Engineering Structures 23 (2001) 319332 www.elsevier.

com/locate/engstruct

Seismic design of R/C buildings with the aid of advanced analytical techniques
Andreas J. Kappos
a

1,a,*

, Alireza Manafpour

Department of Civil Engineering, Aristotle University of Thessaloniki, 54006 Thessaloniki, Greece b Department of Civil and Environmental Engineering, Imperial College, London SW7 2BU, UK Received 24 January 2000; received in revised form 5 May 2000; accepted 8 May 2000

Abstract A seismic design procedure considering performance criteria for two distinct limit states is presented, involving analysis of a feasible partial inelastic model of the structure using currently available powerful tools. The procedure is developed in a format appropriate for incorporation into modern design codes, such as the Eurocode 8, and two alternatives are explored, one involving timehistory analysis for appropriately scaled input motions, and a simpler one involving inelastic static (pushover) analysis. The proposed method is found to lead to better seismic performance than the standard code procedure, at least in the case of regular multistorey reinforced concrete frame structures studied herein, and in addition leads to a more economic design of transverse reinforcement in the members that develop very little inelastic behaviour even for very strong earthquakes. 2000 Elsevier Science Ltd. All rights reserved.

1. Introduction Since inelastic behaviour is typically permitted under the design earthquake, and dynamic characteristics and consequently earthquake-induced actions change with time, use of analytical procedures accounting for these two key features in the design process appears to be a reasonable choice. However, despite the fact that the advantages of using inelastic timehistory analysis for design purposes have been recognised since the early seventies [1], code drafting committees have been reluctant to adopt this type of analysis, even as an alternative to the standard (static or dynamic) elastic analysis procedure. Along with several other seismic codes currently in force, the American UBC [2] was based until 1994 on elastic analysis for equivalent static forces or a design spectrum. The 1997 edition of this code introduced for the rst time some detailed provisions for the application of both elastic and inelastic timehistory analysis, mainly with regard to the selection and scaling of ground
* Corresponding author. Tel.: +30 31 995 743; Fax: +30 31 995 614. E-mail address: ajkap@civil.auth.gr (A.J. Kappos). 1 Formerly, Reader at the Department of Civil and Environmental Engineering, Imperial College, London, UK.

motions. A similar development took place in the recent NEHRP Provisions [3] but in this case timehistory analysis is clearly restricted to structures with seismic isolation. The seismic Eurocode, EC8 [4], recognises that inelastic timehistory analysis might be used in the design procedure, but guidance is only given with regard to the selection of input accelerograms and the way they should be scaled to match the design spectrum. No clear indication is given in EC8 as to what model(s) should be used, or which response quantities should be sought, if such an analysis is used. On the other hand, the New Zealand Code [5] is clear in specifying that the purpose of using inelastic timehistory analysis might be either to calculate strength requirements in yielding members, or assess inelastic demands and/or capacity actions. However, guidance on the possible type of modelling that may be adopted and on the way the results should be interpreted is minimal (even in the otherwise very detailed Commentary). A signicant step forward regarding the use of inelastic timehistory analysis in a practical context was the publication of the NEHRP Guidelines for Seismic Rehabilitation (i.e. Strengthening) of Buildings [6]; this document not only promotes the use of inelastic (static as well as dynamic) analysis methods for assessment of existing structures, but, importantly, provides detailed

0141-0296/01/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 1 4 1 - 0 2 9 6 ( 0 0 ) 0 0 0 5 2 - 3

320

A.J. Kappos, A. Manafpour / Engineering Structures 23 (2001) 319332

guidance regarding both inelastic modelling and evaluation of results in terms of recommended values for local inelastic deformations. These guidelines do not address the issue of designing new buildings. The main objective of the present study is to suggest a new procedure that offers some advantages with regard to previous ones. Both the new procedure and that of EC8 are then applied to a multistorey reinforced concrete (R/C) frame, and the designs are evaluated using an assessment procedure including multiple criteria, representing different limit-states. 2. Evaluation of existing procedures The fact that code-drafting committees have long been reluctant towards adopting inelastic analysis for design purposes did not discourage the development of various methodologies in this direction [711]. The potential application of a design method based on inelastic timehistory analysis, which is the most sophisticated analytical tool currently available, depends heavily on the computational cost, expressed both in terms of computation time and (even more importantly nowadays) on the time required for data preparation and checking, as well as for interpreting the results. A key point in assessing the efciency of a design method is whether it is an explicit or an implicit one. By explicit it is understood that the method provides a set of closed-form equations and/or a one-cycle procedure for selecting cross-section dimensions and (in R/C structures) reinforcement of members. It is noted that, even with currently applicable code procedures, a second analysis is usually carried out to optimise the design in an approximate and practical way, and this should not be deemed to imply that the method is not explicit. A method is implicit if a number of iterations are required to select member cross-sections and/or reinforcement, starting from an initial design (typically of the code-type). With one possible exception [9], all the aforementioned methods are of the implicit type, and obviously they are associated with high computational costs; this appears to be one of the main reasons why these methods have not been adopted by existing codes. Available commercial software packages make it possible for current seismic design practice to tackle realistic structures by means of rather sophisticated (typically 3D) analytical models. Therefore, it is believed that new and/or improved design methods should be built around these sophisticated models, rather than rely on the construction of additional models, whether simplied or not. In other words structural design should preferably be carried out using essentially the same model for all limit states and types of loading. Under this perspective, earlier methods involving partial models consisting of one or more storeys of a frame [7] offer no real advantage.

In addition to convenience and practicality considerations, the ability of the selected model to adequately describe the dynamic response of the structure to be designed should also be addressed. Early methods [1] involved the dynamic analysis of SDOF inelastic systems, and these models form also part of some modern methodologies, generally falling within the displacement-based design category [12,13]. While there is no doubt that SDOF models are well-suited for simple structures such as isolated bridge piers, it is equally clear that they are not suited for structures with more than one signicant modes, such as buildings with stiffness irregularities in plan and/or in elevation. On the basis of the foregoing discussion, it seems appropriate to state that a practical design method intended for general use should be able to treat most of the commonly encountered structural congurations, consequently it should be built around models capable of describing the dynamic response in the fundamental, as well as in higher modes. Finally, an important consideration in evaluating a design method should be the way performance criteria, in particular the degree of seismic damage, are checked. There are at least two issues that have to be addressed in this respect. The rst issue is whether damage is checked explicitly, using some appropriate damage index based directly on inelastic deformations (and/or hysteretic energy dissipation), as suggested by Hatamoto et al. [10], or, alternatively, performance is checked based on quantities indirectly related to damage, such as global displacements (e.g. interstorey drifts) or even forces (e.g. ratios of elastic force demands to member resistances). A second, related, issue is whether the method permits checking of performance locally, as well as globally. Whereas macroscopic quantities such as interstorey drifts are good indicators of both local and global damage in regular structures, this is not generally the case in more complex and/or irregular structures. In all cases it is important that the design procedure recognise the weak links and the designer be given the opportunity to remedy the situation through appropriate detailing. 3. Suggested procedure A procedure based on timehistory analysis and developed in a format appropriate for incorporation into modern design codes, was rst presented by the rst author [14] specically for the capacity design of columns in R/C buildings, and detailed comparisons with several existing capacity design procedures were made. This procedure forms a critical part of the method presented in the following, which however constitutes a comprehensive proposal for performance-based design of R/C buildings; moreover, a simpler alternative of the method, based on pushover analysis is also presented, and both methods are evaluated through a case-study involving a multistorey R/C frame.

A.J. Kappos, A. Manafpour / Engineering Structures 23 (2001) 319332

321

Some key features of the proposed method are: Use of a reduced (or partial) inelastic model, which, although departing from standard approaches (at least for xed base structures), is nevertheless feasible, even in the case of 3D structures, and can be used in a design ofce environment. Explicit treatment of two limit states (or performance levels), with corresponding performance criteria at global, as well as local, level. Depending on the conguration of the structural system, simple (equivalent static) or more rened (set of input accelerograms) loading patterns can be used.

3.1. Outline of the method The various steps in the suggested methodology are summarised below; it is assumed that the structure has already been designed to satisfy code requirements under normal (gravity, wind, environmental) loading. 1. Flexural design of the beams of the structure for the seismic action under which the structure is required to remain essentially in the elastic range, combined with the appropriate (quasi-permanent) gravity loading. For usual buildings this action can be taken as a fraction n0 (varying from about 2/3 to 3/4) of the serviceability earthquake having a 50% probability of exceedance in 50 years; lower probabilities are appropriate for critical facilities. The n0 factor is intended to provide (in combination with the minimum reinforcement requirements) a level of strength to the structure adequate for satisfying the serviceability criteria of step 5. Design moments are calculated from an elastic analysis based on the fundamental mode or multiple modes, depending on the structural system, and stiffnesses of members are estimated assuming moderate amount of cracking. If required, some moment redistribution is carried out with a view to optimising beam design. If structural walls are present, their critical regions (typically including the rst one or two storeys) are also designed for exure, using moments (and axial loads) calculated from the foregoing analysis. 2. Detailing of the exural reinforcement of beams (and wall critical regions), taking into account minimum requirements and convenience of construction. This step establishes a basic strength level for the structure, as the strength of the remaining members strongly depends on that of beams (see following steps). 3. Selection of an appropriate set of input accelerograms, using techniques similar to those described in modern seismic codes [2,4,6]. Either articial, spectrum-compatible, or (preferably) actually recorded,

motions, can be used, and a minimum of three records is recommended. 4. Construction of a model of the structure wherein beams are modelled as yielding elements, with their strength based on the reinforcement actually present (including that in the adjacent slab), and with due consideration of factors such as stiffness degradation; for beams designed according to modern seismic design practice strength degradation is normally not an issue, at least for the limit states considered herein. In the same model, columns, as well as portions of walls (when present) intended to remain elastic, are modelled as elastic members. With regard to initial stiffness assumptions, xed percentages of the gross section rigidity EIg for each member type may be used for convenience, but somewhat more sophistication can be justied in selecting the stiffness of beams, whose reinforcement is already known. In practice, the same model can also be used for the rst set of analyses (step 1), with beams assumed to be of innite strength (i.e. to respond elastically). 5. Timehistory analysis of the model described in the previous step for the selected set of input motions scaled to the intensity of the serviceability or immediate occupancy [6] earthquake; this earthquake is normally associated with a probability of exceedance of 50% in 50 years. Checking of the following performance criteria: Max drifts do not exceed the limits corresponding to damage requiring repair in the non-structural elements (specic values for normal buildings are recommended later in the paper). If the drift criterion is not satised at any storey, stiffening of the structure is necessary; this can be done by increasing the cross-section dimensions and/or reinforcement (in the second case the effective stiffness of the beams should be calculated using expressions involving the reinforcement ratio). Plastic rotations in beam critical regions do not exceed the value corresponding to non-tolerable cracking (typically that requiring repair). If the specied ductility limits are exceeded in some members, the corresponding reinforcement is increased. It is emphasised that both criteria have to be satised, as their role is complementary, the one mainly referring to damage in the non-structural elements and the other to damage in R/C members. It is also noted that care is required in dening this serviceability limit state which is not the same as the corresponding one under gravity loading; the latter refers primarily to durability requirements, whereas the former involves some visible damage in both structural and non-structural elements, but the structure generally does not require structural repair and also can be occupied immediately after the earthquake.

322

A.J. Kappos, A. Manafpour / Engineering Structures 23 (2001) 319332

6. Timehistory analysis of the same model (with beam reinforcement revised if required during the previous step) for the selected set of input motions scaled to the intensity of the repairable damage or life safety [6] earthquake. For normal buildings (i.e. not for essential facilities) this event is typically taken as the one corresponding to a probability of exceedance 10% in 50 years (this is the design earthquake in most current codes [2,4,5]). This analysis provides the critical moment (M) and axial load (N) combinations for each column (and wall) critical section. 7. Design and detailing of longitudinal reinforcement in columns (and walls) of the structure. For a column subjected to biaxial loading, consideration of the following three combinations will be sufcient for most practical purposes (My and Mz are the moments acting along the two main axes of a column): maxMy, and corresponding Mz and N maxMz, and corresponding My and N minN (compression) or max N (tension), and corresponding My and Mz. For uniaxial loading two combinations will sufce. The foregoing are valid on condition that the axial load limitations imposed by modern codes are respected; these limitations should be checked using the min N (max compression) calculated in the time history analysis. 8. Design and detailing of all members for shear, using the shear values calculated in step 6, multiplied by a gR factor (slightly larger than unity) to account for an earthquake intensity higher than that having a 10% probability of exceedance. 9. Detailing of all members for connement, anchorages and lap splices, using design equations involving the level of inelasticity expected in each member. Target ductility of non-dissipating members (columns and upper portions of walls) is taken as that of limited ductility structures (e.g. ductility class L in EC8).

fundamental periods of the structure under consideration, appropriately correlated to the SIm of the design earthquake [16]. More specically, it is recommended to calculate SI starting from the initial (elastic) period of the structure and going up to the anticipated prevailing period during the inelastic response; as a general guide the latter can be taken as 1.3 times the initial period, since relatively minor yielding is expected under the limit states considered in the proposed method. For very tall and/or exible buildings higher modes might be dominating the response in some storeys, and use of the standard Housner SI or a revised range of periods covering at least the rst three modes is more appropriate. In the absence of a set of response spectra corresponding to given probabilities of exceedance, the design spectrum specied by the applicable code for checking the serviceability earthquake can be used. The current version of EC8 implies that the SLS is checked for 1/2 to 1/2.5 the ULS seismic forces, but this is currently under revision. Some more detailed guidance for selecting earthquake forces corresponding to different probabilities of exceedance is given in the NEHRP Guidelines for Existing Buildings [6].

3.2. Practical application of the method The following additional information is necessary for the implementation of the method described in the previous section: 3.2.1. Denition of earthquake loading The use of actually recorded ground motions is recommended, to avoid inconsistencies relating to articial spectrum-compatible records [15]. Whenever elastic response spectra for the site are available for given probabilities of exceedance, the input motions will be scaled to match the spectrum intensities (SI) of the corresponding pseudovelocity spectra. The suggested procedure for scaling the records is based on a narrow-band spectrum intensity (SIm) calculated in the region of expected

3.2.2. Modelling of members The sophistication of the basic analytical model should be compatible with the purpose of the analysis, as well as feasible for practical design purposes. Hence, inelastic beams should preferably be modelled as lumped plasticity elements with point hinges at their ends; whenever concentrated loads and/or large spans are involved, the results from the inelastic analysis should be checked (generally by hand calculations) to detect possibility of hinging within the beam span. In cases of signicant span hinging, subdivision of the member into two point hinge elements might be required. The effective rigidity assumed for each member of the structure should be consistent with its intended behaviour and the models used. It is recommended [6,16,17] to use 3050% of the gross exural rigidity (EIg) for beams, 5080% EIg for the walls, and 6080% EIg for columns (mainly in compression), to account for member cracking, which is different in each type of member; the higher values for columns and walls apply for high axial compression. Simple equations involving the longitudinal reinforcement ratio are also available for cracked section properties of beams and columns, while section analysis using appropriate stressstrain curves for steel and concrete [17,18] can also be used and element stiffness be set to the secant rigidity at yield EIef=My/fy (yield moment divided by yield curvature), hence ignoring tension-stiffening effects. It is pointed out that using fully cracked section properties for the entire member, rather than some average value, might lead to underestimation of ductility demands [16,18].

A.J. Kappos, A. Manafpour / Engineering Structures 23 (2001) 319332

323

3.2.3. Global and local deformation criteria The recommended interstorey drifts to be considered for the serviceability check range from 0.2 to 0.5% the storey height, depending on the type of partitions used; values close to 0.2% are recommended for brick masonry inlls in contact to the surrounding frame [19] whereas 0.5% is more appropriate for plywood, plaster, gypsum and similar light panels [20]. The recommended plastic hinge rotations of yielding members (beams, and, where applicable, walls) under the same earthquake can be taken from the NEHRP Guidelines [6] as 0.005 rad for beams, and 0.0010.005 rad for walls, depending on geometry, axial stress and shear stress levels. Alternatively, the check can involve rotational ductility factors, with recommended values between 1.0 and 2.0. It is recalled herein that adoption of the foregoing local deformation criteria implies that some minor yielding will take place in some locations and some cracks of about 1mm will also appear. It is clear that although this type of damage does not prevent immediate use of the building after the earthquake (possibly after some clean-up), it is nevertheless incompatible with durability requirements; in the latter case maximum crack widths generally do not exceed 0.30.4 mm for normal environmental conditions. Therefore, the fact that a building satises the previously described serviceability check does not preclude the possibility that some non-structural and/or cosmetic repairs might be required, should such an earthquake occur. 3.2.4. Ductility-dependent detailing The nal step of the suggested procedure is the detailing of members for ductile behaviour on the basis of equations directly accounting for the anticipated level of inelasticity. An example of such equation is the Eurocode 8 wd formula for connement reinforcement in columns (and wall critical regions) that involves as a key parameter the target curvature ductility factor mf. It is beyond the scope of this paper to present a complete set of design equations for connement, anchorage, and splicing in terms of ductility factors; some proposals intended primarily for assessment (rather than design) were put forward by Priestley [21] but a lot of work remains to be done in this area. In the case study included herein, the Eurocode 8 equations for buckling prevention, and bar anchorage and splicing pertaining to Ductility Class M are used for simplicity, but the need for developing more rened, ductility-dependent, expressions is emphasised. 3.3. Potential advantages of the method The main advantages of the previously described method compared with procedures specied in codes currently in force appear to be the following:

It provides, through the model used for the timehistory analysis, for many of the uncertainties regarding the strength hierarchy at beamcolumn joints, which are treated in codes by introducing empirical overstrength (gRd) factors, which are generally (but not necessarily) conservative [14]. Unlike current code methods, the proposed procedure accounts explicitly for phenomena such as the shift of the contraexure point in columns due to higher mode effects (in the inelastic structure) and/or non-uniform yielding of top/bottom beams, and the actual uctuation of the axial loading. In structures with complex dynamic behaviour, for which modal analysis is imposed by the codes, the proposed method offers the advantage of providing realistic combinations of moment and axial force in the columns, hence avoiding the need for code-type combination rules which are often cumbersome and tend to penalise columns excessively. It is interesting to note that inclusion of the vertical component of the earthquake is perfectly feasible in the suggested methodology, and its combined effect with the horizontal component(s) can be assessed in a more realistic way than in existing code procedures. As a result of the previously mentioned factors, the capacity design of columns becomes more effective, as will be shown in the case study of Section 4. Possible advantages of the suggested method compared with some existing procedures intended to account properly for inelastic deformations (see Section 2) are: Reliable and costeffective inelastic models for R/C beams are currently available and can be used in the framework of the suggested procedure, while this is not generally the case with available models for columns and walls (especially those of low slenderness). The proposed method requires an explicit account of inelasticity in these members only in a few critical sections (at the base of the structure) and is not sensitive to the type of modelling adopted for these few sections. Beam-type models are already available in numerous 2D-analysis programs, as well as in some 3D programs, both academic and commercial. Hence it can be claimed that the method can be applied in the case of realistic buildings. In contrast, all methods reviewed in Section 2 involve dynamic analysis of a complete inelastic model of the structure (columns included), hence they are effectively restricted to 2D frames; even in this case the reliability of the available models for columns is quite low in many cases [22]. Unlike other procedures currently under development, such as some displacement spectra-based methods, in the inelastic dynamic analysis-based methodology proposed herein no need exists for reducing the struc-

324

A.J. Kappos, A. Manafpour / Engineering Structures 23 (2001) 319332

ture to an equivalent SDOF system, or for constructing additional simpler models; in fact the basic analytical model used is essentially the same one used for ordinary code design, and directly provides design actions for all members. Last but not least, an advantage of the proposed procedure over existing ones, both code and non-code type, is the efcient selection of combinations of limit states and performance criteria. Unlike existing procedures, the proposed method requires more explicit checks for the serviceability earthquake, which the authors have found more difcult to control than earthquakes involving yielding in selected parts of the structure. Note that no specic verication of local deformations is necessary for the life safety earthquake, while incipient collapse or collapse prevention [6] limit states do not have to be checked explicitly. As will be made clear from the case studies presented here, dening reliably the plastic mechanism and controlling plastic deformation requirements under low earthquake levels, typically ensures that performance criteria are also met for higher intensities. 3.4. Limitations of the method As with all methods, there are some restrictions and some delicate points in the proposed method, namely: Although the method is based on an analytical procedure going beyond those of existing codes, it does not represent a total break from standard code approaches that relate the level of ductile detailing to the level of force used for designing the structure; in other words some detailing provisions of existing codes are adopted at the current stage of development of the method. These can and should be revised (to achieve better economy of materials) in future developments. An appropriate selection of (three to four) input accelerograms is a prerequisite for the successful application of the method. A large amount of work has been done in this direction, and it is not the purpose of this paper to assess or even to review it. Nevertheless, attention is drawn to recent works where it is shown that for realistic structure-specic damage computation (in probabilistic terms), it is sufcient to conduct dynamic analyses for only a few records [23,24]. The cases studied so far have indicated that the proposed method is not particularly sensitive to the number and even the characteristics of the selected accelerograms (see Section 4). The corollary of one of the main advantages of the method, namely the use of simplied inelastic beam models, is that the method depends on the assumption regarding the initial stiffness of these simple inelastic

members. While xed percentages of EIg may be used for convenience, it is understood that more rened assumptions would increase the reliability of the analytical model. For assessment purposes the recommended procedure is to rst carry out an analysis based on average stiffness values (as recommended previously), which typically lead to maximum demands for forces and ductilities, and then proceed with a second analysis based on fully cracked properties (and possibly due allowance for bond-slip and shear deformations), which will usually lead to critical demands with regard to displacements [18,24]. Given that this would almost double the required analytical effort (which is already higher than in current code methods), for the design of new structures it is recommended to use a single analysis (for the design earthquake intensity) using the higher stiffness values. The number of MdNd combinations to be used in section design is a problem for this, as well as for all other methods; even in direct displacement-based design, values of forces are required for exural design of members. The very simple rules given in step 7 of the previously described procedure have been found to be quite effective, at least in the case of regular multistorey buildings (see next section).

3.5. Inelastic static alternative Although the recommended use of the method involves timehistory analysis for a number of records, most of its essential features can be maintained if a simpler static pushover analysis is used instead. This is generally appropriate in the case of regular low- and medium-rise buildings with minor contributions from higher modes, in which case dynamic moment patterns can be reasonably captured and the prevailing plastic mechanisms be identied. Application of pushover analysis, with due consideration given to its limitations, can be considered as lying halfway between existing elastic methods and nonlinear time history analysis. Hence a carefully performed pushover analysis will provide insight into both methods in terms of level of reliability vs. complexity. In the pushover analysis-based procedure, step 3 of the suggested methodology is replaced by selection of a loading pattern for the analysis (e.g. triangular, uniform, exponential, or modal). Then at steps 5 and 6 a pushover analysis is performed up to a target displacement limit value for deformation demands corresponding to serviceability or immediate occupancy, and repairable damage or life safety earthquakes, respectively. Steps 7, 8 and 9 are applied as before. Having dened the loading pattern for the pushover analysis, target values of displacement are needed to establish a link with the various levels of seismic action

A.J. Kappos, A. Manafpour / Engineering Structures 23 (2001) 319332

325

that pertain to different limit states. Various methods have already been suggested for the estimation of an appropriate top storey displacement or drift ratio for the structure [6,12,21,25]. The procedure adopted in the recent NEHRP recommendations for seismic rehabilitation [6], based on appropriate correction factors applied to the elastic spectral displacement calculated for the effective lateral stiffness of the building (secant stiffness at effective yield) was found to be an appropriate option for the current study. It is recalled that the target displacement value is still a rather controversial parameter in the literature related to pushover analysis. Although the NEHRP procedure is generally applicable to different performance levels, it has mostly been calibrated and used for life safety earthquakes.

4. Trial application of the method To illustrate the potential as well as the limitations of the foregoing procedure, a case study is presented, involving a multistorey reinforced concrete building with a lateral load resisting frame system. The building that was rst designed to the elastic analysis-based EC8 procedure, was redesigned to the previously described method, and then assessed by subjecting it to various appropriately scaled input accelerograms, including both the ones used in the design (to the new method) and additional ones. 4.1. Structure studied The six-storey frame studied is part of an R/C building within importance category III according to EC8, assumed to be the central one in a series of frames, equally spaced at a distance of 3-m. The ductility class considered for the design according to EC8 is DC M. The frame was designed for a design ground acceleration of 0.25 g assuming class A soil (rock or stiff deposits) for the site of the building. The structure is symmetric in terms of stiffness in elevation. The materials used in the structure are C20/25 (characteristic cylinder strength of 20 MPa) concrete and S400 steel (characteristic yield strength of 400 MPa). The total dead and live loads on the oor slabs are assumed to be 4.0 and 2.0 kN/m2, respectively. The geometric characteristics of the system are shown in Fig. 1 (square cross sections are used for columns). At rst the frame was designed and detailed according to EC8 design provisions (EC8 design). The results concerning the exural design of columns are given in Table 1 together with details of the transverse reinforcement. In line with common practice, it was decided to keep members cross sections the same in (at least) every two stories.
Fig. 1. Initial geometry of 6-storey frame structure (units: mm).

4.2. Analytical modelling Inelastic analyses of the structure have been carried out using IDARC 4.0 [26], a computer program that adopts a member by member modelling approach. Inelastic beam and column members are modelled as inelastic elements with spread plasticity and yield penetration. However, as mentioned in Section 3.2, for practical design purposes one might use simpler models of the lumped plasticity (point-hinge) type. The cyclic behaviour of end cross-sections is represented by a degrading momentcurvature relationship built on a nonsymmetric trilinear envelope curve, which is created using a simple bre model and cross-sectional and material specications provided as input. In constructing the model as described in point 4 of Section 3.1, columns are modelled as elastic elements with an effective rigidity EIeff =0.80EIg, where EIg is the gross section value. The base of the columns (bottom section of ground storey columns) is also modelled allowing for inelasticity to occur. This is because these are potential plastic hinge regions. Mean values of strengths of the materials are used for calculating the resistance of inelastic members. Also for reinforcing steel a value of 594 MPa for ultimate strength, 1% for the strain at the initiation of strain hardening, and a strain hardening modulus equal to 1/60 the initial modulus of elasticity, are assumed. The effect of slab reinforcement lying within the effective width of the anged beams is taken into account in the design of columns, assuming a 0.2% reinforcement ratio parallel to the beam. For the hysteretic behaviour, the nominal stiffness and energybased strength degradation of IDARC4.0 with no pinch-

326

A.J. Kappos, A. Manafpour / Engineering Structures 23 (2001) 319332

Table 1 Results for design of columns in EC8 frame Exterior columns Storey b (mm) h (mm) Longitudinal Re. As=As (cm2) Arrangement of reinforcement Transverse Re. 1 and 2 350 350 2f16+1f18 6.57 3 and 4 350 350 2f16+1f18 6.57 5 and 6 300 300 2f16+1f14 5.56 Interior columns 1 and 2 450 450 4f20 12.57 3 and 4 450 450 2f22+1f20 10.74 5 and 6 400 400 2f20+1f16 8.83

f8@90

f8@90

f6@70

f8@120

f8@120

f8@110

ing effect were selected, reecting the good hysteretic behaviour of members designed to modern code provisions. It is worth pointing out that although IDARC and similar programs are at present used primarily for research purposes, and are regarded with scepticism by some designers, the explicit introduction of the inelastic analysis option in recent guidelines such as the 1997 NEHRP [6] has led to integration of the inelastic beam and column models in commercial software that is increasingly being used in design ofces, for the time being mainly for the assessment of existing structures and the verication of major new design projects. 4.3. Input motions Following the discussion in Section 3.2, natural earthquake ground motions were selected and scaled as described in the following. Among the various parameters affecting the selection of a set of suitable records, one that can provide information on the relative frequency content and duration of ground motions as well as structural response characteristics, is the ratio of peak ground acceleration to velocity, a/v, suggested by Zhu et al. [27]. Therefore, appropriate records to be used for design and assessment were chosen from the Imperial College Strong Motion Database (ICSMD) mainly considering the following parameters: Magnitude: about 6.57. Site condition: since class A soil according to EC8 was assumed for the site, records on alluvia sites were not considered. a/v ratio: considering a moderate range of period for structures under study it was attempted to use the records lying within the medium range of a/v (about 0.81.2). PGA: generally more than 0.25 g (design earthquake level) up to 0.50 g (about twice the design one).

The selected records are shown in Table 2. The rst three records were used for the design of the structure, while the rest (47) were subsequently used for assessing its performance. Three is the minimum number of records recommended by current design codes for time history analysis [2]. As discussed in Section 3.2, a modied Spectrum Intensity method [16] was chosen for scaling the selected earthquake records to the intensity of the EC8 elastic spectra anchored to the appropriate level of peak ground acceleration. The results for the 0.25 g spectra, corresponding to life safety earthquake level, are shown in Table 2. While the life safety earthquake is described by the unreduced code spectrum, the serviceability earthquake was taken as 1/2.5 the code spectrum, along the lines suggested in EC8. Hence the intensity used for designing the yielding regions (step 1 of the method) was taken as 2/3 of the previous value, i.e. the elastic EC8 spectrum divided by 3.75; this conveniently corresponds with the behaviour factor (q=3.75) used for designing the structure to EC8 ductility class M, and allows more meaningful comparisons between the proposed method and the standard EC8 design (i.e. essentially the same level of beam strengths is obtained in both designs). Also shown in Table 2 is the effective duration of each record used for inelastic time history analysis. This duration is similar to that resulting from the Trifunac and Brady [28] procedure. 4.4. Critical aspects of design 4.4.1. Inelastic dynamic method Due to the way in which the serviceability earthquake was dened, the exural reinforcement in the beams and at the base of the ground storey columns was the same as in the EC8 design. A rst series of analyses involving the partial inelastic model were then carried out by applying the rst three records of Table 2, scaled to the intensity of the serviceability earthquake. The maximum

A.J. Kappos, A. Manafpour / Engineering Structures 23 (2001) 319332

327

Table 2 Input earthquake hground motions for design and assessment Earthquake File name Station Comp. Mag. ED PGA (Ms) (km) (g) 7.1 6.7 6.5 7.2 6.5 7.1 6.9 8 6 27 25 13 0.348 0.344 0.313 0.307 0.274 0.201 0.504 PGV (cm/s) 33.50 40.40 32.09 31.69 26.03 32.42 76.27 PGA/PGV Duration Site (s) geology 1.04 0.85 0.96 0.97 1.05 0.62 0.66 26.5 16.0 10.0 13.5 18.5 19.5 10.5 Scaling factor

1 Imperial Valley, ELC 15/5/1940 2 Northridge, 17/7/1944 NORE 3 Friuli, Italy, 6/5/1976 FRIT 4 Kobe, Japan, 17/1/1995 5 San Fernando, 9/2/1971 6 Loma Prieta, 17/10/1989 7 Erzincan, Turkey, 13/3/1992 KOBL SFERT LPRL ERZN

El Centro Aleta Fire St. TolmezzoAmbiesta 1 Kobe Uni. Old Ridge Road Presidio Metrological St.

S00E N90E N90E N90E N69W N90E NS

Stiff soil 0.53 Stiff soil 0.45 Rock 0.67 Rock Rock Rock 0.60 0.72 0.80

Stiff soil 0.39

Fig. 2. Maximum response under three serviceability level earthquakes for partial inelastic modelling of 6-storey frame. Drift values are shown for each earthquake.

response in terms of rotational ductility factors, damage index and interstorey drift are shown in Fig. 2 and the corresponding peaks are summarised in Table 3. Rotational ductility (mq) is dened in the usual way as the ratio of the maximum to yield rotation at the end sections of the member: mq qmax qp 1 qy qy (1)

has not yielded, mq is dened as the ratio of maximum exural moment to yield moment (Mmax/My). A modied ParkAng damage model [26] is also used as a damage index: qmaxqr Eh DIPA b quqr Myqu (2)

where qmax, qy and qp are the maximum, yield, and plastic rotation of the section, respectively. For a section that

where qu is the ultimate rotation capacity of the section, qr is the recoverable rotation when unloading and Eh is the dissipated energy in the plastic hinge. Element damage is taken as the largest damage index of the two end

Table 3 Maximum responses of partial inelastic model under three serviceability level earthquakes Earthquake Max. interstorey drift (%) Top displacement (mm) Global drift mqmax of beams (%) (+) ELC NORE FRIT
a

mqmax of columnsa

Damage index

( ) 0.60 0.57 0.72 0.49 0.46 0.64

Beams 0.024 0.021 0.034

Columns 0 0 0.041

Global 0.018 0.016 0.027

0.14 0.13 0.23

22.32 21.28 33.13

0.12 0.12 0.18

0.64 0.60 0.84

for the base of columns.

328

A.J. Kappos, A. Manafpour / Engineering Structures 23 (2001) 319332

sections. Storey and overall damage indices are computed using weighting factors based on the dissipated energy at component and storey levels, respectively [26]. Interstorey drift is dened as the ratio of relative horizontal displacement of two adjacent oors (d) and corresponding storey height (h). It can be seen from Fig. 2 and Table 3 that maximum value for interstorey drift ratio does not exceed 0.23%, which is acceptable for most inll types [19,20]. Moreover, maximum rotational ductility factors in beams and the base of the columns are less than 0.9, which means that under serviceability-level earthquakes yielding will not occur in the members. Hence, no modication of the original design was necessary and the design process continued into step 6 (see Section 3.1) of the proposed method. Critical values for exural design of columns are calculated, as described in step 7, by means of an in-house developed post-processing program that derives the critical values from the time history outputs produced by IDARC4.0 for columns. Flexural design of columns is based on the conventional design procedure using design values for strength of materials. It is noted that, since the behaviour of the structural system is deemed to have been modelled in a realistic way by introducing realistic material specications for beams (mean values for strengths, strain hardening for steel, and hysteretic behaviour), no further

overstrength (gRd factor) is introduced in the design of columns. Critical shear demands from the life safety level earthquake set, resulted from step 6 of the procedure, are shown in Table 4 for columns and critical regions of beams (these were then multiplied by the gRd=1.10 factor). Shear demands for beams were generally larger than corresponding values from the code procedure (note that mean beam strengths are directly considered in the inelastic analysis, as opposed to characteristic strengths in the EC8 procedure), however only small changes in the design of interior beams at lower stories were needed with respect to the original EC8 design. Longitudinal and transverse reinforcement design for columns is summarised in Table 5. As can be seen from Tables 1 and 5, design of columns to the proposed procedure resulted in an increase in longitudinal reinforcement, which is larger for interior columns than exterior ones, compared to the original EC8 design. However there was no need to revise the cross sectional dimensions. On the other hand, application of steps 8 and 9 of the proposed method results in considerable relaxation of transverse reinforcement in columns compared with conventional code-based design (see Tables 1 and 5). The most critical factor in this respect was the decision to comply only with the EC8 ductility class L (low) transverse steel requirements in the columns (with the exception of their base).

Table 4 Maximum shear forces (kN) in columns and beams for partial inelastic model of frame from time history and pushover analysis for life safety level Storey Timehistory analysis Columns Exterior 1 and 2 3 and 4 5 and 6 81.75 74.89 47.97 Interior 177.7 160.4 104.6 Beams Exterior 127.40 108.76 81.54 Interior 157.00 131.70 83.05 Pushover analysis Columns Exterior 85.55 74.03 43.51 Interior 184.52 157.69 96.05 Beams Exterior 128.13 111.20 81.84 Interior 158.23 132.40 83.00

Table 5 Results for design of columns after dynamic analysis of partial inelastic model Exterior columns Storey b (mm) h (mm) Longitudinal Re. As=As (cm2) Arrangement of reinforcement Transverse Re.
a

Interior columns 3 and 4 350 350 2f18+1f20 8.23 5 and 6 300 300 2f16+1f18 6.56 1 and 2 450 450 4f24 18.10 3 and 4 450 450 4f22 15.20 5 and 6 400 400 4f20 12.57

1 and 2 350 350 3f18 7.63

f6@120a

f6@140

f6@110

f8@180a

f6@180

f6@160

Except the base of the columns.

A.J. Kappos, A. Manafpour / Engineering Structures 23 (2001) 319332

329

4.4.2. Inelastic static method For the six-storey frame considered here the inelastic static procedure was applied using the same analytical model as described for the dynamic method. The calculated base shear vs global drift relation for the partially inelastic model of the frame is depicted in Fig. 3, together with its bilinear approximation. For this structure, having an initial elastic period T0=0.70 s, the effective fundamental period (Tef) was found to be 0.77 s, which results in a target displacement value [6] of about 75 mm (corresponding to a global drift ratio of 0.41%) for the life safety earthquake. The frame was then pushed to a top drift value of 0.41% and the force demands in the columns were calculated. The proportioning of column reinforcement and detailing of the members were carried out in a similar fashion as in the dynamic inelastic method. Shear demands from pushover analysis (see Table 4) slightly exceeded the values from the elastic static analysis (EC8) and were in agreement with those from inelastic dynamic analysis; they resulted in a small increase (with respect to EC8 design) in transverse reinforcement for beams, especially the internal ones. The cross-section dimensions of interior columns at the rst and second stories were increased by 50 mm. This was due to an attempt to keep the reinforcement ratios similar in all designs, hence a change in column dimension was made whenever the required ratio for main longitudinal reinforcement exceeded about 2%. Following a change of dimensions, ideally a second analysis is needed to take into account the effect of this change on the force distribution of columns and also other response parameters in beams. However, the resulting difference in column demands from small cross-section changes is typically not signicant and is often disregarded in practice. The pushover analysis alternative of the proposed method has produced higher demands in the columns of the lower stories than the upper ones compared with the dynamic one. This should be attributed to higher mode effects, which are not usually captured in simple pushover analysis. Transverse reinforcement for columns in this method is more or less similar to that from the dynamic analysis-

based method and again shows considerable relaxation compared with conventional code-based method. 4.5. Assessment of designed structures To assess the seismic performance of the frames designed to the new procedure, as well as the conventional one (EC8), all the frames were modelled as full inelastic systems and analysed for the seven earthquake

Fig. 3.

Base shear vs global drift for 6-storey frame.

Fig. 4. Interstorey drift ratio for frames designed based on conventional and proposed methods under seven life-safety (0.25 g) and collapse-prevention (0.50 g) earthquakes.

330

A.J. Kappos, A. Manafpour / Engineering Structures 23 (2001) 319332

ground motions of Table 2 with different levels of intensity. To check the reliability of the proposed procedure, an additional stronger seismic intensity was also considered, associated with a less frequent earthquake having a probability of exceedance of about 2% in 50 years; this may be referred to as survival or collapse prevention seismic level. No attempt was made to explicitly dene this earthquake in the current study and since the main purpose was the comparative assessment of different structures, the simple practice of doubling the lifesafety earthquake intensity was selected [14,17,18]. The effect of variation of axial load on the strength and stiffness of the columns is neglected in IDARC and the moment curvature relationship of the columns is evaluated at the average axial loading corresponding to grav-

ity loads. Effect of connement on concrete strength and ductility is taken into account using the conned concrete model suggested by Kappos [16]. Strength values based on mean material properties are assumed for all members, as typically done in deterministic assessment studies. The results for the serviceability (immediate occupancy) performance level showed that the maximum interstorey and global drift values were 0.24% and 0.19% respectively. No yielding occurred in either beams or columns. Interstorey drift values are compared in Fig. 4 for three different frames under the life-safety and collapseprevention events (seven records considered). Although all frames show almost similar behaviour under life-

Fig. 5. Storey damage indices for beams and columns in frames designed based on conventional and proposed methods under seven life-safety (0.25 g) and collapse-prevention (0.50 g) earthquakes.

A.J. Kappos, A. Manafpour / Engineering Structures 23 (2001) 319332

331

safety earthquakes, under the collapse-prevention earthquakes there exists more similarity between the two alternatives of the proposed method than between any of them and the code-based one. The proposed method has resulted in slightly less critical drift values than the conventional one. Fig. 5 summarises storey damage indices for beams and columns in the alternative designs, for two levels of earthquake. It is noticed that in the frames designed based on the proposed procedure (dynamic or static alternative) slightly more damage has occurred in the beams compared with the EC8 procedure. On the other hand, average damage values in columns designed to the new method are almost half the corresponding values for the conventional one. This performance is in agreement with the capacity design concept for frame structures and demonstrates the relative advantage of the proposed method. Of particular importance are the very low requirements in the columns under twice the design earthquake, indicating the feasibility of drastically reducing the connement reinforcement above the base of the structure. Probably the most notable differences can be seen in the distribution of plastic hinges throughout the frames under collapse-prevention earthquake level. It is clear from Fig. 6 that the frame designed according to the proposed dynamic method has a larger safety margin against the collapse limit state than its simpler pushover alternative, while both of them show improved performance compared with the EC8-based design. The hinge patterns are shown for the two most critical earthquakes one of which, the Loma PrietaPresidio record (see Table 2) had not been considered in the design of the structure.

5. Conclusions A new methodology of seismic design for R/C buildings has been presented, based on a feasible partial inelastic model of the structure, and the explicit consideration of performance criteria for two distinct limit states. The suggested new procedure was found to lead to a very satisfactory seismic performance, at least for regular multistorey frame buildings to which it has been applied. Behaviour under the life-safety earthquake was found easier to control than that under the serviceability earthquake, mainly due to the fact that the performance criteria adopted for the latter involved ductility requirements of members (beams) that are not capacity-protected. Performance under the collapseprevention earthquake was also assessed and found to be quite satisfactory. The simpler version of the proposed method, based on pushover analysis, offers a number of practical advantages, the main one that it does not involve selection and scaling of earthquake records for timehistory analysis. For the regular frame structures studied herein its results are not substantially different from those of the dynamic procedure, nevertheless this remark can not be generalised without further studies. The suggested procedure can readily be incorporated in the framework of a modern design code, such as EC8 or the UBC, and was found to lead to generally better seismic performance than the standard (elastic analysisbased) EC8 procedure. Although at its present state of development the new method does not explicitly incorporate optimisation, it offers the possibility to reduce cost of materials com-

Fig. 6. Plastic hinge distribution for frames designed based on conventional and proposed methods under the two more critical collapse-prevention (0.50 g) earthquakes.

332

A.J. Kappos, A. Manafpour / Engineering Structures 23 (2001) 319332

pared to that resulting from standard code methods. Of particular practical importance is the conrmation that, at least for the regular frame structures addressed herein, it is feasible to drastically reduce the amount of connement reinforcement in the columns above the base of the structure, without jeopardising their seismic performance.

References
[1] Blume JA. Elements of a dynamic-inelastic design code. Proceedings of 5th World Conf on Earthq Engng, vol. 2, Rome, Italy, 1973. pp. 225665. [2] Int Conf. of Building Ofcials. Uniform Building Code 1994 and 1997 Editions, vol. 2: Structural Engineering Design Provisions. CA: Whittier, 1994, 1997. [3] FEMA (Federal Emergency Management Agency, US). NEHRP Recommended Provisions for Seismic Regulations for New Buildings and other Structures (1997 Edition): Part 1 Provisions; Part 2 Commentary. Washington, DC, BSSC, 1998. [4] CEN Techn. Comm. 250/SC8. Eurocode 8: Earthquake Resistant Design of Structures Part 1: General rules (ENV 1998-11/2/3). Brussels: CEN, 1994, 1995. [5] Standards New Zealand. Code of Practice for General Structural Design and Design Loadings for Buildings (NZS 4203:1992). Wellington, 1992. [6] FEMA. NEHRP Guidelines for the Seismic Rehabilitation of Buildings, FEMA-273, Washington, DC, 1997. [7] Bertero VV, Zagajeski SW. Optimal inelastic design of seismicresistant reinforced concrete framed structures. In: Nonlinear Design of Concrete Structures (Int Symposium, Univ of Waterloo, Ontario, Canada), 1979. pp. 21972. [8] Fintel M, Ghosh SK. Explicit inelastic dynamic design procedure for aseismic structures. Journal of the ACI 1982;79(2):1108. [9] Capatina D, Titaru E. Possible extension of the structural seismic analysis. Proceed of 8th World Conference on Earthquake Engineering, vol. 4, San Francisco, CA, 1984. pp. 4518. [10] Hatamoto H, Chung SY, Shinozuka M. Seismic capacity enhancement of R/C frames by means of damage control design. Proceed 4USNCEE, vol. 2, Palm Springs, CA, 1990. pp. 27988. [11] Stone WC, Taylor AW. ISDP: integrated approach to seismic design of reinforced concrete structures. Journal of Structural Engineering, ASCE 1994;120(12):354866.

[12] Moehle JP. Displacement-based design of RC structures subjected to earthquakes. Earthquake Spectra 1992;8(3):40328. [13] Kowalsky MJ, Priestley MJN, MacRae GA. Displacement-based design of RC bridge columns in seismic regions. Earthquake Engineering and Structural Dynamics 1995;24(12):162343. [14] Kappos AJ. Inuence of capacity design method on the seismic response of R/C columns. Journal of Earthquake Engineering 1997;1(2):34199. [15] Naeim F, Lew M. On the use of design spectrum compatible time-histories. Earthquake Spectra 1995;11(1):11127. [16] Kappos AJ. Analytical prediction of the collapse earthquake for R/C buildings: suggested methodology. Earthquake Engineering and Structural Dynamics 1991;20(2):16776. [17] Pauley T, Priestley MJN. Seismic design of reinforced concrete and masonry buildings. New York: Wiley, 1992. [18] Penelis GG, Kappos AJ. Earthquake-resistant concrete structures. London: E & FN SPON (Chapman & Hall), 1997. [19] Valiasis TN, Stylianidis KC. Masonry inlled R/C frames under horizontal loading experimental results. European Earthquake Engineering 1989;3(3):1020. [20] Freeman SA. Racking tests of high rise building partitions. J Struct Div ASCE 1977;103(ST8). [21] Priestley MJN. Displacement-based seismic assessment of reinforced concrete buildings. Journal of Earthquake Engineering 1997;1(1):15792. [22] CEB Task Group III/6. Behavior and analysis of reinforced concrete structures under alternate actions inducing inelastic response, vol. 2: frame members. Lausanne: Bull. dInf. CEB, 1994. [23] Bazzuro P, Cornell CA. Seismic hazard analysis of nonlinear structures I: methodology. J Struct Engineering, ASCE 1994;120(11):32034. [24] Dymiotis C, Kappos AJ, Chryssanthopoulos MC. Seismic reliability of R/C frames with uncertain drift and member capacity. J Struct Engineering ASCE 1999;125(9):103847. [25] Miranda E. Estimation of maximum inerstorey drift demands in displacement-based design. In: Proc. Int. Workshop on seismic design methodologies for the next generation of codes, Bled, Slovenia. Rotterdam: Balkema, 1997:25364. [26] Valles R E, et al. IDARC2D version 4.0: a computer program for the inelastic damage analysis of buildings. NCEER, State Univ. of New York at Buffalo, NCEER-96-0010, 1996. [27] Zhu TJ, Heiderbrecht AC, Tso WK. Effect of peak ground acceleration to velocity ratio on the ductility demand of inelastic systems. Earthquake Engng Struct Dyn 1988;16:6379. [28] Trifunac MD, Brady AG. A study on the duration of strong earthquake ground motion. Bull Seismol Soc Am 1975;65(3):581626.