You are on page 1of 8

Storey damage identification using system dynamic properties

Kimberly CAPEDING
BSMS Civil Engineering
De La Salle University
Manila, Philippines

Kristian CARINGAL
BSMS Civil Engineering
De La Salle University,
Manila, Philippines

kimberlycapeding@yahoo.com

kristian_caringal@yahoo.com

Kimberly Capeding is a structural


engineering major. She is
particularly interested in studies
about Structural Health
Monitoring.

Kristian Caringal is the President


of the Civil Engineering Society.
He was recently recognized as the
Most Outstanding Civil
Engineering Student by the
Philippine Institute of Civil
Engineers.

Summary
A damage identification method is presented for the detection of storey damage. The proposed
method utilizes the dynamic properties of the system, particularly the natural frequencies and the
mode shapes. A residual force method is introduced and described to locate the damaged floor in the
structure. The validity of the method is demonstrated by the identification of the damaged story of a
structure with known properties using its measured natural frequency. The proposed method
showed success in determining the location of the damaged storey.
Keywords: damage detection, damage identification, structural health monitoring, inverse
engineering problem

1. Introduction
Buildings are man-made vertical structures that are constructed to provide support or shelter to an
intended occupancy. In the span of its service life, it is subjected to different loads which in turn
introduce different forms of stresses to the structure. While buildings are designed to be able to
resist all of these stresses when they occur, overtime the strength capacity of a structure decreases
due to service wear and aging. Old buildings become more susceptible to damage, especially during
catastrophic events such as earthquakes, typhoons, tsunamis, or other cases when the building is
subjected to extreme loading.
Over the past two decades, the health monitoring of old structures have been the subject of many
research studies. This is owing to the realization that an early condition assessment of a structure
can enhance the overall safety and reliability of a structure, thus extending its service life. Detecting
structural damages that can compromise a structures performance is of great interest in the field of
civil engineering.
Damage is defined as any change in the physical and geometric properties of a structure from its
original characteristics. To this respect, the detection of damage requires a comparison between the
healthy undamaged state of a system to its damaged condition. This case study addresses the use of
a structures dynamic properties to identify and locate damaged storeys. Following the logic that the
dynamic properties (period, natural frequencies, mode shapes) are a function of the physical
characteristics of the building (mass, damping, and stiffness), any change in the physical properties
will manifest in the detectable and measurable building dynamic response.

Structural Health Monitoring (SHM) is a process aimed at increasing safety, maintainability, and
reliability of structures by identifying, locating, and quantifying damages using measured building
responses. The damage information is then used to assess and monitor the health and performance
of a structure.
The amount of literature related to damage detection using shifts in resonant frequencies is quite
large. Salawu [1] presents an excellent review on the use of modal frequency changes for damage
diagnostics. The observation that changes in structural properties cause changes in vibration
frequencies was the impetus for using modal methods for damage identification and health
monitoring. A procedure of matrix optimization using measured modal data was proposed by
Baruch [2], and Berman and Nagy [3], to solve for the least change in their values when damage
occurred. Cawley and Adams [4] proposed a sensitivity based method to detect and estimate
damage using only the measured frequencies of the damaged structure. To locate the damage,
theoretical frequency change ratios of two dierent modes were calculated by introducing damage
at selected positions on a nite element model.
Su. et. al. [5] located damaged storeys in a shear building using its sub-structural frequencies. An
autoregressive with exogenous input model of the building substructure in healthy and unhealthy
condition is established from the substructural dynamic responses. Using a wavelet transform, the
substructural natural frequencies were obtained and were compared to determine the damaged
storey.
There are many existing methods on damage detection that utilize the systems dynamic properties
to come up with damage information. However, these methodologies require extensive
computations that lead to longer processing time. The study reduces the amount of computation
needed by analyzing the study as a lumped mass model. The method is tested by applying the
procedure to detect damage in a 2d steel frame.

2. Damage Detection
In locating where damage is present in the structure, an equation of motion for an undamped case
given n Degree of Freedom (DOF) is given by
M u + Ku=F (t )

(1)

where M and K are the n x n mass and stiffness matrices, and u and u are the n x 1 acceleration
and displacement vectors, respectively, F(t) is the n x 1 excitation force vector. Under free vibration
(i.e., F(t) = 0), a characteristic equation can be derived from Eq. (1) and written as
( K i M ) i=0
(2)
Where

is the ith mode eigenvalue or square of the natural frequency

eigenvector or the natural mode shape with respect to the

and

is the

Eq. (2) can be used to determine the natural frequencies and mode shapes of the structure and it
forms the basis for many damage detection methods that make use of modal properties in damage
detection and identification. A residual force method used by Ge and Lui [6] is applied to locate
2

damage. The method is based on identifying the difference in modal properties between the original
(undamaged) and damaged structures. A subscript d is used to denote the damaged structure and Eq.
(2) becomes

( K d di M d ) v =0

(3)

The local damage in the structure will cause a change in the stiffness and mass matrices,
M . This will result into two equations for the damaged matrices expressed as
K d =K u + K

(4a)

M d=M u + M

(4b)

and

where the subscript u represents the undamaged structure.


By substituting Eq. (4a) and (4b) to Eq(2), an equation can be obtained in the form of
K di di M di =(K u + di M u ) di

(5)

The left hand side of the equation is defined as the residual force vector for mode i as:
Ri= K di di M di
(6)
Eq. (5) can be rewritten as
Ri=(K u + di M u ) di

(7)

A close inspection of Ri reveals that if the measured frequencies and mode shapes are uncorrupted
by noise, the value of Ri will be zero if none of the elements associated with the floor is damaged,
but it will assume a non-zero value if any element is associated with this degree-of-freedom is
damaged. This observation can be readily deduced from Eq. (6).
Thus any residual force value that return a non-zero value is said to be damaged. This determines
the part of the matrix where damage is located.

3. Numerical Study
3.1

Methodology

There were two sets of calculations done in the case study in determining the location of damage in
the structure. The first part is a confirmation of the damage location model with the residual force
vector. As an example, a steel shear building as shown in figure 1 was analyzed with a lump massed
model having 3 degrees of freedom (DOF) shown in figure 2. The methodology is presented in
figure 3.

Fig 1: Shear Building Model

Fig 2: Lumped Mass Model

Fig 3: Methodology for confirmation of damage location model


3.2

Input Calculations

The properties of the undamaged structure are listed in Table 1. The lumped masses at the first,
second, and third level are taken as 2.2 x 106, 2.0 x 106, and 1.8 x 106 in kilograms.
Table 1: Properties of Undamaged Structure
Properties
Section
I (mm4)
A (mm2)
L (mm)
K (N/m)

Ground
W360x110
331000000
14100
3000
58844444

Storey
2nd
W360x101
301000000
12900
3000
53511111

3rd
W310x79
177000000
10100
3000
31466667

Using these properties, the stiffness and mass matrices Ku and Mu of the structure are obtained.
4

k 1+ k 2 k 2
0
Ku= k 2 k 2+k 3 k 3
0
k 3
+k3

M1 0
0
Ku= 0
M2 0
0
0
M3

(8)

(9)

Three different variations of local damage are applied to the undamaged structure. The simulated
damage scenarios were:
Case 1:10% stiffness reduction in member 1
Case 2:40% stiffness reduction in member 1
Case 3:90% stiffness reduction in member 1
Table 2: Summary of stiffness values
Stiffness, k
Undamaged Case
(N/m)
58844444
Ground

10%
55902222

Damaged Cases
40%
47075556

90%
32364444

Second

53511111

53511111

53511111

53511111

Third

31466667

31466667

31466667

31466667

Change in mass and stiffness parameters

and

were obtained by simply subtracting

the parameters of the damaged state from the undamaged state. This change is quantified as the
damage imposed to the structure. The summary of the stiffness values k x per story x in several
damage cases is written in Table 2.
Using these properties, Eq. (2) is used to compute for the three eigenvalues and eigenvectors
corresponding to the 3 dofs per model.
Table 3: Summary of mode shapes for each damage scenario
Mode
1
2
3

Undamaged
eigenvalues
5.423084946
30.85697845
74.76101287

10% stiffness

Damaged eigenvalues
40% stiffness

90% stiffness

reduction
5.288666084
30.33584982
74.07917517

reduction
4.831296316
28.71327576
72.14701672

reduction
3.838544184
25.86672974
69.29944965

The residual force vector was then calculated using Eq (6). The summary of results is shown in
Table 4. All of the cases showed that there are no values in the vector R value for rows 2 and 3
5

reflecting that there are no damage present in storey 2 and storey 3. The non-zero R value in the
first row shows that the damage is present in storey 1. The method successfully identified which is
the damaged storey in the 2d frame. However localization of the damage or knowing which specific
member is subjected to damage cannot be determined through the model since a lumped mass
approach was used.

Table 4: Residual force vector results from Eq. (6)


Damaged Case

Mode
1
2
3
1
2
3
1
2
3

1
2
3
3.3

R in kN
2942
0
0
11769
0
0
26480
0
0

Validation of Method

The second set of calculations is done with an objective of determining the location of damage
using the same residual force vector given only the eigenvalue and eigenvector parameters of the
damaged state and the stiffness and mass matrices of the undamaged state. The methodology is
presented in Fig 4.

Fig 4: Methodology for locating damage


The same dynamic properties of the damaged structure from the first part of the methodology were
used as an input for calculating the residual force vector using Eq (7). In this case, it is assumed that
the input values are obtained from a measured response and not from a local damage application.
The summary of results is shown in table 2. The results are exactly the same showing that either
forms of equation 5 can be used in locating the damage in a structure.

Table 5 Residual force vector results from Eq. (7)


Case
1
2
3

4.

Mode
1
2
3
1
2
3
1
2
3

R in kN
2942
0
0
11769
0
0
26480
0
0

Conclusions

It was presented in this paper that knowing the eigenvalues and eigenvectors of the damaged
structure through a measured response along with the structural properties of the undamaged state
such as stiffness and mass parameters can easily determine the location of damage in a structure by
its storey using the residual force vector equation.
The researchers found out that the rows in which R-values are present is the corresponding story
location of the damage in the structure. In a case that there would be two R-values present, the
negative values should be neglected, as it is just a result of the formation of the stiffness matrix of a
lumped mass model.
It was identified through the calculations that residual force equation alone will not be enough to
locate the damage in a localized aspect because the matrices were formed in a lumped mass
approach.
The researcher recommends a modification of the study that addresses the localization of damaged
member for future researches.
5.
[1]
[2]

References
SALAWU O.S., Detection of Structural Damage Through Changes in Frequency: A
Review, Engineering Structures, Vol. 19, No. 9, 1997, pp. 718-723.
BARUCH M., Optimal correction of mass and stiffness matrices using measured modes,
Am Inst AeronautAstronaut J, Vol. 20, No.11, 1982, pp. 16231636.
7

[3]
[4]
[5]
[6]

BERMAN A, NAGY E.J., Improvement of a large analyticalmodel using test data, Am


Inst Aeronaut Astronaut J, Vol. 21, No.8, 1983, pp. 11681173.
CAWLEY P, ADAMS R.D. The predicted and experimental natural modes of free-free
CFRP plates, J Compos Mater, Vol. 12, 1978, pp. 336347.
SU W.C., et. al. Locating damaged storeys in a shear building based on its sub-structural
natural frequencies, Engineering Structures, Vol. 39, 2012, pp. 126-138
GE M., LUI E.M. Structural damage identification using system dynamic properties,
Computers and Structures Vol. 83, 2005, pp. 2185-2196.