EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS Earthquake Engng Struct. Dyn. 2006; 35:1187–1202 Published online 12 April 2006 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/eqe.572
Seismic vulnerability assessment using regional empirical data Ahmet Yakut∗; † , Guney Ozcebe and M. Semih Yucemen Department of Civil Engineering; Earthquake Engineering Research Center; Middle East Technical University; 06531 Ankara; Turkey
SUMMARY This article presents a procedure developed for the seismic performance assessment of low- to mid-rise reinforced concrete buildings in Turkey. The past performance of reinforced concrete buildings during major earthquakes have been compiled and analysed comprehensively using statistical procedures in order to study the empirical correlation between the signicant damage inducing parameters and the observed damage. A damage database of nearly 500 representative buildings experiencing the 1999 Kocaeli and Duzce earthquakes have been used and discriminant functions expressing damage score in terms of six damage inducing parameters have been developed. In order to extrapolate the procedure to other regions that are likely to be subjected to major earthquakes a new approach that takes into account dierent local soil conditions, site-to-source distance and the magnitude of the earthquake has been introduced. The procedure has been applied to a pilot area in Istanbul to estimate expected damage distribution under a credible scenario earthquake. Copyright ? 2006 John Wiley & Sons, Ltd. KEY WORDS:
performance assessment; seismic vulnerability; reinforced concrete; building safety
INTRODUCTION Seismic performance assessment of existing buildings has gained signicant attention during the last decade due to the poor performance of reinforced concrete buildings worldwide. Detailed procedures relying on comprehensive data collection are generally employed for the assessment of individual buildings, as they require sophisticated modelling, analysis and aim at determining whether the building needs rehabilitation. For a building stock, however, preliminary assessment procedures utilizing limited data and simple analysis are preferred because they require less time and thus they are more practical and inexpensive. The majority of the existing procedures developed for the performance assessment of building structures have primarily focused on the structural system, building capacity, layout and certain response parameters [1–6]. These parameters would provide realistic estimates of the expected performance if the as-built features of the structural system were same as the ∗ Correspondence
to: Ahmet Yakut, Department of Civil Engineering, Earthquake Engineering Research Center, Middle East Technical University, 06531 Ankara, Turkey. †E-mail:
[email protected]
Copyright ? 2006 John Wiley & Sons, Ltd.
Received 28 September 2005 Revised 27 December 2005 Accepted 25 January 2006
1188
A. YAKUT, G. OZCEBE AND M. S. YUCEMEN
designed structural and architectural features. In general, the construction practice in Turkey leads to structures that usually have dierent congurations and detailing than the ones shown in the design drawings. Thus for these buildings the common vulnerability assessment procedures do not reect adequately the observed behaviour. Several recent attempts were made to utilize limited eld data in determining the seismic vulnerability, which basically focused on determining the boundaries of demarcations for certain performance levels. The obvious need for procedures that are based mainly on the observed performance had been the motivation for the pioneering work that established the basis for the study presented in this paper [7, 8]. In these studies, a statistical analysis technique, which is called discriminant analysis, was used to develop a preliminary evaluation methodology for assessing seismic vulnerability of existing low- to mid-rise reinforced concrete buildings located in Turkey. The main objective was to identify the buildings that are highly vulnerable to the earthquake eects. Damage scores determined based on certain building attributes were obtained from the derived discriminant functions, and were used to classify existing buildings as buildings in ‘low-’, ‘moderate-’ and ‘high seismic risk group’. The discriminant functions were generated based on six basic damage inducing parameters, namely number of stories (n), minimum normalized lateral stiness index (mnlst), minimum normalized lateral strength index (mnlsi), normalized redundancy score (nrs), soft storey index (ssi) and overhang ratio (or). A detailed denition as well as a comprehensive discussion about the inuence of these parameters on the observed damage is given elsewhere [7, 8]. These parameters that are described next need to be compiled through eld surveys that aim at obtaining the structural layout and plan of the building showing orientation and size of vertical load resisting components as well as relevant architectural features. The number of stories (n) represents the total number of individual oor systems above the ground level. The lateral rigidity of the ground storey, which is usually the most critical storey, represents the lateral stiness of the storey and is taken into account through mnlst which is equal to the minimum of the indexes (Inx ; Iny ) computed for the two orthogonal directions using Equation (1).
Inx Iny
(Icol )x + (Isw )x = × 1000 Atf (Icol )y + (Isw )y = × 1000 Atf
(1)
where (Icol )x and (Icol )y represent the moment of inertias of the columns about the orthogonal x and y axes, respectively. (Isw )x and (Isw )y show the moment of inertias of the structural walls about the x and y axes, respectively. Atf corresponds to total storey area above ground level. Minimum normalized lateral strength index (mnlsi) reects the base shear capacity of the critical storey. The contributions of the columns, structural walls and unreinforced masonry ller walls are considered. The mnlsi is equal to the minimum of the normalized lateral strength indexes calculated in the two orthogonal directions (Anx ; Any ) from Equation (2).
(Asw )x + 0:1 (Amw )x × 1000 Atf (Acol )y + (Asw )y + 0:1 (Amw )y = × 1000 Atf
Anx = Any
(Acol )x +
Copyright ? 2006 John Wiley & Sons, Ltd.
(2)
Earthquake Engng Struct. Dyn. 2006; 35:1187–1202
SEISMIC VULNERABILITY ASSESSMENT
1189
For rectangular columns with longer side along the x-direction, (Acol )x and (Acol )y are taken as equal to the 23 and 13 of the column cross-sectional area, respectively. For square columns, both (Acol )x and (Acol )y are equal to one-half of the cross-sectional area of the column. For structural concrete walls and masonry inll walls, only full areas in the direction of the wall are considered in Equation (2). The normalized redundancy score (nrs) takes into account the degree of the continuity of multiple frame lines to distribute lateral forces throughout the structural system. The normalized redundancy ratio (nrr) of a frame structure is calculated by using the following expression: nrr =
Atr (nfx − 1)(nfy − 1) Agf
(3)
where Atr is the tributary area for a typical column, and is taken as 25 m2 if nfx and nfy are both greater than or equal to 3. In all other cases, Atr is taken as 12:5 m2 . The terms nfx and nfy are the number of continuous frame lines in the critical storey (usually the ground storey) in x and y directions, respectively. Agf is equal to the area of the ground storey, i.e. the footprint area of the building. The value of nrs is 1 if nrr computed from Equation (3) is between 0 and 0.5, and it is taken as 2 if nrr falls into the range 0.5–1.0, for nrr greater than 1, nrs is assigned a value of 3. The soft storey index (ssi) is dened as the ratio of the height of rst storey (i.e. the ground storey), H1 , to the height of the second storey, H2 . A typical feature of the buildings in Turkey is that the area extends out of the frame lines on all sides, dened as the overhang area. This eect is taken into account through overhang ratio (or) computed as the summation of the overhang area of each storey, Aoverhang , divided by the area of the ground storey, Agf . A comprehensive damage database comprising 484 reinforced concrete buildings was compiled in Duzce after the 1999 Duzce earthquake. The database included all the buildings, undamaged as well as damaged, in the studied area. The observed building performances were classied into ve damage groups, as none, light, moderate, severe and collapse. It is quite important to mention that the soil conditions throughout the area surveyed that covers approximately a grid of 4 km × 4 km were uniform (quaternary alluvial deposits) and no soil induced damage such as liquefaction or landslide were observed [9]. Therefore, all buildings in the surveyed area can be considered to have experienced the same level of earthquake intensity. Detailed information about the damage database and the eld study performed for its compilation can be found elsewhere [7, 8]. The central focus of this article is the extension of the statistical vulnerability assessment procedure described in References [7, 8] to other regions that have dierent soil conditions and site-to-source distances than Duzce. An approach based on the attenuation of seismic intensity with distance and the soil type is used along with the spectral displacement that is believed to correlate well with damage. Although the original procedure is based on the data compiled under a xed earthquake magnitude, the modications proposed are valid for other magnitudes at other sites. This procedure has been applied to Zeytinburnu, a district of Istanbul, to assess expected seismic performance of the building stock. For the completeness of the presentation of the proposed extension, a brief summary of the statistical procedure which provides basis for the study presented in this paper is given in the following sections. Copyright ? 2006 John Wiley & Sons, Ltd.
Earthquake Engng Struct. Dyn. 2006; 35:1187–1202
1190
A. YAKUT, G. OZCEBE AND M. S. YUCEMEN
LOCAL STATISTICAL PROCEDURE The damage database employed was modied by reducing the damage states into three; none and light damage states were mapped into one category (N + L), moderate damage level was retained (M) and severe damage and collapse states were combined (S + C). It is possible to evaluate structures at dierent performance levels according to dierent objectives. If the main concern is to identify the buildings that are likely to experience severe damage or collapse, the rst three damage states (i.e. N, L and M) can be considered as one group and the severe damage state and collapse cases as the other group, reducing the distinct damage states into two. Since the main objective is the identication of severely damaged or collapsed buildings for life safety purposes, this classication can be referred to as ‘life safety performance classication’ (LSPC). Similarly, if the main concern is to identify the structures which suer no damage or light damage during an earthquake, the rst two damage states (N and L) can be considered as one group and remaining damage states (M, S and C) as the other group, reducing the distinct damage states again into two. This identication is named as ‘immediate occupancy performance classication’ (IOPC) since the main concern is to identify the buildings that can be occupied immediately after a strong ground motion. The procedure developed using the Duzce damage database is based on damage scores that are determined using two discriminant functions. The unstandardized estimate of discriminant function based on six damage-inducing parameters is obtained for life safety performance classication by utilizing the SPSS [10] software and the database compiled after 1999 Duzce earthquake. Here, DILS denotes the damage index or the damage score corresponding to the LSPC and the other parameters are as described above. DILS = 0:620n − 0:246mnlst − 0:182mnlsi − 0:699nrs + 3:269ssi + 2:728or − 4:905
(4)
In the case of immediate occupancy performance classication, the unstandardized discriminant function, where DIIO is the damage score corresponding to IOPC, based on these variables is DIIO = 0:808n − 0:334mnlst − 0:107mnlsi − 0:687nrs + 0:508ssi + 3:884or − 2:868
(5)
In order to evaluate the performance of the buildings for which damage scores are calculated using the equations derived above, a cuto value needs to be determined. This cuto value is used to rank the vulnerability of buildings by comparing the damage scores computed with the cuto value determined. The cuto functions corresponding to the two types of classication are given in Table I. In this table, LSCVR and IOCVR denote the life safety and immediate occupancy cuto values and n shows the number of stories. In the proposed classication procedure, rst the damage scores are obtained by using Equations (4) and (5) for the cases of LSPC and IOPC, respectively. Then by comparing these damage scores with the storey-dependent cuto values obtained from Table I, the building under evaluation is assigned an indicator variable of ‘0’ or ‘1’. The indicator variable ‘0’ corresponds to none, light or moderate damage in the case of LSPC and none or light damage in the case of IOPC. Similarly, the indicator variable ‘1’ corresponds to severe damage or collapse in the case of LSPC and moderate or severe damage or collapse in the case of IOPC. Copyright ? 2006 John Wiley & Sons, Ltd.
Earthquake Engng Struct. Dyn. 2006; 35:1187–1202
1191
SEISMIC VULNERABILITY ASSESSMENT
Table I. Number of storey-based cuto values (CVR). n
LSCVR
IOCVR
3 4 5 6 7
0.383 0.430 0.495 1.265 1.791
−0:425 −0:609 −0:001 0:889 1:551
In the nal stage of the classication procedure, depending on the values of the indicator variables, the building is rated as either one of the following: • ‘building in low seismic risk group’ (i.e. ‘none or light damage’), • ‘building in high seismic risk group’ (i.e. ‘severe damage or collapse’), or • ‘building in moderate seismic risk group’.
The building is classied as in the ‘low seismic risk group’ if both indicators from LSPC and IOPC are 0, the classication results in ‘building in high seismic risk group’ when both indicators are 1, and for other combinations of the indicators the building is classied as being in the ‘moderate seismic risk group’. For the buildings in the moderate seismic risk group, the nal decision on the seismic safety of the building is left for a more comprehensive seismic evaluation.
EXTENSION OF THE PROCEDURE TO OTHER REGIONS The procedure summarized above is based on the Duzce damage database and therefore is valid for regions that have similar site and source characteristics to that of Duzce. In other words, the magnitude, site-to-source distance, building characteristics and the soil conditions must be similar to that of Duzce. Despite similar construction practice in dierent parts of the country, these constraints lead to limited use of the procedure in other regions having dierent seismicity and site characteristics. For this reason, modications were introduced to the original procedure to allow for its use in other regions of the country. The purpose of the improvement is to capture the relative variation in the ground motion intensity with the magnitude, distance-to-source and the soil type. The spectral displacement (Sd ) value was selected as the damage inducing ground motion parameter, as it is a widely used parameter for expressing the vulnerability of buildings [11–13]. The general trend of a damage curve suggests that the variation of damage with Sd follows the form of an exponential function [14]. This inference is used to link the change in Sd to the change to be imposed on the cuto values obtained for Duzce. The spectral displacement can be obtained from elastic site spectra computed using available attenuation relations. A number of relations, available in the literature, can be employed to relate inelastic spectral displacement to the elastic one [11, 13]. Although the expressions seem quite dierent, their inuence on the cuto modications is shown to be insignicant, especially in the range considered in this study [15]. For this reason, equal displacement rule is considered to be reasonable. Copyright ? 2006 John Wiley & Sons, Ltd.
Earthquake Engng Struct. Dyn. 2006; 35:1187–1202
1192
A. YAKUT, G. OZCEBE AND M. S. YUCEMEN
The proposed procedure is developed on the basis of several assumptions, which are listed below: • Attenuation relations are believed to represent the variation of the ground motion ade-
quately. • Construction practice does not show regional variations. • Damage pattern observed in the reference site would be the same for other sites that
have same distance-to-source and soil type. • The assessed building performance does not include the eect of soil-induced damage
such as liquefaction, foundation settlement and landslide. The steps involved in this procedure can be outlined as follows: Step 1: Obtain site-specic response spectra using an appropriate attenuation model. Step 2: Calculate spectral displacement at the fundamental periods of interest. Step 3: Plot spectral displacement=n as a function of the fundamental period (or n), n representing number of stories considered in the Duzce study. Step 4: Convert spectral displacement to a damage index (cuto value) by assuming an exponential relation. Step 5: Normalize all damage indexes at dierent sites and distances with the damage index obtained for the reference site, i.e. Duzce. Step 6: Modify Duzce cuto values by multiplying with the cuto modication coecients, i.e. normalized values calculated in Step 5. Inclusion of site characteristics Two major parameters used for site classication are the ‘distance-to-source (d)’ and the ‘soil type (ST)’. The sites were characterized by a pair of d and ST bins. Five d bins were selected considering the variation in the response spectra with the distance. ST bins were determined based on the shear wave velocity (Vs ) of the soil types employed by the Turkish Seismic Code [16]. Twenty dierent site classes were obtained from the combination of d and ST bins, which are illustrated in Table II. Note that type B2 represents the reference site (Duzce). This way, any region with a certain d and ST is assigned a site class according to Table II, excluding the sites located farther than 50 km from the source. The number of sites can easily be increased by incorporating other distance ranges and soil types (i.e. Vs ¿1000 m= s). Several attenuation relationships that are suitable for the source mechanism of the North Anatolian Fault (NAF) can be considered. These models are presented in References [17–20] Table II. Site classication. Distance to source (km) Soil type A B C D
Shear wave velocity (m=s)
0–4
5–8
9–15
16–25
26–50
¡200 201–400 401–700 ¿700
A1 B1 C1 D1
A2 B2 C2 D2
A3 B3 C3 D3
A4 B4 C4 D4
A5 B5 C5 D5
Copyright ? 2006 John Wiley & Sons, Ltd.
Earthquake Engng Struct. Dyn. 2006; 35:1187–1202
1193
SEISMIC VULNERABILITY ASSESSMENT
M=7.0, Distance=10 km, Rock site
0.9
Abrahamson and Silva Gulkan and Kalkan Boore et al Ambraseys et al
Spectral Acceleration (g)
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0
0.5
1
1.5
2
Period (s) M=7.0, Distance=50 km, Rock site
0.2
Abrahamson and Silva
0.18
Spectral Acceleration (g)
Gulkan and Kalkan 0.16
Boore et al Ambraseys et al
0.14 0.12 0.1 0.08 0.06 0.04 0.02 0 0
0.5
1
1.5
2
Period (s)
Figure 1. Comparison of response spectra for M = 7:0 at rock sites.
and can be used to provide site-specic response spectra for all 20 sites included in Table II. Boore et al. [17] and Gulkan and Kalkan [18] are the most convenient ones because they use the shear wave velocity directly to account for the soil type. For Abrahamson and Silva [19], however, NEHRP amplication functions need be applied on the rock motion to obtain site response spectra. Among these models, Gulkan and Kalkan’s model was developed based on the local data recorded in Turkey. These models are compared at dierent distances as shown in Figure 1. It is evident from these gures that Gulkan and Kalkan and Boore et al.’s models provide close results. In the analyses, only Boore et al.’s model has been employed because Gulkan and Kalkan’s relationship has jagged behaviour that does not represent a smooth transition between dierent periods. Copyright ? 2006 John Wiley & Sons, Ltd.
Earthquake Engng Struct. Dyn. 2006; 35:1187–1202
1194
A. YAKUT, G. OZCEBE AND M. S. YUCEMEN
Table III. Period versus number of stories for Duzce seismic damage database. Number of stories 2 3 4 5 6
Period (s) 0.275 0.355 0.433 0.504 0.529
In addition to the selected site categories, a range of magnitudes (Mw = 6:0–7.5) to cover moderate and severe earthquakes has been selected. The analyses were carried out for increments of 0.25 to include the intermediate magnitude values. Number of storey and period relationship Since the reference cuto values were obtained as a function of the building height (number of stories), modication factors were also intended for the discrete height levels included in the database. Hence, a relationship between number of stories and the fundamental period was established based on the Turkish Seismic Code formulae [16]. The mean values of the period and the number of stories obtained for the buildings contained in the Duzce seismic damage database are given in Table III. Although the variation and dispersion of the period with number of stories is large for the buildings in the database, this would not signicantly aect the modication factors as will be shown later. Calculation of spectral displacement Spectral displacement is considered to be a reasonable response quantity that correlates well with the expected damage of structures [12, 13]. Spectral displacement is employed here as a ground motion intensity measure; so it reects the variation of the ground motion and the dynamic properties (period of vibration) of the buildings. Assuming that the variations in the spectral ordinates are insignicant within the distance bins that were selected, the spectral displacement values obtained from the calculated spectral accelerations at all periods given in Table III for each of the 20 site classes represent the expected building response. The spectral displacement normalized with number of stories (corresponding to the building period) is plotted against the number of stories as shown in Figure 2. This normalization was done to obtain a similar term that would represent the average drift. The change of Sd with the site class is also evident from these plots. When a linear regression is used to represent data a constant line develops, this is the simplest and the most convenient choice because it leaves out the number of stories. Calculation of modication factors Once Sd values for all sites are computed, they are translated into damage terms. In the vulnerability assessment procedure developed for Duzce, there is a reverse relationship between the cuto value and the damage score of the evaluated building. In other words, as the cuto value is raised the number of buildings in the high seismic risk group decreases. In view of Copyright ? 2006 John Wiley & Sons, Ltd.
Earthquake Engng Struct. Dyn. 2006; 35:1187–1202
1195
SEISMIC VULNERABILITY ASSESSMENT
16.0
y = 2.1306x
14.0
B1
Sd (cm)
12.0
B3
C1
R2 = 0.895
10.0 8.0 6.0 4.0 2.0 0.0 1
0
2
3
4
5
6
7
5
6
7
Number of Stories 3.0 2.5
2.1306 Sd/n
2.0
1.5382 1.5
0.9522
1.0 0.5 0.0 0
1
2
3
4
Number of Stories
Figure 2. Normalized Sd versus number of storey.
this relation, the change of the cuto value (CV) with the normalized spectral displacement was assumed to follow an exponential trend similar to the one between the damage and Sd =n [14]. Although other forms of relationships, such as logarithmic [12, 21] between damage and Sd =n can be employed, authors preferred exponential form that has been observed to provide reasonable results [14]. Thus, the following function is assumed to reect the relation between the CV and the normalized spectral displacement (Sd =n): 1 CV = f (6) 1 − e−Sd =n Since the objective is to obtain cuto modication coecients (CMC) to be applied on the reference cuto values (CVR), the variable of the function in Equation (6) can be used to get CMC values. The CMC values are presented in Tables IV–VI for the magnitudes of 6.5, 7.0 and 7.4. Close inspection of these tables reveal that the CMC value for reference site class B2 is 1.0 because of the normalization with respect to this site. Obviously, at better site conditions and farther distances cuto values should be larger. These CMC values were multiplied with Copyright ? 2006 John Wiley & Sons, Ltd.
Earthquake Engng Struct. Dyn. 2006; 35:1187–1202
1196
A. YAKUT, G. OZCEBE AND M. S. YUCEMEN
Table IV. Cuto modication coecients (CMC) for Mw = 6:5. Distance (km) Vs (m=s)
0–4
5–8
9–15
16–25
26+
0–200 201–400 401–700 701+
0.871 1.115 1.349 1.577
1.012 1.411 1.727 2.116
1.250 1.809 2.396 2.964
1.674 2.577 3.461 4.309
2.469 4.033 5.471 6.832
Table V. Cuto modication coecients (CMC) for Mw = 7:0. Distance (km) Vs (m=s)
0–4
5–8
9–15
16–25
26+
0–200 201–400 401–700 701+
0.802 0.943 1.093 1.246
0.880 1.135 1.351 1.614
1.028 1.403 1.809 2.201
1.310 1.930 2.548 3.138
1.852 2.939 3.945 4.898
Table VI. Cuto modication coecients (CMC) for Mw = 7:4. Distance (km) Vs (m=s)
0–4
5–8
9–15
16–25
26+
0–200 201–400 401–700 701+
0.778 0.864 0.970 1.082
0.824 1.000 1.180 1.360
0.928 1.240 1.530 1.810
1.128 1.642 2.099 2.534
1.538 2.414 3.177 3.900
the respective reference cuto values (given in Table I) to obtain the cuto values for other site classes. Modied cuto values are computed merely from Equation (7), which can handle negative as well as positive values of reference cuto values. CV = CVR + ABS(CVR ) × (CMC − 1)
(7)
The cuto modication coecients for other discrete magnitude values have been obtained using the procedure described above. It is obvious from the results that there is a clear trend between the site-to-source distance, shear wave velocity, magnitude and the CMC values. Variation of the CMC values with the distance and magnitude is illustrated in Figures 3 and 4, respectively for the soil classication employed. The eects of both distance and magnitude have been incorporated through regression analyses that yielded the relationship given in Equation (8). The coecients ci have been obtained from regression of the data obtained for each site class (Vs and d) and magnitude (Mw ) value. The coecients of the equation are presented in Table VII for each soil type listed in Table II. The computed CMC values are plotted against the ones predicted using Equation (8) in Figure 5. A reasonable Copyright ? 2006 John Wiley & Sons, Ltd.
Earthquake Engng Struct. Dyn. 2006; 35:1187–1202
1197
SEISMIC VULNERABILITY ASSESSMENT
6.00 5.00
Type A
Type B
Type C
Type D
y = 1.2877e 0.0374x R 2 = 0.9639
CMC
4.00 3.00 2.00 1.00 Mw =7.0 0.00 0
10
20 Distance, km
30
40
Figure 3. Variation of CMC with distance. 12.00 y = 513.84e-0.6614x R 2 = 0.9957
10.00
Type A
Type B
Type C
Type D
CMC
8.00 6.00 4.00 2.00 d=12 km 0.00 5.5
6
6.5 7 Magnitude, Mw
7.5
8
Figure 4. Variation of CMC with magnitude. Table VII. Coecients of Equation (8) for each soil type. Soil type
Vs (m=s)
c1
c2
c3
c4
A B C D
¡ 201 201–400 401–700 ¿ 700
3.2697 5.4712 7.1840 9.6300
0.4950 0.5945 0.6337 0.6572
0.0565 0.0319 0.0340 0.0348
1.9680 2.3923 2.6023 2.6552
correlation, as reected by the coecient of determination value of R2 = 0:97 was obtained. This equation can be used to obtain the cuto modication coecients for a given site under a given earthquake. The modied cuto values are then used to assess the seismic performance of a given building population by applying the procedure described above. CMC = c1 e(−c2 Mw +c3 d+c4 ) Copyright ? 2006 John Wiley & Sons, Ltd.
(8)
Earthquake Engng Struct. Dyn. 2006; 35:1187–1202
1198
A. YAKUT, G. OZCEBE AND M. S. YUCEMEN
7
Predicted CMC
6 5 4 3 2 1 0 0
1
2
3 4 Computed CMC
5
6
7
Figure 5. Comparison of CMC predictions with computed values.
APPLICATION OF THE PROPOSED PROCEDURE TO ZEYTINBURNU Zeytinburnu, with a population of 240 000 and a building stock of more than 16 000 buildings, is one of the most vulnerable districts of Istanbul. It has been designated by the Municipality as a pilot area for the implementation of an earthquake master plan developed for Istanbul. In this context, Earthquake Engineering Research Center of Middle East Technical University (METU-EERC) led the team of experts in determining the seismic vulnerability of the building stock in Zeytinburnu. First, all buildings within the district boundaries were surveyed by inspection teams through walk down procedures both to obtain building inventory and to rank the buildings according to their apparent attributes that are known to aect seismic vulnerability. The complete building inventory comprising 16 030 buildings revealed that 13 885 buildings were made of reinforced concrete. The remaining buildings were comprised of 853 masonry, 143 steel, 14 wood and 135 mixed-type structures. Since the majority of the buildings were found to be reinforced concrete structures, the assessment of only this construction type is believed to represent the vulnerability of the entire building stock. The classication of the surveyed RC buildings with respect to their number of stories is given in Table VIII. The distribution of normalized redundancy score (nrs) is similar for the Duzce and Zeytinburnu databases. It has been determined as 1 for 67 and 72% of the buildings in Duzce and Zeytinburnu databases, respectively. The nrs took a value of 3 for 15% of the buildings in both databases. The range and average values of the remaining four damage inducing parameters for the subject RC buildings in Zeytinburnu are compared with the ones for Duzce buildings in Table IX. The results in Table IX reveal that the procedure developed using Duzce database is applicable to the buildings in Zeytinburnu because the ranges of damage-inducing parameters in Duzce generally cover the buildings located in Zeytinburnu. The discrepancies between the average values are not crucial because the discriminant functions used in the procedure rely on the range of the parameters rather than their averages. Additionally, not only the construction practice is similar between the two regions, but the structural features of the buildings as reected by the damage-inducing parameters are also similar. The rst level screening has been applied to RC buildings for determining their relative vulnerability under a scenario earthquake of Mw = 7.5 that has been stated to be a credible earthquake to take Copyright ? 2006 John Wiley & Sons, Ltd.
Earthquake Engng Struct. Dyn. 2006; 35:1187–1202
1199
SEISMIC VULNERABILITY ASSESSMENT
Table VIII. Distribution of buildings by height in Zeytinburnu. Number of stories
¡2
3
4
5
6
7
7+
Total
Number of buildings (all RC buildings) Number of buildings (subset of 3036)
1964
1455
2699
4262
2304
1050
151
13 885
9
90
332
929
1147
529
0
3036
Table IX. Damage inducing parameters for Duzce and Zeytinburnu databases. mnlsfti Duzce
Z.burnu
mnlsi Duzce
ssi Z.burnu
Duzce
or Z.burnu
Duzce
Z.burnu
Range 0.012–0.250 0.015–0.150 1.085–4.848 1.003–4.730 0.926–1.560 0.900–1.710 0.012–0.250 0.080–0.400 Average 0.085 0.043 2.428 1.790 1.119 1.115 0.110 0.190
Figure 6. Scenario earthquake on the Marmara Sea Segment of NAF.
place in the Marmara Sea segment of the North Anatolian Fault (NAF) [21, 22]. The expected fault rupture geometry for the expected scenario earthquake and the location of Zeytinburnu are shown in Figure 6. The rst-level assessment recommended a second-level screening to be applied on approximately 3036 buildings, which have been examined thoroughly using the procedure outlined earlier [23]. The Zeytinburnu building statistics for the subset of 3036 buildings are also given in Table VIII. The location of each building in the subset was used to determine its closest distance to the expected fault rupture shown in Figure 6. The soil conditions of Zeytinburnu were obtained from the geotechnical investigations carried out within the context of the pilot implementation of Istanbul Master Plan Project [22]. The assessment of the seismic safety of Zeytinburnu based on the proposed procedure provides a breakdown of the buildings in low-, moderateand high seismic risk groups as presented in Table X per number of stories. These results reveal that around 70% of the buildings are highly vulnerable to the expected earthquake action. The taller buildings have been estimated to be more vulnerable which agrees well with the observations made after recent major earthquakes in Turkey. It is important to note Copyright ? 2006 John Wiley & Sons, Ltd.
Earthquake Engng Struct. Dyn. 2006; 35:1187–1202
1200
A. YAKUT, G. OZCEBE AND M. S. YUCEMEN
Table X. Estimated seismic safety of buildings in Zeytinburnu. High-risk group Number of stories 63 4 5 6 7 Total
Moderate-risk group
Low-risk group
Total
Number
%
Number
%
Number
%
Number
10 180 713 808 387 2098
10 54 77 70 73 69
18 81 170 262 119 650
18 24 18 23 23 21
71 71 46 77 23 288
72 21 5 7 4 10
99 332 929 1147 529 3036
Figure 7. Distribution of buildings in high seismic risk group.
that these buildings were already assessed by rst-level procedures and were classied as vulnerable based on their visual attributes. The spatial distribution of buildings in high seismic risk group is shown in Figure 7. This distribution is strongly aected by the soil conditions, their proximity to the expected fault rupture and the damage inducing parameters that are used in the proposed procedure. CONCLUSIONS A preliminary seismic performance assessment procedure that is applicable to all seismically active regions in Turkey that have similar construction practices has been developed for RC Copyright ? 2006 John Wiley & Sons, Ltd.
Earthquake Engng Struct. Dyn. 2006; 35:1187–1202
SEISMIC VULNERABILITY ASSESSMENT
1201
buildings. This procedure is not applicable to regions where the soil-induced damage such as liquefaction and landslide is expected. The procedure uses data compiled from recent earthquakes in Turkey employing an empirical approach to correlate the observed damage with the selected damage-inducing parameters. A rational approach is introduced to generalize the procedure so that it could be applied to other regions in the country for rapid and reliable seismic performance assessment of a large building stock. The variation of ground motion parameters that have known relationship to the damage of buildings are captured using attenuation models that reect the properties of the sites, i.e. the distance to source and soil type, and that take into account the eect of dierent magnitudes. The procedure can be applied to other regions to identify reliably high-risk areas and vulnerable locations. This would help determine the regional vulnerability and the mitigation priorities, especially for the mega city of Istanbul for which a large earthquake is due. An application of the procedure to Zeytinburnu revealed high-risk areas and the buildings with high risk that need immediate attention. The results have also shown good agreement with the observed performance during the recent earthquakes. REFERENCES 1. Kircher C, Reitherman RK, Whitman RV, Arnold C. Estimation of earthquake losses to buildings. Earthquake Spectra, EERI 1997; 13(4):703–720. 2. Hwang HHM, Huo JR. Generation of hazard-consistent fragility curves for seismic loss estimation studies. Technical Report No. 94-0015, State University of New York at Bualo, 1994. 3. Wen YK, Hwang H, Shinozuka M. Development of reliability-based design criteria for buildings under seismic load. Technical Report No. 94-0023, State University of New York at Bualo, 1994. 4. Hassan AF, Sozen MA. Seismic vulnerability assessment of low-rise buildings in regions with infrequent earthquakes. ACI Structural Journal 1997; 94(1):31–39. 5. Tankut T, Ersoy U. A proposal for the seismic design of low-rise buildings. Turkish Engineering News, Turkish Chamber of Civil Engineers, No. 386, November 1996; 40–43 (in Turkish). 6. Gulkan P, Sozen MA. Procedure for determining seismic vulnerability of building structures. ACI Structural Journal 1999; 96(3):336–342. 7. Yucemen MS, Ozcebe G, Pay AC. Prediction of potential damage due to severe earthquakes. Structural Safety 2004; 26(3):349–366. 8. Ozcebe G, Yucemen MS, Aydogan V. Assessment of seismic vulnerability of existing reinforced concrete buildings. Journal of Earthquake Engineering 2004; 8(5):1–2. Ulusay R, Kumyar H, Tuncay E. Site investigation and engineering evaluation of the Duzce9. Aydan O, Bolu earthquake of November 12, 1999. Report No: TDV=DR09-51, Turkish Earthquake Foundation, Ankara, 2000; 220. 10. SPSS Inc. SPSS Base 11.0 User’s Guide. SPSS Inc.: Chicago, IL, 2001. 11. American Society of Civil Engineers (ASCE). Prestandard and commentary for the seismic rehabilitation of buildings. Report No. FEMA-356, Prepared for the SAC Joint Venture, published by the Federal Emergency Management Agency, Washington, DC, 2000. 12. FEMA-National Institute of Building Sciences. HAZUS-99, Earthquake Loss Estimation Methodology, Washington, DC, 1999. 13. Miranda E. Estimation of inelastic deformation demands of SDOF systems. ASCE Journal of Structural Engineering 2001; 127(9):1005–1012. 14. Erduran E, Yakut A. Drift based damage functions for reinforced concrete columns. Computers and Structures 2004; 82(2–3):121–130. 15. Yakut A, Aydogan V, Ozcebe G, Yucemen MS. Preliminary seismic vulnerability assessment of existing reinforced concrete buildings in Turkey—Part II: inclusion of site characteristics. NATO Workshop, 2003. 16. Turkish Ministry of Public Works and Settlement. Turkish Seismic Code, Ankara, 1997. 17. Boore DM, Joyner WB, Fumal TE. Equations for estimating horizontal response spectra and peak acceleration from Western North American earthquakes: a summary of recent work. Seismological Research Letters 1997; 68(1):128–153. Copyright ? 2006 John Wiley & Sons, Ltd.
Earthquake Engng Struct. Dyn. 2006; 35:1187–1202
1202
A. YAKUT, G. OZCEBE AND M. S. YUCEMEN
18. Gulkan P, Kalkan E. Attenuation modeling of recent earthquakes in Turkey. Journal of Seismology 2002; 6:397–409. 19. Abrahamson NA, Silva WJ. Empirical response spectral attenuation relations for shallow Crustal earthquakes. Seismological Research Letters 1997; 68(1):94–127. 20. Ambraseys NN, Simpson KA, Bommer JJ. Prediction of horizontal response spectra in Europe. Earthquake Engineering and Structural Dynamics 1996; 25:371–400. 21. Japan International Co-operation Agency and Istanbul Metropolitan Municipality. The study on a disaster prevention=mitigation basic plan in Istanbul including seismic microzonation in the Republic of Turkey. Final Report, Tokyo–Istanbul, 2002. 22. Istanbul Metropolitan Municipality Construction Directorate Geotechnical and Earthquake Investigation Department. Earthquake Master Plan for Istanbul, Istanbul, 2003; 1344. 23. Ozcebe G, Sucuoglu H, Yucemen MS, Yakut A. In defense of Zeytinburnu. NATO Workshop on Advances in Earthquake Engineering for Urban Risk Reduction, Istanbul, NATO Science Series IV (66), 95–117, Springer, 2006.
Copyright ? 2006 John Wiley & Sons, Ltd.
Earthquake Engng Struct. Dyn. 2006; 35:1187–1202
