Analytical Vulnerability Assessment of the Moderately Damaged RC Building Stock after 1998 Ceyhan , Turkey EQ

Taking 145 lives and causing extensive damages to structures, 1998 Ceyhan earthquake occurred on June 27 with Ms 6.2 in the southern province Adana of Turkey hitting major districts and their vicinities. Most of the observed structural damages were assessed in traditional rural buildings, however many recently built multi-story residential buildings and industrial facilities also suffered from heavy damages or even collapsed. This study is an attempt to investigate a building stock consisting of 22 low -to mid-rise reinforced concrete buildings which experienced moderate damage during this shake. These buildings have been inspected and their detailed site investigations had been conducted by ITU-Building and Earthquake Application and Research Center. In order to perform an analytical study for this building ensemble, nonlinear dynamic and static analyses are carried out utilizing DRAIN-2DX computer program. A set of artificial ground motions is employed during the analyses and the demand parameters in terms of different structural response parameters are obtained. Finally, evaluations are performed in order to designate the boundaries for moderate damage grading in terms of specific structural demand parameters and figure out the failure mechanism of the investigated moderately damaged buildings.


INTRODUCTION
Over the past quarter-century, Turkey was exposed to several moderate to severe destructive earthquakes which resulted in significant amounts of casualties and economic losses.Those events indicated that this poor in stiffness and strength building stock, mostly consisting of low -and mid-Rise reinforced Concrete (RC) structures, have experienced severe damages and even partial or total collapse.The damages to existing RC buildings can be attributed to some deficiencies that increase the damage potential of these buildings: • Poor reinforcement detailing practice • Lack of confinement zones • Heavy and large-span cantilevers • Indirect supporting preventing the formation of regular structural frames Therefore, the site-observed damages has emphasized the need for a detailed assessment of the existing stock in order to predict the probable damages during the future earthquakes.One of the recent major earthquakes of Turkey was the Ceyhan earthquake that struck the southern region of Turkey on 27 June 1998 with an epicenter located between downtown Adana Province and Ceyhan District about 30 km north of the coast of the Mediterranean Sea.Although the focal distance to Ceyhan was 32 km, fault rupture was directed towards Ceyhan, causing heavy damage in the city as a result of the strong motion with a Peak Ground Acceleration (PGA) of 0.28 g (Sucuoğlu et al., 2000).Official estimates placed the death toll at about 145 and the number of injured at about 1500 (Ulusay et al., 2000).Site-assessments demonstrated that remarkable structural damages occurred in traditional rural buildings, however many recently built multi-story residential buildings and industrial facilities also suffered from heavy damages or even collapsed (Wenk et al., 1998).The most affected settlement areas are announced to be Ceyhan and Yüreğir districts with a total of 1025 buildings experiencing total collapse or heavy damages beyond repair.Within these two districts, 51% of the moderately damaged 15801 buildings were located in Yüreğir district and its villages (Tezcan and Boduroğlu, 1998).Amongst them, 22 low-to mid-rise RC buildings from downtown Yüreğir for which the detailed investigations are carried out by ITU-Building and Earthquake Application and Research Center teams, are taken into consideration in this study (Fig. 1).
In order to investigate the damage state of these buildings selected to have limited torsional irregularity, In order to demonstrate the damage ranges and the failure mechanisms of the buildings, statistical evaluations are performed oriented to the damage state of the building ensemble.Finally, upper and lower bounds to estimate the probable moderate damage region for shear critical and ductile buildings are proposed.

Analytical modelling of the building ensemble
Twenty two moderately damaged buildings with various story numbers are selected from the archives of ITU-Building and Earthquake Application and Research Center, which were previously investigated by means of their structural layout plans, existing reinforcement details and material qualities during site assessments.Table 1 tabulates some characteristics of these buildings, where T 1 is the dominant vibration period in both directions for the undamaged state and is the total weight of the building.
In the entire set, number of stories for the buildings varies from 3-to 6-therefore the set consists of low and mid-rise structures.All buildings with frame App.Sci. Eng. Technol., 10(10): 1236-1244, 2015 1237 earthquake, the settlement of the record station and Yüreğir district analytical studies are performed in terms of nonlinear dynamic and static analyses.Planar structural models are established by considering the contribution of non-2DX (Prakash et al., 1993) computer program is utilized during the computations.A set of artificial ground motions consisting of 7 earthquakes which are compatible with the design spectrum for 0.40 g effective acceleration coefficient 2 soil class defined in the Turkish Earthquake 07) is employed during the analyses.The structural demand parameters story displacement ratio and base shear ratio are obtained for the earthquake ensemble and after defining the base shear capacities of each building; the Ratios (DCR) are calculated.In order to demonstrate the damage ranges and the failure mechanisms of the buildings, statistical evaluations are performed oriented to the damage state of the building ensemble.Finally, upper and lower e probable moderate damage region for shear critical and ductile buildings are

Analytical modelling of the building ensemble:
moderately damaged buildings with various story numbers are selected from the archives of Building and Earthquake Application and Research Center, which were previously investigated by means of their structural layout plans, existing ils and material qualities during site assessments.Table 1 tabulates some characteristics of is the dominant vibration period in both directions for the undamaged state and W , number of stories for the buildings therefore the set consists of lowrise structures.All buildings with frame structural systems are investigated in details by means of structural material quality, reinforcement amount and detailing of bars and local site conditions.compressive strength is found to be f ck = 7.0~16.5MPa with an average value of 11.5 MPa for the ensemble, while the reinforcing steel is S220 (characteristic yield strength f yk = 22 the entire stock.The distribution of the concrete strengths and the number of floors of the ensemble is given in Fig. 2. The modified version by Ascheim (2005) of DRAIN-2DX computer program is utilized for the analysis, because this version is capable of handling the stiffness and strength degradation through Takeda hysteretic model.Nonlinear behavior of columns and beams is assumed to be represented by a stiffness degrading hysteresis model and element type DRAIN-2DX program is assigned.For the entire building ensemble, a structural damping of 5% and a strain hardening of 3% are taken into account.The structural model for non-structural partition walls is established as in Al-Chaar and Lamb (2002), where walls are modeled by two diagonal compression struts as shown in Fig. 3 having the effective width of defined by Eq. ( 1) and element type the analytical model of the infill walls: 1), t is the thickness; length and θ is the diagonal slope angle of the partition wall.Considering the wall height as the moment of inertia of the neighboring columns as and the modulus of elasticity for the partition composed of hallow bricks as E w structural systems are investigated in details by means of structural material quality, reinforcement amount and detailing of bars and local site conditions.Concrete compressive strength is found to be varying between = 7.0~16.5MPa with an average value of 11.5 MPa for the ensemble, while the reinforcing steel is S220 = 220 MPa) class for the entire stock.The distribution of the concrete strengths and the number of floors of the ensemble is The modified version by Ascheim (2005) of 2DX computer program is utilized for the ion is capable of handling the stiffness and strength degradation through Takeda hysteretic model.Nonlinear behavior of columns and beams is assumed to be represented by a stiffness degrading hysteresis model and element type-7 of igned.For the entire building ensemble, a structural damping of 5% and a strain hardening of 3% are taken into account.The structural partition walls is Chaar and Lamb (2002), where iagonal compression struts as shown in Fig. 3 having the effective width of a w defined by Eq. ( 1) and element type-9 is assigned for the analytical model of the infill walls: is the thickness; r w is the diagonal is the diagonal slope angle of the partition wall.Considering the wall height as h w ; story height H; the moment of inertia of the neighboring columns as I c and the modulus of elasticity for the partition walls w = 1000 MPa and  Henceforth, strut widths a w are computed for each nonstructural wall by Eq. ( 1).
For the entire building ensemble, the structural system is modeled as planar frames connected to each other with elastic tension/compression link elements representing rigid diaphragm effect and element type is assigned for these elements.
Figure 4 shows the backbone curves employed to represent for the hysteretic behavior of the structural and non-structural elements during the nonlinear dynamic analysis.In the left side of Fig represent the yield and ultimate moment capacities and χ y and χ u are the corresponding curvatures.initial stiffness; k 2 and k 3 are the postin the positive and negative directions, where strain-hardening ratio and k 4 is the unloading stiffness.In the model for the partition walls shown in the right side of the same figure, k 1 represents the axial stiffness of the un-cracked section and u 1 and u 2 and ultimate deformations, respectively.
Ground motion ensemble: Utilizing TARSCTHS (Papageorgiou et al., 2000) computer program which is capable of simulating earthquake motions compatible with a pre-defined target spectrum, a set of artificial ground motions is achieved by generating 7 acceleration-time histories.According to the TERDC, if a set of 7 strong motions or more are used, then the mean values of the structural responses are permitted to represent the structural demands.App.Sci. Eng. Technol., 10(10): 1236-1244, 2015 1239 varying between 0~16.5 MPa, respectively.
are computed for each For the entire building ensemble, the structural system is modeled as planar frames connected to each other with elastic tension/compression link elements agm effect and element type-9 Figure 4 shows the backbone curves employed to represent for the hysteretic behavior of the structural structural elements during the nonlinear side of Fig. 4, M y and M u represent the yield and ultimate moment capacities and are the corresponding curvatures.k 1 is the -yield stiffnesses in the positive and negative directions, where α is the is the unloading stiffness.In the model for the partition walls shown in the right represents the axial stiffness 2 are the cracking and ultimate deformations, respectively.Utilizing TARSCTHS ., 2000) computer program which is capable of simulating earthquake motions compatible defined target spectrum, a set of artificial ground motions is achieved by generating 7 time histories.According to the TERDC, if a set of 7 strong motions or more are used, then the mean values of the structural responses are permitted to Either recorded or simulated earthquake motions should satisfy the following conditions: • The duration of strong motion acceleration record shall neither be less than 5 times the first natural vibration period of the building nor less than 15 sec.• Mean value of spectral acceleration for each recorded or simulated acceleration record with 5% damping ratio at zero period ( less than the spectral acceleration is the effective ground acceleration coefficient and g is the acceleration of gravity.• Mean of the spectral acceleration values com for each recorded or simulated acceleration record with 5% damping ratio within a period range of 0.2T 1 ~2T 1 , where T 1 is the dominant vibration period of the structure, shall not be less than 90% of the elastic spectral acceleration When the locations and local site conditions of the investigated moderately damaged buildings are considered, the effective ground A 0 = 0.40 g for seismic zone-1 and of Z2 with characteristic periods and T B = 0.40 sec are taken into account.The artificial 7 strong ground motions, each with durations of 25 sec, are generated compatible with the design spectrum with a return period of 475 years as defined in the TERDC, which has the probability of exceedance of 10% with Table 2: Characteristic properties of the earthquake ensemble strong ground motions ----------------------------------------------------------------------------------------SimEQ- Either recorded or simulated earthquake motions the following conditions: The duration of strong motion acceleration record shall neither be less than 5 times the first natural vibration period of the building nor less than 15 Mean value of spectral acceleration for each leration record with 5% damping ratio at zero period (T = 0) shall not be less than the spectral acceleration A 0 ×g; where A 0 is the effective ground acceleration coefficient and Mean of the spectral acceleration values computed for each recorded or simulated acceleration record with 5% damping ratio within a period range of is the dominant vibration period of the structure, shall not be less than 90% of the elastic spectral acceleration S ae (T).
When the locations and local site conditions of the investigated moderately damaged buildings are ground acceleration of a local soil class periods of T A = 0.15 sec are taken into account.The artificial 7 strong ground motions, each with durations of 25 sec, are generated compatible with the design spectrum with a return period of 475 years as defined in the TERDC, which has the probability of exceedance of 10% within ------------------------------------------------------------------------------------------------------- 2 summarizes the engineering parameters; such as peak ground acceleration strong motion duration t SM defined as the duration for the accelerations greater than 5% of g; effective duration t eff calculated as the time interval between the 5 to 95% of Arias Intensity curve.As it can be observed from Table 2, the duration of the strong motion is calculated as 18.54 which is consistent with condition (i) above.Mean of the PGA is computed as 417.2 cm/sec 2 .As well known, this value corresponds to the zero period spectral acceleration which satisfies the above condition (ii) with a design spectral acceleration of 0.40 cm/sec 2 ) for seismic zone-1.Condition (iii) is also satisfied depending on the dominant vibration periods of the buildings which are already given in Table 1.
Figure 5 shows the acceleration response spectra for the earthquake ensemble compared with the 'target' design spectrum for a 5% damping ratio.Mean tripartite elastic response spectrum, as well as the corresponding acceleration, velocity and displacement sensitive spectral regions and separating periods with values T c = 0.46 sec for acceleration to velocity sensitive regions and T d = 3.67 sec displacement sensitive regions, are also shown in the same Fig. 5.

EVALUATION OF THE ANALYTICAL RESULTS
As it is targeted to figure out the upper and the lower bounds for defining the (probable) moderate damage range to existing RC buildings during a future earthquake event, the selected moderately damaged buildings are also analytically investigated.Although each of the 22 buildings experienced this damage rank during a specific shake with a PGA of 0.28 g; a detailed vulnerability analysis is carried out by scaling the PGA values of the 7 artificial motions to 15 different levels.
App. Sci. Eng. Technol., 10(10): 1236-1244, 20151240 Elastic response spectra of ground motions and the design spectra in the TERDC (left) and mean tripartite elastic response spectra of the earthquake ensemble (right) years.Table 2 summarizes the engineering parameters; such as peak ground acceleration PGA; defined as the duration for the accelerations greater than 5% of g; effective calculated as the time interval between the As it can be observed from Table 2, the duration of the strong motion is calculated as 18.54 sec averagely, which is consistent with condition (i) above.Mean of .As well known, this value corresponds to the zero period spectral acceleration which satisfies the above condition (ii) with a design spectral acceleration of 0.40 g (= 392.3 1. Condition (iii) is also isfied depending on the dominant vibration periods of the buildings which are already given in Table 1.
Figure 5 shows the acceleration response spectra for the earthquake ensemble compared with the 'target' design spectrum for a 5% damping ratio.Mean partite elastic response spectrum, as well as the corresponding acceleration, velocity and displacement sensitive spectral regions and separating periods with for acceleration to velocity for velocity to displacement sensitive regions, are also shown in the

EVALUATION OF THE ANALYTICAL RESULTS
As it is targeted to figure out the upper and the lower bounds for defining the (probable) moderate buildings during a future earthquake event, the selected moderately damaged buildings are also analytically investigated.Although each of the 22 buildings experienced this damage rank during a specific shake with a PGA of 0.28 g; a detailed analysis is carried out by scaling the PGA values of the 7 artificial motions to 15 different levels.
Starting from 0.1 g and scaling the PGA up to 0.8 g for each ground motion with an increment of 0.05 g, 4620 nonlinear dynamic analyses (7 buildings, 2 directions) are achieved for the building ensemble.After the calculation of structural responses for each building subjected to each scaled ground motion, the demand parameters in terms of top story displacement ratio (U top /H) and base shear are collected from the outputs, which are summarized in Fig. 6.Furthermore, for determining the Demand/Capacity Ratios (DCR), base shear capacities (V r ) are also calculated by conducting Eq. (2): 0.2 for each building, where, V r = The shear force resisting capacity f ck = The characteristic compression concrete A c = The effective shear area of resist the shear force From Fig. 6, it can be seen that the base shear ratio is approximately varying between 0.9 to 7.9% for the weakest shake with 0.1 g of PGA and is varying between 3.5 to 17.3% for the strongest shake with a PGA of 0.8 g.Additionally, top story drift ratio is varying between 1.2 to 3.9% and 11.7 to 39.1% for the weakest and the strongest shakes, respectively.On the other hand, number 8, 10 and 43 buildings exhibit distinct behavior from the rest due to their structural system which has locally extremely poor design and average quality, or rather regular and better quality, respectively.
Results in terms of top story drift ratios for both directions of the buildings in the ensemble under the effect of the earthquake ground motions scaled from 0.1 to 0.8 g is given in Fig. 7.In the figure, red curve represents the mean distribution for each PGA level tripartite elastic response Starting from 0.1 g and scaling the PGA up to 0.8 g for each ground motion with an increment of 0.05 g, 4620 nonlinear dynamic analyses (7×15 motions, 22 buildings, 2 directions) are achieved for the building ensemble.After the calculation of structural responses for each building subjected to each scaled ground motion, the demand parameters in terms of top story ) and base shear ratio (V b /W) are collected from the outputs, which are summarized in Fig. 6.Furthermore, for determining the Demand/Capacity Ratios (DCR), base shear capacities ) are also calculated by conducting Eq. ( 2): (2) The shear force resisting capacity compression strength of area of the columns that 6, it can be seen that the base shear ratio is approximately varying between 0.9 to 7.9% for the weakest shake with 0.1 g of PGA and is varying between 3.5 to 17.3% for the strongest shake with a PGA of 0.8 g.Additionally, top story drift ratio is g between 1.2 to 3.9% and 11.7 to 39.1% for the weakest and the strongest shakes, respectively.On the 10 and 43 buildings, respectively exhibit distinct behavior from the rest due to their structural system which has locally RC shear walls, extremely poor design and average quality, or rather regular and better quality, respectively.
Results in terms of top story drift ratios for both directions of the buildings in the ensemble under the otions scaled from 0.1 is given in Fig. 7.In the figure, red curve represents the mean distribution for each PGA level  building's results obviously differ from the ensemble, the results of the building are ignored during the evaluation of statistical parameters such as mean and standard deviation values.Furthermore, the analytical results for x-x direction of 08-MD-ST5 building also are not taken into consideration in consequence of conspicuous behavior of the building under the effect of the earthquakes.This finding is comprehensible when we consider the structural system of the building which has locally RC shear walls in the x-x direction at basement level.
Within the scope of representing the failure mechanism of the building stock, a parameter called Demand/Capacity Ratio (DCR) is defined in this research considering the base shear demands under the effect of ground motions and the obtained base shear capacities.Figure 8 illustrates the results in terms of DCR parameter for each structural direction of the buildings under the effect of scaled ground motions.Similarly, the colored curves represent the same values explained in the previous paragraph.
In accordance with the statistical works done for top story drift ratio parameter, the results of the buildings signified above are not considered during the determination of mean and standard deviation parameters for DCR parameter as well.Furthermore, although 10-MD-ST3 building's shear capacity does not vary among the other buildings in the ensemble, base shear demands are found to be in smaller ranks.For this reason, the results of this building are also ignored during the calculation of statistical parameters.
As shown in Fig. 1, the location of Yüreğir district in which the buildings are settled, is approximately 26 km away from the epicenter of the earthquake.The record station is 32 km away from the epicenter and it can be assumed that the peak ground acceleration values might be similar at a distance of 26 km.However it is more likely that the PGA values might be amplified due to the local site conditions, since the district settles near Seyhan River and is very close to this river's reservoir lake.Within this context, it can be suggested that the PGA value that had effected the building stock might vary between the ranges of 0.25~0.45g.In another aspect of a similar research conducted by Booth et al. (2004), where a building database composed of Turkish low-rise and mid-rise RC frame structures which have experienced moderate damages during 1999 Kocaeli Earthquake, it is argued that the PGA values might be between the ranges 0.35~0.55g to cause moderate damage for the investigated buildings.
The obtained ranges for top story drift ratio and demand/capacity ratio parameters that represent the moderate damage level are tabulated in Table 3.The top story drift ratio range is definitely consistent with the interval given in Booth et al. (2004).
In order to figure out the failure mechanisms of the structures in the building database, nonlinear static pushover analyses are performed for each direction of the building assuming that lateral load distribution is the first mode loading pattern and the structural capacities are calculated.Using the bi-linearization technique given in FEMA-356 (2000) for the obtained capacity curves, yielding forces (V y ) are determined for each structure in the ensemble.As it is expected, flexural failure is observed in the buildings that have V r /V y ratio greater than or equal to 1, while shear failure is observed in the buildings that have V r /V y ratio smaller than 1 when the analyses results are taken into consideration by means of the stated intervals in Table 3.The V r /V y distributions of the shear critical and ductile buildings in the stock is shown in Fig. 9.
Depending on the analyses results of the building ensemble, if we consider the lower bound of the "Moderate Damage" zone which is equal to 0.93% in terms of top story drift ratio, it is observed that 64% of the building stock is exposed to shear failure even under the effect of slight to moderate intensity earthquakes.

CONCLUSION
The significant amounts of casualties, injures and economic losses during the devastating earthquakes of Turkey over the last quarter-century are regrettably far too much when compared to those for similar intensity earthquakes recorded elsewhere in the world.Siteobserved damages after the earthquakes indicate that main structural design and construction mistakes cause partial or total collapse of RC buildings in Turkey.Therefore a detailed and a fast estimation of the building stock in order to predict the probable damage during future earthquakes is required by means of organizing the emergency management services efficiently.
Within the scope of this research, 22 RC buildings which were moderately damaged during the 27 June 1998 Adana-Ceyhan Earthquake (M s 6.2) are investigated analytically by means of the site-observed devastations.Nonlinear dynamic and static analyses are carried out utilizing DRAIN-2DX computer program and a set of artificial ground motions is employed during nonlinear dynamic time history analyses by scaling the PGA values of the earthquake ensemble with an increment of 0.05 g from 0.1 to 0.8 g.

Fig. 1 :
Fig. 1: The epicenter of 1998 Ceyhan earthquake analytical studies are performed in terms of nonlinear dynamic and static analyses.Planar structural models are established by considering the contribution of non structural walls and DRAIN-2DX (Prakash computer program is utilized during th A set of artificial ground motions consisting of 7 earthquakes which are compatible with the design spectrum for 0.40 g effective acceleration coefficient and Z2 soil class defined in the Turkish Earthquake Resistant Design Code (TERDC) (2007) is employed during the analyses.The structural demand parameters in terms of top-story displacement ratio and base shear ratio are obtained for the earthquake ensemble and after defining the base shear capacities of each building; the Demand/Capacity Ratios (DCR) are calculated.In order to demonstrate the damage ranges and the failure mechanisms of the buildings, statistical evaluations are performed oriented to the damage state of the building ensemble.Finally, upper and lower bounds to estimate the probable moderate damage region for shear critical and ductile buildings are proposed.

Fig. 2 :
Fig. 2: The distribution of the concrete classes and story numbers of the building ensemble

Fig. 5 :
Fig. 5: Elastic response spectra of ground motions and the design spectra in the TERDC (left) spectra of the earthquake ensemble (right) 50 years.Table2summarizes the engineering parameters; such as peak ground acceleration strong motion duration t SM defined as the duration for the accelerations greater than 5% of g; effective duration t eff calculated as the time interval between the 5 to 95% of Arias Intensity curve.As it can be observed from Table2, the duration of the strong motion is calculated as 18.54 which is consistent with condition (i) above.Mean of the PGA is computed as 417.2 cm/sec 2 .As well known, this value corresponds to the zero period spectral acceleration which satisfies the above condition (ii) with a design spectral acceleration of 0.40 cm/sec 2 ) for seismic zone-1.Condition (iii) is also satisfied depending on the dominant vibration periods of the buildings which are already given in Table1.Figure5shows the acceleration response spectra for the earthquake ensemble compared with the 'target' design spectrum for a 5% damping ratio.Mean tripartite elastic response spectrum, as well as the corresponding acceleration, velocity and displacement sensitive spectral regions and separating periods with values T c = 0.46 sec for acceleration to velocity sensitive regions and T d = 3.67 sec displacement sensitive regions, are also shown in the same Fig.5.

Fig. 6 :
Fig. 6: Base shear ratio versus top story drift ratio distributions of the buildings

Table 2 :
Characteristic properties of the earthquake ensemble Artificial strong ground motions

Table 3 :
Top story drift ratio and demand/capacity ratio ranges for