Sie sind auf Seite 1von 17

Engineering Structures 22 (2000) 12441260 www.elsevier.

com/locate/engstruct

Rened force reduction factors for seismic design


B. Borzi, A.S. Elnashai
*

Imperial College of Science, Technology and Medicine, Department of Civil and Environmental Engineering, Imperial College Road, London, SW7 2BU, UK Received 13 April 1999; received in revised form 19 July 1999; accepted 29 July 1999

Abstract Whereas seismic design based on deformations is a concept that is gaining ground, existing codes are fundamentally force-based, with a nal check on deformations. A central feature of force-based seismic design is the response modication factor (R or q). Many studies have attempted to quantify the potential of structural systems to delimit the level of force imposed by virtue of their ductility and energy absorption capacity. This paper employs a well controlled and evenly distributed earthquake data-set (in magnitude, distance and site characterization spaces) to derive values for force reduction factors needed for the structure to reach, and not exceed, a pre-determined level of ductility. It is observed that the force modication factors are only slightly inuenced by the shape of the hysteretic model used in their derivation and even less sensitive to strong motion characteristics. A linear representation is recommended for use in a benchmark for demand considerations and given in an easy-to-use parametric form. 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Force reduction factors; Seismic design; Ductility; Design spectrum

1. Introduction Conventional seismic design, as employed in codes of practice, is entirely force-based, with a nal check on structural displacements. Force-based design is suited to design for actions that are permanently (or persistently) applied. Members are designed to resist the effects of these actions at levels of stress constrained by their plastic capacity. The deformations corresponding to the plastic member capacity are not normally excessive, and evaluating them is not an onerous task. Since seismic design was developed as an extension to primary load design, it followed the same procedure, noticing though that inelastic deformations may be utilised to absorb quantiable levels of energy leading to reduction in the forces for which structures are designed. This lead to the creation of the response modication (or behaviour) factor; the all-embracing parameter that purports to account for over-strength, energy absorption and dissipation as well as the structural capacity to re-distribute

* Corresponding author. Tel.: + 44-171-594-6058; fax: + 44-171594-6053. E-mail address: a.elnashai@ic.ac.uk (A.S. Elnashai).

actions from inelastic highly stressed regions to other less stressed locations in the structure. Problems of evaluating behaviour factors that are generally applicable to various structural systems, materials, conguration and input motion are well documented and the inherent weakness in code-specied factors is widely accepted. However, the majority of existing studies are concerned with the capacity of structural systems to absorb energy and hence levels of force well below the elastic values are recorded. In other words, studies of the supply response modication factors abound. Much less so are studies that aim at quantifying the demand imposed by earthquake motion using a veried and well distributed natural records data-set. In this work a well controlled data-set was employed for the denition of inelastic constant ductility acceleration spectra. This data-set was selected by Bommer et al. [1] and already used for the denition of attenuation relationships for elastic displacement spectra with different damping values. Inelastic spectra were derived herein using two models: an elastic perfectly plastic representation and another more complex system which has a yield point, a maximum force point and a post-ultimate branch which may represent hardening as well as softening. The strong motion records were further used to

0141-0296/00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 1 4 1 - 0 2 9 6 ( 9 9 ) 0 0 0 7 5 - 9

B. Borzi, A.S. Elnashai / Engineering Structures 22 (2000) 12441260

1245

derive values for response modication factors (q or R) for force-based seismic design. Moreover, the q-factor values herein obtained were compared with the formulation of reduction coefcient proposed in the technical literature. The simplied trilinear representation proposed hereafter is shown to provide uniform reliability across the period range.

2. Input motion

The dataset employed for the denition of inelastic constant ductility spectra was assembled by Bommer et al. [1] for the derivation of frequency-dependent attenuation equations for ordinates of displacement response spectra. All records were ltered individually at Imperial College, London, and used to derive displacement spectra for different levels of damping, from 5% to 30%. The dataset has been adapted from the one employed by Ambraseys et al. [2] to derive attenuation relationships for ordinates of elastic acceleration response spectra. Some weaker records were excluded whilst records obtained after the aforementioned work was undertaken were added. This is a high quality dataset in terms of both accelerograms that have been individually corrected and information regarding the recording stations and earthquake characteristics. The accelerograms of the dataset were recorded during 43 earthquakes of a magnitude between 5.5 and 7.9, at a distance from the nearest point on the fault of up to 260 km. While the source distance and the surfacewave magnitude are available for all the accelerograms, for three records the local site geology is unknown. For the remaining 180, the percentages of distribution in the three site groupings of rock, stiff and soft soil are 25.0%, 51.1% and 23.9% respectively. For two records only one component of the motion is available. The total number of used records is 364. In Fig. 1 the distribution of records comprising the data-set with regard to magnitude, distance and site classication are shown. The gures demonstrate that the data is well-distributed with respect to all three parameters, hence results of analysis will not have signicant bias. The attenuation model used in this work is that of Ambraseys et al. [2]. In this attenuation model three soil types are dened as a function of the shear wave velocity. When the shear wave velocity exceeds 750 m/s then the soil is classied as rock. Shear wave velocity less than 360 m/s leads to categorising the soil as soft. Stiff soil conditions are assumed in the intermediate range of shear wave velocity. Further details of the dataset, attenuation relationship and the regression model are given in Bommer and Elnashai [3].

Fig. 1.

Distribution of records for (a) rock; (b) stiff and (c) soft soil.

3. Structural models 3.1. Elastic perfectly-plastic model In order to determine the inuence of magnitude, distance and soil condition on inelastic response spectra, attenuation relationships have been dened using an elastic perfectly-plastic response model (EPP). The EPP model was employed since it is the simplest form of inelastic force-resistance as well as being the basis for early relationships between seismic motion and response modication factors. Moreover, by virtue of its two parameters denition: level of force-resistance and stiffness. Few structural characteristics are included, hence the inuence of strong-motion records may be better visualized. The stiffness is corresponding to the period of vibration for which the spectral ordinate has to be calculated and the resistance is derived iteratively. In this work inelastic constant ductility spectra were obtained. Therefore the resistance of the system corresponds to the resistance for which the system has a required ductility equal to the target ductility. The ensuing inelastic spectra would reect solely the characteristics of the input motion.

1246

B. Borzi, A.S. Elnashai / Engineering Structures 22 (2000) 12441260

3.2. Hysteretic hardening-softening model In order to investigate the inuence of the response characteristics of structures on inelastic spectra, a hysteretic hardening-softening model (HHS) was used [4]. The structural model is characterised by the denition of a primary curve, unloading and reloading rules. The primary curve for a hysteretic force-displacement relationship is dened as the envelope curve under cyclic load reversals. For non-degrading models the primary curve is taken as the response curve under monotonic load. In this model the primary curve is used to dene the limits for member strength. Two points on the primary curve have to be dened. It is essential to dene cracking and yield loads (Vcr and Vy) and the corresponding displacements (cr and y), as shown in Fig. 2. If, for example, this model was used to describe the hysteretic behaviour of reinforced concrete members, the cracking load would correspond to the spreading of cracks in the concrete and the yielding load would be related with the load at which the strain in bars is equal to the yield strain of steel. Unloading and reloading branches of the HHS model have been established through a statistical analysis of experimental data [5,6]. The load reversal rules are briey described below. Structural members exhibit stiffness degradation under cyclic loading. When the number of cycles or the magnitude of inelastic deformation increases, the system becomes softer. Furthermore, the hysteretic behaviour is affected by pinching. The axial load is an important parameter in predicting pinching effects (due to the onset of crack closure). The slope of reloading branches increases beyond the crack load. The slopes of the lines connecting the origin to the cracking point (K1 in Fig. 2) and the yield point to the cracking point in the opposite quadrant (K2 in Fig. 2) are used to dene the unloading branches under cyclic loads. The latter slope depends on deformation and force

levels attained at the beginning of unloading. Experimental results indicate that if unloading starts between the cracking and the yield load, and the yield load has not been exceeded in the relevant quadrant, then unloading stiffness is bounded by K1 and K2. In this model a linear variation between these limits was proposed as a function of displacement ductility. If the unloading load exceeds the yield load, the unloading curve changes the slope to a value close to the cracking load. In the current investigation, second order effects have not been considered but the effect of axial load on pinching has been accounted for. Fortunately, in most situations, particularly in regions where large seismic forces need to be considered for design, P- effects do not markedly affect the force and deformation supply of the structures [7]. The response of well-designed structures of intermediate periods will not be affected by P- effects during inelastic excursions. On the other hand, after very large displacements, due to large velocity pulse, the frame may not be restored to its original undeformed condition. However, this is more likely to occur in structures without seismic detailing. It is therefore reasonable to ignore second order effects in the course of deriving inelastic spectra for general use. 4. Procedural considerations In order to dene inelastic acceleration spectra and behaviour factors, displacement ductility of 2, 3, 4 and 6 are considered. Ductility levels higher than 6 are not included because they constitute global displacement ductility and structures very rarely have a local ductility supply commensurate with global ductility above 6. The inelastic spectra have been dened between 0.05 s and 3 s. The following period steps have been adopted: T = 0.05 s 0.2 s T = 0.01 s

Fig. 2.

HHS model for structural members.

B. Borzi, A.S. Elnashai / Engineering Structures 22 (2000) 12441260

1247

T = 0.2 s 0.5 s T = 0.02 s T = 0.5 s 1 s T = 0.05 s T = 1 s 3 s T = 0.1 s For the HHS model the initial elastic period can be considered either as the stiffness before Vcr or the secant stiffness. In the current work an initial elastic period corresponding to the secant stiffness was considered. This is because the stiffness before Vcr is not representative of the structural behaviour. A low damping value of 1% was included. This viscous damping is representative of the non-hysteretic dissipation, since hysteretic damping is already included. To justify the chosen low damping value, experimental results on reinforced concrete buildings are taken into account [8]. A damping value up to 2% is adequate to describe sources of damping other than hysteretic for reinforced concrete. Furthermore, sensitivity analyses were undertaken and have indicated that further renement of this assumption is unwarranted. The input parameters for the HHS model described above are the monotonic curve and the relationship between axial compressive force and nominal concentric axial capacity. In order to dene the inelastic constant ductility spectra the magnitude of the monotonic curve is not an input parameter. It is dened in an iterative way forcing the relationship between maximum and yield displacements to satisfy the target ductility. To obtain the inelastic spectra and response modication factor (R or q) an approximation of the primary curve with three linear branches has been assumed (Fig. 3). Consequently, the input parameters dening the shape of the primary curve are: 1. the relationship between the cracking and the yielding load (Vcr/Vy); 2. the relationship between the stiffness before the cracking load and the secant stiffness (Kcr/Ky); 3. the slope of the post yield branch. To select the values of parameters to be employed,

extensive analysis of the inuence of each parameter on the inelastic spectra was undertaken. The results of parametric investigation indicate that the parameter with the strongest inuence on inelastic spectra is the slope of the post yield branch. Hence xed ratios between Vcr and Vy and between Kcr and Ky were considered. From the experimental results of Paulay and Priestley [7], Priestley et al. [9], Calvi and Pinto [10], and Pinto [11], it is reasonable to consider the secant stiffness at the yield point equal to 50% of the stiffness before Vcr; The latter is taken equal to 30% of Vy. The ratio between the cracking and the yield load inuences the pinching behaviour that does not occur often for structures with loads higher than approximately 30% of the yielding load Vy. The considered representative slopes of the structural behaviour are: K3 K3 K3 K3 = = = = 0 (elastic perfect plastic behaviour) 10% Ky (hardening behaviour) 20% Ky (softening behaviour) 30% Ky (softening behaviour)

Only a level of axial load equal to 10% of the nominal axial load is assumed, since the model does not account for second order effects and is not strongly affected by the chosen level of axial load. The above characteristics have been veried to cover both new structures with full seismic detailing, and existing structures with no special detailing [12,13]. An iterative procedure was utilized for the denition of spectral ordinates corresponding to a target ductility. Sometimes it has not been possible to obtain a convergent solution with the HHS model, due to instability emanating from steeply descending post-peak load response. In order to dene the spectral ordinates when the solution is not convergent, the following approach was adopted: For elastic-perfectly plastic and hardening behaviour it was almost always possible to obtain ductility values higher and lower than the target ductility. In cases where the difference of both values from the target ductility value did not exceed 2, the average solution was taken. Otherwise the closest solution was assumed when it corresponded to an obtained ductility with a difference from the target of less than one; For the softening cases, the abrupt change in the instantaneous force capacity leads to unstable solutions, as depicted in Fig. 4. This is the underlying reason why cases with descending post-ultimate branches had a higher percentage of failed convergence. This is dealt with by one of the following options: If the target and obtained ductility are bounded and differ by 1 or less, the median value is adopted;

Fig. 3.

Shape of primary curve used in this work.

1248

B. Borzi, A.S. Elnashai / Engineering Structures 22 (2000) 12441260

Fig. 4.

Failure of iterative procedure for softening behaviour.

Otherwise, the closest obtained ductility is adopted provided that its variation from the target ductility is 0.5 or less; In a case where no such solution existed, the spectral ordinates were not taken into consideration when conducting the regression analysis. The percentage of spectral ordinates that have not been considered in the regression analyses are reported in Table 1. It is observed that the number of spectral ordinates to be excluded from parametric analysis for the slope of the third branch equal to 30% Ky and ductility equal to 4, is very high. The attenuation relationship for this combination of parameters was therefore not considered. In the softening cases it was also not possible to obtain a convergent solution for periods less than 0.1 s. Inelastic response spectra start therefore from 0.1 s for the softening cases. The above observations (non-convergence) are fully justied by noting that highly degrading systems are inherently of low ductility. Therefore the decisions taken do not affect the generality of the reported spectra and response modication factors.

5. Inelastic acceleration spectra The inuence of magnitude, distance and local site conditions have been studied for elastic response spectra in previous work. The objective of this part of the work is to investigate whether inelastic response spectra exhiTable 1 Percentage of ordinates excluded from regression analysis K3 0 10% Ky 20% Ky 30% Ky m=2 0.16% 0.03% 1.67% 2.45% m=3 0.58% 0.13% 5.12% 8.16% m=4 1.23% 0.27% 12.00% 30.98% m=6 3.16% 0.80%

bit the same dependence to the above-mentioned parameters. In order to investigate this, the coefcients of the attenuation relationship used by Ambraseys et al. [2] have been calculated for inelastic constant ductility spectra. To isolate the inuence of hysteretic behaviour and input motion characteristics, the attenuation coefcients of inelastic spectra dened for the EPP model were considered. In order to obtain the relationship between elastic and inelastic acceleration spectral ordinates (response modication factors) a regression analysis was performed to dene an elastic acceleration spectrum for a damping value of 1%. Fig. 5 shows the inuence of ductility, magnitude, distance and soil conditions on inelastic acceleration spectra. These gures demonstrate the strong inuence of input motion parameters on inelastic acceleration spectra. To compare the nature of inuence that the input motion parameters have on inelastic and elastic acceleration spectra, the attenuation coefcients calculated by Ambraseys et al. [2] were invoked. These were obtained using a data-set from which the data-set of Bommer et al. [1] was selected. The records have a surface wave magnitude between 4 and 7.9 and a focal depth less than 30 km. The inuence of magnitude and distance on elastic acceleration spectra, calculated with the attenuation law presented above, is shown in Fig. 6. Considering this gure and the one related to the inelastic spectra, it is possible to observe that the input parameters have the same inuence on elastic and inelastic spectra. Fig. 7 depicts the inuence of different structural characteristics on inelastic acceleration spectra for all the ductility levels considered. It indicates that the hysteretic behaviour does not have a strong inuence on acceleration spectra. This conrms that in a force-based method the hysteretic behaviour does not signicantly change the level of force for which the structure has to be designed, in order to reach a xed level of displacement ductility. In this work inelastic constant ductility spectra were derived. It was therefore assumed that the damage is only a function of the ductility requirement, neglecting the inuence of energy dissipation. If a damage indicator

B. Borzi, A.S. Elnashai / Engineering Structures 22 (2000) 12441260

1249

Fig. 5. Inuence of (a) ductility, (b) magnitude, (c) distance and (d) site condition on enelastic acceleration spectra evaluated for the EPP model.

Fig. 6.

Inuence of (a) magnitude and (b) distance on elastic acceleration spectra obtained by Ambraseys et al. [2].

taking into account the amount of energy that the systems can dissipate is assumed, the hysteretic behaviour becomes more important in terms of level of resistance for which the structure has to be designed. However, the inelastic acceleration spectra can be used in terms of equivalent or reduced ductility instead of displacement ductility. By means of a reduction of the original displacement ductility the amount of the damage due to the dissipated energy can be taken into account [14].

6. Seismic force reduction factors (R or q) 6.1. Response modication factor supply Knowledge of initial period and damping values are insufcient to dene the seismic force intensity for a system exhibiting inelastic behaviour. To obtain the base-shear force, a standard approach is in use, based on the denition of a behaviour factor (q and R for Europe

1250

B. Borzi, A.S. Elnashai / Engineering Structures 22 (2000) 12441260

Fig. 7.

Inelastic acceleration spectra for different hysteretic behaviours (Ms = 6, d = 10 km, soft soil).

and USA, respectively). The response modication or behaviour factor has been used to derive the design acceleration response spectrum from its linear elastic equivalent, allowing the benets offered by the energy dissipation capacity of structures to be availed of, while ensuring that the imposed ductility demand does not exceed the available supply. This factor accounts for the presence of damping and other force reducing effects, such as period elongation (or stiffness degradation). It is evident that there is a strict correlation between the behaviour factor and the ductility resource of the structure. In the behaviour factor both the local ductility of plastic hinges and that of the structural system are considered. This must be taken into account because in order to guarantee a certain level of displacement global ductility, an adequate supply of local ductility in the plastic hinge has to be provided and mechanisms of collapse characterised by low dissipation of energy have to be avoided. A brief review of the behaviour factor in seismic codes in Europe and in the United States is presented below in order to identify areas of possible improvement. The behaviour factor (q) used in Eurocode 8 (EC8) [15] is the ratio between the elastic and inelastic design spectra. In EC8 maximum allowable behaviour factor values are specied for different structural types,

materials and ductility classes. The behaviour factors of EC8 have also to be interpreted as the lower bound of the real ductility capacity of structures designed according to the code. The period-dependent response modication function proposed by EC8 is: T (hb 1) T1 0 when T T1, q q q T 1 + (hb0/q 1) T1 1+ when T T1 where T1 is a characteristic period of the design spectrum (lower-bound period of the constant branch of the EC8 design acceleration spectrum), h relates to the equivalent viscous damping of the structure, (h is 1 when the damping is 5%), b0 is the acceleration amplication factor, (set equal to 2.5), and q is the behaviour factor of the structure. Values of behaviour factor between 1 and 5 are given for different construction materials and structural systems. Until recent research work, the behaviour factor (R) given by United States codes had not been substantially modied since the 1950s, when it was rst introduced. In 1957 a committee of Structural Engineers Association of California (SEAOC) introduced a horizontal force

(1)

B. Borzi, A.S. Elnashai / Engineering Structures 22 (2000) 12441260

1251

factor, the predecessor of the behaviour factor. In the mid-1980s, data from experimental research to dene base shear versus roof displacement curves for steel braced frames and for the formulation of the bahaviour factor [16,17]. It was proposed to split the behaviour factor into three constituents: R RSRmRx (2)

For short period structures the ductility is higher than the behaviour factor and the equal energy approach may be adopted to calculate force reduction. This approach is based on the observation that energies associated with the force corresponding to the maximum displacement reached by an elastic and inelastic systems are similar. The proposed relations for behaviour factor are: Rm 1 when T 0.03 s Rm 2m 1 when 0.12 s T 0.5 s Rm m when T 1 s (4)

where RS is the strength factor (dened as the ratio between the supply resistance and the design one, also considered as an over-strength factor), R the ductility factor and R the damping factor. Other research work [1821] lead to a second formulation of the behaviour factor in which R is replaced by RR. The latter coefcient is known as the redundancy factor, introduced to account for the number and distribution of active plastic hinges. Due to different number of plastic hinges, structures characterised by the same shear resistance have different reliability. This is taken into account by means of the coefcient RR. The new relationship for the behaviour factor is: R RSRmRR (3)

In this formulation R was excluded. This was because of the recognition that although the damping factor can be used to scale the displacement in inelastic structures, it cannot be used to proportionally reduce the strength demand. Substantially larger behaviour factors are proposed in the United States than in Europe. As a consequence, similar structures designed according to these codes are likely to suffer different levels of damage during earthquakes, provided the strength of materials side is the same. However, it is most important to note that higher behaviour factors do not necessarily lead to lighter structures, since there are other loading scenarios that may take precedence. 6.2. Response modication factor demand The behaviour factor demand represents the minimum reduction coefcient corresponding to a specic level of ductility. The relationship between displacement ductility and ductility-dependent behaviour factor has been the subject of considerable research. A few of the most frequently used relationships reported in the technical literature are discussed below. 6.2.1. Newmark and Hall [22] During the seventies the reduction factor was parameterised as a function of ductility [22]. It was observed that in the long period range, elastic and ductile systems with the same initial stiffness reached almost the same displacement. As a consequence, the behaviour factor can be considered equal to the displacement ductility.

6.2.2. Krawinkler and Nassar [23] A relationship was developed for the reduction factor derived from the statistical analyses of 15 western United States ground motions with magnitude between 5.7 and 7.7. The records were obtained on alluvium and rock site, but the inuence of site condition was not explicitly studied. The inuence of behaviour parameters, as well as yield level and hardening coefcient, was taken into account. A 5% damping value was assumed. The equation derived is given as: Rm [c(m 1) 1]1/c where: c(T,a) Ta b a 1+T T (6) (5)

in which a is the hardening parameter of the employed hysteretic model and a and b are regression constants. 6.2.3. Miranda and Bertero [24] The equation for reduction factor introduced by Miranda and Bertero [24] was obtained considering 124 ground motions recorded on a wide range of soil conditions. The soil conditions were classied as rock, alluvium and very soft sites characterised by low shear wave velocity. A 5% of critical damping was assumed. It is given as Rm m1 1 (7)

where assumes different formulations for rock, alluvium and soft sites as shown below: f1 1 1 exp( 1.5(ln(T) 10T mT 2T

0.6)2) for rock site


f=1+ 1 2 exp( 2(ln(T) 0.2)2) for alluvium site 12T mT 5T

f=1+

T1 3T1 exp( 3(ln(T/T1) 0.25)2) for soft site 3T 4T

1252

B. Borzi, A.S. Elnashai / Engineering Structures 22 (2000) 12441260

(8) where T1 is the predominant period of the ground motion. 6.2.4. Vidic et al. [25] The reduction coefcients introduced by Vidic et al. [25] (R) were approximated with a bilinear curve. In the short period range the reduction factor increases linearly with the period, from 1 to a value that is almost equal to the ductility. In the remaining part of the period range the reduction factor is constant. To calculate the reduction factor, a bilinear model and a stiffness degrading Q-model were investigated. A mass proportional damping and an instantaneous stiffness proportional damping were assumed as mathematical models of damping. In this work the standard records form California and Montenegro 1979 were chosen as being representative for standard ground motion (i.e. severe ground motion at moderate epicentral distance, with a duration between 10 and 30 seconds and predominant period between 0.3 and 0.8 seconds). The study of inuence of different structural behaviours was worked out using these records. In order to account the inuence of input motion to the groups of records from California and Montenegro (20 records all together) three other groups of records were used. These are the records from Friuli 1976 (6 records), Banja Luka 1981 (6 records) and Chile 1985 (8 records). The main characteristic of records from Friuli and Banja Luka are the short duration and the short predominant periods, whilst the duration of Chile records is long, but their predominant periods are in the range of standard records period. It should be noted that the groups are rather small. The proposed formulation of reduction factor is: Rm c1(m 1)cR T 1 when T T0 T0 (9)

of records well distributed in terms of magnitude, distance and soil conditions from a wide range of seismo-tectonic environments; Use of comprehensively represented hysteretic models exhibiting hardening-softening behaviour; Using regression curves focusing on uniform distribution of target reliability across the period range and giving simple code-amenable expressions. In this work, following the denition of response modication factor, regression analyses for the evaluation of the ratio between the elastic and inelastic acceleration spectra (q-factor) were undertaken. The inuence of ductility and input motion parameters (magnitude, distance and soil conditions) on the behaviour factor was studied utilising the EPP hysteretic model. These results are presented in Fig. 8. It was observed that the inuence of input motion parameters on elastic and inelastic acceleration spectra is similar (and signicant). However, the effect cancels out for their ratio. Ductility is the most signicant parameter inuencing the response modication factor. Consequently, analyses to dene a perioddependent behaviour factor functions for all the ductility levels and all structural models were undertaken. The average values and the standard deviations were calculated considering various combinations of input motion parameters. The period dependent behaviour factor functions herein calculated were further approximated with a trilinear spectral shape. The reduction coefcient is equal to 1 at a zero period and increases linearly up to a period T1, which is dened as the period at which the behaviour factor reaches the value q1. A second linear branch is assumed between T1 and T2. The value of the reduction coefcient corresponding to T2 is herein denoted q2. For periods longer than T2 the behaviour factor maintains a constant value equal to q2: q (q1 1) T 1 when T T1 T1 T T1 1 when T1 T T2 T2 T1 (11)

Rm c1(m 1)cR 1 when T T0

where T0 is the period dividing the period range into two portions. It is related to the predominant period of the ground motion T1 by means of: T0 c2mcTT1 (10)

q q1 (q2 q1)

q q2 when T T2

The coefcients c1, c2, cR and cT depend on the hysteretic behaviour and damping. 6.3. Constant reliability response modication factors The above formulations were signicant steps forward at the time they were undertaken. Areas of possible improvement, though, are identied as: The data-set can be improved using a large number

The values q1, q2, T1 and T2 that allow the denition of approximate spectra for all relevant ductility levels and hysteretic behaviour, as they are obtained by a piecewise linear regression, are reported in Table 2. Fig. 9 shows that the simplied trilinear curves give an accurate approximation of the behaviour factor spectra throughout the whole period range. To demonstrate this, the standard deviation s of the ratio g between the approximate and the original spectral values is studied. The standard deviation was calculated for all branches

B. Borzi, A.S. Elnashai / Engineering Structures 22 (2000) 12441260

1253

Fig. 8.

Inuence of (a) ductility, (b) magnitude, (c) distance and (d) site condition on q-factor evaluated for the EPP model.

Table 2 Constants for trilinear behaviour factors spectra m=2 T1 EPP K3 = K3 = K3 = K3 = 0.20 0.20 0.21 0.26 0.26 T2 0.79 0.56 0.54 0.26 0.26 q1 2.06 2.20 2.04 2.43 2.42 q2 2.20 2.51 2.33 2.43 2.42 m=3 T1 0.21 0.25 0.27 0.24 0.24 T2 0.78 1.67 1.80 1.76 1.85 q1 2.89 3.10 2.78 2.83 2.76 q2 3.31 4.09 3.62 3.93 3.81 m=4 T1 0.22 0.27 0.29 0.25 T2 0.87 1.55 1.64 1.69 q1 3.59 3.76 3.25 3.25 q2 4.34 5.45 4.56 5.12 m=6 T1 0.25 0.29 0.33 T2 0.99 1.26 1.54 q1 4.81 4.78 3.93 q2 6.13 7.79 6.10

0 10% Ky 20% Ky 30% Ky

of the approximate spectra and across the whole period range. These values are reported in Table 3. It is observed that the dispersion of g is close to the global standard deviation. This has an important consequence from a practical point of view, as the behaviour factor spectra herein proposed correspond to an almost constant seismic design reliability over the whole period range, a feature not previously achieved. Finally, the coordinates of the points that allow the denition of the approximate spectra were expresses as a function of ductility and given here as: T1 bT1 (12)

T2 aT2m bT2 q1 aq1m bq1 q2 aq2m bq2

(13) (14) (15)

where bT1, aT2, bT2, aq1, bq1, aq2 and bq2 are constant values. It was seen that the control periods of the approximate spectral shape do not strongly depend on the hysteretic behaviour. Therefore, it was possible to use common values of the control periods for all the hysteretic models in Eqs. (12) and (13). On the other hand, it was seen that different values of aq1, bq1, aq2 and bq2 correspond to the different hysteretic behaviour

1254

B. Borzi, A.S. Elnashai / Engineering Structures 22 (2000) 12441260

Fig. 9.

Simplied three-linear curve used to t the q-factor period dependent curves.

patterns. The values of the constant in Eqs. (12)(15) are reported in Tables 4 and 5. To demonstrate that the co-ordinates given as a function of ductility in Eqs. (12)(15) still result in a good approximation of the original behaviour factor spectra (characterised by a constant seismic design reliability), the standard deviation s of the ratio g (approximate-toaccurate) are reported in Table 6.

By comparing the errors reported in Tables 3 and 6, it is clear that the parameterised values obtained with Eqs. (12)(15) do not signicantly modify the level of approximation of the original curves. The results obtained in this work conrm the main conclusions of previous studies and give comprehensive and qualitative guidelines for response modication factors (R or q) for seismic design and assessment. The

B. Borzi, A.S. Elnashai / Engineering Structures 22 (2000) 12441260

1255

Table 3 Standard deviations of g% (ratio of approximate to accurate q/R factors) m=2 s1 EPP K3 = K3 = K3 = K3 = 2.1 3.1 3.8 3.6 3.7 s2 1.9 1.6 1.6 4.7 4.9 s3 2.0 2.7 2.8 s 2.0 2.6 2.9 4.4 4.6 m=3 s1 3.4 4.5 6.4 2.6 2.9 s2 2.1 3.6 3.4 3.2 3.1 s3 2.8 3.3 3.5 3.7 4.1 s 2.8 3.9 4.7 3.2 3.3 m=4 s1 4.7 6.2 9.0 2.9 s2 2.5 3.8 3.4 3.1 s3 3.5 3.9 3.4 4.3 s 3.6 4.8 6.1 3.4 m=6 s1 7.0 8.5 2.8 s2 2.8 3.0 1.7 s3 4.2 6.2 2.5 s 5.0 6.4 2.6

0 10% Ky 20% Ky 30% Ky

Table 4 Values of the constants in Eqs. (12) and (13) bT1 0.25 aT2 0.163 bT2 0.60

Table 5 Values of the constants in Eqs. (14) and (15) aq1 EPP K3 = K3 = K3 = K3 = 0.69 0.55 0.32 0.38 0.29 bq1 0.90 1.37 1.69 1.67 1.83 aq2 1.01 1.33 0.96 1.24 1.21 bq2 0.24 0 0.51 0 0

coefcients given in previous investigations. This is shown in Fig. 10. The results obtained in the current work are in good agreement with the previous formulations of reduction coefcient. The current formulation is, however, derived using a much wider data-set with consistent distributions in the magnitude, distance and site condition spaces. Moreover, the idealisation proposed above leads to hazard-consistent or reliability-consistent force reduction factors. They are therefore consistent with the commonly used uniform hazard response spectra. 6.4. Comparison between supply and demand To investigate the relationship between supply and demand in terms of reduction coefcient of lateral seismic loads, two reinforced concrete regular frame structures were considered. In the rst case a 12-story building designed according to EC8 provisions for ductility class H (Fardis [26]) was studied. The design acceleration spectrum proposed by EC8 for soil class B and a peak ground acceleration equal to 0.3 g were assumed to evaluate the seismic loads. The reinforcement (both longitudinal and transverse) and the concrete are classied as S500 and C25/30, respectively. The 12-story building was analysed by Salvitti and Elanshai [27] by means of modal, nonlinear static and dynamic analyses. In the dynamic analyses four acceleragrams articially generated from the EC8 design spectrum were

0 10% Ky 20% Ky 30% Ky

behaviour factor is only slightly dependent on the period in the long period range and almost corresponds to the ductility value An exception to this, briey investigated by the writers but not reported herein, is long duration or multiple earthquake events. On the other hand in the short period range, the behaviour factor is dependent on both ductility and period. A moderate, though not negligible, inuence from the hysteretic behaviour is observed throughout the period range. The behaviour factor spectra calculated in this work were compared with the formulations of ductility-dependent reduction
Table 6 Standard deviations of g(%) for approximate q-factor denition m=2 m=3

m=4

m=6

1
EPP K3 = K3 = K3 = Ky K3 = Ky 0 10% Ky 20% 30% 3.3 3.5 3.6 4.4 4.5

2
3.2 2.3 2.9 2.1 2.2

3
2.1 3.0 3.0 2.8 3.1

4.0 3.5 3.4 4.1 4.5

1
4.1 4.6 6.5 3.5 3.8

2
2.5 2.6 2.5 3.3 4.2

3
3.0 5.1 5.6 3.3 3.7

3.6 4.2 5.4 4.0 5.1

1
5.2 6.2 9.3 3.4

2
3.4 3.8 3.9 3.4

3
3.3 4.1 3.8 4.2

4.0 5.1 7.0 3.7

1
6.7 8.2 13.5

2
4.2 4.9 3.7

3
4.9 6.9 4.7

5.8 9.0 11.2

1256

B. Borzi, A.S. Elnashai / Engineering Structures 22 (2000) 12441260

Fig. 10.

Reduction coefcients proposed (a) in this work and (b), (c), (d) and (e) in previous works.

employed. Peaks of acceleration equal to and twice the design value were adopted. The aims of the dynamic analyses were to evaluate the behaviour factor and establish how the structure performed in terms of both local and global damage. The second selected case is a 7-story building. The building was founded on rock. It is located in Los Angeles where it was affected by both the San Fernando and Northridge earthquakes of 1971 and 1994,

respectively. The building was analysed by the ATC project team (FEMA 273 and 274, 1997) [28] to demonstrate techniques for evaluation of strength and ductility factors by means of inelastic static collapse analysis. The main aim of the research in the aforementioned work [27,28] was to compare the behaviour factor proposed by seismic codes design with that obtained by means of sophisticated inelastic analyses. The aim of the

B. Borzi, A.S. Elnashai / Engineering Structures 22 (2000) 12441260

1257

Fig. 11.

Measured and idealised force-displacement curve of the structure analysed by (a) Salvitti et al., 1997 and (b) ATC project team.

current work is to investigate whether the behaviour factor, as given in the codes, corresponds to an adequate ductility supply. In order to dene the ductility demand, the behaviour factors proposed by the European and United State codes are compared with the behaviour factor spectra corresponding to the hysteretic model that best describes the global behaviour of the real structure. To calculate the ductility supply of both structures, the force-displacement curves obtained by the nonlinear static analysis were used. The latter curves were also employed to evaluate the hysteretic model best describing the structures. In order to dene the characteristics of the equivalent hysteretic model (substitute structure), a secant stiffness approach as proposed by Paulay and Priestley [7] was adopted. Hence, the initial elastic stiffness was evaluated as the secant stiffness for the point corresponding to a base shear value equal to 75% of the maximum force (Fig. 11). The collapse criteria assumed by Salvitti and Elnashai [27] are based on either a maximum interstory drift of 3% or the attainment of a conned concrete limiting strain. On the other hand, the ATC research team [28] used a maximum rotation capacity in the vertical structural elements of 0.005 rad to dene failure. The hysteretic model associated with the global structural behaviour of the structures is an HHS model with elastic-perfectly plastic behaviour. Values for the ratio of shear force to total weight of the structure (V/W) and for the yield and maximum top story displacement (y and m, respectively) are reported in Table 7. Overstrength (Rs) and ductility values were also
Table 7 Results from static nonlinear analyses y (mm) Salvitti and Elnashai [22] ATC [27] 220 127 m (mm) 648 300 V/W 0.260 0.179 Rs 2.11 2.98 m 2.95 2.36

calculated. Herein, overstrength is dened as the ratio between the calculated maximum strength and the design strength. Both structures have a large overstrength factor Rs. This is principally, though not exclusively, due to the fact that in the design process design material resistance values are assumed, while in the analyses nominal values were employed. For the structure analysed by Salvitti and Elnashai [27], large overstrength was also calculated in the nonlinear dynamic analyses. Overstrength factors of 1.93 and 2.35 were evaluated for the peak acceleration values equal to and twice the design peak ground acceleration, respectively. Evidently, the ratio 1.93 at the design acceleration is an instantaneous overstrength, not the true value. The denition of behaviour factors used in the work of Salvitti and Elanshai [27] are:
el q (SA)el c /(SA)y in qD (SA)el c /(SA)d

(16) (17)

In the rst expression the behaviour factor is obtained by relating the elastic spectra at yield (subscript y) to that at collapse (subscript c), while in the second equation, the denominator changes to the inelastic design spectral value. The behaviour factor given by the code is the ratio between the elastic and inelastic spectra. Therefore, Eq. (17) may be written as:
el qD (SA)el c /(SA)d qcode

(18)

By assuming that the response acceleration spectra at collapse and at yield have the same dynamic amplication, the behaviour factors can be expressed as a function of peak ground acceleration as below: q ag(design) ag(yield) (19)

1258

B. Borzi, A.S. Elnashai / Engineering Structures 22 (2000) 12441260

qD

ag(collapse) ag(design) qcode

(20)

Table 8 Ductility supply and ductility demand msupply Salvitti and Elnashai [27] 2.95 ATC [28] 2.36 mdemand 3.79 2.61 Difference 22% 10%

Eq. (19) is the analytical translation of the denition of behaviour factor, but does not account for the effect of overstrength which is signicant for the structures under consideration. The proposed behaviour factor of EC8 for the considered structural type is 5, while the minimum behaviour factor calculated by Salvitti and Elanshai [27] using Eq. (20) is 12.5. The 7-story building analysed by the ATC project team [28] was designed for a behaviour factor equal to 3.4. In order to dene the key components of the behaviour factor as formulated in Eq. (3), the denition of ductility-dependent behaviour factor proposed by Miranda and Bertero [24] is invoked. The ductility-dependent key component is 2.73. The building has an excellent redundancy in the horizontal direction of loading, hence a redundancy factor of 1 is used. A behaviour factor of 8.14 was calculated using Eq. (3). The main conclusion of the aforementioned work was that the behaviour factor values proposed by the codes are too conservative. The overstrength of structures designed according to this code lead to high values of behaviour factor capacity, or supply. For both buildings considered, the ductility corresponding to the proposed behaviour factors is obtained by means of a comparison with the behaviour factor spectra (Fig. 12). The ductility supply, evaluated by nonlinear static analysis, and the ductility demand corresponding to the behaviour factor proposed by the codes are reported in Table 8. The above observations lead to the conclusion that both the European and US standards are too conservative, due to the overstrength shown by the structures. On the other hand, the ductility demand corresponding to the behaviour factor proposed by codes is higher than the ductility supply measured in the nonlinear static

analyses. As a consequence, the shear resistance assumed for the design of both the buildings is inadequate for the required ductility capacity. The above treatment shows that if the latter buildings did not exhibit rather high overstrength, the demand could have exceed the supply in terms of their ability to dissipate energy in the inelastic range thus leading to an unsafe situation in terms of deformation capacity. In other words, had the buildings not been of a much higher strength than their seismic design base shear would imply, their seismic performance would have been demonstrably inadequate. 7. Conclusions Regression coefcients for inelastic acceleration spectra have been calculated using the elastic perfectly-plastic (EPP) and the hysteretic hardening-softening (HHS) structural response models. The data-set employed is carefully selected to exhibit consistent distribution in the magnitude, distance and site condition spaces. A total of 364 records were used. The results obtained with the EPP model were employed to highlight the inuence of input motion parameters on inelastic acceleration spectra, while some different sets of parameters for the HHS model were used to evaluate the inuence of hysteretic behaviour. In the light of the obtained results it is observed that: the inuence of input motion parameters on inelastic spectra is similar to that for the elastic spectra;

Fig. 12. Denition of the ductility demand corresponding to the behaviour factors given by seismic codes design for the structure analysed by (a) Salvitti and Elnashai [27], and (b) ATC project team.

B. Borzi, A.S. Elnashai / Engineering Structures 22 (2000) 12441260

1259

the hysteretic models herein assumed only mildly inuence the inelastic acceleration spectra. Therefore the level of force imposed on structures is not heavily inuenced by their global hysteretic behaviour. A ductility-based behaviour factor (q or R) is dened as the ratio between elastic and inelastic spectral ordinates. This is considered as a demand value, which gives an indication of the minimum level of ductility, and energy dissipation capacity that structures have to possess. Coupled with this is the response acceleration, which is proportional to the required strength of the system. Utilising a well controlled and evenly distributed data-set, period-dependent force reduction factors have been calculated. These are recommended for assessment and design and are associated with the ductility factors indicated. A trilinear representation was derived in a fully parametric fashion for use as a benchmark for coderecommended response modication factors. The applicability of these functions to practice is underlined by the fact that they provide uniform reliability designs, e.g. structures designed with these factors across the whole period range will be subjected to the same probability of their capacity being exceeded. As such, the proposed formulation is consistent with elastic uniform hazard spectra employed in codes of practice. Two RC structures, studied in detail elsewhere, representing typical European and US seismic design practice were investigated, as an application exercise. Their supply and demand force reduction factors were studied. It was concluded that had it not been for their high overstrength (ratio of calculated-to-design horizontal force resistance), both structures would not have possessed sufcient deformational supply to meet the imposed demand, the latter estimated from the proposed spectra.

References
[1] Bommer JJ, Elnashai AS, Chlimintzas GO, Lee D. Review and development of response spectra for displacement-based seismic design. ESEE Research Report No. 983, ICONS, Imperial College, London, 1998. [2] Ambraseys NN, Simpson KA, Bommer JJ. Prediction of horizontal response spectra in Europe. Earthquake Eng and Struct Dyn 1996;25:371400. [3] Bommer JJ, Elnashai AS. Displacement spectra for seismic design. J. of Earthquake Eng. 1999; 132. [4] Ozcebe G, Saatcioglu M. Hysteretic shear model for reinforced concrete members. J of Structural Eng, ASCE 1989;115:13348. [5] Saatcioglu M, Ozcebe G, Lee BCK, Tests of reinforced concrete columns under uniaxial and beaxial load reversals. Dept. of Civ. Engrg., Univ. of Toronto, Toronto, Canada, 1988. [6] Saatcioglu M, Ozcebe G. Response of reinforced concrete columns to simulated seismic loading. ACI Struct. J. 1989; 312. [7] Paulay T, Priestley MJN. Seismic design of reinforced concrete and masonry buildings. John Wiley and Sons, Inc, 1992. [8] Negro P, Verzeletti G, Magonette GE, Pinto AV. Tests on a fourstorey full-scale R/C frame designed according to Eurocodes 8 and 2: Preliminary Report. EUR 15879, Ispra (VA), Italy, 1994. [9] Priestley MJN, Seible F, Calvi GM. Seismic design and retrot of bridges. John Wiley and Sons, Inc, 1996. [10] Calvi GM, Pinto PE. Experimental and numerical investigations on the seismic response of bridges and recommendations for code provisions. ECOEST and PREC8, Report No. 4, 1996. [11] Pinto AV. Pseudodynamic and shaking table tests on R.C. bridges. ECOEST and PREC8, Report No. 5, 1996. [12] Borzi B. Design spectra based on inelastic response. PhD thesis, Politecnico di Milano, 1998. [13] Borzi B, Elnashai AS, Faccioli E, Calvi GM, Bommer JJ. Inelastic spectra and ductility-damping relationships for displacementbased seismic design. ESEE Research Report No. 984, ICONS: Imperial College, London, 1998. [14] Fajfar P. Equivalent ductility factors, taking into account lowcycle fatigue. Earthquake Eng and Struct Dyn 1992;21:83748. [15] Eurocode 8. Design provisions for earthquake resistance of structure. ENV 19981, CEN, Brussels, 1994. [16] Uang CM, Bertero VV. Earthquake simulation tests and associated studies of a 0.3-scale model of a six-story concentrically braced steel structure. Earthquake Engineering Research Center, Report No. UBC/EERC-86/10, University of California, Berkeley, California, 1996. [17] Whitteker AS, Uang CM, Bertero VV. An experimental study of the behavior of dual steel systems. Earthquake Engineering research Center, Report No. UCB/EERC-88/14, University of California, Berkeley, California, 1990. [18] ATC. An investigation of the correlation between earthquake ground motion and building performance. ATC-10 Report, Applied Technology Council, Redwood City, California, 1982. [19] Freeman SA. On the correlation of code forces to earthquake demand. Proceedings of 4th U.S.Japan workshop on improvement of building structural design and construction practices, ATC-15-3 report. Redwood City, California, 1990. [20] ATC. A critical review of current approaches to earthquake resistance design. ATC-34 Report, Applied Technology Council, Redwood City, California, 1995. [21] ATC. Structural response modication factors. ATC-19 Report, Redwood City, California: Applied Technology Council, 1995. [22] Newmark NM, Hall WJ. Earthquake spectra and design, EERI Monograph Series. Oakland: EERI, 1982. [23] Krawinkler H, Nassar AA. Seismic design based on ductility and cumulative damage demand and capacities. In: Fajfar, Krawinkle, editors. Nonlinear seismic analysis and design of reinforced concrete buildings. New York: Elsevier Applied Science, 1992.

Acknowledgements The writers would like to express their gratitude to Dr. J.J Bommer and Mr. G.O. Chlimintzas for their help with the strong-motion data-set. The regression program was kindly provided by Dr. S.K. Sarma, whilst support was given by Mr. D. Lee during the implementation of the hysteretic model. Valuable advice on statistical error measures was given by Dr. M. K. Chryssanthopoulos and Ms. C. Dymiotis. All the above are from Imperial College. Thanks are also due to Professor E. Faccioli (Politecnico di Milano) and Professor G.M. Calvi degli Studi di Pavia) who supervised part of (Universita the work. Funding for the primary author whilst at Imperial College was provided by the EU network programmes ICONS and NODISASTR.

1260

B. Borzi, A.S. Elnashai / Engineering Structures 22 (2000) 12441260

[24] Miranda E, Bertero VV. Evaluation of strength reduction factor for earthquake-resistance design. Earthquake Spectra 1994;10:35779. [25] Vidic T, Fajfar P, Fischinger M. Consistent inelastic design spectra: strength and displacement. Earthquake Eng and Struct Dyn 1994;23:50721. [26] Fardis MN. Analysis and design of reinforced concrete buildings according to Eurocodes 2 and 8, prenormative research in support of Eurocode 8 report, 1994.

[27] Salvitti LM, Elnashai AS. Evaluation of behaviour factor for RC buildings by nonlinear dynamic analysis. Eleventh World Conference on Earthquake Engineering, Acapulco, Mexico, 1996. [28] ATC. Guidelines and commentary for the seismic rehabilitation of buildings, Vol. I and II. FEMA 273 and 274 Report, Washington: Federal Emergency Management Agency, 1997.

Das könnte Ihnen auch gefallen