Professional Documents
Culture Documents
www.elsevier.com/locate/engstruct
Abstract
This paper presents the results of the structural identification of a historical structure damaged during the 17 August 1999 Kocaeli earthquake.
Results of the structural models and methods are cross-validated in order to identify the reliable dynamic characteristics. Spectral analysis of the
eight structural points yields average peak frequencies, respectively at around 2.5, 3.5, 4.3, 5.3 Hz and so on for NW–SE direction. Similarly for
the SW–NE direction, peaks are seen at around 2.6, 3.2, 4.5–5 Hz with small differences in amplitudes between different structural points. Among
the stations, station 3 behaves differently than the other observation points in low and high frequency regions indicating that the data recorded at
station 3 should be examined in more detail.
In parametric analysis, (a) the SDOF system as an SISO-ARX model, and (b) the MDOF system as a MIMO-ARX model are adopted and
modal deformations are obtained using a state-space model with the OKID technique. The SDOF model captures the important peak frequencies
with sufficient accuracy, but magnitudes are arguably small. In MDOF model, responses at the crown stations 1, 2, 3, and 4 display almost the
same peak magnitudes at the same frequency, 2.51 Hz for NE–SW direction, but station 3 (which possesses smaller peak amplitude at a smaller
frequency, 2.2 Hz) is exceptional and indicative of an anomaly. Similar conclusion with different peak frequency for NE–SW direction is also
inferred from the SISO and Spectral analyses, when compared. Station 3 among the crown stations exceptionally yields peak frequency at 1.8 Hz
in spectral analysis and around 1.8 Hz in SISO analysis. Natural frequencies of the structure estimated by SDOF and MDOF systems are close
to each other, and thus, they can be accepted as global properties of the structure, except station 3. Such inconsistency implies that, for indicative
anomaly, there might be a practical avenue through the evaluation of the first mode estimated by spectral, SDOF and MDOF models for a structure
with the members of equal importancy in global dynamic behavior and symmetrical in plan and in elevations.
Calculated elastic response spectrum of the Fatih record has larger amplitudes at the period of 0.55 s (with broadband characteristic) and larger
than the amplitudes of the design spectrum with 10% probability of excedance in 50 years (for new constructions) for North–South direction.
These larger local peaks at 0.55 s; (1) verify the 1.8 Hz of the fundamental frequency of the local ground and (2) suggest that, as the minimum
requirements for the earthquake resistant design of buildings, in accordance with the strengthening applications in current seismic design code,
the 2% probability of exceedance of the 5% damped elastic design acceleration spectrum within a period of 50 years is adequate. First mode of
the structure, 2.4 and 2.5 Hz for NS and EW directions respectively, falls into the range of the dominant broadband period of the seismic loading.
Such a match points out the resonance possibility, which might have occurred during the earthquake. The duration of the earthquake and following
shocks of many small earthquakes in addition to the resonance might be the key contributors to the damage.
c 2007 Elsevier Ltd. All rights reserved.
Keywords: Structural identification; Ambient vibration; Frequency response function (FRF); Nonparametric (spectral) analysis; Parametric analysis; Transfer
function; SIMO (SDOF); MIMO (MDOF); Structural response; Response spectra; Historical structure
0141-0296/$ - see front matter c 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2007.08.005
2166 K. Beyen / Engineering Structures 30 (2008) 2165–2184
Table 1
Important historical earthquakes of İstanbul
Ottaman period 16 October 1489 Mosques built after the conquest and especially their minarets received intensive damages.
22 August 1509 Particularly strong earthquake referred as ‘The Lesser Judgement’. The aftershocks continued
for forty five days. A hundred and nine mosques and more than a thousand dwellings were
destroyed. In records, the death toll varies between 5000 and 15000. Tsunamis hit the city
walls and rushed in from the Marmara sea.
30 April 1557 Fatih mosque received serious damage.
28 June 1648 Ferries and ships hit each other and received serious damage. Great number of dwellings were
collapsed.
6 February 1659 City walls and Rumeli Castle on the Bosphore received damages.
11 July 1690 Occurred in night time. All domes of the Fatih mosque and the main entrance of the city wall
at the Topkapı damaged.
24 May 1719 Many parts of the city walls on the seven castles and the dome of the Mihrimah Sultan
mosque collapsed.
3 September 1754 Domes of the Fatih and Beyazid mosques received damage.
22 May 1766 Intensively occurred on the second day of the religious holiday, Aftershocks took 8 months.
Great number of the dwellings and much of the parts of Fatih mosque destroyed.
23 September 1841 City was shaken 3 times and received intensive damage.
10 July 1894 Earthquake duration is 12 s. Great number of dwellings and tools on historical peninsula and
princes islands lost.
Turkish republic 29 May 1923 Moderate earthquake with 5.5 Magnitude, epicenter is in north of the İzmit.
26 October 1923 Moderate earthquake with 5 Magnitude, epicenter is in Catalca.
13 March 1952 Moderate earthquake with 5 Magnitude, epicenter is in Marmara Ereģlisi.
26 December 1957 Moderate earthquake with 5.2 Magnitude, epicenter is in Izmit.
18 September 1963 Moderate earthquake with 6.3 Magnitude, epicenter is in Cınarcık.
24 April 1988 Moderate earthquake with 5.1 Magnitude, epicenter is in Marmara Ereģlisi.
Feb–April 1995 Small earthquakes with 4.4–4.6 Magnitudes, epicenters are in Marmara Sea.
17 August 1999 Intensive earthquake with 7.4 Magnitude, epicenter is in Golcuk, epicentral distance is 70 km
to the city center of Istanbul.
despite its distance of about 110 km from source zone [1]. affected Fatih mosque in its lifetime as shown in Table 1. In the
Countless number of historical structures in Istanbul are in recent earthquake, the structure received light damage localized
similar situation and they are subjected to the effect of active at the crown of the north arch. Moreover, historical records
faults in either near or far regions with almost equal risks. For show that Istanbul can expect to be severely shaken about once
conservation and restoration of historical structures, significant every hundred years by a major earthquake that takes place in
advances have been made in non-destructive experimental and the vicinity while suffering from smaller shocks about once a
numerical techniques for identifying structural characteristics, decade or so in between as seen in Table 1.
and knowledge on traditional materials [2,3]. To enhance the To detect the onset of damage, loosening of a connection
efforts in rehabilitating vulnerable structures including those or deterioration at the earliest possible stage is important
already damaged, structural identification [4] will serve as a for structural safety. Vibration-based damage detection may
tool in assessing the structural behavior and the local effects. help find changes in modal properties. Properties or quantities
Historical structures made of stone brick with arch and derived from the measured vibrations are being used as
dome components take an important portion of the historical indicators of damage. Damage is mostly a local phenomenon
structure stock. As a historical record, collapse of the east and local response is captured by higher frequency modes
arch of Hagia Sophia in 1347 is a notable example for a post- whereas lower frequency modes tend to capture the global
earthquake failure of a structure that received damage during response of the structure. However, it should be pointed out that
the major earthquake of 1344. During the huge earthquake in contrast to moment resisting shear frame reinforced concrete
of 1766, many domes and arches in the city including Fatih structures, the response of the main structural system such as
mosque collapsed, since the dome was supported on arches an arch + dome type, directly affecting the global response
that rise from heavy corner piers [5,6]. Many earthquakes have may have indicative changes in the lower frequency modes.
K. Beyen / Engineering Structures 30 (2008) 2165–2184 2167
Drawing 2. Plan and elevation of the structure from SouthWest towards NorthEast.
K. Beyen / Engineering Structures 30 (2008) 2165–2184 2169
3. Structural identification of the structure connectivity and global behavior as observed in this structure,
and (5) close eye inspection is available for correlation with the
Steel and reinforced concrete are ductile materials that measured features, then the damage level and location become
have linear elastic properties and good post-elastic strength important definitions to clarify the damage as local or global
characteristics. After yielding, these materials maintain most of one.
their strength while undergoing substantial plastic deformations Both nonparametric and parametric techniques are em-
and they can be analysed with reasonable accuracy using ployed in order to cross-validate the results obtained for the
analytic or computational methods. In contrast, since historical Fatih Mosque. Using recorded input and output datasets; para-
buildings were not designed as moment resisting shear frame metric identification involves the estimation of system param-
reinforced concrete structures to resist cyclic shear or bending eters, while nonparametric identification yields an estimate of
moments at pier and arch connections, after cracks are initiated, the characteristic structural transfer function (i.e. frequency re-
the behavior of brittle, unreinforced materials, such as stone sponse function (FRF)) only. FRFs are the result of the Fourier
or adobe brick, is extremely difficult to predict even with transforms of the responses at each selected structural point.
today’s advanced computational capabilities [11]. In a brittle Processing is performed to estimate the parameters of the char-
material, once yielding occurs, cracks develop and there is a acteristic transfer functions for each structural measurement
complete loss of tensile strength. After cracks have developed, point in cases of the single-input–single-output (SISO) models
the seismic behavior of such a structure is dominated by the and multi-input–multi-output (MIMO) models. The philosophy
interactions among large, cracked sections of the members that behind these models is that the natural frequencies and damp-
rock out-of-plane and collide or rub against each other in- ing ratios do not vary from FRF to FRF and global properties of
plane [11]. a structure can be estimated by averaging out small variations
In general, strength capacities of the arch, namely, sliding in modal properties inferred from several single analyses.
capacity, rocking capacity and shear capacity are reduced. Such
strength reduction in the main arch and at the base level of the 3.2. Instrumentation, calibration tests and data processing
main dome introduces deformations at the lower frequencies.
Similar results found in [12] states that the low frequency modal In order to record the ambient vibrations of the structure,
properties significantly decrease at the progressing damage, a total of 10 single-component seismometers were mounted at
more than those reported for similar structures in literature. A critical points as shown in Drawings 1 and 2 and described
post-earthquake quick ambient vibration survey conducted in in Table 2. Locations of the sensors are key points believed
Hagia Sophia showed a frequency drop in the first two modal to reflect the structural characteristics. The equipment used
vibrational frequencies around the order of 4%–5%, indicating for the measurement and the data acquisition are as follows;
locations where most of the damage might be expected during Analog to digital converter, DT-2827 A/D board with 16 bit
a major earthquake close to Istanbul [13]. Past earthquake A/D converter at a high speed of 100 kHz accommodating both
damages and failure mechanism of the Fatih mosque also verify single-end and differential analog inputs, manufactured by Data
the progressive damage and weak structural points [14]. One Translation Inc., Multiplexer card, DI-64x4-V2, manufactured
compilation mentions the damages on dome and at the top by Quesing Electronics, accepting up to 64 channels at 1000 Hz
of the four piers during 1509 earthquake and about the 1766 sampled data and signal conditioner, SC-1 manufactured by
earthquake, it gives the collapse of the main dome and piers, Kinemetrics. From Mark products, L22 type low frequency
sequentially [14]. seismometers with peak-to-peak 0.15 in. amplitude precision
were utilized for the tests. Digitized data is acquired at a rate
3.1. Methodology for modal analysis of 122 Hz through the A/D converter. In order to quantify
and minimize the effects of changing environmental conditions,
As a faster and reliable method to evaluate the safety of such four sets of ambient vibration measurements were carried out
historical structures after an earthquake, assumed reference to identify the dynamic characteristics of the structure along
(undamaged) structural configuration of the similar members with the free-field measurements in the courtyard. Ambient
and the vibration measurements in the similar structural points vibrations were recorded for each structural point for a duration
may be used for comparison. Difference between the damaged of 5 min in each setup.
structural component at the damaged frequency (i.e. it is not the After the structural measurements were completed, for
classical mode anymore.) and the undamaged frequency might calibration purpose, another test was performed at the base
be an indicative anomaly pointing at two different states of level of the structure for each direction corresponding to
the main system members. To implement this idea in practice, symmetry axes of the rectangular structural geometry in plan.
limited conditions should be followed as if (1) the data are not To eliminate the effects of white noise, hardware problems
available from both the undamaged and damaged structures, and effects of the different cable lengths, baseline correction
(2) damage detection is primarily performed on the equally (linear and, if necessary nonlinear), decimation for eliminating
load-carrying structural components of the main frame system high frequency spikes and filtering between frequencies of
within the symmetry in plan and in both elevations, as seen 0.2 Hz and the Nyquist frequency were performed. In order to
in Drawing 2, (3) location and severity of the damage in the compensate the differences in instrumental transfer functions,
main frame system is important, (4) the damage affects the other calibration tests were performed for logarithmically
2170 K. Beyen / Engineering Structures 30 (2008) 2165–2184
Fig. 1. Displacement time histories for the recorded structural points. Dark lines and lightly colored lines represent the displacement of the structure and the ground,
respectively. Left two columns are for NE–SW direction and right ones is for NW–SE direction.
Fig. 2. Transfer functions of the observation points for one data set for NE–SW direction estimated from spectral analysis.
increasing amplitudes of the swept-sine excitation up to SGolay filter [15] with window length of 21 samples and 5th
frequency at which the seismometers yield almost constant order filter for noise reduction without loss of high frequency
amplitude under the lab conditions. information component. Then, ambient vibration displacement
Data pre-processing forms an important step in structural time histories of the structural points and the ground for two
identification. It includes resampling (from 122 Hz of directions are plotted in Fig. 1.
sampling frequency to 200 Hz), baseline correction (2nd order
polynomial fit in order to remove distortions, if necessary) 3.3. Nonparametric (spectral) analysis
and smoothing. Since Savitzky–Golay filters are optimal in
the sense that they are more effective at preserving the In order to identify the frequency content of the structural
pertinent high frequency components of the signal while behavior, the transfer functions of the test points have been
minimizing the least-squares error in fitting a polynomial calculated for both directions and the one for SW–NE direction
to frames of noisy data [15]. Smoothing is performed by is plotted in Fig. 2. Except station 3, the spectral peaks at
K. Beyen / Engineering Structures 30 (2008) 2165–2184 2171
Fig. 3. Power spectra of the ambient measurements recorded at the structural observation points (a) for NW–SE, (b) for NE–SW directions, and (c) for the basement
for NE–SW directions.
cornice-level stations at four springing points of main arches was applied to ambient measurement dataset to reconstruct the
and those at the crowns of the main arches at the dome-base original signal from the wavelet packet decomposition. Using
occur almost at the same frequencies. However, the spectral different levels of decompositions with wavelets of varying
peak at observation point 3 at the dome base displays a shift decibels, noise in the records are examined satisfying the best
to the lower frequencies unlike other locations. Figs. 3(a) and possible threshold criteria to compress the signal. Details can
(b) display the power spectra of the eight structural points along be found in [15]. Removed noise as a residual is re-evaluated in
the NW–SE and SW–NE directions, respectively. Power spectra frequency domain to see any matching portions with the major
of the basement ambient vibration measurement in Fig. 3(c) portion of the structural data. It is seen that the effects of the
shows the fundamental frequency of the ground at about 1.8 Hz. noise in the analysis is not dominant and reconstructed signals
For NW–SE, spectral peaks are observed at 2.5, 3.5, 4.3, with retained energy of 81.49% is seen dominant.
5.3 Hz. The SW–NE spectra display broadband characteristics;
3.4. Parametric analysis
the peaks are located at 2.6, 3.2, and 4.5–5 Hz with small
differences in amplitudes between different structural points. Structural identification is the process of constructing and
Station 3 behaves differently in low and high frequency calibrating a mathematical model of a physical system using
regions when compared with other observation points along experimental data. The two major categories comprise time-
both directions. After close inspection at the crown region, domain methods that use the time histories, and frequency-
this unique difference may be explained only by the damage domain methods that use the Frequency Response Functions
received during 17 August 1999 Kocaeli earthquake. All these (FRF) of recorded input and output datasets. There are several
evidences indicate that the data recorded at station 3 should algorithms available for system identification within each
be studied in more detail. Wavelet Transform (WT) technique category, involving several parameters and options that must
2172 K. Beyen / Engineering Structures 30 (2008) 2165–2184
Table 3b
Modal frequencies and corresponding damping ratios for both directions during the Kocaeli earthquake
Table 4
Quality indicators of the identified modal characteristics
Modal frequency (Hz) Constitutive modal indicator Modal history prediction Mode singular value
NE–SW NW–SE NE–SW (%) NW–SE (%) NE–SW (%) NW–SE (%) NE–SW (%) NW–SE(%)
2.31 2.42 71.06 73.61 99.54 99.81 99.54 68.81
3.20 3.29 57.89 33.06 99.73 98.16 99.73 39.32
3.38 3.52 22.50 21.46 98.20 99.51 98.20 38.69
4.04 4.41 52.65 36.20 73.59 97.73 73.59 26.51
4.47 4.59 67.43 49.01 97.80 97.72 97.80 26.89
Table 5a
Descriptions for first five mode shapes of both directions from ambient measurements for both directions
a structural component in symmetrical structures in plan and strong nonlinearity is nonunique. In this study, it is proposed
in elevations with the members of equal importancy in global that the weak nonlinear structural member is the one in which
dynamic behavior. the distribution of the normal probability density of the modal
Every measurement is evaluated whether or not the displacements, irrespective of the frequency information, shows
corresponding structure is nominally linear, weakly nonlinear many clips on either plus or minus one. Otherwise, linearity
or strongly nonlinear. The distinction between these categories will be the case if the distributions line up diagonally. For
is somewhat open to interpretation and on how the results are to this purpose, normal probability density functions of the modal
be used. Deviation from linear behavior may be an essential displacements of the structure have been plotted in Figs. 7a
information and the type of nonlinearity may be important and 7b for NE–SW and NW–SE directions, respectively.
for further analysis. Linear behavior is uniquely defined but Distribution of the station 3 seems identical to the clipped
nonlinear behavior is not, therefore the definition of weak or modal displacements of the zero memory nonlinear system [19]
2174 K. Beyen / Engineering Structures 30 (2008) 2165–2184
Table 5b
Descriptions of first ten deformed shapes of the structure under the Kocaeli earthquake for both directions
Fig. 4a. Transfer functions and the phase histories of the Structural points estimated from SDOF (dashed) and MDOF (solid) systems for one set of ambient
measurement for NE–SW.
since the normal probability density function of station 3 at the station 3 presuming a serious anomaly and reasons
exceptionally clips at plus and minus one when compared need to be found. All analysis, namely spectral, parametric
with others. Such a distribution shows that the maximum SDOF and MDOF models yield almost same result for station
modal displacements over all the modal frequencies generated 3 alarming seriously in SW–NE direction. Similar densely
K. Beyen / Engineering Structures 30 (2008) 2165–2184 2175
Fig. 4b. Transfer functions and the phase histories of the Structural points estimated from SDOF (dashed) and MDOF (solid) systems for one set of ambient
measurement for NW–SE.
Fig. 5. All transfer functions of the observation points for one dataset (a) for NE–SW, (b) for NW–SE directions estimated from the analysis of the MDOF system.
At the first modal frequency, four stations located at the crowns of the main arches have identical response except station 3 with less peak frequency and magnitude.
distributed normalized modal displacements with a positive 4. Structural response to the 17 August Kocaeli earthquake
unit value is seen for station 1 for NE–SW direction. It is
arguable, but station 1 also needs visual and instrumental This section aims to (a) estimate the structural responses to
inspection to explain the reason. As a result, normal probability the recent great Turkish earthquake, (b) discuss the findings in
density function of the modal displacements of station 3 the light of today’s knowhow and the code issued in 2007 for
with intensively distributed normalized maxima looking like a post-earthquake safety evaluation and (c) identify engineering
clipped system is a promising indice for detecting an anomaly issues that must be dealt with before the next earthquake
on structural member when compared with the others in strikes the city, Istanbul. During this progress, although it is
ambient vibration studies. important to assess site effects, topographic effects and other
2176 K. Beyen / Engineering Structures 30 (2008) 2165–2184
Fig. 6a. SISO transfer functions of the ARX structural models selected from many candidates based on the estimation performance for NW–SE direction.
Fig. 6b. SISO transfer functions of the ARX structural models selected from many candidates based on the estimation performance for SW–NE direction.
Fig. 7a. Normal probability densities of the modal displacements of the structure for NE–SW direction. Clipped modal displacements at mines and plus unit values
are the values normalized with respect to the maxima of each mode and is strongly seen at station 3.
Fig. 7b. Normal probability densities of the modal displacements of the structure for NW–SE direction. Clipped modal displacements at mines and plus unit values
are the values normalized with respect to the maxima of each mode. Such a property is strongly seen at station 3.
possibility. For this reason, in the study the 17 August 1999 site. The maximum acceleration has been observed to be
Kocaeli earthquake is used as an input to the historical structure. statistically dependent on the magnitude of the earthquake.
Corrected main shock is replotted for 100 s-window for the FAT Hence it is dependent on the severity of the seismic source, and
station in Fig. 8. is also highly dependent on the distance to the epicentre, soil
One of the essential parameters used in engineering characteristics and other local site conditions such as subsurface
applications and characterizing the seismic motion is the irregularity and the topographical effects [20]. Fig. 9 shows the
maximum ground acceleration of the recorded data at a given type of relationship that exists between acceleration magnitude,
2178 K. Beyen / Engineering Structures 30 (2008) 2165–2184
Fig. 10. Arias intensity maps for the main shock recorded at different soil
conditions (a) for NS direction, (b) for EW direction.
Fig. 12. Structural displacement responses to the 1999 Kocaeli earthquake for EW direction.
for NE–SW direction are plotted in Fig. 12 as an example. For been estimated by using the MIMO parametric model coded
a perfectly symmetric structure, all these responses therefore in MatLab package [15] equivalent to the MDOF system.
mass partitions are expected to be uncoupled, whereas due Descriptions are given for first 10 modes (i.e. it is not classic
to either asymmetry of the structural system or the damage modes, fundamental deformed shapes under forced vibration)
pattern, off-diagonal elements in stiffness matrices therefore in Table 5b. Identified first five deformed shapes under the
partitions may be coupled and its indications may be observed Kocaeli earthquake are plotted in Fig. 13 for both directions,
visually in the response time histories, depending on the namely NW–SE in left two columns and NE–SW in right two
particular damage as identified by close inspection. However, columns within two subsets, plan view and isometric view.
displacement responses show almost symmetrical behavior Light solid lines are for undeformed structure and dark one is
with same polarizations at each observation point. for deformed shape. Earthquake vibrated deformed shapes look
On the displacement time history, in terms of the identified like the ambient vibrated mode shapes with small differences
damage, there does not strongly seem to be a substantial when Tables 5a and 5b are compared.
difference in global scale between the behaviors of the damaged
and undamaged structural points. Structural movement in the 5. Post-earthquake safety evaluation
history plots shows that the points are almost in same phase
at each level. This may be justified if the difference in the As a standard approach, post-earthquake safety evaluation
damage patterns simply does not strongly show up in the data can be performed by spectral analysis for the design spectrum
and it is suspected that the damage is local and small (i.e. there seismic loading issued by the building code. It has been
seems to be a small shift in the identified frequencies indicating common practice to represent the design ground motions
the damage possibility in Figs. 5(a) and (b) in this case). using macro acceleration response spectra developed through
However, small shift in peak frequency also can be considered statistical processing of raw spectra from a collection of
to be within the noise tolerance for some cases, and hence earthquake time histories recorded on different site conditions.
cannot be directly attributed to a real damage when no real A deficiency of these spectra is that they do not represent true
physical evidence on the structural member is observed. But values of the same probability of exceedance for a specified
it is obvious that such 10% difference in the first fundamental return period of time. Therefore, the probability of exceedance
peak frequency for station 3 indicates a damage as observed. of the spectral value at any specified period is not well-known.
For both the main directions, fundamental deformed shapes Bearing this deficiency in the mind, for major structures,
of the structure under the Kocaeli earthquake have also to develop a site dependent design spectrum including the
K. Beyen / Engineering Structures 30 (2008) 2165–2184 2181
Fig. 13. Identified dominant First five deformed shapes of the structure under the Kocaeli earthquake in isometric and plan views for the EW (first two columns)
and NS (right two columns). Light solid lines are for undeformed and black solid lines for deformed structure.
effect of local soil and site conditions and distance to the event, corresponding return period shows that the probability
nearest faults may be one way to overcome the problem of of the structure being subjected to such an earthquake during
the design spectra. Such a study needs collection of the past its lifetime has high possibility. It is obvious that the available
earthquakes recorded at the vicinity therefore instrumentation data is very limited to perform such analysis. In this study,
is essential. Contrary to this fact, at the site, there is one instead, the 17 August 1999 Kocaeli earthquake, recorded at
recorded earthquake with about 0.2g peak ground acceleration the backyard, is used as an input excitation to the historical
of the moment magnitude of 7.4 earthquake which occurred structure through the structural identification analysis approach
after a certain period of time like a century. For such a larger in order to provide how the structure behaves at large response
2182 K. Beyen / Engineering Structures 30 (2008) 2165–2184
Fig. 14. Elastic response spectra with 5% damping ratios for NS direction (a) PSV and (b) PSA. Left column for structural responses to the 1999 Kocaeli earthquake,
right one for ambient measurements from the structural points.
levels. In Figs. 14(a) and (b), elastic response spectra of a SDOF the recent earthquake data yields their dominant peaks at about
model with 5% damping ratio for the estimated velocity and the same period, 0.5–0.6 s. This fact points out the resonance
acceleration responses as well as the ambient vibration data in condition building up during the earthquake, receiving high
the NS direction are plotted. In Fig. 14(a), PSV values of the kinetic energy. First period of the structure is also about 0.4 and
structural points and the PSV values of the NS component of 0.45 s in each direction attracting high velocity input energy
K. Beyen / Engineering Structures 30 (2008) 2165–2184 2183
6. Conclusions
For both main directions, deformed shapes of the structure I: theoretical discussions. ASCE Journal of Engineering Mechanics 2003;
under both the earthquake vibrations and ambient vibrations 129(5):477–88.
[5] Goodwin Godfrey. Gaspare Fossati Di Morcote and His Brother
also have been estimated by using the MIMO parametric
Giuseppe. In: A history of Ottoman architecture. London; 1971.
model equivalent to the MDOF system. Modeshapes confirm [6] Sakin Orhan. Tarihsel kaynaklarıyla İstanbul Depremleri, İstanbul, 2002
indicative deformed shape at station 3 for the first mode for the reference: Gregoras Nikephoros Historia, 2, p. 694.
NE–SW direction. Descriptions of the first five mode shapes for [7] Farrar CR, Doebling SW, Cornwell PJ, Straser EG. Variability of modal
both directions are summarized in Table 5a. parameters measured on the Alamosa Canyon bridge. In: Proc. 15th
international modal analysis conf. 1997.
Calculated elastic response spectrum of the Fatih record
[8] Beyen Kemal. Structural identification of the fatih mosque damaged
has larger amplitudes at the period of 0.55 s (with broadband during the 1999 Kocaeli earthquake. Soil dynamics and earthquake
characteristic) and larger than the amplitudes of the design engineering [in review].
spectrum with 10% probability of excedance in 50 years (for [9] Ekrem Hakki Ayverdi. Fatih era in Ottoman architecture—Osmanli
new constructions) for North–South direction. These larger mimarisinde Fatih devri: 855–886 (1451–1481):III, Baha Matbaasi,
Istanbul, 1973; p. 356–387 [Turkish].
local peaks at 0.55 s; (1) verify the 1.8 Hz of the fundamental
[10] Godfrey Goodwin. A history of Ottoman architecture. London: Thames
frequency of the local ground and (2) suggest that, as the and Hudson; 1997. p. 121–31 [Reprint of 1971].
minimum requirement for the earthquake resistant design of [11] Ginell WS, Leroy Tolles E. Seismic stabilization of historic adobe
buildings, in accordance with the strengthening applications in structures. JAIC 2000;39(1):147–63.
current seismic design code, the 2% probability of exceedance [12] Ramos LF, Guido de Roeck Paulo B, Lourenço Alfredo, Campos-Costa.
In: Lourenço PB, Roca P, Modena C, Agrawal S, editors, Vibration
of the 5% damped elastic design acceleration spectrum within
based damage identification of masonry structures. Structural analysis of
a period of 50 years is adequate. historical constructions. New Delhi; 2006.
First mode of the structure, 2.4 and 2.5 Hz for NS and EW [13] Durukal Eser, Cimilli Serkan, Erdik Mustafa. Dynamic response of two
directions respectively, falls into the range of the dominant historical monuments in Istanbul deduced from the recordings of Kocaeli
broadband period of the seismic loading. Such a match points and Düzce earthquakes. Bulletin of the Seismological Society of America
2003;93(2):694–712.
out the resonance possibility, which might have occurred during
[14] Sakin Orhan. Tarihsel kaynaklarıyla İstanbul Depremleri, İstanbul, 2002
the earthquake. The duration of the earthquake and following reference: Gregoras Nikephoros Historia, 2, p. 694.
shocks of many small earthquakes in addition to the resonance [15] Mathworks. Signal processing toolbox for Matlab, Ver. 6.5 Release 13.
might be the key contributors to the damage. Natick (MA): The MathWorks Inc.; 2002.
Many structurally critical points are discussed and under- [16] Juang J-N, Pappa RS. An eigensystem realization algorithm for modal
parameter identification and model reduction. Journal of Guidance,
lined in detail for structural safety. Especially station 3 located
Control, and Dynamics 1985;8(5):620–7.
at the crown of the NW main arch can be prone to intensive [17] Phan MQ. Observer/Kalman filter identification (OKID): step-by-
damage under a new expected earthquake of Istanbul in future step guide and references. In: A Matlab-based system identification
even with a smaller magnitude if compared with the recent 17 software package. Department of Mechanical and Aerospace Engineering,
August 1999 Kocaeli earthquake. Princeton University; 1997.
[18] Juang J-N. Applied System Identification. Englewood Cliffs (NJ 07632):
Prentice Hall, Inc.; 1994.
References [19] Bendat JullusS. Nonlinear system analysis and identification from random
data. Wiley-Interscience; 1990.
[1] Mustafa Erdik. Report on 1999 Kocaeli and Düzce (Turkey) earthquakes, [20] Beyen Kemal, Erdik Mustafa. Two-dimensional nonlinear site response
Bogazici Univ., Kandilli Observatory and Earthquake Research Institute, analysis of Adapazarıplain and predictions inferred from aftershocks of
Department of Earthquake Engineering; 2000. the Kocaeli earthquake of 17 August 1999. Soil Dynamics and Earthquake
[2] Lourenco̧ PB, Krakowiak KJ, Fernandes FM, Ramos LF. Failure analysis Engineering 2004;24:261–79.
of monastery of jeronimos, Lisbon: How to learn from sophisticated [21] Arias A. A measure of earthquake intensity. In: Hansen RJ, editor. Seismic
numerical models. Engineering Failure Analysis 2007;14:280–300. design for nuclear power plants. Cambridge (MA): MIT Press; 1970.
[3] Herndndez S, Romera LE. Computer modelling of the Basilica of Pilar p. 438–83.
in Zaragoza (Spain). In: C.A BREBBIA Wessex Institute of Technology, [22] Trifunac MD, A G Brady. A study on the duration of strong earthquake
UK, editor. Structural studies, repairs and maintenance of heritage motion. BSSA 1975;65(3):581–626.
architecture VIII. 2003. 864 pages. [23] Ministry of Public Works and Settlement Government of Republic of
[4] Lus H, De Angelis M, Betti R, Longman RW. Obtaining physical Turkey, Specification for structures to be built in disaster areas; 6 March
parameters of mechanical systems from identified state space models. Part 2007 [Turkish].