Você está na página 1de 19

Elsevier Editorial System(tm) for Structures

Manuscript Draft

Manuscript Number: STRUCTURES-D-18-00043

Title: EVALUATION OF VIBRATION-BASED DAMAGE DETECTION TECHNIQUES FOR


REINFORCED-CONCRETE BEAMS

Article Type: Research Paper

Keywords: Change in Flexibility; Curvature Mode Shape; Damage detection;


Finite element method; Modal Strain Energy Change Ratio; Principal
Eigenvector of Modal Flexibility Change

Corresponding Author: Dr. Shamsher Bahadur Singh, Ph.D.

Corresponding Author's Institution: Birla Institute of Technology and


Science Pilani

First Author: Shamsher Bahadur Singh, Ph.D.

Order of Authors: Shamsher Bahadur Singh, Ph.D.; Shubham Aggarwal, Under


graduate; Shreyas Pranav, Undergraduate

Abstract: This study aims at comprehensively exploring and comparing four


different methods of vibration-based damage detection in reinforced-
concrete beams with three different support conditions. The methods
chosen are Change in Flexibility (FC), Curvature Mode Shape (CMS), Modal
Strain Energy Change Ratio (MSECR) and Principal Eigenvector of Modal
Flexibility Change (PE). The parameters considered for this study are
damage detection, localization and quantification, for both single and
multiple damage cases. The objective is to propose the most efficient
method for each of these parameters. A beam of length 3.15m is analyzed
using a self-developed finite element model based program with MATLAB
platform and damage is simulated by reductions in equivalent flexural
rigidity. It is found that the FC Method is most suitable for single
damage detection while the MSECR method is most suitable for multiple
damage detection.

Suggested Reviewers: S. V. R. Madappa PhD


Assistant Professor, Civil Engineering, NIT Puducherry
s.madappa@gmail.com
Dynamic Researcher

K. K. Viswanathan PhD
Universiti Teknologi Malaysia
visu20@yahoo.com
Well known researcher

Opposed Reviewers:
Cover Letter

To

The Chief Editor

Structures

Subj: Submission of a manuscript for possible publication in Structures

Dear Sir,

Enclosed please find herewith the manuscript entitled “EVALUATION OF VIBRATION-BASED DAMAGE
DETECTION TECHNIQUES FOR REINFORCED-CONCRETE BEAMS,” for possible publication in Structures.

Please let me know if you have any constructive comments.

Looking forward to hearing from you soon in this regard

With kind regards


Shamsher
-----------------------------------------------------------------------------------------
Shamsher Bahadur Singh, Ph.D., P.E. (USA), FIE, PDF (USA)
Professor, Department of Civil Engineering

Birla Institute of Technology & Science, Pilani


Pilani Campus

Vidya Vihar
Pilani 333031, Rajasthan, India

Phone: +91 9414648283


Fax: +91 1596 244183
Email: sbsingh@pilani.bits-pilani.ac.in
Profile Page: http://universe.bits-pilani.ac.in/pilani/sbsingh/profile
*Manuscript
Click here to view linked References

1
2
3
4
5 EVALUATION OF VIBRATION-BASED DAMAGE DETECTION
6 TECHNIQUES FOR REINFORCED-CONCRETE BEAMS
7
8 S. B. Singha, S. Aggarwalb and S. Pranavb
9
10 a
Professor, , Birla Institute of Technology and Science, Pilani-333031(sbsinghbits@gmail.com)
11
12 b
Undergraduate Student, Birla Institute of Technology and Science, Pilani-333031
13
14
15
16 Abstract:
17
18 This study aims at comprehensively exploring and comparing four different methods of
19
20 vibration-based damage detection in reinforced-concrete beams with three different support
21 conditions. The methods chosen are Change in Flexibility (FC), Curvature Mode Shape (CMS),
22 Modal Strain Energy Change Ratio (MSECR) and Principal Eigenvector of Modal Flexibility
23
24
Change (PE). The parameters considered for this study are damage detection, localization and
25 quantification, for both single and multiple damage cases. The objective is to propose the most
26 efficient method for each of these parameters. A beam of length 3.15m is analyzed using a self-
27
28
developed finite element model based program with MATLAB platform and damage is
29 simulated by reductions in equivalent flexural rigidity. It is found that the FC Method is most
30 suitable for single damage detection while the MSECR method is most suitable for multiple
31 damage detection.
32
33 Keywords: Change in Flexibility, Curvature Mode Shape; Damage detection; Finite element
34
35 method; Modal Strain Energy Change Ratio; Principal Eigenvector of Modal Flexibility Change
36
37 1. Introduction
38
39 It is of vital importance that the damage is detected at an early stage in structures so that the
40 maintenance cost, chances of sudden failure and fatalities due to failure are minimized.
41 Vibration-based damage detection in civil engineering structures has been a topic of research
42
43 interest for a long time, and different aspects of this field have been explored to different depths.
44 The present study attempts to supplement the existing knowledge base in the field of damage
45 detection by performing a comprehensive study on four different methods namely, Curvature
46
47 Mode Shape (CMS-1991) [1], Modal Strain Energy Change Ratio (MSECR-1998) [2], Change
48 in Flexibility (FC1994) [3] and Principal Eigenvector of Modal Flexibility Change (PE-2016)
49 [4]. Rytter [5] categorized damage detection methods based on whether the method is effective
50
51 in predicting the following parameters: Presence of damage, localization of damage,
52 quantification of the damage severity and determining the structure’s remaining service life. The
53 present study considers only the first three parameters, for both single and multiple damage
54
55 cases.
56
57 An overall comparative study is done, proposing which method is most appropriate for which
58 parameter, followed by suggesting the effectiveness of the modern PE method as compared to
59 the other methods.
60
61
62
63
64
65
1
2
3
4 A beam is often subjected to vibrations in the field, and the response of the beam (in terms of
5
6 natural frequency and mode shapes) can be calculated by simple eigenvalue analysis, using a
7 finite element model (FEM) of the beam. All the four methods, in essence, are based on the
8 vibration mode shapes of the beam. Three kinds of beams, namely cantilever, simply supported,
9
10 and overhanging are considered to decisively conclude the effectiveness of each method. This
11 also helps in suggesting the most suitable method for each of the three beams. The values of
12 Curvature Mode Shape, Modal Strain Energy, Flexibility, and PE curvature of Flexibility are
13
14 obtained from the mode shapes and compared for damaged and intact cases in the corresponding
15 four methods. Since damage in a reinforced-concrete beam causes alteration in its cross-section
16 and elastic modulus, damage is introduced at different locations by reducing the flexural rigidity
17
18
of the beam, thus altering the stiffness matrix. The mass matrix is assumed to remain unaffected
19 as is consistent with [6] Cawley and Adams.
20
21
22
23 2. A brief description of the methods
24
25 2.1 Change in Flexibility Method (FC Method)
26
27 The method essentially involves finding out the flexibility matrix of the structure for both the
28 damaged and intact cases, and comparing the two graphically. A distinct peak or a slope change
29
30
is observed at the damage location.
31
32 Eq. (1) can be used to find the flexibility matrix F of a structure:
33
34 (1)
35
36
37 where the mode shape matrix Φ is normalized to unity (ΦTM Φ=I), ϕi is the ith mode shape
38 vector(that is, ith column of the mode shape matrix Φ), ωi is the ith modal frequency, Ω=diag
39
40
( and n is the number of degrees of freedom (DOFs).
41
42
This flexibility matrix is calculated for both damaged and undamaged cases, and difference of
43 the flexibility matrices Δ is calculated according to Eq. (2).
44
45 (2)
46
47 where Fi and Fd are the flexibility matrices for the undamaged and damaged cases, respectively.
48
49 If represents the maximum absolute value of elements in column j of Δ, Eq. (3) for is used
50
51 as a measure of change in flexibility at each degree of freedom:
52
53 (3)
54
55 where is the element corresponding to row i and column j of Δ.
56
57 From these values of , the values corresponding to vertical displacement DOFs are chosen, a
58
59 plot between the modified and the corresponding node number is made, and the results are
60 analyzed.
61
62
63
64
65
1
2
3
4 2.2 Principal Eigenvector of Modal Flexibility Change Method (PE Method)
5
6 In this method, the principal eigenvector of the change in flexibility matrix is computed, and the
7
8 variation of this quantity with node number is used as a measure of damage detection.
9
10 The flexibility matrix is obtained from Eq. (1) for both damaged and undamaged cases and the
11 difference matrix ΔF is found from Eq. (2).
12
13 The principal eigenvector (PE) of ΔF is obtained by solving the eigenvalue problem in Eq. (6) as
14 considered in [4]:
15
16
17 (6)
18
19
20 where λ1 is the principal eigenvalue and is the principal eigenvector of ΔF, respectively.
21
22 Now, curvature of is calculated using the central difference method according to Eq. (7) as,
23
24
25
(7)
26
27
28
29 where is the ith coefficient of , h is the length of one finite element and is the ith
30 curvature coefficient of Now, from these values of , the values corresponding to
31
32 vertical displacement DOFs are chosen, a plot between this modified and the
33 corresponding node number is made, and the results are analyzed.
34
35 2.3 Modal Strain Energy Change Ratio Method (MSECR Method)
36
37 This method involves finding out the change in modal strain energy for each finite element
38
39
before and after damage. Modal strain energy of jth element for ith mode shape can be calculated
40 using Eq. (8),
41
42 = and = (8)
43
44
45 where and are the modal strain energy (MSE) values for undamaged and damaged
46 elements respectively, and represent the ith mode shape vectors for undamaged and
47
damaged element respectively, and represents the undamaged stiffness matrix of the jth
48
49 element. This technique involves no change in the stiffness matrix.
50
51 Further, Modal Strain Energy Change Ratio (MSECR) for ith mode and jth element is calculated
52 according to Eq. (9),
53
54 –
55 = (9)
56
57
58 Also, MSECRj for a given element j is defined as the ratio of the average of values for
59
60 all m modes, to the largest value of each mode. This is as shown in Eq. (10).
61
62
63
64
65
1
2
3
4
5 (10)
6
7
8 Now, a plot of these MSECR values is made with element number on the horizontal axis, and the
9 results are analyzed.
10
11 2.4 Curvature Mode Shape Method (CMS Method)
12
13 This method involves computation of the Curvature Mode Shape Change (CMSC) as the
14 absolute difference in the curvature mode shapes of damaged and intact beams. Central
15
16 difference approximation is used to compute curvature mode shape κ from the vertical
17 displacement mode shapes, according to Eq. (11),
18
19
20 (11)
21
22 where, υi is the vertical displacement mode shape of the ith element and h is the length of one
23 finite element.
24
25 Finally, the CMSC values are plotted against corresponding node number and the results are
26
27 analyzed. One thing worth noticing in this method is that the CMSC values for multiple modes
28 do not aggregate to give a single plot. Rather, multiple plots are obtained corresponding to the
29 number of modes considered. Out of these plots, only a few locate damage properly. The present
30
31 study considers only one plot which is the best one out of the plots corresponding to the first
32 three modes.
33
34
35
36 3. Analytical Model
37
38 A reinforced-concrete beam of length 3.15m having uniform cross-section (0.3mx0.2m), density
39 2400 kg/m3 is considered for evaluating the four methods discussed in Section 2 and comparing
40
41 them. The beam is modeled using the finite element method as having 21 one-dimensional
42 elements each of length 0.15m. Two degrees of freedom namely, the displacement along the Y
43 axis and rotation about the Z axis are used at each node in the analysis. Three different boundary
44
45 conditions (Cantilever, Simply-Supported and Overhanging) are chosen to give a general idea of
46 the effect of different boundary conditions on the efficiency of the four methods to detect, locate
47 and quantify damage. Figure 1 shows the finite element models of these three beams. The
48
49
overhanging beam is restrained in the vertical direction at nodes 8 and 15, dividing the beam into
50 three equal parts of length 1.05m each.
51
52
53
54 4. Parametric Study
55
56 The numerical computing language MATLAB is used to simulate the four methods,
57 parametrically study them and identify their advantages and shortcomings. The parametric study
58
59
considers both single and multiple damage cases, in Sections 4.1 and 4.2 respectively. As
60 discussed in Section 1, damage detection entails a study of the following: 1.) Detecting the
61
62
63
64
65
1
2
3
4 presence of damage, 2.) Precision in locating the damage and 3.) Quantifying the intensity of
5
6 damage. The most efficient method is the one which best predicts these parameters.
7
8 The results of the MATLAB program for all the four methods are seen to be consistent with
9 those obtained by other researchers [1 - 4].
10
11 4.1 Single Damage Detection
12
13 A single damage is introduced at Element 11 in all the three beams, with successive reductions
14 of 10%, 20%, 40%, 60%, 80% and 90% in the flexural rigidity. Figures 2 to 4 show the results
15
16 obtained for all the four methods. Each figure shows the ability of the method to detect damage
17 at the precise location and to predict its severity.
18
19 Every method is able to clearly detect damage at element 11 for all the beams as the peaks show.
20
21 The FC method in case of a cantilever beam shows the presence of damage by a distinct change
22 in the slope of ΔF at the defect location. This is also true for overhanging beams when the
23
24 damage is located not between the two restraints. For simply supported beams, the damage
25 location is represented by a peak. Methods other than the FC method indicate damage by a sharp
26 peak at the damage location.
27
28 The ability of each method to predict severity is quantified by calculating the relative change in
29
30 peak value (or slope in the particular cases discussed above) corresponding to ‘p’% damage
31 intensity, with the peak corresponding to 10% as reference (p represents damage intensities of
32 20%, 40%, 60%, 80% and 90%). This is shown in Eq. (12).
33
34
35
36
37 (12)
38
39
40
41 In other words, severity is the percent increase in peak value in any general damage case with
42
43
respect to the peak value corresponding to 10% damage intensity. The severity values for single
44 damage detection, calculated from Eq. (12), are summarized in Tables 1 to 3, respectively for all
45 the three beam types. Each table contains these values for all the four methods. So, for example,
46
from Table 1, the value 0.4448% indicates that the peak value corresponding to 20% damage is
47
48 0.448% higher than the peak corresponding to 10% damage.
49
50 The higher the severity value, the more efficient (or sensitive) the method is, in predicting the
51 intensity of damage applied. It is thus seen from Tables 1-3 that the FC Method is most efficient
52 in distinctly identifying severity of damage, while the PE method is least capable of identifying
53
54 the damage severity.
55
56 The severity values progressively increase as the damage increases in all cases as the tables
57 show. By taking into account the values corresponding to 90% damage, a definite trend for each
58 method can be observed from Tables 1 to 3. The severity values range between the following
59
60 values:
61
62
63
64
65
1
2
3
4 PE ~ 10-20%
5
6 MSE ~ 15-30%
7
8 CMS ~ 3500-4500%
9
10 FC ~ 8000-13000%
11
12
13
A similar order of values is observed for all damage cases, i.e. the values for FC method are
14 always the highest and those for PE are always the lowest. So, Tables 1 to 3 give an idea of the
15 effectiveness of each method in predicting the severity of damage. The FC method appears best
16 in this regard while the PE method appears the least capable, because of the high and low
17
18 severity values, respectively.
19
20 4.2 Multiple Damage Detection
21
22 Three damages are introduced at elements 6, 12 and 18 in all the three beams with intensities of
23 30%, 60% and 90%. Three cases are considered as shown in Table 4. Each case comprises of a
24 different permutations of these three damage intensities. Case-1 introduces 30%, 60% and 90%
25
26 damage, respectively at elements 6, 12 and 18, Case-2 has 60%, 90% and 30% damage while
27 Case-3 has 90%, 30% and 60% damage at the respective locations. Such a methodology has been
28 devised to affirm the consistency of a particular method in detecting multiple damages.
29
30 The results obtained from this analysis are shown in Figures 5 to 7. Individual enlarged images
31
32 (5(c2), 6(c2) and 7(c2)) are shown for the FC method because of scale issues in the combined
33 image. Each figure showcases the ability of the method to detect damage at the precise locations
34 and to predict their relative severities. Damage is represented by peaks, or slope changes as in the
35
36 case of single damage. If the peak height or slope value progressively increases for increasing
37 damage intensities as in Case-1 (and according to the corresponding order in Case-2 and Case-3),
38 then it can be concluded that the method is capable of predicting relative damage severity
39
40 effectively. Figures 5(b), 6(b) and 7(b) show that the MSECR method is consistently able to
41 detect and locate damage at the precise locations. It is also able to predict the relative severity of
42 damage at different locations.
43
44 The PE method detects and locates damage precisely in all cases as Figures 5(d), 6(d) and 7(d)
45
46
show (although with lesser sharpness as compared to the MSECR method), but is not reliable in
47 predicting relative severity clearly. Only the simply supported beam case gives accurate results
48 for relative severity.
49
50 It can be seen from Figures 5(c1, c2), 6(c1, c2) and 7(c1, c2) that the FC method seldom detects
51
52
more than one damage; although the slope does change at the damage locations, it is very minute
53 in most cases. But it does not indicate damage at incorrect locations. The simply supported beam
54 case shows only one damage location consistently. Even severity cannot be predicted accurately.
55
56 The CMS method fails to detect multiple damages in quite a few cases as Figures 5(a), 6(a) and
57
7(a) show. High intensity damage (90%) is detected consistently, but it masks the further
58
59 damages having lower intensities from being detected. Many random peaks are also observed.
60 The severity results are also unreliable.
61
62
63
64
65
1
2
3
4 5. Conclusions
5
6
7
 All the methods are efficient in single damage detection and localization.
8  The following is the order of capability of clearly identifying single damage severity.
9
10 FC > CMS > MSE > PE
11
12  The efficiency of the methods in multiple damage detection is as follows:
13
14 MSE > PE > FC > CMS
15
16  Erroneous peaks are obtained only in the CMS method when there are multiple damages.
17
18
 Only the MSE method predicts relative severities accurately. The other methods are
19 unreliable in this regard.
20  The CMS method gives multiple plots, each one suggesting damage at different locations
21
22
(especially for higher modes), and therefore this method is less reliable for multiple
23 damage detection using higher modes.
24  The modern PE method, which was expected to outshine the older methods, does not
25
26
seem to be a clear winner, although it gives decent enough results.
27
28
29
30 6. References
31
32 [1] A. K. Pandey and M. Biswas 1991 Journal of Sound and Vibration 145(2), 321-332. Damage
33 Detection from Changes in Curvature Mode Shapes.
34
35 [2] Z. Y. Shi, S. S. Law and L. M. Zhang 1998 Journal of Sound and Vibration 218(5), 825-844.
36 Structural Damage Localization from Modal Strain Energy Change.
37
38 [3] A. K. Pandey and M. Biswas 1994 Journal of Sound and Vibration 169(1), 3-17. Damage
39
40 Detection in Structures Using Changes in Flexibility.
41
42 [4] Cui-Hong Li, Qiu-Wei Yang and Bing-Xiang Sun 2016 Algorithms. Structural Damage
43 Localization by the Principal Eigenvector of Modal Flexibility Change.
44
45 [5] Rytter, A. 1993. Vibration based inspection of civil engineering structures. Ph.D.
46 Dissertation, Department of Building Technology and Structural Engineering, Aalborg
47
48 University, Denmark.
49
50 [6] P. Cawley and R. Adams 1979 Journal of Strain Analysis 14(2), 49-57. The location of
51 Defects in Structures from Measurements of Natural Frequencies.
52
53
54
55
56
57
58
59
60
61
62
63
64
65
Figure

(a)

(b)

(c)

Fig. 1. Finite element model of (a) Simply supported, (b) Cantilever and (c) Overhanging beam.
(a) (b)

(c) (d)

Fig. 2. Single damage detection results for simply supported beam by (a) CMS, (b) MSECR, (c)
FC and (d) PE methods.
(a) (b)

(c) (d)

Fig. 3. Single damage detection results for cantilever beam by (a) CMS, (b) MSECR, (c) FC and
(d) PE methods.
(a) (b)

(c) (d)

Fig. 4. Single damage detection results for overhanging beam by (a) CMS, (b) MSECR, (c) FC
and (d) PE methods.
(a) (b)

(c1) (c2)
(d)
Fig. 5. Multiple damage detection results for simply supported beam by (a) CMS, (b) MSECR,
(c1, c2) FC and (d) PE methods.

(a) (b)
(c1) (c2)

(d)

Fig. 6. Multiple damage detection results for cantilever beam by (a) CMS, (b) MSECR, (c1, c2)
FC and (d) PE methods.
(a) (b)

(c1) (c2)
(d)

Fig. 7. Multiple damage detection results for overhanging beam by (a) CMS, (b) MSECR, (c1,
c2) FC and (d) PE methods.
Table

Table 1. Severity values for simply supported beam case.

Method PE MSE CMS FC


20% 0.4448% 2.60% 122.12% 126%
40% 1.7439% 7.39% 470.43% 515%
60% 4.1935% 11.02% 1094.90% 1335%
80% 10.5313% 14.68% 2538.06% 4135%
90% 19.9273% 16.50% 4309.85% 11264%

Table 2. Severity values for cantilever beam case.

Method PE MSE CMS FC


20% 0.1816% 3.68% 121.44% 125%
40% 0.7220% 8.01% 463.70% 499.8%
60% 1.7832% 13.28% 1062.17% 1249.6%
80% 4.8175% 17.24% 2371.06% 3498.9%
90% 10.2833% 18.83% 3836.18% 7997.6%

Table 3. Severity values for overhanging beam case.

Method PE MSE CMS FC


20% 0.4661% 2.33% 121.95% 127%
40% 1.8239% 8.79% 468.69% 522%
60% 4.3702% 16.85% 1086.10% 1372%
80% 10.8713% 25.87% 2488.80% 4411%
90% 20.2749% 29.91% 4156.70% 12683%

Table 4. Three permutations for multiple damage cases.


Case No. Element 6 Element 12 Element 18
1 30% 60% 90%
2 60% 90% 30%
3 90% 30% 60%

Você também pode gostar