Engineering Structures 32 (2010) 23942406

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Assessment of RC moment frame buildings in moderate seismic zones: Evaluation of Egyptian seismic code implications and system configuration effects
S.S.F. Mehanny a, , H.A. El Howary b
a b

Structural Engineering Department, Cairo University, Egypt Structural Engineer, DAR Al-Handasah, Cairo, Egypt

article

info

abstract
Building code restrictive seismic design provisions and building systems type and configuration have remarkable implications on seismic performance of reinforced concrete moment framed structures. Seismic assessment of ductile versions of low- to mid-rise moment frames located in moderate seismic zones is carried out through comparative trial designs of two (4- and 8-story) buildings adopting both space and perimeter framed approaches. Code-compliant designs, as well as a proposed modified code design relaxing design drift demands for the investigated buildings, are examined to test their effectiveness and reliability. Fragility curves for the frames are generated corresponding to various codespecified performance levels. Code preset lower or upper bounds on either design acceleration or drift, respectively, that would control the final design are also addressed along with their implications, if imposed, on the frames seismic performance. The trial design study demonstrates that built-in static overstrength is generally larger for space frames than for perimeter frames, whereas the force reduction attributable to inelastic dynamic response differs from one frame type to the other for various investigated heights and for different target performance levels. Nonetheless, all trial designs are shown to meet the minimum performance implied by building code provisions but with varying margins. However, the study suggests that more consistent reliability for designed structures can be achieved by disaggregating the force reduction factor into its static and dynamic parts and that code default values of this factor for some building types would be better reduced for a more reliable performance. 2010 Elsevier Ltd. All rights reserved.

Article history: Received 4 November 2009 Received in revised form 21 February 2010 Accepted 4 April 2010 Available online 7 May 2010 Keywords: RC moment frames Ductile Moderate seismic zones Codes Response modification factors

1. Introduction Performance-Based Design (PBD) is now widely recognized as the pre-eminent seismic design methodology for structures. The advent of PBD methodologies now requires that engineers develop code-compliant structures that also achieve specific performance objectives. Accordingly, it is necessary to develop efficient designs with predictable seismic response. To this day, the seismic designs of most general and some complex building structures are performed with Force-Based Design (FBD) method. This method is conceptually straightforward and thus appealing, but relies heavily upon unique, semi-empirical, force reduction factors and displacement equivalences for a selected lateral force resisting structural system. These factors are largely based on consensus opinion of code committees. The FBD methodology may

Corresponding author. Tel.: +20 12 444 8008. E-mail addresses: sameh.mehanny@stanfordalumni.org, smehanny@darcairo.com (S.S.F. Mehanny). 0141-0296/$ see front matter 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.engstruct.2010.04.014

yield life-safe designs in most cases; however, its ability to deliver designs that achieve specific performance objectives remains in question. These issues of life safety and predictable response are addressed in this paper through an investigation of a modern-day FBD code. Earlier efforts in this direction include but are not limited to the work by Mehanny et al. [1] and Rivera et al. [2]. The former [1] was mainly geared towards calculating estimates of force modification and displacement amplification factors (R and Cd , respectively, known as R and Rd in ECP 201 [3], and q and qd in EC8 [4]) for composite RCS and Steel moment frames designed as per US standards (e.g., [5,6]), and comparing them to their corresponding values specified in the adopted design codes in order to assess how such provisions were successful to deliver safe, reliable and economic designs. On the other hand, the recent work by Rivera et al. [2] focused on trying to furnish an answer to the question that naturally arises: Are FBD provisions of modern seismic codes compatible with PBD objectives?. They therefore investigated the predictability of response and margin of safety of trial designs of regular medium ductility RC moment

S.S.F. Mehanny, H.A. El Howary / Engineering Structures 32 (2010) 23942406

2395

framed structures designed according to [4]. Their assessment was performed by comparing the design displacements and forces for these frames to those obtained from nonlinear time history analysis. The current paper is an additional effort along the same frontier looking into semi-empirically based key factors R and Rd used for FBD procedures. The research focuses though on investigating only low- to mid-rise ductile RC moment resisting frames located in moderate seismic zones (0.25 g), and further studying the implications that the frames configuration (perimeter versus space frames) may have on the overall response. Seismic provisions of interest for this study are the emerging Egyptian seismic provisions [3] that are largely compatible with EC8 main directions. The ultimate goal is to evaluate the current code-specified R and Rd factors, and to eventually improve the reliability of constructed facilities designed using FBD methodologies. Four Code-Compliant-Design (CCD) versions of RC ductile moment resisting frame buildings (4-, and 8-story, adopting perimeter and space frames configurations) are developed using ECP 201-FBD provisions. Using nonlinear analyses involving inelastic static pushover analysis and incremental dynamic time history analysis under a suite of 20 multi-level scaled records, static and dynamic contributions to inelastic force reduction are identified and compared to code/regulations-specified assumptions. Fragility curves for the frames are also developed corresponding to various universally code-specified performance levels encompassing, for example, Immediate Occupancy (IO), Life Safety (LS) and Collapse Prevention (CP) as identified by FEMA 356 [7]. Generated information facilitates retrieving relevant actual inherent R and Rd factors and comparing them to code pre-specified values adopted earlier in the FBD process. A Modified Code Design (MCD) procedure relaxing design drift demands for the investigated buildings (and hence overcoming a specific deficiency in the current requirements of the ECP 201 seismic provisions as will be demonstrated in what follows) is proposed in the current research and is further examined to test its effectiveness and reliability. 2. Outlines and specifics of the seismic design procedures The main design requirements specified in [3] are the nocollapse and the damage limitation requirements. Satisfying the no-collapse requirement depends mainly on the strength of the designed elements to resist all expected stress resultants that occur due to the seismic actions. Design seismic actions correspond to the reference seismic hazard associated with a reference probability of exceedance of 10% in 50 yrs (or a reference return period of 475 yrs). In a complementary step, and in line with EC8 regulations [4], the structure shall be also checked to withstand a seismic action having a larger probability of occurrence (minor earthquake) than the design seismic action associated with the no-collapse requirement, without occurrence of damage to structural and non-structural elements. Such seismic action is used to verify the damage limitation requirement. It has a probability of exceedance of 10% in 10 yrs (or a return period of 95 yrs) and is almost equal to half of the design seismic action for the nocollapse limit state taking into account the importance factor of the building. As per code, the damage limitation requirement is satisfied if the interstory drifts are limited to a given fraction of the story height depending on the type and fixation form of the non-structural elements. The interstory drift associated with the design seismic action for the no-collapse limit state has thus to be first reduced to take into account the lower return period of the seismic action associated with the damage limitation requirement. Implicit in the use of this reduction is the assumption that the response spectrum of the seismic action for the nocollapse requirement has the same shape as the spectrum of

the seismic action for damage limitation requirement (i.e., the latter is a scaled down replica of the former). For buildings investigated herein, this reduction factor, , is taken equal to 2.0 [3] and the interstory drift limit is set to 0.5% associated with nonstructural elements of brittle materials that are attached to the structure. It is worth pointing herein that in other similar seismic provisions commonly adopted worldwide especially in the US practice (such as in [5,8,9]), instead of performing the drift checks for a minor earthquake with a larger probability of occurrence (10% in 10 yrs) than the design level earthquake used for strength checks (i.e., the 10% in 50 yrs event), and accordingly reducing the interstory drift limit or capacity (e.g., 0.5%), they rather perform the drift (and strength) check(s) for one same design level earthquake of 10% in 50 yrs but with a magnified interstory drift limit. This magnified limit is roughly equal to the limit set by Eurocode (as a ratio of the story height) times the factor mentioned above. In other words, even though different codes apparently approach the same task from different perspectives, they are basically more-orless heading towards the same target. Note that, furthermore, in order to avoid excessively low design acceleration values (and hence potentially non-conservative designs in terms of lateral strength/resistance) for medium- to long-period structures that may arise from inaccurate modeling, and again similar to Eurocode directions in that concern, ECP 201 is imposing a constant minimum design acceleration of 0.2ag . Such enforced lower bound sometimes introduces too much conservatism into the design which will be examined in the course of this research. Two seismic design scenarios are performed in this paper on four case study buildings. The buildings consist of 4- and 8-story moment framed ductile RC structures adopting either space or perimeter frames systems. The two seismic design procedures are depicted below: 1. Code-Compliant Design (CCD): It is a design procedure where (1) no-collapse in terms of satisfying strength of different structural elements considering second-order effects and (2) damage limitation in terms of satisfying code interstory drift limits under reduced hazard requirements are jointly satisfied. Code Design Response Spectrum (DRS) modified by the response modification factor, R, as shown in Fig. 1(a) and featuring the constant acceleration lower bound of 0.2ag is adopted. 2. Modified Code Design (MCD): It is a modified (more relaxed) seismic design procedure through ignoring the code pre-specified constant acceleration lower bound when checking drift demands. This concept is not uncommon in well established international seismic design provisions (e.g. [5,8,9]). In other words, checking drift is carried out for a scaled down version of the code acceleration Elastic Response Spectrum (ERS) associated with 10% in 50 yrs hazard as shown in Fig. 1(a) by directly dividing its ordinates by the R factor, as well as by a reduction factor = 2.0 [3] accounting for the lower return period (corresponding to a 10% in 10 yrs hazard) of the seismic action associated with the code damage limitation requirement, then magnifying it back by a displacement behavior factor, Rd , approximately equal to 0.7R [3]. The resulting Modified Elastic Response Spectrum, MERS [=ERS (1/) (Rd/R)] used for checking drift and developed in the context of this step is shown in Fig. 1(a) for comparison purposes. This proposed step entirely discards any effect on seismic design drift demands that may arise from the lower bound of 0.2ag on the design acceleration specified by code and reflected into the code DRS. However, the no-collapse requirement is still verified for the code acceleration DRS with the lower bound on the design acceleration. MCD procedure, despite being a code non-compliant design procedure, is promoted herein since it provides potentially economic versions of the case study buildings yet without risking safety as will be demonstrated later.

2396

S.S.F. Mehanny, H.A. El Howary / Engineering Structures 32 (2010) 23942406

(a) Acceleration response spectra as per ECP 201.

(b) Displacement spectra.

Fig. 1. ECP 201 elastic and design acceleration response spectra and interpretation of design spectral displacement curves with and without an upper bound limit on displacement.

3. Egyptian code (ECP 201) versus Eurocode (EC 8) seismic design provisions It is worth noting that ECP 201 is more liberal than EC8 in selecting the R factor that is set to a value of 7 in the former for ductile RC moment resisting frames. EC8 assumes instead a value of 5.85 for the behavior, q, factor (equivalent to ECP 201 R-factor). This value of 5.85 that accounts for energy dissipation capacity as per Eurocode wording is calculated for a multi-story multi-bay moment resisting frame system pertaining to the high ductility class (DCH). On the other hand, both codes rely on linear analysis to estimate the actual expected displacement, ds , induced by the design seismic action at a given point within the structure through the following simplified expression: ds = Rd de (1)

hazard) as per Eq. (3). Applying this upper bound, or limitation, is more-or-less compatible for medium- to long-period structures with the MCD procedure promoted in this paper since the latter directly adopts the code acceleration ERS or more precisely its modified version MERS (instead of the DRS) for checking drift and accordingly ignores the 0.2ag lower bound as previously stated. Conversely, the second curve shown in Fig. 1(b) refers to the situation when this limit on ds is instead ignored; this is equivalent to likewise applying Eq. (2) to calculate the spectral displacement but overlooking the upper bound value of SD,ERS as is currently the case in [3]. This results in unbounded, and hence unrealistic, values for the spectral displacements for medium- to long-period structures. Note that Eq. (2) is equivalent to simply stating that Rd /R 1. SD,DRS = Rd Sa,DRS (T /2 )2 SD,ERS SD,ERS = Sa,ERS (T /2 ) .
2

(2) (3)

where de is the displacement of the same point as determined by a linear analysis based on the design response spectrum (DRS); and Rd is a displacement behavior factor. Rd (or qd in [4]) is approximately assumed by the Eurocode equal to R (or more precisely q) in line with the commonly recognized equal displacement rule, whereas ECP 201 is assigning to Rd a reduced value of 0.7R. Such reduction could be justified by the large (relaxed) value of R (=7) assumed to determine the design base shear used throughout the strength design of the frames. The validity and effectiveness of these values to mimic actual response will be assessed in what follows when retrieved relying on pushover nonlinear inelastic static analyses and incremental dynamic analyses performed in the current research. Another point to highlight is the reason for promoting the MCD procedure introduced in this work as a potentially useful and corrective modification to ECP 201 seismic design provisions. MCD procedure, outlined in the previous section, draws its threshold from the note spelled out in [4] stating that the value of ds does not need to be larger than the value derived from the elastic spectrum. The fundamental role of this note for buildings with relatively medium to long periods is to avoid the effect of the 0.2ag lower bound enforced in the code acceleration DRS that produces increasingly larger spectral displacements compared to the case of simply adopting the ERS [2], which is physically not possible. For illustration purposes, one may refer to Fig. 1(b) showing the trends of the spectral displacement, SD , curves and their interpretation for the following two cases. The first is if this limit on ds is equally imposed in ECP 201 as in EC8; this is equivalent to applying Eq. (2) below while computing SD,DRS (that is the spectral displacement based on the code acceleration DRS corresponding to the 10% in 50 yrs hazard, Sa,DRS ) but with the limitation that the resulting value should not exceed SD,ERS calculated based merely on the code acceleration ERS (also associated with the 10% in 50 yrs

From another perspective, one may therefore consider the MCD procedure introduced in the previous section as directly using the ERS corresponding to the 10% in 10 yrs hazard (obtained by reducing the code acceleration ERS associated with the designbasis earthquake of 10% in 50 yrs hazard by a factor of 2 as illustrated above) while estimating design drift demands. Such calculated drift is yet to be further reduced by a factor of 0.7 [3] in order to get the actual expected displacement, ds , induced by the design seismic action. As such the Equal Displacement Rule (EDR) is not activated (conversely to [4]) and the estimation of actual drift is hence still along the same base line of [3]. Note that the notion for reducing the displacement factor, Rd , as adopted in [3] is likewise followed by other world-widely recognized seismic provisions and standards such as [5,9] setting this factor equal to approximately 0.69 (=[Cd = 5.5]/[R = 8]) times R for RC special (i.e., ductile) moment resisting frames. Although ECP 201 [3] draws most of its background from EC8 [4], it may nevertheless have set Rd = 0.7R (and thus not literally following the EDR adopted by EC8 provisions that simply set qd = q) to counter-account for the larger value of R = 7 used a priori for the design of ductile RC moment frames versus the smaller value of q = 5.85 set by EC8 for the seismic design of the same type of frames. Among the major objectives of this paper is therefore to assess the effect that such upper bound note stated in [4] (or, the proposed MCD procedure) regarding drift estimation may have on the reliability/safety versus economy of ductile RC moment frames of either the space or the perimeter type when located in moderate seismic zones. 4. Case study building design The case study building (see structural plan in Fig. 2) is developed according to the general layout of a theme structure

S.S.F. Mehanny, H.A. El Howary / Engineering Structures 32 (2010) 23942406

2397

6000

6000 24000

6000

6000

Fig. 2. Layout (plan) of the case study buildings case of PFB (dimensions shown are in millimeters).

proposed for this research work. It is designed as a 4- and 8-story building in moderate seismic region according to appropriate portions of relevant codes and standards [3,10]. RC ductile framed designs are developed first employing a space frame configuration where the lateral system consists of five moment frames in each direction. Another framed design is then developed using a perimeter frame approach as common, for example, in the US practice [11] where only perimeter frames constitute the lateral resisting system. A typical floor height of 3 m is adopted whereas the ground floor is 5 m high. Buildings layout is essentially bi-symmetric in plan, square in shape with a typical bay width of 6 m in both directions, and is representative of benchmark typical office buildings in current practice in Egypt. For gravity load design, dead loads include the self-weight of the structure, a typical floor cover of 1.5 kN/m2 and partition (wall) loads intensity of 1.5 kN/m3 including plastering and assuming typical walls thickness of 250 mm. A live load of 5.0 kN/m2 is also considered. On the other hand, for seismic design purposes, a total seismic mass including self-weight and floor cover plus 50% of live load is considered. The seismic design has been carried out assuming a soil type B as per [3,4] referring to dense/stiff soil, an importance factor of 1.0 and a seismic zone 5 (as per Egyptian zoning system) with a design ground acceleration, ag , of 0.25 g associated with the code reference probability of exceedance of 10% in 50 yrs. Code acceleration ERS type 1 is adopted [3] known as type 2 in [4] and is shown in Fig. 1(a) for the case study buildings. For comparison purposes, also shown in Fig. 1(a) is the code DRS used for elastic analysis of the buildings after introducing a lateral force reduction factor R of 7. A solid slab is used at all floors with a designed constant thickness of 140 mm. All columns and beams dimensions and reinforcement are as shown in Tables 1 and 2. Reinforcing steel used has a minimum guaranteed (i.e., nominal) yield strength of 360 MPa, and concrete has a minimum specified cube characteristic strength in compression of 30 MPa. For design purposes using FBD methodology and linear elastic analysis, cracked members properties are adopted as per recommendations in [3]; 70% of the gross inertia is used for columns while 50% of the gross inertia is used for beams. Furthermore, the current

trial designs have considered the first interior frame in the Space Frame Building (SFB) configuration (i.e., the one adjacent to the edge frame) believed to be the vulnerable frame of interest worth to be studied when only a two-dimensional analysis of a single representative critical frame is sought for this space frame building. This decision was taken due to the larger tributary gravity loads and seismic mass of this selected interior frame relative to the edge one residing along the perimeter of the building, along with an associated larger contribution from the design accidental eccentricity relative to the most inner frame located on the axis of symmetry of the buildings footprint with basically no eccentricity effects. As a result, and as per the recommendations in [3] (similar to requirements set in [4]) for the design for minimum accidental eccentricity, the design base shear for this selected frame has been increased by 15%, while the design base shear for the perimeter frame in the Perimeter Frame Building (PFB) configuration also investigated in this research has been increased by 30%. Such factor will add to the intrinsic (actual) overstrength of the various moment frames considered herein but with different magnitudes and effects thereof as will be highlighted in what follows. All case study moment frames designed satisfy the minimum strength, stiffness (drift), and strong columnweak beam requirements specified in [3]. Members (columns and beams) sizes in both CCD and MCD procedures were controlled nearly exclusively by drift requirements whereas only the design of the 4-story SFB frame has been marginally controlled by the strength requirements under gravity loads fundamental ultimate combination. This resulted in having the 4-story MCD-SFB frame a replica of the 4-story CCD-SFB frame. Calculated fundamental period of vibration (and second lateral mode period) along with the associated modal mass ratios relative to the total considered seismic mass for all case study frames are given in Table 3 for both CCD and MCD procedures for future relevance in the seismic assessment study. From design data reported in Tables 1 and 2, it may be observed that CCD procedure constantly results in heavier mid-rise frames compared to the MCD procedure (1.43 times heavier for 8S-PFB and 1.2 times for 8S-SFB). This is a straight outcome of relaxing design drift demands for the proposed MCD technique. This ratio is however reduced for investigated low-rise frames scoring only a value of 1.12 for 4S-PFB and a straight 1.0 for 4S-SFB (literally speaking CCD and MCD result in exactly the same frame for the 4S-SFB building). From another perspective, PFB frames are consistently heavier than their equivalent SFB frames irrespective of the adopted design technique (CCD versus MCD), especially for low-rise buildings investigated herein. For example, 4S-CCD-PFB and 4S-MCD-PFB frames are 1.35 and 1.2 times heavier than 4S-CCD-SFB and 4S-MCD-SFB frames, respectively. This could be inferred from the relevant larger lateral seismic mass assigned to the perimeter frames compared to the equivalent space frames for a particular building configuration. This ratio is reduced to only 1.2 for mid-rise 8S-CCD buildings. It is further noted though that 8S-MCD-PFB frame (as designed) has almost same overall weight relative to the 8S-MCD-SFB frame. Moreover, referring to Table 3, and in harmony with the observations made above in this paragraph, it may be noted that MCD technique furnishes frames that are fairly more flexible (i.e., having longer period of vibration) than their equivalent CCD frames. In addition, whether adopting the CCD or the MCD procedure, PFB frames consistently feature longer periods than their equivalent SFB frames for both low- and mid-rise buildings investigated herein. 5. Selected ground motion records A bin of 20 ground motions is selected for the seismic evaluation study presented in this paper. The 20 records pertain to the large database of records gathered in [12] and are originally

6000 6000 6000 24000

6000

2398

S.S.F. Mehanny, H.A. El Howary / Engineering Structures 32 (2010) 23942406

Table 1 Sizes and reinforcement of structural members of CCD case study moment framesa . Building type Outer columns Story # 4S-PFB 4S-SFB 18 8S-PFB 18 8S-SFB 500 500 12 16 500 500 14 14 Size (mm) 600 600 400 400 Reinf. 12 20 12 16 12 16 Inner columns Story # 12 34 12 34 13 46 78 13 46 78 Size (mm) 400 1100 400 1000 700 700 600 600 500 1500 500 1400 400 1300 800 800 700 700 600 600 Reinf. 7 20 6 20 16 20 12 20 8 25 7 25 5 25 20 20 16 20 12 20 Beams Story # 14 14 14 58 14 58 Size (mm) 300 1000 250 800 300 1100 250 900 250 1000 250 900 Reinf. 4 25 4 20 6 25 6 20 5 20 4 20

a Reinforcement shown in tables for all columns with a square cross-section represents the total number of re-bars to be distributed equally along the 4 sides, while that for columns with a rectangular cross-section represents the number of main re-bars per each of the two opposite shorter sides of the cross-section, i.e., in the direction resisting the bending moment in the frame direction; additional secondary re-bars are placed along the longer sides. Reinforcement given for beams represents the number of re-bars used per each side (top and bottom) of the beams cross-section. Beams have symmetric reinforcement to accommodate expected reversible bending moments during seismic events.

Table 2 Sizes and reinforcement of structural members of MCD case study moment framesa . Building type Outer columns Story # 4S-PFB 4S-SFB 8S-PFB 14 Size (mm) 600 600 Reinf. 12 20 Inner columns Story # 12 34 Size (mm) 400 900 400 800 400 1100 300 900 300 600 800 800 600 600 400 400 Reinf. 6 20 6 20 8 20 7 16 5 16 24 20 16 20 12 16 Beams Story # 14 Size (mm) 300 900 Reinf. 7 20 8 20 7 20 7 20 6 20

8S-SFB

A replica of the 4S-CCD-SFB moment resisting frame 18 500 500 12 16 13 46 78 18 500 500 12 16 13 46 78

14 58 14 58

300 1000 250 700 250 800 250 700

a Reinforcement shown in tables for all columns with a square cross-section represents the total number of re-bars to be distributed equally along the 4 sides, while that for columns with a rectangular cross-section represents the number of main re-bars per each of the two opposite shorter sides of the cross-section, i.e., in the direction resisting the bending moment in the frame direction; additional secondary re-bars are placed along the longer sides. Reinforcement given for beams represents the number of re-bars used per each side (top and bottom) of the beams cross-section. Beams have symmetric reinforcement to accommodate expected reversible bending moments during seismic events.

Table 3 Period and associated modal mass ratio for fundamental and second mode of vibration for case study moment frames from modal analysis. Building type Fundamental mode of vibration Period (s) Modal mass ratio (%) Second mode of vibration Period (s) Modal mass ratio (%)
Sa [g]

1.0 0.8 0.6 0.4 0.2 0.0 0.0 0.5 1.0 1.5 Period, T [sec] 2.0 2.5 3.0

Moment frames as per CCD procedure 4S-CCD-PFB 1.01 92.9 4S-CCD-SFB 0.83 92.9 8S-CCD-PFB 1.52 81.2 8S-CCD-SFB 1.37 86.9 Moment frames as per MCD procedure 4S-MCD-PFB 4S-MCD-SFB 8S-MCD-PFB 8S-MCD-SFB 1.22 0.83 2.01 1.65 93.5 92.9 80.6 82.6 0.31 0.24 0.54 0.49 0.38 0.24 0.80 0.63 5.8 5.9 12.9 9.4 5.4 5.9 14.6 11.2

Fig. 3. Elastic acceleration response spectra for the 20 SMLR unscaled selected records 5% damping.

[1214] as well as to bridges [15]. For specific details of the records including earthquake names, sensor location, magnitude, distance, soil type, faulting mechanism, and peak waveform ordinates, one may be referred to [12]. For completeness, Fig. 3 shows acceleration response spectra curves for these 20 selected unscaled records. 6. Analytical models for case study moment resisting frames In order to perform inelastic nonlinear static pushover and dynamic time history analyses, computer models of the asdesigned buildings are required. The structural analysis platform OPENSEES [16] is used to determine structural response of the case study moment resisting frames. Among the main features of the analytical models adopted in this research are (1) the use of nominal (minimum specified) material property values rather than expected ones; (2) confined concrete response as per the uni-axial KentScottPark model with degraded linear unloading/reloading stiffness according to the early work of Karsan and Jirsa [17]

extracted from the PEER (Pacific Earthquake Engineering Center) Strong Motion Database (PEER Strong Motion Catalog). The ground motions represented by the records are characteristic of non-nearfault motions recorded in California. They all have magnitudes Mw less than 6.5 and have been identified by the PEER database as Small Magnitude (SM) records. Furthermore, the selected records considered herein feature a distance R to the fault that is larger than 30 km, and are hence recognized by the PEER database as Large Distance (LR) records. In brief, this bin of records has been referred to by the PEER as a Small Magnitude Large Distance (SMLR) bin. All twenty ground motions were recorded on NEHRP soil types C or D (stiff soil or soft rock) sites. These records were extensively used in several earlier studies related to building structures

S.S.F. Mehanny, H.A. El Howary / Engineering Structures 32 (2010) 23942406

2399

with no tensile strength, and using confined concrete parameters as illustrated in [18]; (3) steel reinforcement uni-axial bilinear material model with kinematic strain hardening; and (4) hysteretic behavior in the form of distributed plasticity integrated along the length of two-dimensional beamcolumn elements using a fiber-element model available in OPENSEES library. For more data related to these issues, the reader may be referred to [16,19]. Beam-column element with a displacement-based formulation is adopted to model both beams and columns of the two-dimensional moment resisting frames studied herein. P (i.e., second-order geometric) transformation is activated. Displacement-based elements assume cubic displacement shape functions and present distributed plasticity. Such formulation thus approximates the displacement field within the element. In order for this type of element to capture the concentration of plastification (and consequently high curvature gradients at plastic hinge locations), a relatively fine discretization of beamcolumn members should be maintained. Good results are expected for elements size approximately equals to the length of plastic hinge. Displacementbased elements are typically used in Finite Elements applications. After carrying out a sensitivity analysis, it has been found that the most convenient, reliable and fairly accurate discretization of beams and columns of the case study moment resisting frames is achieved through using 10 displacement-based elements to model both beams and columns with 5 sections per each element and at least 15 fibers per section depending on each cross-section dimensions [19]. This number of elements (and sections per element) captures a reasonable balance between computational issues at one hand and convergence and accuracy of calculated demands of interest (maximum displacements) at the other hand. Furthermore, such fairly dense discretization automatically captures P in addition to P effects. Rayleigh mass and stiffness proportional damping is adopted in the current study. A damping ratio of 5% has been assigned to the first two modes of vibration for all case study frames. It could be noted from Table 3 that modal mass ratios associated with these two selected modes constitute an overall value that is fairly above 90% of the total seismic mass for each investigated frame. A dummy column (commonly known by a leaning column) is introduced in all PFB frames to account for P effects from the tributary gravity loads carried by the non-seismically designed interior gravity-only columns in this building configuration. The leaning column technique is based on introducing a set of leaning columns connected to the main perimeter frame using rigid links. These rigid links have a moment release condition at their ends. The main function of these rigid links is to push the leaning columns with the same lateral sway value of the PFB frame under consideration at the different floor levels. The leaning columns themselves have negligible inertia along with a large axial stiffness. 7. Static inelastic displacement-controlled pushover analysis Displacement-controlled inelastic pushover analyses with geometric nonlinearity (P effects) are conducted on twodimensional base line models for the case study frames using OPENSEES. Pushover analysis consists of first applying the distributed gravity load (full dead loads and 50% of the design live load) to the structure and then applying incremental displacements to the top of the frame with a given pre-specified distribution as per [3] at different floor levels until reaching a given target displacement. Note that for the frames of the PFBs, a leaning column as introduced above is modeled in the two-dimensional pushover analysis to account for the interior gravity columns of the PFB that are not part of the lateral resisting system. The percentage of the design base shear for each of the case study frames is plotted versus the Roof Drift Ratio (RDR) defined as the lateral

Table 4 Summary of pushover analysis results and static built-in overstrength factors. Building type

o = o T

IDRmax at RDR = 2%

Moment frames as per CCD procedure 4S-CCD-PFB 2.54 1.30 4S-CCD-SFB 3.28 1.15 8S-CCD-PFB 1.88 1.30 8S-CCD-SFB 2.29 1.15 Moment frames as per MCD procedure 4S-MCD-PFB 4S-MCD-SFB 8S-MCD-PFB 8S-MCD-SFB 1.93 3.28 1.16 2.30 1.30 1.15 1.30 1.15 3.30 3.77 2.44 2.63 2.51 3.77 1.51 2.65 0.042 0.053 0.027 0.069 0.047 0.053 0.082 0.032

drift at the top of the frame divided by the frame total height. The maximum value scored by this percentage simply defines the so-called static actual built-in overstrength, o , for each frame as designed. Fig. 4(a) gives this relationship for 4- and 8-story SFB and PFB frames designed according to the CCD approach. Fig. 4(b) shows the same relationship for the four MCD case study frames for comparison purposes. Note that in order to properly recover base shear from column shear forces when performing displacementcontrolled lateral analysis with P effects included, a correction technique is introduced through a post-processing step [20]. It entails reducing at each displacement step in the analysis the sum of shearing forces retrieved in the columns by the summation of the horizontal components of the axial loads in these columns. This total horizontal component at the base of the frame is simply generally computed by summing the product of the recorded axial forces by the corresponding rotation (small angles approximation) for each column. This rotation is basically obtained by dividing the interstory lateral drift by the story height. SFBs are generally stiffer and less sensitive to P effects in their pre-peak performance compared to PFBs since the gravity loads are tributary directly to the lateral resisting system and not through gravity-only interior columns that are non-seismically designed. On the other hand, the descending branch in the base shearRDR curves (Fig. 4) is though generally steeper for the SFBs relative to same height PFBs for both CCD and MCD approaches. This is mainly due to deeper, and hence more efficient, column sections (in the direction resisting bending moments resulting from lateral loading) for the latter, and accordingly more detrimental and accelerated post-peak strength deterioration for the former due to increasing P effects associated with further lateral pushing of the frames. A summary of o values for all case study frames is given in Table 4. Reported also in Table 4 are maximum Interstory Drift Ratios (IDRmax ) recorded at an RDR of 2%. As a result of having same design base shear for strength calculations in both CCD and MCD procedures in conjunction with the adoption of larger cross sections for the CCD frames to satisfy restrictive code interstory drift requirements as previously mentioned, the MCD approach usually yields more flexible (and further less strong ) frames than these developed using the CCD approach. It may be also generally observed that the actual intrinsic static overstrength, o , for space frames is usually larger than that for perimeter frames as expected by intuition; this is mainly the result of the significant effect of considerable gravity loads mobilizing more detrimental P effects for the case of the perimeter frames (captured through the leaning column technique), thus resulting in a considerable loss in their lateral capacity relative to the space frames of same height. This is as well a proof of the strength dominance in the design, and hence the lateral capacity, of the space frames compared to the perimeter frames for which drift requirements are more importantly controlling the seismic lateral design. Furthermore, it has been noted that when following the proposed MCD procedure, o , of the 8-story PFB largely decreased

2400

S.S.F. Mehanny, H.A. El Howary / Engineering Structures 32 (2010) 23942406

(a) CCD frames.

(b) MCD frames.

Fig. 4. Static inelastic displacement-controlled pushover analysis results for case study frames.

from 1.88 to 1.16 due to the reduced cross sections dimensions, and thereafter the magnified tortuous effects of the P phenomenon. However, on the other hand, o for 8S-SFB remains almost constant for both CCD and MCD approaches. Such observation shows the relatively lower P effects on the lateral capacity of the SFBs compared to the PFBs for these mid-rise RC frames designed for moderate seismic regions, and further reinforces the fact that such proposed relaxed design technique (MCD) does not penalize the built-in overstrength (i.e., the strength reserve) for the SFBs. On the other hand, as shown in Table 4, the inherent static overstrength remarkably increases with decreasing number of stories for the space frame construction when following the same design procedure (i.e., either CCD or MCD), while it increases less sharply for perimeter frames. This observation is justified since it could be again directly related to the decrease in the role of the P effects for low-rise buildings along with the gravity dominance in design relative to lateral drift demands. It is finally worth keeping in mind that the design of almost all frames (except 4S-CCD- and MCD-SFB) is controlled by drift limitations and not by strength requirements. Another general observation to report is that whenever IDRmax scores a large value at large RDR, and the associated significant nonlinear demand is localized in a single story, this story is usually the ground floor. On the other hand, when the large nonlinear demand is distributed among two or more stories for a given frame, IDRmax associated with a fairly large RDR is much lower than its value for the case of localized nonlinear demand. An example could be extracted from Table 4; IDRmax at 2% RDR scores a considerably lower value of 0.032 for 8S-MCD-SFB relative to the value of 0.069 scored for the 8S-CCD-SFB. The latter value is concentrated at the ground floor with other stories scoring much lower values, while the value of 0.032 for 8S-MCD-SFB is almost constant for the first three stories. It is though worth reporting that occasionally unexpected differences in response could be attributed to human factor involving some minor changes in design decisions made throughout the design process. Results of the pushover analysis presented above for studied perimeter and space frames are in line with other recent published data in [21,22] where low- and mid-rise RC moment framed buildings have been thoroughly investigated. In addition, similar to results highlighted in [23] where low-rise 3-story one-way (i.e., perimeter) and two-way (i.e., space) steel frames have been studied, the current research shows that space frames are stiffer (regarding the elastic stiffness) and stronger than their equivalent perimeter frames (Fig. 4). Moreover, Tagawa et al. [23] further reported a calculated factor referred to by DCF giving the first story Drift Concentration Factor that is defined as the first story drift angle divided by the roof drift angle. Similar factor analogous to DCF could be computed from results given in the current paper by dividing the IDRmax (reported in last column of Table 4) by the corresponding Roof Drift Ratio of 0.02 (i.e., 2% RDR), and similar trend to

what has been reported in [23] is then observed. In brief, this quotient for the space frames is consistently larger than for the perimeter frames at this RDR, except for the mid-rise 8-story MCD frame. A point that could be of some importance to justify many of the results presented herein is that the ratio of gravity to lateral tributary area for the space frame in a SFB is 1.0 while this ratio is much smaller (=0.25) for the perimeter frame in a PFB. In the latter configuration, the gravity load of half of the building is mobilizing P effects placing extra demands on this perimeter frame (captured through the leaning column technique). Generally speaking, results of pushover analyses presented in this paper reflect the following sources of overstrength: (1) minimum stiffness (drift) criteria, (2) structural redundancy, (3) strong-columnweak-beam criterion, among other sources commonly identified in the literature and recognized by seismic provisions worldwide. In addition to reported o values in Table 4, there is an inherent overstrength factor included in the design base shear calculation in this research. This factor is attributed to the accidental torsion specified by the code as has been previously pointed out in the literature [1]. Accordingly, the actual overstrength of the frames is higher than the values previously presented. Updated (i.e., adjusted) overstrength value, o , for each case study frame is determined as the product of o and T and is also presented in Table 4. T simply refers to the overstrength introduced by the accidental torsion. This updated o is the actual overstrength and is the one affecting the response since all assessment static pushover and time history analyses herein are based on a two-dimensional configuration (with no torsional effects). This adjustment approach is therefore necessary so that the actual R-values, as defined by the code, can be extracted relying on the time history results presented in what follows. Note that T takes different values for SFBs and PFBs since, in the current research, the first interior frame is the one considered for the former while the perimeter frame is the one investigated for the latter as mentioned earlier. This will further contribute to the final conclusions made in this research. 8. Incremental dynamic analysis, target performance levels and fragility curves Seismic performance is further assessed through nonlinear time history analyses using the set of 20 SMLR acceleration records presented above. For multi-level seismic hazard analyses, it is assumed that the acceleration component of the records can be linearly scaled based on the spectral acceleration computed at the fundamental period of the structure, Sa(T1). Shome and Cornell [24] have demonstrated that, compared to other approaches, scaling based on Sa(T1) will reduce the record-torecord dispersion in the response data and will not bias the results

S.S.F. Mehanny, H.A. El Howary / Engineering Structures 32 (2010) 23942406

2401

especially when the response of interest is the IDR. The spectral accelerations of the scaled earthquake records, Sa(T1), can be related to the maximum interstory drift ratio, IDRmax , depicting the peak response from corresponding time history analyses providing what is referred to in the literature as Incremented Dynamic Analysis (IDA) curves [25]. Fragility Curves (FC) constitute a representation of the relationship between (a) the probability of a set of Performance Levels (PL), or limit states, being reached or exceeded at a prescribed system demand and (b) the system demand itself. For a conventional performance-based seismic analysis, the system demands are typically represented by (or, are corresponding to) various ground motion severities or Intensity Measures (IM). A recently promoted efficient IM is typically represented by the spectral acceleration at the fundamental period of the structure, Sa(T1) as previously highlighted. On the other hand, the structural performance limit states of interest can vary from Immediate Occupancy (IO), to Life Safety (LS), to Collapse Prevention (CP) as per FEMA 356 definitions [7], and even up to complete failure of the structure. PLs are generally represented in the literature by a given Engineering Demand Parameter (EDP). The EDP adopted in the current study is a semi-global parameter given by IDRmax . The fragility function is basically thus giving the probability that a particular PL is exceeded (reflected by recorded IDRmax exceeding a prespecified IDRtarget set by specialized seismic provisions as will be clarified in what follows) conditioned on Sa(T1) (simply referred to by P [IDRmax IDRtarget |Sa(T1)]). Characterized by a lognormal distribution, fragility curves developed in this paper represent a cumulative distribution function defined by the median IM correa (T1), and the dispersion given sponding to exceeding a given PL, S by the standard deviation of the natural log, (lnSa(T1)), both of which are obtained from IDA data. In brief, FCs in this research are based on the two-parameter lognormal distribution function to get an S-shape curve. This approach was used by several researchers in the literature (e.g. [14,26]) and proved to give precise results. Qualitative structural performance levels: IO, LS, and CP mentioned above are reported in [7]. For RC frame structures, FEMA 356 recommendations further give a quantitative format for these PLs through assigning to them deterministic interstory drift limits of 1%, 2%, and 4% of the story height for IO, LS, and CP performance levels, respectively. Although these suggested limits are approximate, they are deemed fairly reasonable for buildings designed for seismic loading [27]. For the sake of the current study of low- to mid-rise ductile RC moment resisting frames located in moderate seismic zones, it is practically adequate to further assume that checking conformity with IO performance level corresponding to IDRmax = 0.01 could be associated with the 10% in 10 yrs hazard of [3,4]. Similarly, the 10% in 50 yrs (i.e., the design level earthquake as set by most seismic codes worldwide) and the 2% in 50 yrs events could hence represent the hazard associated with LS and CP performance levels corresponding to IDRmax = 0.02 and 0.04, respectively. Fig. 5 shows samples of developed FCs for 8-story SFB and PFB frames designed according to both CCD and MCD approaches for various PLs introduced above. For similar data related to the 4-story case study frames not given herein for space limitations one could refer to [19]. Note that the smaller (i.e., milder) slopes of FCs depict more uncertainty in the system. From Fig. 5, it could be thus easily observed that there is much more uncertainty associated with CP limit state, as expected, followed by LS performance level, and then finally by IO reporting the least uncertainty. Moreover, it is obvious from Fig. 5 that MCD frames are more vulnerable to damage than their corresponding CCD frames at a given IM reflected by a particular Sa(T1), especially at severe PLs such as CP; and similarly, PFB frames are more prone to damage than their equivalent SFB frames for either CCD or MCD approaches. One should however note that FCs of the type shown in

Fig. 5 only reflect (and consider) record-to-record variability and do not account for modeling uncertainties and other aspects of the ground motions such as consideration of the spectral shape ( parameter) introduced in the literature by Baker and Cornell [28]. FCs accounting for these additional sources of uncertainties have been developed in the literature by different researchers either for RC ductile and ordinary moment frames [21,22] or for actively and passively controlled structures composed of Steel ductile moment resisting frames [29]. These studies concluded the importance of the inclusion of such uncertainty sources for an accurate performance prediction of structures located in high seismic zones. However, in a very recent study [30] on the seismic fragilities of RC frames in regions of moderate seismicity, it has been concluded that fragilities that are developed under the assumption that all structural parameters are deterministic and equal to their median (or mean) values are sufficient for purposes of earthquake damage and loss estimation in regions of moderate seismicity. Not accounting for modeling uncertainties in the fragility analysis in the current research may be therefore justified since all case study frames are located in moderate seismic zones. Moreover, considering capacity and modeling uncertainties in the FCs only changes the slope of the curve while its center (i.e., the median corresponding to P [PL|Sa(T1)] = 0.5) remains unaltered [29] which keeps our conclusions based on median values, and presented in the next section, generally unaffected. This is basically the classic first-order assumption identified in [31] which may be considered a valid approximation although not entirely true; a recent study by Liel et al. [32] shows that there are some situations where the epistemic uncertainty may shift the median. 9. Evaluation of R and Rd values endorsed by ECP 201 [equivalent to q and qd values of EC 8] In this section, two previously highlighted hazards are considered and are provided in Fig. 6 for illustration purposes. A brief important description of these two hazard levels is presented below while Table 5 reports the actual numbers depicting these hazard levels in terms of Sa(T1). Sa(T1) will be referred to from now on by Sa for brevity. The first hazard level is Sa(10% in 50 yrs) that is specified by [3] and other similar worldwide building codes seismic provisions as the design-basis earthquake in which the performance objectives for a given building are limited to structural damage (i.e., Life Safety). The value of Sa(10% in 50 yrs) is directly extracted from ERS in [3] for each case study frame. The code R-value is eventually based upon this hazard level; therefore, the product of o d (or, more precisely o d ) shaping R at this level should be consistent with the code, where d is the dynamic component of R as demonstrated in Fig. 6. On the other hand, the second hazard level of interest is Sa(2% in 50 yrs) as specified in US seismic provisions (e.g. [5]) referring to the maximum capable earthquake in which the performance objectives for a building are near collapse (or in brief, the associated implied behavior is that the structure will maintain Collapse Prevention for this 2% in 50 yrs hazard). Values of Sa(2% in 50 yrs) are derived for case study frames based on the work in [33,34] relating the two above mentioned hazard values for Egypt as follows; Sa(10% in 50) 0.55 Sa(2% in 50). This value of 0.55 is the equivalent of the 2/3 factor adopted in [5] for the US and further endorsed by California practice [1]. Relying on information provided by IDA results and generated a (LS) values corresponding to IDRmax reaching or exFCs, median S a (CP) values correceeding IDRtarget = 0.02, as well as median S sponding to IDRmax reaching or exceeding IDRtarget = 0.04, may

a (IO) values that could be easily determined. For completeness, S be related to IDRtarget = 0.01 (and hence associated with a 10% in 10 yrs hazard) could be also retrieved, although relating this IO

2402

S.S.F. Mehanny, H.A. El Howary / Engineering Structures 32 (2010) 23942406

(a) 8S-CCD-PFB frame.

(b) 8S-CCD-SFB frame.

(c) 8S-MCD-PFB frame.

(d) 8S-MCD-SFB frame.

Fig. 5. Fragility curves for investigated 8-story CCD- and MCD-PFB and SFB frames. Table 5

a (PL)/Sa(hazard) ratios for case study frames for various performance levels. Summary of S
Building type IO performance level Sa (10% in 10) (g) Moment frames as per CCD procedure 4S-CCD-PFB 0.118 4S-CCD-SFB 0.141 8S-CCD-PFB 0.077 8S-CCD-SFB 0.082 Moment frames as per MCD procedure 4S-MCD-PFB 4S-MCD-SFB 8S-MCD-PFB 8S-MCD-SFB 0.097 0.141 0.059 0.069 2.97 2.84 1.95 3.05 2.49 2.70 1.19 2.31 0.211 0.252 0.137 0.147 0.173 0.252 0.105 0.123 2.84 2.98 2.92 3.06 2.77 2.98 1.14 2.44 0.383 0.457 0.249 0.267 0.314 0.457 0.191 0.223 3.13 2.84 3.41 3.45 3.12 2.84 1.15 2.29
a (IO) S Sa(10% in 10)

LS performance level Sa (10% in 50) (g)

a (LS) S Sa(10% in 50)

CP performance level Sa (2% in 50) (g)

a (CP) S Sa(2% in 50)

performance level to a specific earthquake hazard and to a particular IDRtarget is not as apparent as LS and CP levels. To examine the implications of these data on the safety/reliability of trial designs developed in this research, three limit state ratios are rea (IO,) S a (LS), and S a (CP) to the hazported in Table 5 comparing S ard accelerations Sa(10% in 10), Sa(10% in 50), and Sa(2% in 50), a (PL)/Sa(hazard) ratios simply furnish respectively. Note that the S margins against satisfying various code expected performance levels associated with different hazards of interest. These ratios fairly exceed 1.0 for all frames which indicates that all case study frames perform well in excess of code expectations at the various PLs investigated herein. One should realize that a lower value for this ratio (or for this margin) for any particular frame at one PL relative to the other (e.g., LS versus CP) indicates that this PL governs the design of the frame. A general note worth pointing is that referring to

Table 5, all case study frames (irrespective of the height, frame type or design procedure) feature almost consistent value of that margin at the three PLs of interest, except for one or two frames. Such exception even only occurs for the IO performance level whereas consistency in that margin still holds for the other two (LS and CP) levels. Such observation points to a certain homogeneity and uniformity preserved by the code seismic design procedure as well as by the MCD procedure proposed in this paper. Moreover, relying on such more-or-less uniformity in these reported margin values, one could further draw some global conclusions based on average values of these margins calculated for the three considered performance levels (IO, LS and CP) for each case study frame. One could then notice that the lowest average margin is scored for the 8S-MCD-PFB (1.16); whereas it was on average 2.76 for the same frame in the CCD procedure (i.e., roughly

S.S.F. Mehanny, H.A. El Howary / Engineering Structures 32 (2010) 23942406

2403

a (CP or LS) median where W is the seismic weight of the building. S values are easily determined for each frame from the IDA results and fragility curves referred to above. Furthermore, Rd can be determined as
Rd =

inelastic corresponding to LS IDR elastic corresponding to LS IDR

R.

(6)

Fig. 6. Schematic of the R-factor components and hazard levels considered.

2.4 times larger). This reflects the effect of the significant reduction in the concrete dimensions due to the MCD procedure, and hence the magnified tortuous P effects. Enforcing ECP 201 restrictive limitations on design interstory drifts (by imposing the 0.2ag lower bound to the code DRS while checking drifts in the design procedure whereas ignoring the upper bound on ds similar to that set in [4] and explained in Section 3) will apparently resolve this issue but this is not the correct approach. It has been previously demonstrated (refer to SD,DRS in Eq. (2) and Fig. 1(b)) that such lower bound on acceleration in absence of the upper bound on ds results in unrealistic spectral displacement values for mediumto long-period structures. Furthermore, these displacements seem even unbounded with the increase of the period of vibration which is physically impossible. Therefore, it is still beneficial to follow the proposed MCD procedure (i.e., pursuing same trend as illustrated by SD,ERS in Eq. (3) and Fig. 1(b)) but performing alternatively the design considering a priori lower R values. Such step is recommended in order to guarantee a reasonably larger margin of safety against reaching or exceeding LS and CP performance levels for these mid-rise perimeter framed buildings. On the other hand, other case study frames (pertaining to 4-story SFB and PFB buildings and 8-story SFB buildings) are observed to have margins against satisfying IO, LS and CP performance levels more than double those scored by the 8S-MCDPFB frame (Table 5). These case study frames further reveal that following the MCD procedure with code-specified R value of 7 is still a good option that keeps a more economic design than the CCD procedure yet with an adequate margin of safety against exceeding these different performance levels. Moreover, it may be noted from a (PL)/Sa(hazard) ratios associated with LS and CP Table 5 that S performance levels remain practically unaltered for the studied 4-story frames (either SFB or PFB) irrespective of the adopted design method (CCD or MCD); this is an evidence that lowrise structures are less prone to damage that may result from detrimental P effects. In a complementary step for evaluation of codes design procedure implications, estimates of R and Rd are then obtained (i.e., extracted) from the detailed nonlinear inelastic static (pushover) and time history (IDA) analyses supplemented by FCs. For instance, an estimate for R may be obtained as R= Base shear corresponding to a given PL (CP or LS) Design base shear, Vdesign (4)

and adopting a SDOF approximation, this relation might be simplified as R=

a (CP or LS) S
Vdesign /W

a (CP or LS) S
Sa,design

(5)

IDRinelastic in Eq. (6) due to a given record can be generally easily extracted from the IDA results; it is simply equal to 0.02 in the context of this study since this value corresponds to the limit set a priori for the LS performance level. On the other hand, IDRelastic may be calculated by performing an elastic time history analysis of the structure under the same record scaled to Sa (T1 ) corresponding to LS performance level (i.e., corresponding to IDRinelastic = 0.02). values Retrieved Rd /R (calculated as the ratio of the median, IDR, of IDRinelastic and IDRelastic for the set of 20 records as per Eq. (6)) is given in Table 6 for each of the case study frames. Note that Rd /R values, if calculated for CP hazard level, would hold less significance since checking drifts in all worldwide seismic provisions is seldom related to such a high hazard. In the current study, Rd /R ratio retrieved from the two-dimensional dynamic analysis is therefore only evaluated at the LS performance level. Referring to Table 6, overall average Rd /R ratio for all MCD and CCD ductile moment resisting frames studied herein could be calculated and is found to range between 1.1 and 1.2. This ratio could be directly compared to ECP 201 recommended value of 0.7 [3], or to the value of 1.0 (supporting the concept of the Equal Displacement Rule) proposed in EC8 [4]. Note that the 1.1 value in the range reported above refers to the Rd /R ratio averaged only for the four case study frames dimensioned according to the CCD procedure; while the 1.2 represents the value of the Rd /R ratio averaged for the same four frames but when designed according to the proposed MCD technique. Both average values for this ratio support applying the Equal Displacement Rule, such as in [4], for more compatible, representative and realistic estimates of the inelastic displacement demand based on the elastic demand. An Rd /R value of 1 shall therefore be promoted instead of keeping the current ECP 201 endorsed value of 0.7 [3]. Such recommendation is basically also to ensure some conservatism in the design estimates of the inelastic displacement demands when applying seismic design procedures in [3]. It is also worth reporting that regular low- to mid-rise structures studied herein (featuring uniform distribution of mass and stiffness) have been found to show a first mode dominated behavior with limited localization of drift demands. As a result, Rd /R ratio scores practically close values whether estimated (using Eq. (6)) based on maximum IDR or roof drift values. This result is particularly also due to the fact that Rd values are basically retrieved in the current research at LS performance level at which only little plastification occurs and hence limited localization takes place. Rd /R ratio retrieved per Eq. (6) for the types of frames investigated herein accounts for both global (i.e., roof drift) versus local (i.e., IDR) effects (though with a marginal resulting difference) together with elastic versus inelastic effects. For completeness, same steps have been implemented to estimate as well Rd /R ratios corresponding to the IO performance level associated with IDRinelastic = 0.01, and a value of about 1.1 still holds, on average, for all (CCD and MCD) case study frames. This result further confirms that it is certainly more suitable to apply the EDR in order to deliver representative and conservative estimates of inelastic displacement of these ductile moment frames relying on elastic displacement values obtained from a linear analysis. Alternatively, as mentioned before, the R value may be rather simply disaggregated into the product of its static, previously defined, part o (or, more precisely o ) and its other dynamic

2404

S.S.F. Mehanny, H.A. El Howary / Engineering Structures 32 (2010) 23942406

Table 6 Summary of retrieved R and Rd values for the case study frames for the LS performance level. Building type Sa,ULT (g)

d (LS)

R(LS) = d (LS) o

R (LS) = d (LS) o

Rd /R

Moment frames as per CCD procedure 4S-CCD-PFB 0.165 4S-CCD-SFB 0.190 8S-CCD-PFB 0.122 8S-CCD-SFB 0.133 Moment frames as per MCD procedure 4S-MCD-PFB 4S-MCD-SFB 8S-MCD-PFB 8S-MCD-SFB 0.126 0.190 0.075 0.133 3.63 3.94 3.27 3.39 3.83 3.94 1.59 2.25 9.2 12.9 6.2 7.8 7.4 12.9 1.9 5.2 12.0 14.9 8.0 8.9 9.6 14.9 2.4 6.0 1.0 1.3 0.8 1.3 1.1 1.3 1.4 1.0

component, d as shown schematically in Fig. 6. d is calculated as follows

d =

a (CP or LS) S
Sa,ULT

(7)

where Sa,ULT is given (Table 6) as Vu /W following the same SDOF approximation previously mentioned (refer to Fig. 6) and can be extracted from the pushover analysis results. Now that all parameters have been properly defined, estimates of R are computed for each of the case study frames at each of the two previously defined hazard levels of interest, LS and CP, and values are presented in Tables 6 and 7, respectively. In an attempt to reconcile the code-specified R value adopted a priori in the design with expected (and modeled) behavior using data extracted from the static and dynamic inelastic analyses described above, R values reported in Tables 6 and 7 referring to LS and CP performance levels, respectively, are compared to a value of 7 and 12.7, respectively. Implied are the facts that (1) the code [3] specified R value of 7 is calibrated for the LS performance level corresponding to the design-basis earthquake; while as (2) the value of 12.7 associated with CP level is based on the fact that the 2% in 50 yrs level is simply (1/0.55=) 1.82 times that of the 10% in 50 yrs level [34]. Referring to Table 4, note that R values based on o (instead of simply R values based on o ) and reported in Tables 6 and 7 are more realistic values that rationally represent the actual behavior and that are more adequate to be compared to codespecified R values. It may be accordingly observed that R (LS) values computed for 4-story (i.e., low-rise) case study frames of either the SFB or the PFB type designed according to either the CCD or the MCD procedure show a great deal of conservatism when compared to the code value of 7 (Table 6). Level of conservatism in these versions of the case study frames is even more pronounced when looking into the CP performance level (Table 7) with all reported R values significantly higher than 12.7. It is though worth reporting that the dynamic part, d , of R is more contributing to the final value of R than its static counterpart, o . On the other hand, it may be noted that the performance of the 8-story (mid-rise) CCD case study frames either of the SFB or the PFB type just marginally exceeds that implied by the code-specified R of 7 calibrated for LS performance level. Retrieved performance of the 8S-CCD-PFB and 8S-CCD-SFB frames shows R (LS) values of 8.0 and 8.9, respectively. The margin to code R value is though improved for these two frames at the CP level as shown in Table 7 where retrieved R (CP) values of 17.0 and 18.2, respectively, are recorded compared to a corresponding target value of R of 12.7 at this PL. The only exception to the so far quite adequate performance guaranteed by code pre-specified R values takes place for midrise 8-story frames designed according to the MCD technique proposed in this research. To be more specific, retrieved R values of 2.4 and 6.0 (Table 6) associated with LS level for 8S-MCD-PFB and 8S-MCD-SFB, respectively, emphasize the non-conservative

current code value of R = 7.0. Same poor results persist for the same frames at the CP level as shown in Table 7. It is therefore highly recommended that whenever this MCD procedure is promoted, the relevant code provisions should assign lower values of R for mid-rise moment resisting frame buildings especially of the perimeter framed type. It may however be more prudent to propose this reduction explicitly in the overstrength component of R rather than inherently in the aggregated R value that includes combined contributions from overstrength, ductility, besides other factors of relevance. In addition, an investigation for generating adequate (and efficient) P factors to be considered in the design phase in line with the MCD technique presented herein is undergoing by the authors. The recommendation for lower R values may be also preferably equally applicable to mid-rise CCD moment frame buildings (either of the SFB or PFB type) especially if it is required that code R values maintain a certain level of conservatism in the design. To illustrate the need for such recommendation, one should note that if the values of R (LS) = 8.0 and 8.9 retrieved for 8S-CCD-PFB and 8S-CCD-SFB, respectively, and reported in Table 6, were not calculated based on median values of Sa(LS), but were instead estimated from IDAs and FCs based on an Sa(LS) value corresponding to a lower conditional probability of exceeding the IDRtarget associated with LS level (i.e., corresponding to P [IDRmax IDRtarget = 0.02|Sa(T1)] < 0.5 referring to the relevant FC), we would have ended up with smaller d (LS), and hence lower R (LS). This would have probably lead to R (LS) values less than 7.0 thus revealing a non-conservative code R value for the design of these CCD mid-rise moment frames. In general, the FCs make the approach presented herein more versatile and capable of encompassing wide range of estimates for retrieved R values corresponding to different probabilities. Considering lower conditional probabilities as highlighted above and studying their consequent effects is under scrutiny by the authors. Nonetheless, it is important to highlight that (1) the adoption of only rectangular sections along all beams in the current computer models of the investigated frames when performing inelastic pushover and dynamic analyses rather than L- or T -sections where deemed appropriate based on the continuously changing direction of the bending moment demands, along with (2) the use of minimum specified (i.e., nominal) materials strength properties instead of expected values, may have contributed to a relatively ), and hence low value of actual built-in overstrength, o (or o to an underestimation of actual retrieved R values. The first point was adopted in analysis to avoid overly complicated models along with further ad hoc assumptions related to the effective flange width to be considered with L- or T -beams, that at best would have added extra uncertainty to the subject instead of improving modeling and actual structure realization. In addition, regarding the second point related to material properties, it was decided by the authors to rather rely on minimum specified values to account for occasionally poor production of construction materials

S.S.F. Mehanny, H.A. El Howary / Engineering Structures 32 (2010) 23942406 Table 7 Summary of retrieved R values for the case study frames for the CP performance level. Building type Sa,ULT (g)

2405

d (CP)
7.27 6.84 6.96 6.93 7.81 6.84 2.92 3.82

R(CP) = d (CP) o 18.5 22.4 13.1 15.9 15.1 22.4 3.4 8.8

R (CP) = d (CP) o

Moment frames as per CCD procedure 4S-CCD-PFB 0.165 4S-CCD-SFB 0.190 8S-CCD-PFB 0.122 8S-CCD-SFB 0.133 Moment frames as per MCD procedure 4S-MCD-PFB 0.126 4S-MCD-SFB 0.190 8S-MCD-PFB 0.075 8S-MCD-SFB 0.133

24.0 25.8 17.0 18.2 19.6 25.8 4.4 10.1

in the local market. The decisions made herein for these couple of modeling issues will add some conservatism to the current estimation of R and R values reported in this paper which should be generally accepted for design codes evaluations/studies. 10. Summary and conclusions Different trial designs of ductile low- to mid-rise (4- and 8-story) RC moment resisting frames located in moderate seismic zones (0.25 g) have been implemented using both space and perimeter frame configurations according to emerging Egyptian seismic code that is in line with Eurocode 8 seismic provisions. Codes controlling design criteria (including strength versus stiffness criteria, as well as currently imposed constant lower bound on design acceleration and missing upper bound on calculated actual expected displacement) have been addressed along with their implications on the frames seismic performance. A series of static inelastic pushover analyses have been performed on the case study frames. In addition, incremental dynamic analyses have been carried out for each investigated frame under a bin of 20 small magnitude large distance ground records in a multi-level analysis context. Fragility curves are hence developed for the case study frames corresponding to the probability of exceeding various performance levels of interest including Immediate Occupancy, Life Safety and Collapse Prevention levels conditioned on an efficient Intensity Measure defined by Sa(T1). Code-Compliant Designs as well as a proposed Modified Code Design relaxing design drift demands are examined to test their effectiveness and reliability. A major contribution of this paper is the evaluation of the structural response modification factors, R and Rd , using advanced inelastic static and time history analyses along with prespecified hazard levels, and comparing them to code proposed values for ductile moment resisting frames. The paper shows that either CCD or MCD procedure yields life-safe moment resisting frames that perform well relative to seismic performance objectives and hazard levels implied by current design codes. Moreover, retrieved margins against satisfying various code expected performance levels associated with different hazards of interest (IO, LS, and CP) exceed 1.0 for all case study frames. These margins further show consistency across all performance levels of interest for each investigated frame thus pointing to reasonable uniformity and homogeneity furnished by the two design procedures adopted herein. Moreover, results of the inelastic static and dynamic analyses permit disaggregation of the R value into its component parts provided by static overstrength, o , and inelastic dynamic response, d . It may be noted that d values are generally larger than o values for nearly all cases investigated herein at either the LS or CP performance levels. It has been further noted that current code R value may be non-conservative for some of the buildings investigated in this paper. In other words, the one size fits all approach in the selection of response modification factors may result in significant underestimations of internal forces and seismic demands of

moment frames of various types or height. For example, it has been shown that lower R values would be better assigned by the code for the design of mid-rise (8-story and most probably higher) moment frames either of the space or of the perimeter type, with a special emphasis to reduce this R value for the perimeter frame configuration particularly if the proposed MCD procedure is adopted. It may be somewhat conceptually more appropriate for the case of MCD technique to propose such reduction explicitly in the overstrength component of R rather than intrinsically in the aggregated R value. The analyses also indicate that inelastic displacements would be more accurately (and conservatively) estimated for ductile moment resisting frames designed per either the CCD or the proposed MCD procedure using an equal displacement rule (implying that Rd = R), rather than through the inelastic displacement factor given as 0.7 in the current provisions (i.e., Rd = 0.7R) [3]. Such recommendation shall be well perceived by the design community and users of ECP 201 seismic provisions especially since there is a precedent for adopting this equal displacement rule in EC8 [4] which is a parent code for ECP 201 as well as in other major building codes worldwide and in other previous published researches (e.g. [1]). It is however worth reporting that Rd /R results presented in the current research are mainly applicable (and promoted) to first mode dominated structures with no irregularities in either mass or stiffness and with limited expected localization of drift demands up to Life Safety performance level. Furthermore, it has been proven that it would be better to impose the constant lower bound of 0.2ag on the design acceleration only when designing for strength of different structural elements. If this bound on the design acceleration response spectrum is equally enforced while checking drift, it will result in unrealistically large values of drift demands unless an upper bound to the expected (actual) displacements is implemented. In other words, the expected actual displacement, ds , need not be larger than the value derived from the elastic spectrum. Such upper bound on displacement set in [4] is moreor-less equivalent to the MCD procedure proposed in this research and is also of primary importance when estimating actual drifts since it guarantees realistic displacement values for medium- to long-period structures especially in case the 0.2ag constant lower limit to the design acceleration response spectrum is enforced in the design procedure. It is nonetheless important to realize that this study and the conclusions thereof are so far only valid for low- to mid-rise moment resisting framed buildings up to 8-story high located in moderate seismic zones (0.25 g). Extrapolation of the results presented herein to other seismic zones, to near-fault motions and/or to higher moment framed buildings shall be the subject of a similar effort before being either applied or denied. References
[1] Mehanny SS, Cordova PP, Deierlein GG. Seismic design of composite moment frame buildings case studies and code implications. In: Engineering foundation conference: composite construction IV. Banff (Alberta, Canada): ASCE; 2000. p. 11.

2406

S.S.F. Mehanny, H.A. El Howary / Engineering Structures 32 (2010) 23942406 [19] El Howary HA. A probabilistic framework for assessing seismic performance of reinforced concrete moment frame buildings in moderate seismic zones. M.Sc. thesis. Structural Engrg. Dept., Egypt: Cairo University; 2009. [20] FEMA-451. NEHRP Recommended provisions: design examples, prepared by the building seismic safety council for the federal emergency management agency. National Institute of Building Sciences, Washington (DC, USA); 2006. [21] Haselton CB. Assessing seismic collapse safety of modern reinforced concrete moment frame buildings. Ph.D. thesis. Department of civil and environmental engineering. CA (USA): Stanford University; 2006. [22] Liel AB. Assessing the collapse risk of Californias existing reinforced concrete frame structures: metrics for seismic safety decisions. Ph.D. thesis. Dept. of Civil and Env. Engrg., Stanford Univ., CA (USA); 2008. [23] Tagawa H, MacRae G, Lowes L. Probabilistic evaluation of seismic performance of 3-story 3D one- and two-way steel moment-frame structures. Earthquake Eng Struct Dyn 2008;37:68196. [24] Shome N, Cornell CA. Probabilistic seismic demand analysis of nonlinear structures. Rep. no. 35. Reliability of marine structures program. Dept. of Civil and Env. Engrg., CA: Stanford University; 1999. [25] Vamvatsikos D, Cornell CA. Incremental dynamic analysis. Earthquake Eng and Struct Dyn 2005;31(3):491514. [26] Shinozuka M, Feng MQ, Kim HK, Kim SH. Nonlinear static procedure for fragility curve development. J Eng Mech, ASCE 2000;126(12):128795. [27] Ramamoorthy SK. Seismic fragility estimates for reinforced concrete framed buildings. Ph.D. dissertation. Texas (USA): Civil Engineering Dept., Texas A&M University; 2006. [28] Baker JW, Cornell CA. A vector-valued ground motion intensity measure consisting of spectral acceleration and epsilon. Earthquake Engineering and Structural Dynamics 2005;34(10):1193217. [29] Taylor E. The development of fragility relationships for controlled structures. M.Sc. thesis. Dept. of Civil Engineering. Saint Louis (MO, USA): Washington University; 2007. [30] Celik OC, Ellingwood BR. Seismic fragilities for non-ductile reinforced concrete frames role of aleatoric and epistemic uncertainties. Struct Safety 2010; 32(1):112. [31] Cornell CA, Jalayer F, Hamburger RO, Foutch DA. The probabilistic basis for the 2000 SAC/FEMA steel moment frame guidelines. J Struct Engrg ASCE 2002; 128(4):52633. [32] Liel AB, Haselton CB, Deierlein GG, Baker JW. Incorporating modeling uncertainties in the assessment of seismic collapse risk of buildings. Struct Safety 2009;31:197211. [33] Ove Arup & Partners international. Seismic hazard assessment for LNG project at Damietta port in Egypt. London (UK): Ove Arup & Partners International, Ltd., London Operating Centre; April, 2002. [34] Mehanny SSF, Gendy AS, Seif SP. Seismic assessment of bridges over the river Nile. fib Symposium on Concrete Structures in Seismic Regions. May 69, 2003. Athens (Greece); 2003.

[2] Rivera JA, Petrini L, Lai CG. Do the Eurocode 8 force-based design provisions lead to the safe and predictable seismic response of RC frame buildings? In: The 14th world conference on earthquake engineering. Oct. 1217, Beijing, China; 2008. [3] ECP 201. Egyptian code of practice for loads on buildings and bridges. Ministry of Housing, Utilities and Urban Communities of the ARE. Housing and Building Research Center, Egypt; 2003. [4] Comit Europen de Normalisation, Eurocode 8 CEN. Design of structures for earthquake resistance. Part 1: general rules, seismic actions and rules for buildings. In: European Committee for Standardization. Doc. EN 1998-1, Brussels, Belgium; 2005. [5] International code council. 2003 International building code, Falls Church, Virginia (USA); 2003. [6] American institute of steel construction. AISC seismic provisions for structural steel buildings. 2nd ed. Chicago (IL, USA); 1997. [7] FEMA-356. Prestandard and commentary for the seismic rehabilitation of buildings. In: Federal emergency management agency. Washington (DC, USA); 2000. [8] International conference of building officials (ICBO). 1997 Uniform building code (UBC). vol. 2. Whittier (CA, USA); 1997. p. 574. [9] American society of civil engineers. Minimum design loads for buildings and other structures. ASCE/SEI 7-05. ASCE standard no. 7-05. Virginia (USA). 2005. p. 424. [10] ECP 203. Egyptian code of practice for design of reinforced concrete structures. Ministry of housing, utilities and urban communities of the ARE. Housing and building research center, Egypt. 2007. [11] Mehanny SS, Deierlein GG. Modeling and assessment of seismic performance of composite frames with reinforced concrete columns and steel beams. J.A. Blume Earthq. Eng. Center. report no. 135. Stanford CA (USA). 2000. [12] Medina R. Seismic demands for non-deteriorating frame structures and their dependence on ground motions. Ph.D. thesis. Depart. of Civil and Env. Engrg. CA (USA): Stanford University; 2002. [13] Gupta A, Krawinkler H. Seismic demands for performance evaluation of steel moment resisting frame structures (SAC Task 5.4.3), John A. Blume Earthq. Eng. Center. Report no. 132. CA (USA): Stanford University; 1999. [14] Chenouda M, Ayoub A. Probabilistic collapse analysis of degrading multi degree of freedom structures under earthquake excitation. Eng Struct 2009; 31(12):290921. [15] Mackie KR, Stojadinovi B. Fragility basis for California highway overpass bridge seismic decision making. PEER Report 2005/02. Berkeley (CA, USA): University of California; 2005. [16] OpenSees. http://opensees.berkeley.edu Web page. [17] Karsan ID, Jirsa JO. Behavior of concrete under compressive loading. J Struct Div Proc ASCE 1969;95(ST12):254363. [18] Paulay T, Priestley MJN. Seismic design of reinforced concrete and masonry buildings. NY (USA): John Wiley & Sons, Inc.; 1992.