Journal of Constructional Steel Research 60 (2004) 739–749

www.elsevier.com/locate/jcsr

Performance criteria for MR steel frames

in seismic zones
Daniel Grecea a,
, Florea Dinu b, Dan Dubină a
a
Department of Steel Structures and Structural Mechanics, Civil Engineering and Architecture Faculty,
‘Politehnica’ University of Timisoara, Stadion 1, RO-1900, Timisoara, Romania
b
Laboratory of Steel Structures, Centre of Advanced and Fundamental Technical Sciences, Romanian
Academy, Timisoara Branch, M. Viteazul 24, RO-1900, Timisoara, Romania

Abstract

A parametrical inestigation on different types of steel moment resisting frames (MRF)

with rigid and semirigid joints subject to seismic motions is presented in order to establish
their performance criteria. Performance criteria of MR steel frames are evaluated through
the global and local characteristics, i.e. behaviour factor and damage index. The behaviour
factor is related essentially to the maximum inelastic base shear force of the structure
obtained from inelastic dynamic analyses, but also related to other methods from the litera-
ture. Damage is attributed to plastic rotations at members and joints. A linear damage
accumulation low and the rainflow method for counting of cycles are applied. The para-
metric study is performed via a non-linear dynamic time-history analysis.
# 2003 Elsevier Ltd. All rights reserved.

Keywords: Global performance; Moment-resisting frame; Seismic zones; Ground motion levels; Building
performance levels; Global ductility; Local ductility

1. Introduction

During the last years, a large number of devastating earthquakes have occurred
throughout the world. A large number of structures designed and built in accord-
ance with current building codes have been subjected to strong ground-motions,
exceeding the levels for which they have been designed. Damage assessments dur-
ing these events have enabled engineers to learn and improve building code design
provisions, as well as construction techniques for buildings located in regions of
high seismic hazard. For example, steel moment resisting frames (MRF), which are

Corresponding author. Tel./fax: +40-256-403-932.
E-mail address: [email protected] (D. Grecea).

0143-974X/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0143-974X(03)00140-8
740 D. Grecea et al. / Journal of Constructional Steel Research 60 (2004) 739–749

Nomenclature

Du fatigue deformability (plastic rotaion)

N number of plastic rotation range cycles
m slope of the fatigue plastic rotation range curves
log a constant
ni number of cycles of deformation range Dui
Ni number of cycles of the same deformation range that cause failure
V(e,th) theoretical elastic base shear force
V(inel) inelastic base shear force
V(e) base shear force corresponding to first yielding
ku acceleration multiplier for limit state
ke acceleration multiplier for first yielding
a acceleration
ad acceleration for damageability limit state
pr returning period
prd returning period for damageability limit state

widely used as lateral load resisting systems for low- to medium-rise buildings, suf-
fered a surprising amount of damage during the 1994 M6.7 Northridge and 1995
M7.2 Hanshin-Awaji (Kobe) earthquakes.
In several countries, seismic design is in the process of fundamental change. One
important reason for the need for change is that although buildings designed to
current codes performed well during recent earthquakes from a life safety perspec-
tive, the economic loss was unexpectedly high. After Ghobarah [1], conventional
methods of seismic design are providing for life safety (strength and ductility) and
damage control (serviceability drift limits). The design criteria are defined by limits
on stresses and member forces calculated from prescribed levels of applied lateral
shear force. The actual codes are presenting some uncertainties between the seismic
demand and the seismic capacity of the structure. Performance based design is a
more general design philosophy in which the design criteria are expressed in terms
of performance objectives, like lateral deflections, interstorey drifts, element duc-
tility, and element damage indices, when the structure is subjected to different
levels of seismic hazard. To reduce high costs, due to loss of use and repair of
heavily damaged structures, different levels of performance objectives need to be
taken into account.

2. Performance based design and evaluation

The concept of performance based design and evaluation procedures for building
is not new, but is in continuous development in the field of seismic design. For
many years, the seismic design provisions contained in the building codes have
 D. Grecea et al. / Journal of Constructional Steel Research 60 (2004) 739–749 741

been implicitly based on multiple performance objectives. However, most of the

codes consider explicitly only one limit state, defined as protection of occupants
lives in case of a major earthquakes.
Performance based seismic design is defined as a selection of design criteria and
structural systems such that at the specified levels of ground motion and with
defined levels of reliability, the structure will not be damaged beyond certain limit
states or other useful limits [4,5].

2.1. Performance objectives and levels

The basic specification of performance that is used as a basis for performance

based design is the performance objective. It comprises two parts, a design hazard
level and a design performance level. The design hazard level is a quantification of
the severity and character of ground shaking that a structure has to resist. The
design performance level is a quantification of the permissible types and distri-
bution of damage to the structure, given tat design hazard level shaking is experi-
enced. This concept is not new because for many years, the seismic design
provisions contained in the building codes have been implicitly based on multiple
performance objectives:

– resist to minor earth quakes without damage;

– resist to moderate earthquakes without structural damage but with some non-
structural damage;
– resist to major earthquakes with significant structural and non-structural dam-
age; and resist to the most severe earthquakes ever likely to affect the building
without collapse.

Unfortunately, these performance specifications are quite vague with regard to

definition of both the hazard and anticipated performance, making the attainment
of the desired performance difficult.
The coupling of a performance level with a specific level of ground motion pro-
vides a performance design objective [2,6]. In the United States, the most impor-
tant provisions, referring to performance based design were offered by FEMA [3]
and SEAOC [2]. In this perspective, the SEAOC Vision 2000 Committee proposes
four performance levels: fully operational, operational, life safe and near collapse.
Fully operational is a state in which the facility continues in operation with negli-
gible damage to non-structural elements only. In the operational state, the facility
continues in operation with minor damage to both structural and non-structural
elements and minor disruption in non-essential services. Structures in the life safety
condition are significantly damaged, but are expected to be repairable, although
perhaps not economically. Structures in the near collapse condition still guarantee
the safeguard of the human lives, but corresponding to potential complete econ-
omic losses [6]. In Fig. 1, these four performance levels as a function of building
destination and earthquake frequency are presented.
742 D. Grecea et al. / Journal of Constructional Steel Research 60 (2004) 739–749

Fig. 1. Performance objectives according to Vision 2000 provisions.

It can be observed in Fig. 1 that for buildings belonging to category, under a fre-
quent earthquake, structure will not suffer any damage, and under a rare or very
rare earthquake, the level of damages will be extended, but life protection and pre-
vention of collapse to be assured. However, it seems more reasonable not to ask
design engineers to perform many verifications and to introduce only three levels:
serviceability (verification of rigidity), damageability (strength verification) and
ultimate (verification for ductility) limit states.
2.2. Design methods based on performance

Structures designed against earthquakes have to comply with specific criteria

such as stiffness, strength and ductility [9]. However, the majority of modern codes
consider explicitly only one performance objective, defined as protection of occu-
pants against injury or loss of life in case of major earthquakes. In the present
paper, three limit states, which are referring to conditions of drift, residual drift
and rotation capacity of elements, are introduced:

2.2.1. Serviceability limit state—SLS (stiffness criterion)

In case of frequent earthquakes (return period of 20 years), the building can be
used without interruption, the non-structural elements present minor damages and
the structure remains in elastic range. The limit state is defined as the situation
where the interstorey drift exceeds 0.6% of the relevant storey height.

2.2.2. Damageability limit state—DLS (strength criterion)

In case of rare earthquakes (returning period of 475 years), the building presents
important damages of non-structural elements and moderate damages of structural
elements, which may be although repaired after earthquake without high costs or
special technical difficulties. Structure is responding to the seismic motion in elasto-
plastic range and the determinant criterion is the resistance of member sections.
This criterion may be considered as a state index of the building after a strong
 D. Grecea et al. / Journal of Constructional Steel Research 60 (2004) 739–749 743

seismic motion. The limit state is defined as the situation where the residual non-
recoverable part of the interstorey drift exceeds 1% of the relevant storey height.

2.2.3. Ultimate limit state—ULS (ductility criterion)

In case of very rare (the strongest possible ground motion—returning period of
970 years), the building presents major damages (of both non-structural and struc-
tural elements) but safety of people is guaranteed. Damages are extended so that
structure cannot be repaired and demolition is unavoidable. Structure is in the
elasto-plastic range and the determinant criterion is the local ductility (rotation
capacity of elements and connections). The development of cracks in the beam-
to-column joint regions may be associated to the exhaustion of the low-cycle-
fatigue strength. This strength may be expressed in terms of plastic rotation. The
low-cycle fatigue resistance curve is expressed in terms of the plastic rotation
according to [12]:
logN ¼ loga  mlogDu ð1Þ
where Du is the fatigue deformability (plastic rotation), N the number of plastic
rotation range cycles, m the slope of the fatigue plastic rotation range curves and
log a a constant.
Reference values for the damage evaluation may be provided from the results of
monotonic tests. Monotonic loading corresponds to one-half of a cycle of a speci-
men deformed to up to Du ¼ umon and unloaded to zero plastic deformation. That
means that monotonic loading corresponds to a pair Nmon ¼ 1=2 and Du ¼ umon .
Accordingly, if the rotation capacity under monotonic loading /mon is known, the
number of cycles for a certain range of plastic rotation may be determined from:
!m
1 umon
N¼ ð2Þ
2 Dup

Studies from experimental investigations reveal that the value of the slope m is
approximately equal to 2. Accordingly a value m ¼ 2 was adopted for the fatigue
curve. For the rotation capacity under monotonic loading umon, a value equal to
0.04 radians was adopted for the fatigue curve. For variable ranges of plastic
rotation, the damage assessment is performed in accordance to the linear Palmg-
ren–Miner cumulative law in accordance with:
X ni
D¼ ð3Þ
Ni
where ni is the number of cycles of deformation range Dui and Ni the number of
cycles of the same deformation range that cause failure.
For the determination of the design spectrum in the fatigue assessment, the rain-
flow or reservoir method for counting the cycles for a certain deformation history
has been employed. The limit state is defined as the situation where the damage
index becomes equal to 1.
744 D. Grecea et al. / Journal of Constructional Steel Research 60 (2004) 739–749

Table 1
Characteristic value limits associated to the performance levels
Performance Description of damage state Limit drift Limit Plastic
level/limit state (%) residual rotations
drift (%) (rad)
SLS Light damages in structure and 0.6 – –
non-structural elements
Continuity in building occupancy
DLS Damages are moderate but structure – 1% –
is stable
Building can be evacuated after earthquake
because repairs are necessary
Repairs are possible
ULS Severe damages but the collapse of – – 0.04a
structure is prevented
Building repair is no longer possible
a
The corresponding drift is 3% approximately.

In Table 1 the performance levels, together with the description of damage state
and maximum characteristic values (interstorey drifts, residual interstorey drift,
plastic rotations) are described.
2.3. Simplified methods of evaluation for global and local seismic performances of
MR frames

The most suitable approach for seismic design based on performance appears to
be deformation-controlled design. However, today codes are based on force-
controlled design, using the base shear concept. The most important parameter in
this approach is the behaviour q-factor [13], which is based on the maximum ca-
pacity of structure to dissipate energy during the plastic deformations correspond-
ing to ULS criterion, without possibility to verify other performance levels.
Aribert and Grecea [7,8] have introduced a new definition of the q-factor based
on the reduction of the base shear force of a structure (Eq. (4)). Values for the new
q-factor have been established for different types of steel structure with rigid and
full strength joints or with semirigid and/or partial strength joints
V ðe;thÞ V ðeÞ ku
q¼ ðinelÞ
¼ ðinelÞ ð4Þ
V V ke
It is very important to be underlined that this new definition may be applied to
any level of performance mentioned above and settled by Fig. 1, such as as
strength, drift and rotation capacity. Today codes give a maximum q-factor of con-
stant value, including both effects of ductility and overstrength. The ULS corre-
sponding ductility cannot be attained, if higher levels of performance are required.
In that case, a reduced ductility corresponding to a partial q-factor is attained by
structure. This partial q-factor may be computed in any case, at any level of per-
formance wished by design engineer, applying this new method. Thus, the use of
 D. Grecea et al. / Journal of Constructional Steel Research 60 (2004) 739–749 745

Table 2
Geometric properties of the frames under consideration
Type of frame

Frame L (m) H (m) Structure period (sec) Beams Columns

Semirigid Rigid
1 5 3 0.73 0.67 IPE300 HEB180
2 4 4 1.27 1.16 IPE330 HEB240
3 4 3 1.39 1.25 IPE360 HEB280

this partial q-factor gives the possibility to implement the multiple performance
design in the actual code methodology.

3. Parametric study

The parametric study refers to frames with different geometric conditions (see
Table 2). Two types of behaviour for the beam-to-column joints were examined,
rigid and semirigid, with an amount of semirigidity of 0.4K (according to the defi-
nitions of Eurocode 3 [11]), where K ¼ 25Elb =Lb expresses the stiffness of the
beam. The joints were considered as full resistant. If the column web is slender, it
is possible to obtain beam-to-column joints with properties approaching of the
parameters considered in this study. The frames were subjected to Kobe (1995)
seismic record. The response spectrum of Kobe record is shown in Fig. 2.

Fig. 2. Acceleration response spectra of Kobe (1995) record.

746 D. Grecea et al. / Journal of Constructional Steel Research 60 (2004) 739–749

Fig. 3. Characteristic of ground motion: acceleration vs. return period [11].

To be used in design, performance levels must be translated into seismic action,

represented by magnitudes or accelerations. In case of using the recurrence period,
the level of acceleration depends on it. The selection of ground motion acceleration
for the three limit states was determined as a function of the return periods. For
DLS and ULS, respectively, there are no contradictions concerning the return peri-
ods (475 and 970 years, respectively). Contrary to this, for SLS there are different
proposals (ranging from 10 to 75 years), due to the difficulties in choosing a
rational criterion for non-damage limit states. If the acceleration for damageability
limit state ad is considered as a basic value for ground motion acceleration, for the
other periods the corresponding accelerations are determined from the equation
[10,15]:
 0:28
a pr
¼ ð5Þ
ad prd
based on ATC 40 [16] proposal. For the SLS and ULS, the corresponding accel-
erations are as follows (see Fig. 3):
as ¼ 0:412ad ð6aÞ
au ¼ 1:22ad ð6bÞ

Table 3
Acceleration multipliers for the three limit states

Frame 1 Frame 2 Frame 3

Rigid Semirigid Rigid Semirigid Rigid Semirigid

SLS 0.31 0.29 0.36 0.30 0.32 0.29

DLS 0.46 0.44 0.53 0.62 0.74 0.77
ULS 0.46 0.43 0.53 0.63 0.57 0.62
 D. Grecea et al. / Journal of Constructional Steel Research 60 (2004) 739–749 747

Fig. 4. Limit acceleration multipliers au for rigid and semirigid frames.

The analysis was performed using the general purpose DRAIN-2DX software
package [14].
In the first step, the maximal accelerations a, for which the structures meet the
specified performance criteria, were determined by appropriate scaling (see Table 3) if
the additional rigidity brought by interaction between structural and non-structural
elements is considered (structure is in elastic range and non-structural elements are
undamaged), this difference could be reduced).
In Fig. 4, the variation of acceleration multipliers for the limit states considered
in the analysis is presented. One of the most important problems that must be ful-
filled in case of multi-level design is the optimisation of solutions. It may be
observed that in case of DLS and ULS, the input accelerations are very close,
which means the both requirements are simultaneously satisfied. If the additional
rigidity brought by interaction between structural and non-structural elements is
considered (structure is in elastic range and non-structural elements are unda-
maged), this difference could be reduced.
In the second step, in order to evaluate the global ductility of the frames under
consideration, the q-factor was calculated using the new approach described by Eq.
(4) (see Table 4). In case of SLS, a value of q-factor of 1.0 was imposed, consider-
ing the structure in elastic range. Inelastic base shear forces were evaluated at a
level of acceleration specified in the first step (see Table 5).

Table 4
q-factors for the three limit states

Frame 1 Frame 2 Frame 3

Rigid Semirigid Rigid Semirigid Rigid Semirigid

SLS 1.00 1.00 1.00 1.00 1.00 1.00

DLS 2.30 1.80 2.20 2.60 3.80 3.30
ULS 2.70 2.10 2.60 3.20 3.60 3.20
748 D. Grecea et al. / Journal of Constructional Steel Research 60 (2004) 739–749

Table 5
Base shear forces for the three limit states
Frame 1 Frame 2 Frame 3
Rigid (kN) Semirigid Rigid (kN) Semirigid Rigid (kN) Semirigid
(kN) (kN) (kN)

SLS 81.0 96.8 160.7 176.5 304.4 348.4

DLS 131.2 131.3 271.8 278.2 583.1 613.0
ULS 131.1 131.1 270.7 277.9 583.9 617.9

In Fig. 5, the variation of acceleration multipliers for the limit states considered
in the analysis is presented. Analysing values of q-factor presented in Table 5, it
is observed again that in case of DLS and ULS the base shear forces are quite
equal, both for rigid and semirigid frames. However, q-factors are increasing from
SLS to ULS, mainly due to different levels of base shear force V(e) associated
with SLS.

4. Conclusions

Performance criteria of MR steel frames were evaluated through global and local
characteristics. Three performance levels/limit states were introduced, i.e. SLS,
DLS and ULS.
Acceleration multipliers and q-factors for the three limit states were obtained via
a non-linear dynamic time-history analysis.
One of the most important problems to be fulfilled in case of multi-level design
is the optimisation of solutions. It was observed that in case of DLS and ULS, a
proper choice of limit characteristics (inter-storey drift and plastic rotation
capacity) could lead to a similar level of input accelerations, which represents the
optimal case. The only difference exists in case of SLS, but taking into account the
interaction between structural and non-structural elements, this difference could be
reduced.

Fig. 5. q-factors for rigid and semirigid frames.

 D. Grecea et al. / Journal of Constructional Steel Research 60 (2004) 739–749 749

Concerning the q-factor, it is very important to underline that this new definition
may be applied to any level of performance. Usually, classical methods give a
maximum q-factor, which cannot be used if we apply for a performance-based
design, and a partial q-factor, less than the maximum one, should be used.

References
[1] Ghobarah A. Performance-based design in earthquake engineering: state of development. Eng
Struct 2001;23:878–84.
[2] Structural Engineers Association of California (SEAOC). Vision 2000 a framework for perform-
ance-based engineering. Sacramento (CA): Structural Engineers Association of California; 1995.
[3] Applied Technology Council (ATC). NEHRP guidelines for seismic rehabilitation of buildings.
Report No. FEMA-273. Washington (DC): Federal Emergency Management Agency; 1997.
[4] Bertero RD, Bertero VV. Application of a comprehensive approach for the performance-based
earthquake resistant design buildings. 12th World Conference on Earthquake Engineering, Auck-
land, 30 January–4 February. 2000 [CD-ROM 0847].
[5] Bertero RD, Bertero VV. Redundancy in earthquake-resistant design. J Struct Eng 1999;125(1):
81–88.
[6] Mazzolani FM, Montuori R, Piluso V. Performance based design of seismic-resistant MR frames.
In: Mazzolani FM, Tremblay R, editors. Proceedings of the Third International Conference
STESSA 2000, Behaviour of Steel Structures in Seismic Areas, 21–24 August, Montreal, Canada.
Rotterdam: A.A. Balkema; 2000, p. 611–8.
[7] Aribert JM, Grecea D. A new method to evaluate the q-factor from elastic-plastic dynamic analysis
and its application to steel frames. In: Mazzolani FM, Akiyama H, editors. Proceedings of the
Second International Conference STESSA 1997, Behaviour of Steel Structures in Seismic Areas, 3–
8 August, Kyoto, Japan. Napoli: Edizioni 10/17, 1997. p. 382–93.
[8] Aribert JM, Grecea D. Numerical investigation of the q-factor for steel frames with semi-rigid and
partial-strength joints. In: Mazzolani FM, Tremblay R, editors. Proceedings of the Third Inter-
national Conference STESSA 2000. Behaviour of steel structures in seismic areas, 21–24 August
2000, Montreal, Canada. Rotterdam: A.A. Balkema; 2000, p. 455–62.
[9] Vayas I, Dinu F. Influence of joint flexibility in the seismic performance of moment resisting steel
frames. NATO Advanced Research Workshop ‘‘The Paramount Role of Joints into the Reliable
Response of Structures, From the Rigid and Pinned Joints to the Notion of Semi-rigidity’’,
Ouranoupolis, Greece, 21–23 May. p. 240–248.
[10] Lee LH, Lee HH, Han SW. Method of selecting design earthquake ground motions for tall build-
ings. Struct Des Tall Build 2000;9:201–13.
[11] Eurocode 3 Part 1.1. Design of steel structures, general rules and rules for buildings. CEN, European
Committee for Standardisation. pr. EN 1993-1-1. 1992.
[12] Eurocode 3 Part 1.9. Fatigue strength of steel structures. CEN, European Committee for Standardi-
sation. pr. EN 1993-1-9. 2000.
[13] Eurocode 8 Part 1.1. Design provisions for earthquake resistance of structures. CEN, European
Committee for Standardisation. ENV 1998-1-1. 1994.
[14] Kannan, A, Powel, G, DRAIN-2D. A general purpose computer program for dynamic analysis of
inelastic plane structures. EERC 73-6 and EERC 73-22 reports. Berkeley, USA. 1975.
[15] Gioncu V, Mazzolani FM. Ductility of seismic-resistant steel structures. London: SPON Press;
2002.
[16] ATC. Seismic evaluation and retrofit of concrete buildings, vol. 1. ATC-40. Redwood City: Applied
Technology Council; 1996.