13th World Conference on Earthquake Engineering

Vancouver, B.C., Canada

August 1-6, 2004
Paper No. 947

GENERAL CONSIDERATIONS ON STRUCTURAL REGULARITY

Aurelio Ghersi1

SUMMARY

In the last decades the problem of structural regularity has been analyzed in a large number of papers,
which pointed out the negative effects of the lack of regularity on the elastic and inelastic seismic
response of structures and suggested design approaches able to limit the risks connected to it. Nearly all
the seismic codes include general definitions of structural regularity and provisions aiming at limiting
negative effects of irregularity. Nevertheless, in many cases (as in the recent version of the European
seismic code, Eurocode 8) the approach appears to be generic and oversimplified, not taking into account
the outcome of research.
The basic idea presented in this paper is that a building should be considered regular when its seismic
response may be foreseen by means of standard geometrical and mechanical models and standard methods
of analysis. In order to discuss regularity all the different problems connected to the definition of the
model and to the selection of the method of analysis have to be analyzed. Each problem gives rise to
specific criteria for regularity, which have to be met in order to allow a simplification of the model or of
the method of analysis.
The first part of the paper discusses the different models that may be used in describing the building, the
local problems of modeling, and the influence of the stiffness of the floor-diaphragm and of the non-
structural elements. The second part examines the approaches used for evaluating its elastic response,
discussing the use of static and modal analysis and the use of corrective eccentricities for asymmetrical
buildings. The third part tackles the problem of the inelastic response and the influence of stiffness and
strength distribution, both in plan and along the height of the building.

INTRODUCTION

Every structure to be erected in a seismic region has to be designed and constructed in such a way to meet,
with an adequate degree of reliability, specific requirements connected to the return period of seismic
action. Each seismic code should define a set of return periods of seismic action and the corresponding
required performances, ranging from a damage limitation requirement to a no-collapse requirement.
In the first case the structure will remain in the elastic range, while in the last one it will undergo large
inelastic deformation.
In order to check the structural performance, it is necessary:

1
Full professor, DICA, University of Catania, Italy. Email: aghersi@dica.unict.it
 to define a geometrical and mechanical model of the building, which may include only the structural
elements or also the so-called non-structural elements;
to evaluate the seismic response of the building in the elastic range;
to evaluate, or to estimate, the seismic response of the building in the inelastic range.
If the building, and its structure, has been conceived respecting the general principles of a good
conceptual design (structural simplicity, uniformity, symmetry and redundancy, bi-directional resistance
and stiffness, torsional resistance and stiffness, diaphragmatic behavior at story level, adequate
foundation, e.g. see Eurocode 8) it is possible to use standard modeling while checking the structural
performance. The above-mentioned principles grant adequate reliability of the numerical analysis and, at
the same time, promote a good behavior under seismic actions more severe than the design ones.
It is thus possible to define regular a building, and a structure, the seismic response of which may be
foreseen by means of standard geometrical and mechanical models and standard methods of analysis. In
order to discuss regularity all the different problems connected to the definition of the model and to the
selection of the method of analysis have to be analyzed. Each problem gives rise to specific criteria for
regularity, which have to be met in order to allow a simplification of the model or of the method of
analysis.
Note: The progress of research, together with the availability of more and more powerful computing
devices, allows the use of more complex geometrical and mechanical models and methods of analysis.
Although this may influence the above-proposed approach to the problem of regularity, it must be
reminded that a well-designed structure should always require the simplest numerical approach possible.

MODELING

Standard and simpler models of the building

A widely adopted model of the building, which is usually able to adequately represent the distribution of
stiffness, is the three-dimensional frame with rigid floor diaphragms:
Beams and columns of the building are modelled as one-dimensional members, mutually connected in
points named nodes.
The floor is considered as a diaphragm, infinitely rigid in its plane.
Non-structural elements, as partition walls, are usually not included in the model.
Fixed restraints are often considered at the base of the three-dimensional frame, thus assuming very
stiff soil and foundation; elastic base restraints, or beams on elastic soil, are sometimes used to model
more flexible soil and foundation.
When the vertical resisting elements are arranged along two orthogonal directions, a slightly simplified
model may be used, which describes the structure as a three-dimensional set of plane frames. All the other
hypotheses (one-dimensional members, rigid floor diaphragms, neglected non-structural elements, fixed or
elastic base restraints) remain unchanged.
A much simpler model may be in some cases used: the plane frame. It is a poor model, because it cannot
give information on the deck rotations, which are generally present even in geometrically symmetric
schemes because of the possibility of non-symmetrical distribution of variable loads. Anyway, it may give
in some cases useful information on the structural response, provided that proper integrations or
corrections are applied.

Local problems
The three-dimensional frame model considers one-dimensional members connected in nodes. In actual
buildings, the axes of beams and columns may not converge in single points; the dimensions of cross-
section may be not small, compared to the length of the member, and may abruptly vary from one member
to another. This may cause problems in action transfer, which are not pointed out by the model.
Problems in action transfer may arise also when vertical members are not continuous, e.g. when a column
is missing at a story. The three-dimensional frame model is able to evaluate the effect of such a
discontinuity. On the contrary, the three-dimensional set of plane frames model may give unreliable
results, if the column belongs to different plane frames, because the model does not consider the vertical
compatibility of nodes belonging to different plane frames.
A building is considered regular, under this aspect, when:
The axes of beams and columns are properly aligned, so as to converge in single nodal points.
The dimensions of the cross-sections of beams and columns are small, compared to their length, and do
not abruptly vary in adjacent elements.
The vertical members (columns, walls, cores, etc.) are continuous, from the foundation to the top of the
building.
The lack of regularity imposes:
Local checks, with models able to provide more detailed information about the stress distribution in the
non-regular zones.
More detailed constructive provisions for the non-regular zones.

Floor diaphragm
In the model, the floor slabs are considered as mutual restraints among the nodes they connect. In actual
buildings the stiffness or the strength of the floor may be not adequate. This may depend on the floor
typology: e.g., in reinforced concrete structures a floor with reinforcements in two orthogonal directions is
by itself able to provide large stiffness and strength; on the contrary, a floor with bricks and one-
directional main reinforcements requires an adequately thick slab, with bi-directional reinforcement, in
order to be considered a rigid diaphragm. It may depend also on the shape of the floor: very elongated in-
plan shapes, large re-entrant corners, large openings in the floor for lifts or technical plants and the
discontinuity of horizontal slab in correspondence of the stairs may reduce the stiffness and may give rise
to localized high values of stress. Finally, it may depend on the distribution of the main resisting elements:
e.g., abrupt variation of size or location of steel bracings, or r.c. walls, from one story to another may cause
a relevant re-distribution of horizontal actions, which may lead to very high values of stress and strain in
the floor.
A building is considered regular, under this aspect, when:
The floor is adequately conceived in order to bear horizontal actions.
The floor shape is compact.
The main horizontal-action resisting elements are uniformly distributed in plan and their stiffness and
location do not abruptly change from one story to another.
Note: In some cases, the first two requirements may be achieved by subdividing the entire building into
dynamically independent units by means of seismic joints
Although simple geometrical conditions may be suggested by codes in order to check the above regularity
conditions, it is recommended to perform a numerical check of stiffness and strength of the floor
diaphragm, whenever it is deemed necessary:
The floor may be modeled as a plate, or as a grid, subjected to in-plan actions.
The horizontal actions may be obtained by the results of the three-dimensional frame model
(distributed seismic action along the floor, concentrated reactions of the horizontal-load resisting
elements of the above and below inter-story. These actions should be increased in order to comply with
the capacity design criterion, so as to consider the maximum reactions of the horizontal-load resisting
elements.
The stress induced in the floor slab by the above horizontal actions should be smaller than given
reference limits. In most cases, this condition may be satisfied by adopting adequate reinforcements (in
r.c. slabs) or adequate cross-section of the bracings (in steel slabs).
 The deformation of the floor slab should be as small as not to induce relevant changes in the stress
distribution of vertical resisting elements. This condition may be deemed satisfied when, applying to
the elements of the three-dimensional frame relative displacements corresponding to the slab
deformation, the internal actions vary less than 10%.
The lack of regularity imposes:
The use of a more general structural model, which accounts for the deformability of the floor
diaphragm, or of additional simplified models able to estimate, in a safe way, the variation of beam and
column internal actions caused by floor deformation.
More detailed constructive provisions for the weak zones of the floor diaphragm.

Non-structural elements
The model does not include non-structural elements, like partition or external walls. In actual buildings,
these elements participate to the bearing of seismic actions, with aspects that may be either safe or unsafe
for the structure. They reduce the global shear on the structure (safe) but, at the same time, they may
modify the location of stiffness center and stiffness radius of gyration, giving rise to in-plan rotation of the
floor that may increase the shear in the outmost elements (unsafe). They produce relevant changes in the
axial force of adjacent columns, together with local increment of shear force and bending moment in the
contact zone (unsafe). They increase the global stiffness and reduce the fundamental period of vibration of
the building, thus reducing the horizontal displacements (safe) but in most cases increasing the horizontal
accelerations (unsafe). The variation of stresses they induce may lead to unexpected structural collapse
mechanisms, like soft story (unsafe).
A building is considered regular, under this aspect, when:
The non-structural elements (partition and external walls) are not particularly rigid, compared to the
structural elements.
Their in-plan distribution is uniform and not very different from that of the structural elements.
Their stiffness and in-plan location do not abruptly vary from one story to another.
The lack of regularity imposes:
The inclusion of non-structural elements in the model. In the elastic range, they may be modeled as
plates or struts, although the uncertainties about their connection to the structure make not easy to
define whether their contribution starts immediately, or if it must be considered the presence of a gap.
Much more complex is the definition of a model that accounts for their cyclic behavior in the inelastic
range; simplified models, able to give safe results, should be used in this case.
More detailed constructive provisions for the contact zones of the adjacent columns.

Soil and foundation

Usually, the model assumes fixed restraints at the base. In actual buildings, the vertical displacement and
the rotation of the base, due to the flexibility of soil and foundation, may modify the distribution of
internal actions among the elements of the frame. Furthermore, even when the foundation is very rigid, the
rocking due to the soil flexibility or to the limited extension of the foundation may increase the first period
of vibration of the structure.
A building is considered regular, under this aspect, when:
The foundation elements connect at the base all the elements of the frame and have flexural stiffness
higher than that of the elements of the frame.
The soil stiffness and the extension of the foundation are such as to avoid rocking of the building.
Although simple geometrical conditions may be suggested by codes in order to check the above regularity
conditions, it is recommended to perform a numerical check of the stiffness of soil and foundation,
whenever it is deemed necessary:
The analysis of soil and foundation, subjected to the actions given by the over-standing frame, provides
displacements and rotations of the nodes that connect frame and foundation.
 The deformation of the foundation should be as small as not to induce relevant changes in the stress
distribution of beams and columns. This condition may be deemed satisfied when, applying to the
elements of the three-dimensional frame the displacements and rotations, above evaluated, the internal
actions vary less than 10%.
The rotation of the base, considered as a rigid block, should induce small variation of the horizontal
displacement of the frame. This condition may be deemed satisfied when, taking into account the effect
of a rigid rotation of the base, the top displacement vary less than 10%.
The lack of regularity imposes:
The inclusion of soil and foundation in the model, by considering elastic restraints instead of fixed
restraints or by adding more structural elements, e.g. beams on elastic soil.
A careful modeling of the soil, in order to account for the effect of the possible variability of its
mechanical characteristics in different points of the base of the building.

Plane frame model

The plane frame model gives correct results only in the case of balanced schemes, i.e. when mass center
and stiffness center coincide at every story. Obviously, two plane frames have to be considered, in order to
analyze the seismic response of the building to the seismic actions applied along two orthogonal
directions. Anyway, if the structural eccentricity es (distance between mass center and stiffness center) and
the radius of gyration of stiffness rk are constant along the height of the building, it is possible to evaluate
the internal actions due to static forces in a member at a distance x from the stiffness center, by
multiplying the results of the plane frame model by the normalized factor = 1 + es x / rk2. Under the same
conditions, plus the hypothesis that the radius of gyration of mass rm is constant along the building height,
it is possible to obtain correct values of modal analysis, by multiplying the results of this type of analysis,
performed for the plane frame, by a coefficient that represents the normalized displacement of a one-
story three-dimensional system having the same structural eccentricity and radius of gyration of stiffness.
The correction is easier when the building is torsionally rigid: rk has a quite large value; the corrective
factor for static analysis is less relevant; the corrective factor for modal analysis is not so far from the
linear values expressed for static analysis.
A building is considered regular, under this aspect, when:
The mass centers (or the points of application of static forces) and the stiffness centers are located
along two vertical lines, the distance of which is es.
The radius of gyration of stiffness rk and radius of gyration of mass rm are constant along the building
height.
The building is torsionally rigid, i.e. rk is larger than rm. This last condition is not strictly necessary, but
it may give some simplification, avoiding the necessity of an accurate evaluation of rk.
Although some geometrical conditions may be suggested by codes in order to check the above regularity
conditions, it is recommended to perform a numerical check of the alignment of mass and stiffness centers
and of the constancy of the radius of gyration of stiffness and mass, whenever it is deemed necessary:
Knowing the distribution of masses (permanent and variable loads) it is possible to evaluate at each
inter-story the location of the center of masses and their radius of gyration.
Basing on the results of the analysis of the two plane frames subjected to horizontal forces, it may be
defined lateral stiffness of a column (or a r.c. wall, or a sub-frame constituted by elements aligned in
plan) at a given inter-story and in a given direction the ratio between the adsorbed shear force and
the relative displacement of the ends of the element.
It is thus possible to evaluate at each inter-story the location of the center of the above-defined
quantities and their radius of gyration.
The non-alignment of the centers and the differences in radius of gyration should be such as not to
modify the structural response and the distribution of internal actions more than 10%. This may be
 checked, if necessary, by repeating the afore-mentioned correction using both the maximum and
minimum value assumed by each variable quantity (eccentricity or radius) along the height.
Note: Researchers should investigate if the results corresponding to a variation, story-by-story, of
eccentricity and radius are included between those obtained using maximum and minimum values of
eccentricity and radius.
The lack of regularity imposes to use the three-dimensional frame or three-dimensional set of plane frames
model.

Model for masses

In order to adequately represent the distribution of masses, it is necessary to take into account the in-plan
distribution of permanent (dead) and variable (live) loads. Although in actual buildings some masses are
concentrated in points (columns) or distributed along lines (beams, partition and external walls), they are
usually considered uniformly distributed in the floor. Also the value of variable loads, which may differ
from point to point, is considered uniformly equal to a reduced value. The center of masses and their
radius of gyration are thus assumed coincident to those of the floor.
The possible variation of the location of mass center, due to non-uniform distribution of variable loads, is
taken into account by means of a so-called accidental eccentricity ea. Although present seismic codes
suggest for such eccentricity a unique value (ea = 0.05 L, being L the floor dimension perpendicular to the
seismic action), the influence of variable loads should be connected both to the use destination of the
building and to the structural material used. E.g., the accidental eccentricity should be smaller in masonry
building and larger in steel structures.
A building is considered regular, under this aspect, when:
The masses are distributed in the floor in a nearly uniform way.
The entity of permanent and variable loads is properly connected to the value assumed for accidental
eccentricity.
The lack of regularity imposes:
A more detailed evaluation of the location of mass center and of radius of gyration of masses.
A more detailed evaluation of the variation of the mass-center location, due to the most unfavorable
disposition of variable loads.

ELASTIC RESPONSE

Standard and simpler models

Nowadays, modal analysis may be considered the standard approach in evaluating the elastic response of a
structure. Modal analysis provides a correct estimate of the elastic response of any structure in the elastic
range, provided that:
The effect of each horizontal component of seismic action is evaluated using a more accurate
combination procedures, like Complete Quadratic Combination (CQC), instead of Square Root Sum of
Squares (SRSS) combination whenever the periods of two vibration modes differ less than 10%, i.e.
for nearly all three-dimensional schemes.
The effect of the contemporaneous presence of two orthogonal components of seismic action is
properly evaluated, e.g. by combining the separate effects of the two components by means of the
SRSS combination.
Note: I believe modal analysis always able to provide a correct estimate of the elastic response. Anyway, if
some researcher will point out cases in which it is unsafe and it is necessary to use a step-by-step
response analysis regularity conditions necessary to avoid such a more complex procedure may be added
here. Furthermore, an alternative approach might consider standard a modal analysis with SRSS
combination and without combination of the effects of orthogonal components. In this case, regularity
conditions should be added to define when this approach is safe and when, on the contrary, CQC and
combination of orthogonal actions are necessary.
An alternative, widely used, approach is the static analysis. Although this is a poor approach, which gives
no information on the periods of vibration of the scheme and on the effects of higher modes of vibration,
its simplicity and physical evidence makes it an important tool for evaluating seismic elastic response,
provided that the effect of the first mode of vibration is prevailing in the scheme and proper integrations or
corrections are applied when necessary.
A building is considered regular, under this aspect, when:
The structural typology, the structural material used, the entity and location of non-structural elements
allow a reliable estimate of the first period of vibration.
The effect of the first mode of vibration is prevailing in the scheme. This condition may be deemed
satisfied if the estimated period of vibration is smaller than given values, related to the expected
response spectrum, i.e. to the mechanical characteristics of the soil.
The conditions described in the following sections, in function of the geometrical characteristics and
the model used for the building, are respected.
Note: Limits of the first period of vibration, for the application of static analysis, should be explicitly
added here.
Although seismic codes should provide formulations able to allow an easy and reliable estimating of the
first period of vibration, it is recommended to perform a numerical check of the first period, whenever it is
deemed necessary:
Knowing value and distribution of masses along the height, it is possible to evaluate the total mass M
and the elevation z of it center.
Basing on the results of the static analysis, it may be defined lateral stiffness K of the frame the ratio
between the base shear-force and the displacement of the frame at the elevation z.
It is thus possible to evaluate the first period as 2 M / K .
The lack of regularity imposes the use of modal analysis.

Plane frames
Static analysis of plane frames provides safe results, compared to the modal one, if mass and stiffness do
not abruptly vary along the height. It is anyway well known that, for frames having more than one story,
the base shear-force used in static analysis is quite larger than the corresponding value obtained by means
of modal analysis (from 10 to 30% more). For a better correspondence between the results of the two
analyses, many codes conventionally prescribe to increase the results of modal analysis, so as to obtain at
the base the same shear force used in static analysis (some codes accept a smaller shear force when the
scheme is regular). With a theoretically more correct approach, although less safe, Eurocode 8 (2002)
allows a 15% reduction of the results of static analysis, provided that the first period of vibration does not
exceed given limits. This tendency to equalize the two analyses is acceptable, and theoretically desirable,
but it reduces the safety intrinsically connected to static analysis and requires a more careful check of its
applicability.
It must be furthermore taken in mind that static analysis gets the effect of horizontal acceleration only. In
the case of very large span or of discontinuity of the vertical elements (columns) the effect of vertical
acceleration cannot be neglected.
A building is considered regular, under this aspect, when in addition to the regularity conditions of section
3.1:
There is no abrupt variation of mass or stiffness along the height.
The vertical members (columns, walls, cores, etc.) are continuous, from the foundation to the top of the
building.
The span of the horizontal elements (beams) does not exceed given limits.
The lack of regularity imposes:
 The use of modal analysis.
The evaluation of the effects of vertical acceleration, if the last two conditions are not satisfied.

Balanced buildings analyzed by means of plane frame model

As mentioned before, the plane frame model should be used only in the case of balanced schemes, i.e. if
the decks translate without any rotation when the seismic forces are applied at mass centers of the
building (or, equivalently, if mass center and stiffness center coincide at every story). Even in this case,
the possibility of non-symmetrical distribution of variable loads (i.e. the above-defined accidental
eccentricity ea) requires an estimation of the in-plan rotation of the floors. It is possible to account for the
static effect of this eccentricity, in a member at a distance x from the stiffness center, by multiplying the
results of the plane frame model by the normalized factor = 1 + ea x / rk2. In torsionally rigid schemes
(for which static and modal analysis do not differ too much) this approach gives results approximately
valid also for modal analysis; further simplifications may be used in this case: e.g., Eurocode 8 suggests to
use the expression = 1 + 12 ea x / L2, which is equivalent to assume rk 0.3 L approximately.
A building is considered regular, under this aspect, when in addition to the regularity conditions of
sections 3.1 and 3.2:
The nominal position of mass center (corresponding to a uniform distribution of loads) and the position
of stiffness center coincide at every story.
The radius of gyration of stiffness rk and radius of gyration of mass rm are constant along the building
height.
The building is torsionally rigid, i.e. rk is larger than rm.
The lack of regularity imposes:
The use of modal analysis, if the building cannot be considered balanced or it is not torsionally rigid.
Or:
A more detailed evaluation of the corrective factor , if the only non-satisfied condition is the torsional
rigidity of the scheme (see references in section 3.4).

Single-story buildings
In an asymmetric (or non balanced) single-story scheme, the force applied with static analysis is not very
different from the base shear-force provided by modal analysis. Nevertheless, the horizontal displacements
given by static analysis may be very different from the maximum displacements of modal analysis. In
particular, the first ones vary in-plan in a linear way, while the second ones present a curved shape,
because the maximum displacement at different points is reached in different times. Static analysis is
therefore always safe for some structural elements, unsafe for other. In order to get safe results, it is
necessary to evaluate the internal actions as the envelope of the results of two static analyses, performed
applying horizontal force in points displaced, from the mass center, by a quantity named corrective
eccentricity ec.
Under this aspect, a building can never be rigorously considered regular.
The peculiarity of single-story asymmetric schemes imposes:
The use of modal analysis.
Or:
The use of two static analyses, performed applying the force in points displaced, with respect to the
mass center, of proper values of corrective eccentricity.
Although many seismic codes and research groups provide simple formulations for estimating the value of
corrective eccentricity, it is recommended to use more reliable formulations or procedures. E.g., see
Anastassiadis et al. [1], Calderoni et al. [2].
Multi-story buildings
Multi-story buildings denote the same problem of non-correspondence of the shape of static and modal
deformation discussed for single-story schemes. The safety intrinsically connected to the static analysis of
schemes having more than one story, already recalled for plane frames, may be in many cases (although
not always) sufficient to counterbalance the above-mentioned inadequacy, but it may be nullified by the
attempt of equalizing static and modal base shear-force. For this reason, a correct use of static analysis
should consist, as proposed for single-story schemes, in evaluating the internal actions as the envelope of
the results of two static analyses, performed applying horizontal forces in points displaced, from the mass
center, by proper values of corrective eccentricity ec.
Being corrective eccentricity connected to structural eccentricity es and radius of gyration of mass and
stiffness, rm and rk, in multi-story buildings the proposed approach may be applied only if these quantities
do not vary along the height. Schemes that respect this condition are usually named regularly asymmetric.
E.g., see Hejal and Chopra [3].
A building is considered regular, under this aspect, when in addition to the regularity conditions of
sections 3.1 and 3.2:
The mass centers and the stiffness centers are located along two vertical lines, the distance of which is
es.
The radius of gyration of stiffness rk and radius of gyration of mass rm are constant along the building
height.
Although some geometrical conditions may be suggested by codes in order to check the above regularity
conditions, it is recommended to perform a numerical check of the alignment of mass and stiffness centers
and of the constancy of the radius of gyration of stiffness and mass, whenever it is deemed necessary:
Knowing the distribution of masses (permanent and variable loads) it is possible to evaluate at each
inter-story the location of the center of masses and their radius of gyration.
Basing on the results of the analysis of the three-dimensional frame model subjected to a set of
horizontal forces Fi, applied to the mass centers, and to a set of couples Mi, obtained by multiplying the
forces Fi by a whatsoever eccentricity e0 (e.g., e0 = ea), it is possible to evaluate at every story the
quantities:
2
F ,i v F ,i F ,i
e s , i = e 0 rk ,i =e 0
M ,i e 0 M ,i M ,i
The procedure may be applied if the variations along the height of the above quantities do not modify
the structural response and the distribution of internal actions more than 10%. This may be checked, if
necessary, by repeating the analysis using both the minimum and the maximum value assumed by each
variable quantity (eccentricity or radius) along the height.
E.g. see Calderoni et al. [2], Moghadam and Tso [4]. Recent studies show that static analysis with
corrective eccentricities may give safe results even when the structural eccentricity and radius of gyration
of stiffness vary in a more relevant way along the height. It has been proposed the use of other parameters
in order to check the applicability of static analysis. See, for example, Bosco et al. [5].
The lack of regularity imposes the use of modal analysis.

INELASTIC RESPONSE

Standard approach
Nowadays, the standard design approach to the problem of the inelastic response of buildings is the so-
called force-based design. It is assumed that the inelastic response of the building shall be acceptable, i.e.
the building will not collapse, if the structure is able to withstand (by means of an elastic analysis, without
exceeding allowable stresses or the ultimate limit state of each cross-section) horizontal forces evaluated
by a design spectrum. This one is obtained by reducing the elastic response spectrum by a proper
coefficient (named R in UBC, q in Eurocode 8), related to the dissipative characteristics of the scheme and
of the cross-sections. Each seismic code should provide standard values of this coefficient, usually
calibrated for plane frames with different ductility levels, together with rules for improving both global
and local ductility of the structure.
A building is considered regular, under this aspect, when its inelastic behavior does not differ in a relevant
way from that of the ductile plane frame, to which the reduction factors proposed by codes are related.
Specific regularity conditions are described in the following sections, in function of the geometrical
characteristics and the model used for the building.
The lack of regularity may impose, in function of the geometrical characteristics and the model used for
the building:
The use of smaller values of the reduction factor.
Or:
The use of additional load conditions.
Or:
The use of step-by-step inelastic response analysis.

Plane frames
Abrupt variations of stiffness along the height of the frame may lead to high values of locally required
ductility, thus causing a worsening of the inelastic response and making necessary to provide more
strength. Typical example is the so-called soft story, i.e. a story having stiffness really smaller than that of
the storys above and below it; notice that this situation may be caused also by an improper distribution of
non-structural elements, partition or external walls, even in buildings that present a uniform distribution of
stiffness of structural elements.
An analogous problem may be produced by a strong increase in stiffness of on or few vertical elements at
a story, if it is not coupled to a proportional increase in strength. Typical examples are the presence of a
shorter column or of an inclined beam connecting a floor to the mid-height of a column (as in a typology
of stairs commonly used in Italy).
A building is considered regular, under this aspect, when:
The total lateral stiffness of structural and non-structural elements does not vary in a relevant way from
one story to another.
No element has a lateral stiffness much higher than that of other elements of the same story, unless it
has a proportionately higher strength.
The lack of regularity imposes:
The use of smaller values of the reduction factor. When the irregularity is clearly referred to a limited
portion of the frame (e.g. to a single story) it is possible to perform two numerical analyses: the first
one with a standard value of reductive coefficient and the second one with a smaller value of it, using
the results of the latter only for the portion of frame in which the irregularity is shown.
Or:
The use of step-by-step inelastic response analysis.
Note: Although seismic codes provide simple indications for defining this irregularity and for decreasing
the reductive coefficient, my opinion is that the research has not, up to now, well clarified this problem
and that much more research work has to be done about it.

Single-story buildings
The inelastic response of an asymmetric, or non-balanced, three-dimensional scheme presents some
peculiarities, with respect to that of a plane frame. Indeed, many researchers pointed out that the inelastic
response of a non-balanced single-story three-dimensional scheme is less rotational than its elastic
response. Elastic analysis is therefore safe for some structural elements, unsafe for other. This seems to be
confirmed also in the case of schemes subjected to bi-directional ground motions, because of the un-
correlation of the orthogonal components of the motion. E.g. see Ghersi and Rossi [6], Ghersi and Rossi
[7].
Under this aspect, a building can never be rigorously considered regular.
The peculiarity of single-story asymmetric schemes imposes:
The use of two elastic analyses, performed with reference both to the nominal position of the mass
center and to a conventional, less eccentric, locations of it. A simple possibility for the latter is to
consider mass center coincident to stiffness center, i.e. to analyze a scheme subjected to pure
translation. Alternative, more reliable, formulations for this design eccentricity may be found in
references.
Or:
The use of step-by-step inelastic response analysis.

Multi-story buildings
The inelastic response of non-balanced multi-story buildings presents, in addition to the problems
discussed for plane frames, the peculiarities above mentioned for single-story three-dimensional schemes.
For this reason, a correct use of elastic analysis should consist, as proposed for single-story schemes, in
evaluating the internal actions as the envelope of the results of two elastic analyses, performed with
reference both to the nominal position of the mass center and to a conventional, less eccentric, locations of
it.
A building is considered regular, under this aspect, when in addition to the regularity conditions of section
4.2:
The mass centers and the stiffness centers are located along two vertical lines, the distance of which is
es.
The radius of gyration of stiffness rk and radius of gyration of mass rm are constant along the building
height.
The lack of regularity imposes:
The use of step-by-step inelastic response analysis.
Note: The inelastic response of multi-story buildings has been object of research only in the last few years
and it cannot be considered fully investigated. In particular, scarcely known is the behaviour of buildings
that are not regularly asymmetric.

CONCLUSIONS

This paper tries to provide guidelines on the subject structural regularity, which may be used both from
researchers and practical engineers. The basis of the approach is to start from single problems, pointing
out the conditions necessary to simplify their treatment, instead of starting from general definitions of
regularity, which might be confusing because related to many different problems. Another basic aspect of
the approach is the attempt of providing, whenever possible, indications on how to perform a posteriori
a numerical check of the regularity conditions, starting from the results of the analysis performed.
The paper is specifically devoted to buildings. The possibility of extending its considerations to other civil
engineering works has to be carefully discussed.

REFERENCES

1. Anastassiadis, K., Athanatopoulos, A. and Makarios, T., 1998, Equivalent static eccentricities in the
simplified methods of seismic analysis of buildings, Earthquake Spectra, 14, 1-34.
2. Calderoni, B., DAveni, A., Ghersi, A. and Rinaldi, Z., 2002, Static versus modal analysis of
asymmetric buildings: effectiveness of dynamic eccentricity formulations, Earthquake Spectra.
3. Hejal, H. and Chopra, W.K., 1987, Earthquake response of torsionally-coupled buildings, Report
No. UBC/EERC-87/20, University of California at Berkeley, Berkeley.
4. Moghadam, A.S. and Tso, W.K., 2000, Extension of Eurocode 8 torsional provisions to multi-story
buildings, Journal of Earthquake Engineering. 4, 1, 25-41.
5. Bosco, M., Ghersi, A., Marino, E. and Rossi, P.P., 2002, Effects of in-elevation irregularity on the
elastic seismic response of in-plan asymmetric buildings, Third European Workshop on the Seismic
Behaviour of Irregular and Complex Structures, Florence.
6. Ghersi, A. and Rossi, P.P., 2000, Formulation of design eccentricity to reduce ductility demand in
asymmetric buildings, Engineering Structures, Elsevier Science Ltd, vol. 22, pp. 857-871.
7. Ghersi, A. and Rossi, P.P., 2001, Influence of bi-directional ground motions on the inelastic
response of one-story in-plan irregular systems, Engineering Structures, Elsevier Science Ltd, vol.
23, pp. 579-591.