The Maule (Chile) earthquake of February 27, 2010: Development of hazard, site specific ground motions and back-analysis of structures

Soil Dynamics and Earthquake Engineering 42 (2012) 229–245

Contents lists available at SciVerse ScienceDirect

Soil Dynamics and Earthquake Engineering journal homepage: www.elsevier.com/locate/soildyn

The Maule (Chile) earthquake of February 27, 2010: Development of hazard, site specific ground motions and back-analysis of structures Amr S. Elnashai a, Bora Gencturk n,b, Oh-Sung Kwon c, Youssef M.A. Hashash a, Sung Jig Kim d, Seong-Hoon Jeong e, Jazalyn Dukes f a

Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, Newmark Civil Engineering Lab, 205 N Mathews Ave, Urbana, IL 61801, USA Department of Civil and Environmental Engineering, University of Houston, N107 Engineering Building 1, Houston, TX 77204-4003, USA c Department of Civil Engineering, University of Toronto, 35 St. George St., Toronto, ON, Canada, M5S 1A4 d Department of Architectural Engineering, Keimyung University, 1095 Dalgubeoldaero, Dalseo-Gu, Daegu 704-701, South Korea e Department of Architectural Engineering, #2S413, Inha University, 253 Yonghyun-Dong, Nam-Gu Incheon, 402-751, South Korea f School of Civil and Environmental Engineering, Georgia Institute of Technology, 790 Atlantic Drive, Atlanta, GA 30332, USA b

a r t i c l e i n f o

absrat ct

Article history: Received 16 February 2011 Received in revised form 21 February 2012 Accepted 8 June 2012 Available online 13 July 2012

The Maule (Chile) earthquake of 27 February, 2010 has caused severe disruption and economic losses. With a magnitude of 8.8, it has been recorded as one of the largest earthquakes of the last century. The ground motion records from large subduction earthquakes, such as the Chile earthquake, are sparse. The number of accelerograms that recorded the strong ground motion was relatively few and only a few of these ground motions were released to engineering community. One of the objectives of this paper is to develop site specific ground motions that take into account the particular characteristics of this major earthquake. These are proposed to the engineering community as representative ground motions based on the best available data. The second objective of the paper is to investigate, using numerical tools, some typical failures observed in the engineered buildings and bridges. Although, in general engineered structures performed very well and the majority of failures, hence losses, were to non-engineered structures, some repeated deficiencies in structural design were observed. The developed hazard and site specific ground motions are used as inputs for inelastic dynamic analysis of advanced finite element building and bridge models. The results are processed to explain quantitatively the structural deficiencies observed in the field. & 2012 Elsevier Ltd. All rights reserved.

1. Preamble On February 27, 2010 at 03:34 am local time, a powerful earthquake of magnitude 8.8 struck central Chile. The epicenter of the earthquake was approximately 8 km off the central region of the Chilean coast. With an inclined rupture area of more than 80,000 square km that extends onshore, the region of Maule was subjected to a direct hit, with an intense shaking duration of at least 100 s, and peak horizontal and vertical ground acceleration of over 0.6 g. Over 800,000 individuals were directly affected through death, injury and displacement. According to the Ministry of Interior of Chile, the earthquake caused the death of 521 persons, with almost half of the fatalities caused by the consequential tsunami. More than a third of a million buildings were damaged to varying degrees, including several cases of total collapse of major structures. The transportation system was dealt a crippling blow,

n

Corresponding author. Tel.: þ1 713 743 4091; fax: þ 1 713 743 4260. E-mail address: [email protected] (B. Gencturk).

0267-7261/$ - see front matter & 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.soildyn.2012.06.010

with 830 failures registered with the Ministry of Public Works on roads in both the public and private transportation networks. Out of 130 hospitals in the effected region, four became uninhabitable, 12 had greater than 75 percent loss of function, only eight were partially operational after the main shock, and 80 hospitals needed repairs. A total of 4013 schools (representing nearly half of the schools in the affected areas) suffered significant damage. Severe disruption of commerce as well as the rescue and response effort resulted from the damage to roads, embankments, bridges, ports and airports. According to the Ministry of Treasury, the economic losses are estimated to be $30 billion (loss of infrastructure: $20.9 billion, loss of production: $7.6 billion, other costs such as nutrition and debris removal: $1.1 billion) which is equivalent to approximately 17 percent of the GDP of Chile. Only a few acceleration records were released to the engineering community as of January 2011; this withholding of such information of great importance to detailed studies that benefits society at-large is regrettable. The available records confirm the severe shaking that resulted from the earthquake and the long duration of the strong-motion part of the records. The dearth of

230

A.S. Elnashai et al. / Soil Dynamics and Earthquake Engineering 42 (2012) 229–245

available records led the authors to generate spectrum compatible signals. Candidate attenuation relationships that are suitable for large subduction earthquakes are selected and amongst these, the ones that give the best correlation to the available data are identified. The identified attenuation relationships are combined to develop site specific hazard spectra. In order to generate acceleration time histories for back-analysis of a building and two bridges, ground motions are selected from the available digital recordings and spectrum matching is utilized to obtain the site specific time histories at the locations of the back-analysis structures. The objective of conducting back-analysis of structures is to explain some typical failures to engineered structures observed in the field. The selected building is a three story flat-slab reinforced concrete structure and the back-analysis aims to explain the particular damage pattern observed only at the first story columns of the exterior frames of the buildings. The most commonly observed bridge damage during the Chile earthquake was due to unseating or displacement of superstructure, especially for skewed bridges. This resulted in severe yielding of seismic restrainers, which are extensively used in bridge construction in Chile as opposed to other parts of the world, and the destruction of shear keys in abutments. One of the bridge case studies focuses on explaining the effect of seismic restrainers on the bridge performance. And the other bridge case study investigates the reasons for excessive displacements of the superstructure that resulted in unseating and collapse of several bridges. A very typical bridge has been selected for the second case study and seismic fragility functions are also developed using hazard specific ground motions from different parts of Chile.

2. Engineering seismology 2.1. Seismo-tectonic environment The Maule earthquake struck Chile on 27 February 2010 at 03:34 a.m. local time. The magnitude of the earthquake is estimated as Mw 8.8. The epicenter was located offshore at 35.9091S, 72.7331W with the following distances to major cities: Chilla´n 95 km, Concepcio´n 105 km, Talca 115 km, and Santiago 335 km. The hypocenter was 35 km deep [27]. The average slip over the approximately 81,500 km2 rupture area was 5 m, with slip concentrations down-dip, up-dip and southwest, and up-dip and north of the hypocenter. Relatively little slip was observed up-dip/offshore of the hypocenter. The average rupture velocity was estimated to be in the range of 2.0–2.5 km/s. The Global Centroid Moment Tensor (GCMT) solution yielded a seismic moment of 1.84  1022 Nm, a centroid location of 35.951S, 73.151W, and a best double couple fault plane geometry with strike and dip angles of 181, and a rake angle of 1121 [14]. The size of the fault zone varies depending on the calculation method. According to method proposed by the National Earthquake Information Center (NEIC), the size of the fault zone is 189  530 km. The slip amplitude reached 9 m at peak location. As a comparison, the maximum value of slip in the 2004 Sumatra and recent Haiti earthquakes are about 20 m and 5 m respectively. The uplift reached as high as 2 m and settlements of 0.4 m were observed. The coast of Chile moved west, into the ocean as much as 6 m at some locations. The earthquake nucleated on the subduction zone that runs along the entire 5000 km length of the western coastline of South America, known as the Peru–Chile trench. Earthquakes in this region are due to stress buildup resulting from the movement of the oceanic Nazca plate eastward and downward towards the South American plate at a rate of approximately 70 mm per year [25]. Due to the close proximity to the Nazca-South America subduction zone, Chile has long been subjected to earthquakes

Fig. 1. Recent major earthquakes in Chile and seismic gaps, reproduced from [6].

of large magnitude. On average, a magnitude 8.0 earthquake occurs every decade and a magnitude 8.7 earthquake or greater is observed within a century. The map in Fig. 1 shows recent major earthquakes in Chile alongside seismic gaps. The white circles indicate the rupture for individual events, the red circles show the epicenters and the yellow dots are the aftershocks. From north to south, the Peru–Chile trench has ruptured in several incidents except for three locations that are shown with red lines. The segment of the subduction zone, Concepcio´n gap [24] in Fig. 1, between the 1985 Valparaı´so and the 1960 Valdivia earthquakes, which last produced an earthquake in 1835, is the zone that ruptured during the 2010 earthquake. Another segment, the Arica gap has been relatively inactive since 1877 and has the potential to produce an earthquake of magnitude 8.0 to 8.5. There has been energy release with the Valparaı´so (1906 and 1985), La Serena (1943) and Vallenar (1922) earthquakes in the La Serena seismic gap but it is uncertain if all of the energy accumulated in this region has been released with these events. 2.2. Shaking intensity and recorded strong ground motion Ground motions were recorded by two departments at the University of Chile. At the time of writing this paper, digital recordings from 10 stations were available through the

A.S. Elnashai et al. / Soil Dynamics and Earthquake Engineering 42 (2012) 229–245

Seismological Service at the Department of Geophysics. Additionally, plots of recordings (both acceleration time histories and response spectra) from 9 stations were available through Red Nacional de Acelerografos Departamento de Ingenieria Civil (RENADIC). No information was available regarding the soil properties where the stations are located. The recordings are listed in Table 1, along with peak ground acceleration of each component, distances to fault based on different measures and maximum duration of horizontal components. The finite fault model by U.S. Geological Survey [26] is used to calculate the closest distance to the rupture plane. As described earlier, the size of the fault zone is 189  530 km and the fault strike, dip and rake angles are approximately 17.51, 181 and 1121, respectively. The durations are based on the bracketed (with a threshold¼0.05 g) and significant duration definitions. Fig 2(a) shows the surface projection of the fault and the location of the stations along with zones defined by the Chilean seismic code, NCh 433 [8]. Fig 2(b) plots the acceleration time histories and the respective Arias intensities for the station labeled as CCSP. As given in Table 1, the peak ground accelerations (PGA) of the horizontal components from CCSP and MELP stations are 0.65 g and 0.78 g, respectively. These are the highest PGAs recorded during the earthquake. The significant duration of CCSP record, based on Arias intensity, is 67.6 s. The bracketed duration, with a threshold of 0.05 g, is much longer, yielding 113 s of strong ground shaking for the CCSP record. The existing attenuation relationships for strong motion duration [10,22] show that the bracketed duration decreases with increasing distance from the source, while the significant duration increases. This is because the bracketed duration uses the absolute threshold of the amplitude, while the significant duration utilizes the relative threshold and thus, is related to the geometry of the accelerogram, regardless of its absolute amplitude. Therefore, to compare the duration of the strong ground motion from the Chile earthquake, the bracketed durations (with a 0.05 g threshold value) of large earthquake records from the Pacific Earthquake Engineering Research (PEER), Next Generation Attenuation (NGA) project database (http://peer.berkeley.edu/nga) are utilized. Relatively large earthquakes (Mw Z6.9) with a peak ground acceleration of

231

0.1 g or more and source distances of less than 100 km are selected. The durations are compared in Fig. 3, which indicates that the strong ground shaking during the Chile earthquake was significantly longer compared to other earthquakes. The long duration of the Chile earthquake is an important feature that needs to be taken into account in the development of site specific ground motions that are discussed in Section 2.3. The 5 percent damped elastic spectra for the ground motions are shown in Fig. 4. The spectra are separated based on the zones specified in the Chilean seismic code. The recommended design spectra for soil types I to III are also shown in the plots. Note that soil type IV is omitted because none of the records considered here are on such soil. Although soil type information is not available for the recording stations, it is seen that the recorded spectra exceed the design spectra for soil types I and II for the intermediate period range. Additionally, the spectra at stations MAIP, CURI, CCSP and MELP exceed significantly the design spectra for periods shorter than 1 s. For these records, the peak spectral ordinates normalized by individual PGAs are 4.26, 4.06, 3.39, and 3.67, respectively, which indicate the severity of the ground shaking, particularly when compared with the amplification factors from the Chilean seismic code, which vary from 2.76 to 3.09 depending on the soil type. Such a feature could result in relatively high demand imposed on short period structures that are designed to conform to code requirements. Overall, the comparison between the unnormalized spectra of the individual records from the 2010 earthquake and the code spectra suggests a need to amplify the latter. The available data from the 19 stations in Chile are used to select appropriate attenuation relationships for horizontal ground motions. Attenuation relationships that are developed for subduction zones with thrust mechanisms and that utilize a large and uniformly processed database of large magnitude events are selected here. Five candidate attenuation relationships are identified: Zhao et al. [31], Atkinson and Boore [2], Campbell and Bozorgnia [3], Gregor et al. [7], and Youngs et al. [30]. Note that most other attenuation relationships available in literature are not developed to account for large magnitude earthquakes such as the 2010 Chile earthquake. Due to the lack of soil type information for the ground motion records, rock site attenuation relationships are used, to generate acceleration response

Table 1 Information on stations and the recorded strong ground motions. Station

Distance (km)a

PGA (g)

Duration (sec)b

ID

Name

NS

EW

Vert

de

dh

Seismological CCSP CSCH MELP ANTU STL LACH CLCH OLMU SJCH ROC1

service Colegio San Pedro, Concepcio´n Casablanca Melipilla Campus Antumapu, Santiago Cerro Santa Lucia Colegio Las Americas Cerro Cala´n, Santiago Olmue´ San Jose´ de Maipo Cerro El Roble

0.65 0.29 0.57 0.23 0.24 0.31 0.21 0.35 0.47 0.19

0.61 0.33 0.78 0.27 0.34 0.23 0.23 0.25 0.48 0.13

0.58 0.23 0.39 0.17 0.24 0.16 0.11 0.15 0.24 0.11

109.1 311.7 283.0 323.0 334.2 339.1 343.8 353.7 332.5 361.6

114.6 313.6 285.1 324.9 336.0 340.9 345.6 355.4 334.4 363.3

0.0 20.9 0.0 25.3 32.5 39.8 43.1 62.2 49.8 67.9

36.4 48.5 52.5 66.1 69.2 72.9 74.8 78.6 78.8 85.4

RENADIC MMVM CEVM MAIP CURI SRSA UCSA MMSA LTSA VALD

˜ a del Mar (Marga Marga) Vin ˜ a del Mar (Centro) Vin CRS MAIPU RM Hosp. Curico´ Hosp. So´tero de Rı´o Universidad de Chile, Santiago Estacio´n Metro Mirador Santiago Hosp. Luis Tisne RM Hosp. Valdivia

0.35 0.22 0.56 0.47 0.27 0.17 0.24 0.30 0.09

0.34 0.33 0.48 0.41 0.26 0.16 0.17 0.29 0.14

0.26 0.19 0.24 0.20 0.13 0.14 0.13 0.28 0.05

336.7 337.8 321.3 170.4 325.2 331.6 329.5 332.3 437.8

338.5 339.6 323.2 174.0 327.1 333.5 331.3 334.2 439.2

47.0 48.5 19.1 13.0 29.7 30.0 30.3 33.4 182.8

60.8 61.0 64.0 65.1 68.0 68.0 68.2 69.6 192.7

a b

dsp

drup

Db

Ds

113.1 50.3 60.2 45.9 58.9 54.9 49.3 48.9 71.2 34.9

67.6 32.2 31.9 37.7 41.2 37.5 42.4 32.1 38.6 35.5

Distance—de: epicentral distance, dh: Hypocentral distance, dsp: distance to surface projection of the fault, and drup: distance to rupture plane. Duration—Db: bracketed duration with a threshold of 0.05 g, Ds: significant duration (5–95% of Arias Intensity).

unavailable

232

A.S. Elnashai et al. / Soil Dynamics and Earthquake Engineering 42 (2012) 229–245

Acceleration (g)

0.8

PGA: 0.65 g

NS

0.4 0 -0.4 -0.8

Acceleration (g)

0.8 EW

0.4 0 -0.4 PGA: 0.61 g

-0.8

Arias Intensity (%)

Acceleration (g)

0.8 PGA: 0.58 g

0.4

Vert

0 -0.4 -0.8 100

NS EW Vert

50

0

0

20

40

60

80

100

Time (sec) Fig. 2. (a) Fault plane and the seismic zones from the Chilean seismic code and the locations of the stations (yellow and red filled circles indicate records from Seismological Service and RENADIC, respectively); (b) acceleration time histories and Arias intensity at station CCSP. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Bracketed Duration (>0.05g, sec)

120

Cape Mendocino (1992), Mw= 7.0 Chi-Chi, Taiwan (1999), Mw= 7.6

100

Denali, Alaska (2002), Mw= 7.9 Duzce, Turkey (1999), Mw= 7.1

80

Hector Mine (1999), Mw= 7.1 Imperial Valley (1940), Mw= 7.0 Irpinia, Italy (1980), Mw= 6.9

60

Kern County (1952), Mw= 7.4 Kobe, Japan (1995), Mw= 6.9

40

Kocaeli, Turkey (1999), Mw= 7.5 Landers (1992), Mw= 7.3 Loma Prieta (1989), Mw= 6.9

20

Manjil, Iran (1990), Mw= 7.4

0

St Elias, Alaska (1979), Mw= 7.5

0

10

20

30 40 50 60 70 80 Closest Distance to Rupture Plane (km)

90

100

Tabas, Iran (1978), Mw= 7.3 Chile (2010), Mw = 8.8

Fig. 3. Comparison of strong ground shaking during the Chile earthquake with other notable earthquakes.

spectra. Relationships for rock sites are observed to provide the best fit to the available data points (see Fig. 5). The predictions of the candidate attenuation relationships for the various soil classifications are provided in Elnashai et al. [4]. The PGA and spectral acceleration values predicted by the attenuation relationship by Zhao et al. [31] and Campbell and Bozorgnia [3] correlate best with the measured data (see Fig. 5). Therefore, these attenuation relationships are selected to generate the ground motions required for back-analysis of structures. 2.3. Development of suites of ground motion for structural analysis No recordings exist at the locations of the case study structures, therefore, a suite of spectrum compatible records that are

representative of the hazard at the locations of the case study structures are generated. Most of the stations provided by RENADIC are located north of the rupture plane and none of the stations are within 100 km from the case study structures. Additionally, no digital records from RENADIC are available. Thus, the digital recordings available through the Seismological Service at the University of Chile are used as seed signals. Locations of the case study structures and stations are shown in Fig 2(a). The acceleration response spectra which take into account the distance from the fault are considered for generating the spectrum compatible records. Particularly, the selected attenuation relationships are combined accounting for their standard deviation as follows: RSA ¼ RSAz ðmz 7 asz ÞW 1 þRSAc ðmc 7 asc ÞW 2

ð1Þ

A.S. Elnashai et al. / Soil Dynamics and Earthquake Engineering 42 (2012) 229–245

2.5

2.5 ANTU STL

Zone 2 Soil I Soil II Soil III

1.5

CLCH MAIP CURI SRSA

1.5

0.5

1

1.5

2

OLMU SJCH MMVM

1

MMSA LTSA

0.5

MELP

ROC1

UCSA

0

Soil I Soil II Soil III

2

LACH

1

CCSP CSCH

Zone 3

Sa (g)

Sa (g)

2

0

233

CEVM VALD

0.5

2.5

0

3

0

Period (sec)

0.5

1

1.5

2

2.5

3

Period (sec)

Fig. 4. Comparison of measured spectra (5 percent damping) with design spectra from the Chilean seismic code; the thick and thin lines show the NS and EW components respectively.

Zhao et al. (2006)

Atkinson and Boore (2003) Seis Service

Youngs et al. (1997)

Campbell and Bozorgnia (2003)

2 Sa (g) at T=0.20 sec

2

0.5

1 0.5

0.1

0.1

2

2 Sa (g) at T=1.00 sec

Sa (g) at T=0.40 sec

PGA (g)

1

1 0.5

0.1 10

Gregor et al. (2002)

RENADIC

50

100

200

Closest Distance to Rupture Plane (km)

1 0.5

0.1 10

50

100

200

Closest Distance to Rupture Plane (km)

Fig. 5. PGA and response spectra attenuation relationships for rock.

where: RSAz is the attenuation relationship by Zhao et al. [31], RSAc is the attenuation relationship by Campbell and Bozorgnia et al. [3], m is the mean value, a is the weighting constant for standard deviation, here selected as 0, 7 0:5, 71:0 and7 1:5, s is the standard deviation, and W is the weight on the spectral acceleration attenuation relationship, here selected as 0.5. Fig. 6 shows the comparison between the combined attenuation relationship according to Eq. (1) and PGA or spectral values of the recorded ground motions. The attenuation relationship is plotted along with selected standard deviation values of 70:5s, 71:0s and 71:5s. The combined attenuation relationship is used to generate the target spectra at the case study sites. The mean spectra and spectra generated with 70:5s are selected to be used in spectrum matching of the case study records as this range is considered to capture the variation in the strong ground motions from the Chile earthquake.

The spectrum-matched records are generated by using WavGen [20] program. WavGen modifies a given seed record to render it compatible with a given spectrum. As mentioned earlier, the seed records are selected from the ground motions recorded by the Seismology Service at the University of Chile. Table 2 lists the seed records for each site and Fig. 7 shows example spectrum compatible acceleration time histories and alongside the target and matched response spectra.

3. Observed structural damage This paper focuses on the damage to a selected building and two bridges. Therefore, the observed damage to these structural types during the 2010 Chile earthquake is very briefly summarized in this section.

234

A.S. Elnashai et al. / Soil Dynamics and Earthquake Engineering 42 (2012) 229–245

mean ± 1.5σ

mean ± σ

mean ± 0.5σ

mean

0.5

1 0.5

0.1

0.1

2

2 Sa (g) at T=1.00 sec

Sa (g) at T=0.40 sec

PGA (g)

1

1 0.5

0.1 10

RENADIC

2 Sa (g) at T=0.20 sec

2

Seis. Service

50 100 Closest Distance to Rupture Plane (km)

200

1 0.5

0.1 10

50 100 Closest Distance to Rupture Plane (km)

200

Fig. 6. Combined PGA attenuation relationship for rock site and spectral acceleration at different periods.

Table 2 Seed records used for generating synthetic ground motions for back-analysis of the case study structures. Site

Fault distance (km)

Seed records

Odontology building ITATA Bridge Bridge on route (Ruta) 5

37.9 53.4 68.9

CCSP-NS, MELP-NS MELP-NS, STL-EW LACH-NS, STL-EW

3.1. Observed building damage According to the Ministry of Housing and Urban Development of Chile, a total of 370,051 houses were damaged from earthquake and the consequential tsunami. The damage distribution is given in Table 3. The cost of damage to private houses is estimated as $3.7 billion. The adobe construction in Maule region suffered the most damage. In Curico´, 90 percent of adobe construction was destroyed. Failure to engineered buildings was due to the common causes of irregularity and limited ductility, with a few cases of damage due to a special provision in the Chilean code that allows structural walls with thin webs. On the whole, the performance of engineered structures was reasonable, taking into account the magnitude and proximity of the earthquake. A study by Rene Lagos using the building permit statistics from National Institute of Statistics Chile indicates that out of 9974 buildings constructed between 1985 to 2009, only four buildings collapsed and 50 buildings need to be demolished [15]. Less than 2.5 percent of engineered structures in Chile suffered damage and out of all casualties, less than 20 died in engineered buildings. 3.2. Observed bridge damage Considering the large earthquake magnitude and rupture area, the damage to bridge structures was less than that could have been expected. The major highways in Chile are constructed and maintained by private companies. Based on the information obtained from researchers in Chile during the field investigations, the highway network is approximately 2200 km in length and has around 2000 bridges with span lengths longer than 10 m. Among

these structures, only one percent (8 highway and 12 pedestrian bridges) collapsed due to the earthquake. Approximately 100 bridges (50 highway and 50 pedestrian bridges) were damaged to a level requiring repair. According to Yen et al. [29], when shorter-span bridges (excluding culverts and pedestrian bridges) are counted, the number of collapsed or damages bridges is approximately 3% of the total number of bridges in Chile. During the field investigations the authors visited several damaged bridges to study failure modes. Additionally, there had been several teams from other organizations, such as Earthquake Engineering Research Institute (EERI) and Japan Society of Civil Engineers (JSCE), which focused specifically on bridge performance. Table 4 presents a compiled list of damaged bridges, their locations, and failure modes based on the data collected from the field investigation and reports available in the public domain [11,28]. The most commonly observed bridge damage was due to unseating or displacement of superstructure, especially for skewed bridges. Even if the centers of mass and stiffness coincide, skewed bridges tend to develop rotation, which results in unseating of the bridge girder or failure of shear keys. Bridges designed according to relatively recent design practice apparently suffered more damage than bridges constructed in the past. In the old construction practice, the integrity of bridge superstructure was high due to diaphragms connecting the girders. On the other hand, bridges constructed in recent years often did not have diaphragms. As a result of the lack of in-plane stiffness and connectivity, bridge girders were damaged due to the pounding of the superstructure onto the shear keys. In the report from another field investigation team [28], it is stated that the seatwidth on bridge bent was not large enough to prevent unseating. In addition, in many bridges, shear keys were not strong enough to resist the lateral forces from the pounding superstructure. The specifications in the United States require that the shear keys, restrainers, and bearing seat width should be properly designed such that the loss of bridge spans due to unseating can be prevented. Elastomeric bearings that are often used in overcrossing bridges in Chile can only be used in single span bridges in the United States. For multi-span continuous bridges, bridge decks

A.S. Elnashai et al. / Soil Dynamics and Earthquake Engineering 42 (2012) 229–245

1.5

Sa (g)

0.3 0 -0.3

1 0.5

-0.6

0

0.6

1.5

Sa (g)

0.3 0 -0.3

1 0.5

-0.6

0

0.6

1.5

0.3 Sa (g)

Acceleration (g)

Acceleration (g)

Acceleration (g)

0.6

0 -0.3 -0.6

235

0

10

20

30

40 50 Time (sec)

60

70

80

90

1 0.5 0

0

1 2 Period (sec)

3

Fig. 7. Spectrum matched records for back-analysis of the case studies.

Table 3 Housing damage distribution.

Coastal Urban adobe Rural adobe Government housing developments Private housing developments TOTAL

Destroyed Major damage

Minor damage

Total

7931 26,038 24,538 5489

8607 28,153 19,783 15,015

15,384 14,869 22,052 50,955

31,922 69,060 66,373 71,459

17,448

37,356

76,433

131,237

81,444

108,914

179,693

370,051

are connected to bent-caps, which prevents rotation and/or unseating of bridge decks during an earthquake. If more stringent requirements for shear keys and bearing seat width were stipulated in the bridge design specifications in Chile, the damage from unseating of superstructure could have been minimized without significantly increasing the bridge construction cost. Only a few bridges reportedly had damage to their substructure. It is speculated that as there are not many bridges with integral superstructure–substructure connection, the large inertial force from superstructure was not transferred to the substructure. The inertial force caused the unseating or large movement of the superstructure, or the failure of shear keys.

4. Case studies In this section, a building and two bridges are modeled and analyzed using finite element software under the site-specific ground motions described in Section 2.3. The numerical models are based on complete design drawings and measurements obtained during the field investigations. The objective here is to build on the field observations and identify the sources (deficiencies in structural designs) that contributed to the observed

failures. The damage to the building was peculiar in that it was limited to the first story columns of the exterior frames. The damage mainly resulted from the short column effect due to the existence of masonry infill walls which most probably were not taken into account in the design. This is a typical failure mode observed in most earthquakes. One bridge case study aims at explaining the effect of seismic restrainers, which are commonly used in Chile, on the dynamic response and the bridge selected for the other case study exemplifies the most commonly used highway overpass design in Chile in the recent years. Several bridges of similar type suffered damage to different extents in the 2010 Maule earthquake. Seismic fragility functions are also developed for this typical bridge to be used as a means to predict damage for similar structures under various earthquake intensities. 4.1. Odontology building of the University of Concepcion 4.1.1. Introduction and building configuration The Odontology building of the University of Concepcion is investigated in this case study. The building is located at the southern east part of Concepcion as shown in Fig. 8(a). The structure was used as the school of dentistry which also included rooms for examination and medical treatment of patients. Due to the requirements of the building occupancy, the reinforced concrete structure has a particular configuration. As shown in Fig. 8(b), the building has three regular stories and three, approximately half height, service stories that serve to utility lines. A typical floor plan of the structure is shown in Fig. 9. The building has a rectangular shape with seven bays in x-direction and five bays in y-direction. Structural core-walls are provided at the center of the building to resist the majority of the lateral loads. Except for minor differences in column dimensions of the exterior frames that have five bays, the structure is symmetrical with respect to both x and y axes.

236

A.S. Elnashai et al. / Soil Dynamics and Earthquake Engineering 42 (2012) 229–245

Table 4 List of damaged and/or collapsed bridges. Bridge name9Roadway carried by the bridge9Location Damage/failure modes and other remarks Miraflores Bridge9Vespucio Norte Express9Independencia, Santiago Lo Echevers Bridge9Vespucio Norte Express9Lo Echevers, Santiago Quilicura Bridge9n/a9n/a Las Mercedes Bridge9Route 5 South93414.32350 S 70145.7530 W Costanera Norte9n/a9n/a Pasarela Peatonal9Route 5, Norte9n/a Perquilauque´n Bridge9Route 5936115.2080 S 71148.82570 W Juan Pablo II Bridge9n/a9Concepcio´n, 36149.4010 S 7315.4770 W Llacole´n Bridge9n/a9Concepcio´n, 36150.0390 S 7314.6230 W Bı´obı´o Bridge9n/a9Concepcio´n, 361 50.4420 S 731 4.1150 W Las Ballenas Bridge9n/a9Suburb of Concepcio´n Rı´o Claro Bridge9n/a9n/a Tubul Bridge9n/a9Arauco Paso Cladio Arrau9n/a9361 39.5360 S 721 19.5450 W Route 5 overpass near Chilla´n9Route 5 (overpass)9Chilla´n 361 35.0830 S 721 6.5340 W

North bridge had diaphragm between girders and concrete shear keys preventing the movement of the superstructure. Damage on concrete shear keys. South bridge had steel shear keys for each girder. Most shear keys were damaged. Skewed bridge. Unseating of superstructure due to large transverse displacement. Skewed steel girder bridge. Unseating of deck. Failure of shear keys. Shear keys at only one side was damaged. Skewed pre-stressed girder bridge. Unseating of deck. No diaphragms joining pre-stressed girders. Skewed bridge. Shear key failure. Large displacement of superstructure at abutment. Pedestrian bridge. Bridge deck was bolted to the bent of the bridge. The bolts were sheared and the superstructure unseated. The bridge bent was undamaged. Consists of two bridges, one built in mid-1990s and the other one recently built. The older bridges suffered less damage than the new one. The damage on the new bridge was large transverse displacement of superstructure. The shear key of the new bridge was very weak and failed. Shear failure on bridge column due to lateral spreading of soil toward river. Indication of lateral spreading toward river. Unseating of several ramps. Built in 1930s. Closed before the earthquake for maintenance. Steel stringer bridge. Unseating of bridge decks. Total collapse. Rupture of elastomeric bearings. Unreinforced masonry bridge built in 1870. Collapsed during the earthquake. South most location of complete bridge collapse. Steel girder bridge. All eight steel girders were unseated. In Ref. [11], it was reported that performance of foundation was insufficient. Minor transverse translation of the superstructure. The bridge was serviceable after the earthquake. Skewed bridge. Large translation and close to unseating of superstructure. Shear key at the abutment was not long enough to provide resistance to the superstructure.

Fig. 8. (a) Location of the Odontology building of the University of Concepcion and the ground motion recording station CCSP, (b) the Odontology building of the University of Concepcion.

As shown in Fig. 10, three floors of the building are designed as flat slabs. Another feature of the building is that masonry infill walls are used at the first floors of the exterior frames. Based on the available information, it is not certain whether the lateral resistance provided by these infill walls is included in design calculations. It is postulated that these walls were considered to be non-structural. The height of the first story is 3.1 m and the height of the masonry infill walls was measured to be approximately 2.1 m.

columns were due commonly observed ‘‘short-column’’ effect that results from increased shear demands on a specific portion of the vertical members due to a decrease in the effective length. Several of the first-story columns of the Odontology building, which were damaged due to short-column effect, are shown in Fig. 11. In the light of the above described observations, it is deemed suitable to model the seven bay exterior frame of the building, as indicated in Fig. 9.

4.1.2. Observed damage The earthquake caused a particular damage pattern. The structural damage was confined to columns of the exterior frames. The rest of the observed damage was non-structural: cracking on the partition walls (inside the building) and the diagonal cracks of the masonry infill walls at the exterior frames. The field observations indicated that the combined axial-shear failure at the top portions of the

4.1.3. Modeling approach The fiber-based finite element analysis software ZEUS NL [5] is used to model the frame shown in Fig. 10. All members except for the masonry infill walls are represented with 3-D elasto-plastic beamcolumn elements described by a cubic shape function [9]. The columns and slabs are modeled using rectangular sections. The effective flange width (for regular floors where a T-beam exists) is

A.S. Elnashai et al. / Soil Dynamics and Earthquake Engineering 42 (2012) 229–245

calculated according to (ACI 318-08 [1]). For the flat-slab floors, the portion of the slab that contributes to the frame analysis is determined by using the formulations proposed by Luo and Durani [16,17]. The concrete material is modeled using the constant confinement model developed by Martinez-Rueda and Elnashai [18]. Steel is modeled with a bilinear elasto-plastic model with kinematic strain hardening. Based on the information on the design drawings, the following material properties are used: a concrete compressive strength of 29.4 MPa (300 kgf/cm2), steel yield and ultimate strengths of 411.6 MPa (4197.2 kgf/cm2) and 617.5 MPa

P5

P6

P5

P5

P5

P5

P5

P6

N

P4

P1

P3

P2

Wall A

P5

P5

Y

Wall C

Wall A

P1

Wall C

P3

P1

P4

P1

P4

P2

P3

Wall B

P5

P5

P10

P10

P5

Modeled Frame

X P6

P2

Wall A

P4

P5

Wall B

P5

P5

P5

P5

P5

[email protected] m = 44.1 m Fig. 9. A typical plan view sketch of the Odontology building.

P6

[email protected] m = 35 m

P2 Wall A

P3

P5

237

(6296.7 kgf/cm2), respectively. In modeling the masonry infill walls the following assumptions are made: infill walls do not carry any vertical loads and they can be represented with diagonal struts that have horizontal resistance only. The modeling of masonry infill walls follow the approach proposed by Mostafaei and Kabeyasawa [19] as described in Kwon and Kim [13].

4.1.4. Analysis results Two building configurations are considered: with and without infill walls. An eigenvalue analysis indicates that the first and second mode periods of the building reduce from 0.83 s and 0.27 s to 0.80 s and 0.26 s respectively if infill walls are included in the modeling. It can be concluded that the shift in building period would not cause a significant change in the force demand on an equivalent single-degree-of-freedom structure; however, the presence of the masonry infill walls increases the shear demand on the first story columns as demonstrated in the following. Inelastic dynamic time history analyses are conducted using a total of eight ground motion records. As described previously in Section 2.3, for the sites of the case study structures three acceleration response spectra are developed using the source characteristics and a set of suitable attenuation relationships. For this case study, the spectrum compatible records are generated based on the original recordings CCSP and MELP. Additionally, for the back-analysis conducted here the original (without spectrum matching) record from the CCSP station (both horizontal components) is also used due to close proximity of the station to the building site (6.7 km) as illustrated in Fig. 8(a). The interstory drift ratio profiles for all eight records and the two configurations (with and without infill walls) are provided in Fig. 12. It is observed that the interstory drift ratios are very

1.25 m 3.85 m 1.65 m 3.85 m 1.87 m 3.10 m

[email protected] m = 44.1 m Fig. 10. Elevation view sketch of the Odontology building.

Fig. 11. Short-column effect observed at the Odontology building.

238

A.S. Elnashai et al. / Soil Dynamics and Earthquake Engineering 42 (2012) 229–245

Spectrum compatible record: CCSP

Or i g i n a l r e c o r d : CCSP

Sp e c t r u m c o m p a ti b l e r e c o r d : MEL P

3.5 3

Story

2.5 2

1.5 1 0

1 2 Interstory Drift Ratio (%)

3 0

1 2 Interstory Drift Ratio (%)

3

0

1 2 Interstory Drift Ratio (%)

3

500

500

400

400 Capacity Demand

300

Shear (kN)

Shear (kN)

Fig. 12. Maximum interstory drift ratio: dotted lines show frame without infill wall, solid lines show frame with infill wall.

200 100 0 10

300 200 100

20

30

40

50

0 10

20

Time (sec)

30

40

50

Time (sec)

Fig. 13. Comparison of shear capacity and demand for rightmost first story column, spectrum compatible mean record with seed CCSP: frame (a) without and (b) with infill walls.

similar for both cases except for the first story. The reduction in the drift ratio at the first story (due to the presence of infill walls) ranges from 44 percent to 51 percent depending on the ground motion. The shear strength of the columns is calculated according to (ACI 318-08 [1]) and compared to the earthquake demand obtained from inelastic dynamic analysis. As an example, the shear capacity and the shear demand on the rightmost first story column of the modeled frame is shown in Fig. 13. It is observed that the presence of masonry infill walls significantly increase the shear demand on the columns and the shear capacity is exceeded at certain times throughout the analysis. The maximum shear demand to capacity ratio is illustrated in the bar plot in Fig. 14. It is concluded that the existence of masonry infill walls significantly increases the likelihood of exceeding the shear capacity. It is concluded that the interstory drift ratio at the first story is significantly reduced when the infill walls are present in the expense of a significant increase in the shear demand. Many columns of the first story are expected to fail due to increased shear demand as a result of the masonry infill walls. Particularly, for the exterior columns of the considered frame the shear demand exceeds the shear capacity considerably which indicates severe damage. The analytical results are in agreement with the observed damage in the field.

4.2. Paso Cladio Arrau The reference bridge (Paso Cladio Arrau) for this case study has four spans and it is 77.5 m long. The bridge has three bents each of which consists of 10 piers as shown in Fig. 15. As observed in most other bridges in Chile, seismic restrainers were installed in the bridge. Girders are supported on elastomeric bearings made of neoprene. The bridge has a skew angle of approximately 50 degrees. The bridge suffered minor damage from the earthquake as shown in Fig. 16. The superstructure displaced in the transverse direction, seismic restrainers yielded and elongated, and the shear keys failed. Even though the shear keys were damaged, it is likely that they prevented the unseating of the bridge superstructure. In addition to the shear keys, the seismic restrainers might have reduced the displacement demands as those can reduce the effect of vertical ground motion by maintaining the contact between the superstructure and neoprene bearings. The 3D model of the bridge was built using ZEUS NL [5] as shown in Fig. 17(a). The stiffness of abutments and hysteretic behavior of the bearings are modeled using lumped springs. The piers, seismic restrainers, and superstructure are modeled using cubic beam-column elements. Due to lack of information regarding the post-tensioning force in the seismic restrainers, it is

Shear Demand/Capacity

Shear Demand/Capacity

Shear Demand/Capacity

A.S. Elnashai et al. / Soil Dynamics and Earthquake Engineering 42 (2012) 229–245

239

1.5 1 0.5 0

C1

C2

C3

C4

C5

C6

C7

C8

C1

C2

C3

C4

C5

C6

C7

C8

C1

C2

C3

C4

C5

C6

C7

C8

1.5 1 0.5 0 1.5 1 0.5 0

Columns Fig. 14. Maximum shear demand to capacity ratio: empty bars indicate frame without infill walls, solid bars indicate frame with infill walls.

Fig. 15. Configuration of bridge bents and abutment.

assumed that they were post-tensioned up to 30 percent of the yield stress. The additional normal force to elastomeric bearings due to the post-tensioning is taken into account in the hysteretic behavior of elastomeric bearings; the hysteretic curve is shown in Fig. 17(b). The deck is considered as a rectangular beam having the equivalent section properties of the girders and the bridge deck. The configuration of the superstructure–substructure connections is illustrated in Fig. 17(b). Spectrum compatible ground motions from Section 2.3 are used for inelastic response history analyses. A total of 18 horizontal components of ground motions are applied in the transverse direction. Two sets of analyses are carried out. In the first set, the seismic restrainers are included in the numerical model as flexural elements. In addition, the normal force on the elastomeric bearings is increased to take into account the posttensioning force. It is assumed that the additional normal force on elastomeric bearings due to post-tensioning remains constant throughout the earthquake excitation. For the second set of analyses, all parameters are kept the same as those in the first set except that the contribution of seismic restrainers is removed from the model.

To illustrate the modeling of gaps at the both ends of the bridge, the trajectories of the bridge deck end points are shown in Fig. 18. If the bridge span moves only in the transverse direction without rotation, Node A in Fig. 17(a) is bounded between two solid lines in Fig. 18, which indicate the maximum displacement limits corresponding to the cases where the gaps at the ends are closed. If there is a rotational component, Node A may move further away from the upper boundary. Since the sample result shown in the figure is from a low-intensity input motion, which does not develop large rotational displacement, the trajectory of Node A stays within the boundary. Fig. 19 compares the maximum transverse displacement of Node A for the two analysis cases as a function of PGA of the input ground motion. As it can be observed from the figure, the models without seismic restrainers generally have larger displacement demand, and the trend is clearer as the PGA of the input ground motions increases. This difference is primarily due to the increased normal force in elastomeric bearings, which increases the energy dissipation capacity of the bearings.

4.3. Las Mercedes Bridge 4.3.1. Configuration, observed damage and numerical model Las Mercedes Bridge exemplifies the most commonly used highway overpass design in Chile in the recent years. Several bridges with similar configuration suffered damage during the earthquake. The damage to the bridge, mainly due to excessive movement and rotation of the superstructure in the transverse direction resulting in the failure of the shear keys and unseating of the deck, is shown in Fig. 20. The bridge has a two-span continuous deck with a span length of 28.5 m. The superstructure consists of pre-stressed girders and cast-in-place reinforced concrete slabs and is supported by a two-column bent at the center and retaining walls at both ends as shown in Fig. 21(a). The longitudinal direction of the bridge is skewed by an angle of 11 degrees from the perpendicular direction to the highway which it passes over.

240

A.S. Elnashai et al. / Soil Dynamics and Earthquake Engineering 42 (2012) 229–245

Fig. 16. Observed damage to Paso Cladio Arrau.

Fig. 17. Details of the numerical model.

Three-dimensional numerical models are generated by using the finite element analysis program ZEUS NL [5] to investigate the seismic response of the bridge. Based on the design specification of a typical bridge on the highway (Route 5) and Schmidt hammer tests at the field, compressive strength of concrete is assumed to be 50 MPa (510 kgf/cm2) for girders and 35 MPa (357 kgf/cm2) for other parts of the bridge including the columns. The yield stress of

reinforcement is assumed as 400 MPa (4079 kgf/cm2). Overview of the numerical model is provided in Fig. 21(b). Elastomeric bearings are placed at both ends of each girder and a total of twelve bearings are used as shown in Fig. 22(a). The force–displacement relationship of the bearing is determined by friction force and friction displacement; see Fig. 22(b). The friction coefficient between the surfaces of the bearing and the girder is

A.S. Elnashai et al. / Soil Dynamics and Earthquake Engineering 42 (2012) 229–245

241

Fig. 18. Trajectory at the end of bridge span.

Fig. 19. Maximum responses of the bridge with (w/) and without (w/o) seismic restrainers.

assumed to be 0.21 which is the friction coefficient between neoprene rubber and concrete. The thickness of bearing pads on the column bent is 37 mm, while the thickness of those on retaining walls is 60 mm. The maximum friction displacement (Db) at which the friction slip begins is assumed to be 40 mm and 25 mm for the tall and short bearings, respectively. Gap elements are used to represent the delayed contact at the construction joints between the bridge deck and approach slabs on the abutments. Based on the design drawings, the clearance of the construction joint is assumed to be 60 mm. A unidirectional spring element is used to represent the conditional force transfer at the construction joint; only compressive forces are transferred when the gap is closed.

4.3.2. Analysis results Under earthquake excitations, the bridge behaves as a semirigid diaphragm on a stiff substructure. The response of the superstructure is mainly determined by slip at the girder-bearing interface. The fundamental period of the bridge is evaluated as 0.80 s and this mode is induced by the deformation of the elastomeric bearing before the friction slip. The first mode shape is determined by the movement of the superstructure in the transverse direction without torsional responses because the bridge is symmetric along its length. In the static pushover analysis, lateral forces are applied in the transverse direction. The locations of shear keys and direction of lateral forces are shown in Fig. 23(a). From the analysis results, it was observed that shear key #1 reaches its ultimate shear capacity earlier than others, followed by severe damage on shear key #2. The reason is that the shear keys on the stiff embankments (#1 and #3) attract

more forces than shear key #2, which is placed on the flexible column bent. Under the lateral forces that are assumed for the static pushover analysis [see Fig. 23(a)], shear key #1 reaches its ultimate capacity earlier than shear key #3 because the center of lateral forces shifts to the left. This shift is caused by a minor distortion of the superstructure due to large lateral forces. Uneven lateral force distribution on shear keys and unbalanced force redistribution can magnify torsion in highly inelastic response range of the bridge. The status of the bridge on the capacity curve in Fig. 23(b) is explained as follows: (1) the superstructure displaces due to lateral deformation of elastomeric bearings; (2) the lower part of a girder hits a shear key; (3) shear key #1 reaches its ultimate shear strength; (4) force redistribution occurs among the shear keys; (5) shear key #2 reaches its ultimate shear strength. As soon as the shear key fails, the girder and elastomeric bearing slip off from the bearing support and once the latter slipoff occurs, the girder hardly moves back in the opposite direction and thus its displacement accumulates unidirectionally. At the same time, the other end of the superstructure moves in the opposite direction, which causes cumulative torsion of the superstructure. A sample plot of dynamic response history analysis by the record matched to þ0.5s spectrum using the original ground motion from the station STL (EW component), see Section 2.3, is shown in Fig. 24. The earthquake loading is applied in the transverse direction of the bridge deck. Shear keys are omitted in the analytical model in order to investigate causes of torsion other than the uneven resistance of shear keys. While the maximum displacement is 90 mm, which is larger than the allowable deformation of the elastomeric bearing, the torsion is negligible, only about 0.0012 degrees. This implies that the large amount of torsion observed at the field had not been caused by dynamic characteristics of the structure. In this reference bridge, the skew angle is too small for the observed torsion to be caused by the pounding response. Based on the above discussion, the large torsion observed at the field is explained as follows: premature failure of a shear key causes uneven and excessive displacement at an end of the superstructure. Then the girder end slips off from the bearing support and this initiates the torsion that is accumulated unidirectionally.

4.3.3. Fragility curve development A multi-component fragility analysis is performed on this case study bridge using an analytical fragility methodology, which includes using joint probabilistic seismic demand models to integrate the responses of several bridge components in the

242

A.S. Elnashai et al. / Soil Dynamics and Earthquake Engineering 42 (2012) 229–245

Fig. 20. Damage to Las Mercedes Bridge.

Fig. 21. (a) Configuration of the Las Mercedes bridge (plan view) and (b) overview of the numerical model.

Fig. 22. Numerical model: (a) locations of joint and gap elements and (b) properties of joint elements for bearings.

analysis. The monitored components include the column ductility, column maximum drift ratio, and abutment movement in the longitudinal and transverse directions. The limit states used for the fragility curve generation are adopted from the research of Nielson and DesRoches [21], and Ramanathan et al. [23]. To address the issue of uncertainty for this analysis, the bridge geometry and material properties are modeled deterministically, and only the suite of ground motions chosen contributed to the variability in the response. This is because the variability of the response of the system is much more susceptible to the ground motion variability than the material uncertainties [12]. The suite of ground motions chosen for this analysis characterizes the

specified region and also provides a range of intensities. The procurement and development of these ground motions is described in Section 2.3. Fig. 25 shows the fragility curves for this bridge based on the peak ground acceleration (PGA). Two fragility curves are shown for the same bridge, one curve using column curvature ductility as the engineering demand parameter for columns, and the other using maximum column drift ratio, Fig. 25(a) and (b) respectively. As is shown, these two parameters give slightly different estimates of the fragility of this bridge in all of the damage states. For example, for a PGA level of 0.5 g, the fragility curve based on curvature ductility predicts around 50 percent chance

A.S. Elnashai et al. / Soil Dynamics and Earthquake Engineering 42 (2012) 229–245

243

5000 Equivalent lateral force Equivalent lateral force

Shear key #3

Base shear (kN)

4000

2000

(1) 0

Shear key #1

(4)

3000

1000

Shear key #2

(5)

(3)

0

(2) 50

100 150 200 Displacement (mm)

250

300

Fig. 23. (a) Direction of equivalent lateral forces and locations of shear keys and (b) capacity curve of the reference bridge.

0.002 Torsional angle (degree)

Displacement (mm)

100 50 0 -50 -100

0.001 0 -0.001 -0.002

0

20

40

60

80

100

0

20

40

60

80

100

Time (sec)

Time (sec)

1

1

0.8

0.8

P(LS|PGA)

P(LS|PGA)

Fig. 24. Selected dynamic response history results: (a) transverse displacement at the center of the bridge and (b) angle of torsion.

Slight Moderate Extensive Complete

0.6

0.4

0.4

0.2

0.2

0

0.6

0

0.5

1 PGA (g)

1.5

2

0

0

0.5

1 PGA (g)

1.5

2

Fig. 25. Fragility curves based on (a) column curvature ductility, and (b) maximum column drift ratio.

of slight damage, while the drift ratio based curve predicts approximately 20 percent probability of slight damage. This difference could be due to the fact that the limit states come from different sources demonstrating that the limit states used in fragility analysis affect the results greatly. These curves show that the bridge has low vulnerability to damage for the more severe limit states.

The damage prediction by the fragility curves provided in Fig. 25 are in accordance with the field observations in that almost no damage was observed in the bridge columns due to limited continuity with the superstructure. The failure resulting from the demolition of the shear keys and the unseating of the deck is addressed in the back analysis described in the previous sections.

244

A.S. Elnashai et al. / Soil Dynamics and Earthquake Engineering 42 (2012) 229–245

5. Summary and conclusions An earthquake of magnitude 8.8 hit central Chile on February 27, 2010. Despite the large magnitude and proximity of the event to large settlements, the majority of engineered buildings performed satisfactorily and the loss of life was limited. However, the transportation system, hospitals and schools were affected significantly causing disruption in the aftermath of the earthquake. The economic losses were major amounting to approximately 17 percent of the GDP of Chile. Large magnitude earthquakes that affected densely populated regions with well-designed engineered structures are sparse. Therefore, the 2010 Maule (Chile) earthquake is an important opportunity for the scientific community to study the strong ground motion produced by such events and the deficiencies in design practices to prevent the impacts of future earthquakes. In this paper, due to the lack of strong ground motion data, hazard and site specific acceleration time histories are developed through spectrum matching of response spectra obtained from selected attenuation relationships. The developed ground motions are used for back-analysis of a building and two bridges to investigate the dynamic response of these structures under the large magnitude earthquake event and to explain the commonly observed modes of failure. For the building, the analysis results confirmed the shear failure of column observed in the field resulting from the increased demand due to presence of infill walls. This type of failure is also known as the short-column effect and it is widely seen in the past earthquakes as well. The bridge case studies aimed at understanding the effectiveness of seismic restrainers (that are peculiar to Chilean bridge construction) and unseating of bridge decks due to shear key failures extensively observed during the 2010 Maule earthquake. The analysis results indicate that the seismic restrainers are an effective way of reducing the displacements of bridge decks and their effectiveness increases with increasing ground motion intensity. It is also found out that the torsional response of bridge superstructure is a result of unidirectional accumulation of displacements that occur after the failure of weak shear keys. It is concluded that if shear keys had been designed properly to resist the loads from superstructure, damage to bridges could be minimized and several of the bridges would be functional after the earthquake, facilitating the rescue and response efforts.

Acknowledgments The field mission to Chile was sponsored by the Mid-America Earthquake Center, Missouri University of Science and Technology, University of Connecticut, Georgia Institute of Technology, and National Research Foundation Grant provided by the Government of Korea (2011-0028552). The MAE Center is a graduated National Science Foundation (NSF) Engineering Research Center, which was funded under NSF Grant EEC-9701785. The authors express their gratitude to the following individuals: Jeffery Roesler, Imad L. Al-Qadi, Angharad Valdivia, Rafael Riddell, Guillermo Thenoux Z., Marcelo Gonza´lez H., Carlos Videla, Mauricio Lo´pez, Mauricio Pradena Miquel, Ramo´n Verdugo, Gregory Pluta, Carolina Cerda, Juan Vargas, Luis Echeverria, Moise´s Vargas Eyzaguirre, Fernando Gonza´lez, and Jonguen Baek.

References [1] ACI. Building code requirements for structural concrete (ACI 318-08) and commentary. Michigan, USA: American Concrete Institute, Farmington Hills; 2008.

[2] Atkinson GM, Boore DM. Empirical ground-motion relations for subductionzone earthquakes and their application to Cascadia and other regions. Bulletin of the Seismological Society of America 2003;93(4):1703–29. [3] Campbell KW, Bozorgnia Y. Updated near-source ground-motion (attenuation) relations for the horizontal and vertical components of peak ground acceleration and acceleration response spectra. Bulletin of the Seismological Society of America 2003;93(1):314–31. [4] Elnashai AS, Gencturk B, Kwon O-S, Al-Qadi IL, Hashash Y, Roesler JR, et al. The Maule (Chile) Earthquake of February 27, 2010: consequence assessment and case studies. Research report 10-04. Mid-America Earthquake Center. University of Illinois at Urbana-Champaign, available from /http://hdl. handle.net/2142/18212S; 2010a. [5] Elnashai AS, Papanikolaou VK, Lee D. ZEUS NL—a system for inelastic analysis of structures, user’s manual, Mid-America Earthquake (MAE) Center. Department of Civil and Environmental Engineeering. Urbana, Illinois, USA: University of Illinois at Urbana-Champaign; 2010. [6] ENS. Large Earthquakes in the Subducting Nazca. Laboratoire de Ge´ologie. Ecole Normale Supe´rieure (ENS). Available from: /http://www.geologie.ens. fr/ vigny/chili-f.htmlS; Accessed on December 1, 2010. [7] Gregor NJ, Silva WJ, Wong IG, Youngs RR. Ground-motion attenuation relationships for cascadia subduction zone megathrust earthquakes based on a stochastic finite-fault model. Bulletin of the Seismological Society of America 2002;92(5):1923–32. ˜ o Sismico de Edificios. Instituto Nacional de [8] INN. Norma Chilena Oficial. Disen Normalizacion, NCh 433.Of96; 1999, (in Spanish). [9] Izzuddin BA, Elnasahi AS. Adaptive space frame analysis, Part II: a distributed plasticity approach. In: Proceedings of the Institution of Civil Engineers, Structures and Buildings, vol. 99; 1993. p. 317–26. [10] Kamiyama M. Effects of subsoils conditions and other factors on the duration of earthquake ground shakings. Eighth World Conference on Earthquake Engineering, San Francisco, vol. 2; 1984. p. 793–800. [11] Kawashima K, Unjoh S, Hoshikuma J, Kosa K. Damage of transportation facility due to 2010 Chile Earthquake, Available from: /http://peer.berkeley. edu/events/pdf/2010/Bridge_Damage_by_Chile_EQ_JSCE_Team.pdfS; Accessed on November 1, 2010. [12] Kwon O-S, Elnashai A. The effect of material and ground motion uncertainty on the seismic vulnerability curves of RC structure. Engineering Structures 2006;28(2):289–303. [13] Kwon O-S, Kim E. Case study: analytical investigation on the failure of a twostory RC building damaged during the 2007 Pisco–Chincha Earthquake. Engineering Structures 2010;32(7):1876–87. [14] Lay T, Ammon CJ, Kanamori H, Koper KD, Sufri O, Hutko AR. Teleseismic inversion for rupture process of the 27 February 2010 Chile (Mw 8.8) earthquake. Geophysical Research Letters 2010;37(13):L13301. [15] Los Angeles Tall Buildings Structural Design Council. Performance of tall buildings during the 2/27/2010 Chile Magnitude 8.8 earthquake—a preliminary briefing, Available from: /http://www.tallbuildings.org/PDFFiles/ LATBSDC%204-27-10%20Chile%20EQ%20Briefing.pdfS; Accessed on December 1, 2010. [16] Luo YH, Durrani AJ. Equivalent beam model for flat-slab buildings—Part I: interior connections. ACI Structural Journal 1995;92(1):115–24. [17] Luo YH, Durrani AJ. Equivalent beam model for flat-slab buildings—Part II: exterior connections. ACI Structural Journal 1995;92(2):250–7. [18] Martı´nez-Rueda JE, Elnashai AS. Confined concrete model under cyclic load. Materials and Structures 1997;30(3):139–47. [19] Mostafaei H, Kabeyasawa T. Effect of Infill Masonry walls on the seismic response of reinforced concrete buildings subjected to the 2003 Bam earthquake strong motion: a case study of Bam telephone center. Bulletin of the Earthquake Research Institute, University of Tokyo 2005;79(3/4):133–56. [20] Mukherjee S, Gupta VK. Wavelet-based generation of spectrum-compatible time-histories. Soil Dynamics and Earthquake Engineering 2002;22(9– 12):799–804. [21] Nielson BG, DesRoches R. Analytical seismic fragility curves for typical bridges in the Central and Southeastern United States. Earthquake Spectra 2007;23(3):615–33. [22] Papazachos BC, Papaiannou CA, Margaris VN , TheodulidisNP. Seismic hazard assessment in Greece based on strong motion duration. Tenth World Conference on Earthquake Engineering, Madrid, vol. 2; 1992. p. 425–30. [23] Ramanathan K, DesRoches R, PadgettJ. Analytical fragility curves for seismically and non-seismically designed multi-span continuous concrete girder bridges in moderate seismic zones. 9th U.S. National/10th Canadian Conference on Earthquake Engineering, Toronto, Canada; July 25–29 2010. [24] Ruegg JC, Rudloff A, Vigny C, Madariaga R, de Chabalier JB, Campos J, et al. Interseismic strain accumulation measured by GPS in the seismic gap between Constitucio´n and Concepcio´n in Chile. Physics of the Earth and Planetary Interiors 2009;175(1-2):78–85. [25] Schellart WP, Freeman J, Stegman DR, Moresi L, May D. Evolution and diversity of subduction zones controlled by slab width. Nature 2007;446(7133):308–11. [26] U.S.Geological Survey. Finite Fault Model: Updated Result of the February 27, 2010 Mw 8.8 Maule, Chile Earthquake, Available from: /http://earthquake. usgs.gov/earthquakes/eqinthenews/2010/us2010tfan/finite_fault.phpS, Accessed on December 1, 2010a. [27] U.S.Geological Survey . Magnitude 8.8—offshore Maule, Chile, Available from: /http://earthquake.usgs.gov/earthquakes/eqinthenews/2010/us2010t fan/S; Accessed on December 1, 2010b.

A.S. Elnashai et al. / Soil Dynamics and Earthquake Engineering 42 (2012) 229–245

[28] YashinskyM, Oviedo R, Ashfod S, Fargier-Gabaldon L, HubeM. Performance of Highway and Railway Structures during the February 27, 2010 Maule Chile Earthquake, Available from: /http://www.eqclearinghouse.org/20100227-chile/ wp-content/uploads/2010/04/Yashinsky-et-al-Bridges.pdfS; Accessed on November 1, 2010. [29] Yen W-HP, Chen G, Buckle I, Allen T, Alzamora D, Ger J, et al. Post-earthquake reconnaissance report on transportation infrastructure: impact of the February 27, 2010, offshore Maule Earthquake in Chile, Publication no. FHWA-HRT-11-030, Federal Highway Administration (FHWA); 2011.

245

[30] Youngs R, Chiou S, Silva W, Humphrey J. Strong ground motion attenuation relationships for subduction zone earthquakes. Seismological Research Letters 1997;68:58–73. [31] Zhao JX, Zhang J, Asano A, Ohno Y, Oouchi T, Takahashi T, et al. Attenuation relations of strong ground motion in Japan using site classification based on predominant period. Bulletin of the Seismological Society of America 2006;96(3):898–913.