Farrar M B Prime and D W ShevitzldquoDamage identification and healthmonitoring of structural and
Shock and Vibration 13
mechanical systems from changes in their vibration character-istics a literature reviewrdquo Research Report LA-13070-MS ESA-EA Los Alamos National Laboratory Los Alamos NM USA1996
[21] M Nobahari and S M Seyedpoor ldquoStructural damage detec-tion using an efficient correlation-based index and a modifiedgenetic algorithmrdquoMathematical and Computer Modelling vol53 no 9-10 pp 1798ndash1809 2011
[22] A Kaveh and S Talatahari ldquoOptimal design of skeletal struc-tures via the charged system search algorithmrdquo Structural andMultidisciplinary Optimization vol 41 no 6 pp 893ndash911 2010
[23] A Kaveh and S Talatahari ldquoParticle swarm optimizer antcolony strategy and harmony search scheme hybridized foroptimization of truss structuresrdquoComputers and Structures vol87 no 5-6 pp 267ndash283 2009
[24] H J Salane and J W Baldwin Jr ldquoIdentification of modalproperties of bridgesrdquo Journal of Structural Engineering vol 116no 7 pp 2008ndash2021 1990
[25] O S Salawu and CWilliams ldquoBridge assessment using forced-vibration testingrdquo Journal of Structural Engineering vol 121 no2 pp 161ndash173 1995
[26] J-M Ndambi J Vantomme and K Harri ldquoDamage assessmentin reinforced concrete beams using eigenfrequencies and modeshape derivativesrdquo Engineering Structures vol 24 no 4 pp 501ndash515 2002
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
12 Shock and Vibration
Table 10 The values of objective function of the 39-element frame for algorithms
Run number1 2 3 4 5 6 7 8 9 10 Success
Scenario 1CSS 44 times 10
minus0346 times 10
minus0387 times 10
minus0462 times 10
minus0411 times 10
minus0366 times 10
minus0437 times 10
minus0480 times 10
minus0429 times 10
minus0398 times 10
minus04 50PSOPC 12 times 10
minus0213 times 10
minus0244 times 10
minus0315 times 10
minus0243 times 10
minus0213 times 10
minus0339 times 10
minus0214 times 10
minus0239 times 10
minus0315 times 10
minus02 10Scenario 2
CSS 19 times 10minus03
13 times 10minus03
17 times 10minus03
12 times 10minus03
23 times 10minus03
28 times 10minus03
19 times 10minus03
11 times 10minus03
21 times 10minus03
16 times 10minus03 30
PSOPC 29 times 10minus02
39 times 10minus02
11 times 10minus02
18 times 10minus02
11 times 10minus02
59 times 10minus03
69 times 10minus02
36 times 10minus02
53 times 10minus02
13 times 10minus02 0
Due to noise the real value of natural frequencies ofstructure will change it can be seen from the diagramsthat in these cases accuracy reduction of the results of theCSS algorithm is very low This indicates the high powerof this algorithm in damage detection considering noise instructure
Charged system search by considering small populationand low iteration cycle can detect the severity and location ofdamage with acceptable accuracy which shows high powerand speed of convergence of this algorithm The proposedmethod by the charged system search algorithm producedbetter results compared with particle swarm optimizationmethod
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] D J Ewins Modal Testing Theory and Practice John Wiley ampSons New York NY USA 1984
[2] S W Doebling C R Farrar and M B Prime ldquoA summaryreview of vibration-based damage identification methodsrdquoShock and Vibration Digest vol 30 no 2 pp 91ndash105 1998
[3] N Hu XWang H Fukunaga Z H Yao H X Zhang and Z SWu ldquoDamage assessment of structures using modal test datardquoInternational Journal of Solids and Structures vol 38 no 18 pp3111ndash3126 2001
[4] O S Salawu ldquoDetection of structural damage through changesin frequency a reviewrdquo Engineering Structures vol 19 no 9 pp718ndash723 1997
[5] N Bicanic and H-P Chen ldquoDamage identification in framedstructures using natural frequenciesrdquo International Journal forNumerical Methods in Engineering vol 40 no 23 pp 4451ndash4468 1997
[6] A K Pandey M Biswas and M M Samman ldquoDamagedetection from changes in curvature mode shapesrdquo Journal ofSound and Vibration vol 145 no 2 pp 321ndash332 1991
[7] J E Mottershead and M I Friswell ldquoModel updating instructural dynamics a surveyrdquo Journal of Sound and Vibrationvol 167 no 2 pp 347ndash375 1993
[8] R Perera andR Torres ldquoStructural damage detection viamodaldata with genetic algorithmsrdquo Journal of Structural Engineeringvol 132 no 9 pp 1491ndash1501 2006
[9] S Fallahian and S M Seyedpoor ldquoA two stage method forstructural damage identification using an adaptive neuro-fuzzyinference system and particle Swarm optimizationrdquo AsianJournal of Civil Engineering (Building and Housing) vol 11 pp797ndash810 2010
[10] B H Koh and S J Dyke ldquoStructural health monitoring forflexible bridge structures using correlation and sensitivity ofmodal datardquo Computers and Structures vol 85 no 3-4 pp 117ndash130 2007
[11] R-S He and S-F Hwang ldquoDamage detection by a hybrid real-parameter genetic algorithm under the assistance of grey rela-tion analysisrdquo Engineering Applications of Artificial Intelligencevol 20 no 7 pp 980ndash992 2007
[12] H Y Guo and Z L Li ldquoA two-stage method to identifystructural damage sites and extents by using evidence theoryand micro-search genetic algorithmrdquo Mechanical Systems andSignal Processing vol 23 no 3 pp 769ndash782 2009
[13] L Yu and X Chen ldquoBridge damage identification by combiningmodal flexibility and PSO methodsrdquo in Proceedings of thePrognostics and System Health Management Conference (PHMrsquo10) Macau China January 2010
[14] Z Tabrizian E Afshari G Ghodrati Amiri M H A Beigyand SM PourhoseiniNejad ldquoAnewdamage detectionmethodBig Bang-Big Crunch (BB-BC) algorithmrdquo Journal of Shock andVibration vol 20 no 4 pp 633ndash648 2013
[15] A Kaveh and S Talatahari ldquoA novel heuristic optimizationmethod charged system searchrdquo Acta Mechanica vol 213 no3-4 pp 267ndash289 2010
[16] S M Pourhoseini Nejad G Ghodrati Amiri A Asadi EAfshari and Z Tabrizian ldquoDamage detection of skeletal struc-tures using particle swarm optimizer with passive congregation(PSOPC) algorithm via incomplete modal datardquo Journal ofComputational Methods in Civil Engineering vol 3 no 1 pp1ndash13 2012
[17] S He Q H Wu J Y Wen J R Saunders and R CPaton ldquoA particle swarm optimizer with passive congregationrdquoBioSystems vol 78 no 1ndash3 pp 135ndash147 2004
[18] X Wang N Hu H Fukunaga and Z H Yao ldquoStructuraldamage identification using static test data and changes infrequenciesrdquo Engineering Structures vol 23 no 6 pp 610ndash6212001
[19] A Messina E J Williams and T Contursi ldquoStructural damagedetection by a sensitivity and statistical-based methodrdquo Journalof Sound and Vibration vol 216 no 5 pp 791ndash808 1998
[20] S W Doebling C R Farrar M B Prime and D W ShevitzldquoDamage identification and healthmonitoring of structural and
Shock and Vibration 13
mechanical systems from changes in their vibration character-istics a literature reviewrdquo Research Report LA-13070-MS ESA-EA Los Alamos National Laboratory Los Alamos NM USA1996
[21] M Nobahari and S M Seyedpoor ldquoStructural damage detec-tion using an efficient correlation-based index and a modifiedgenetic algorithmrdquoMathematical and Computer Modelling vol53 no 9-10 pp 1798ndash1809 2011
[22] A Kaveh and S Talatahari ldquoOptimal design of skeletal struc-tures via the charged system search algorithmrdquo Structural andMultidisciplinary Optimization vol 41 no 6 pp 893ndash911 2010
[23] A Kaveh and S Talatahari ldquoParticle swarm optimizer antcolony strategy and harmony search scheme hybridized foroptimization of truss structuresrdquoComputers and Structures vol87 no 5-6 pp 267ndash283 2009
[24] H J Salane and J W Baldwin Jr ldquoIdentification of modalproperties of bridgesrdquo Journal of Structural Engineering vol 116no 7 pp 2008ndash2021 1990
[25] O S Salawu and CWilliams ldquoBridge assessment using forced-vibration testingrdquo Journal of Structural Engineering vol 121 no2 pp 161ndash173 1995
[26] J-M Ndambi J Vantomme and K Harri ldquoDamage assessmentin reinforced concrete beams using eigenfrequencies and modeshape derivativesrdquo Engineering Structures vol 24 no 4 pp 501ndash515 2002
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
Shock and Vibration 13
mechanical systems from changes in their vibration character-istics a literature reviewrdquo Research Report LA-13070-MS ESA-EA Los Alamos National Laboratory Los Alamos NM USA1996
[21] M Nobahari and S M Seyedpoor ldquoStructural damage detec-tion using an efficient correlation-based index and a modifiedgenetic algorithmrdquoMathematical and Computer Modelling vol53 no 9-10 pp 1798ndash1809 2011
[22] A Kaveh and S Talatahari ldquoOptimal design of skeletal struc-tures via the charged system search algorithmrdquo Structural andMultidisciplinary Optimization vol 41 no 6 pp 893ndash911 2010
[23] A Kaveh and S Talatahari ldquoParticle swarm optimizer antcolony strategy and harmony search scheme hybridized foroptimization of truss structuresrdquoComputers and Structures vol87 no 5-6 pp 267ndash283 2009
[24] H J Salane and J W Baldwin Jr ldquoIdentification of modalproperties of bridgesrdquo Journal of Structural Engineering vol 116no 7 pp 2008ndash2021 1990
[25] O S Salawu and CWilliams ldquoBridge assessment using forced-vibration testingrdquo Journal of Structural Engineering vol 121 no2 pp 161ndash173 1995
[26] J-M Ndambi J Vantomme and K Harri ldquoDamage assessmentin reinforced concrete beams using eigenfrequencies and modeshape derivativesrdquo Engineering Structures vol 24 no 4 pp 501ndash515 2002
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporation httpwwwhindawicom
Journal ofEngineeringVolume 2014
Submit your manuscripts athttpwwwhindawicom
VLSI Design
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation httpwwwhindawicom
Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
DistributedSensor Networks
International Journal of