Sei sulla pagina 1di 10

ACI STRUCTURAL JOURNAL TECHNICAL PAPER

Title no. 107-S38

Nonlinear Analysis of Cross Sections under Axial Load and


Biaxial Bending
by Marco Di Ludovico, Gian Piero Lignola, Andrea Prota, and Edoardo Cosenza

The paper presents a method and an integration procedure The proposed computation method specifically focuses on
specifically targeting the analysis of cross sections under biaxial the analysis of regular or irregular shaped cross sections (for
bending and axial load, whatever the shape—including internal example, made of concrete) internally or externally reinforced
voids—and made of different materials (for example, concrete) by mild steel (reinforced concrete [RC]) and prestressing
internally or externally reinforced by mild steel, prestressing tendons (prestressed concrete [PC]) as well as by fiber-reinforced
tendons, and fiber-reinforced polymers. Surface integration of
meshed cross sections allows moment-curvature relationships to
polymers (FRP). The developed algorithm provides the
be computed using nonlinear constitutive relationships (also with moment-curvature diagrams for cross sections subjected to
softening) and three-dimensional (3D) interaction domains P-Mx-My axial load and biaxial bending using linear and nonlinear
to be drawn. Iteration procedures and convergence criteria are stress-strain relationships, also with softening. Moreover, it
herein proposed to rapidly solve the highly nonlinear problem. allows cross sections to be designed and/or checked by
Numerical tests showed fast convergence of the algorithm and determining the ultimate flexural moments Mx-My given a
good agreement with experimental results elsewhere; comparisons constant axial load P and orientation angle β of the resultant
between the numerical outcomes and the simplified expressions moment; the Mx-My domain for a given axial load and,
available in the literature have been performed to point out the consequently, the 3D failure domain can be plotted.
current limits of the available codes’ formulas. A review of the literature’s methods and a description of
the proposed numerical method, along with model validation
Keywords: biaxial bending; cross section; nonlinear; moment-curvature; by comparison with results found elsewhere, are outlined in
three-dimensional interaction domain. the following sections.

INTRODUCTION RESEARCH SIGNIFICANCE


The study of axially loaded members under biaxial This study deals with reliable and rapid nonlinear analyses
bending is of particular interest in the case of structures in terms of strength and deformability of cross sections of
subjected to earthquake motion. In such cases, the presence regular or irregular shape and made of different materials. The
of biaxial bending moments reduces the cross section provided algorithm is intended as a tool, adopting robust
capacity both in terms of strength and ductility, usually procedures, well suited to study RC cross sections under axial
evaluated only under uniaxial loads. The current seismic load and biaxial bending (for example, seismic actions). In
codes, however, allow cross sections to be designed under particular, the procedure may be used to assess the main
biaxial bending by analyzing such sections as subjected to characteristics of the nonlinear behavior of cross sections,
two uniaxial bendings and axial loads acting separately. namely, cracking, yielding point, flexural strength, ultimate
In particular, FEMA 356 1 and Italian seismic guideline curvature, moment-curvature relationship also with softening,
D.M. 14/01/20082 provide simplified equations to curvature ductility, and 3D interaction domains P-Mx-My).
approximate the effective three-dimensional (3D) failure The results provided by the developed procedure have
surface, P-Mx-My; Eurocode 83 suggests treating biaxial been compared to experimental results on symmetric and
bending by carrying out the checks separately in each direction, irregular cross sections available in the literature. Comparisons
with the uniaxial flexural strength reduced by 30%. between the numerical results and the simplified expressions
These are some of the several approximate methods that available in the literature indicate the current limits of the
have been proposed in recent decades. The use of simplified available codes’ formulas.
approaches is unsafe in some cases and cannot give any
information about fundamental properties (that is, cracking REVIEW OF LITERATURES’ METHODS
and yielding point and curvature ductility) of the cross The design and/or check procedures to analyze cross
section. They are generally justified by the difficulty in sections under biaxial load are iterative and require the integration
manually computing the cross section capacity under biaxial of the stress field over the cross section at each step. To avoid
bending and axial load. Such computation requires the such a time-consuming procedure, several simplified methods
integration of stresses associated with strain-based failure were previously developed to provide an approximate
criteria that can be numerically performed by discretizing the analytical expression for the failure surface. One of these,
cross section and iteratively locating the neutral axis depth. the load contour (LC) method, was provided by Bresler,4
Thus, an appropriate numerical integration procedure that
can be implemented in a computer analysis program, ACI Structural Journal, V. 107, No. 4, July-August 2010.
MS No. S-2008-222.R5 received September 25, 2009, and reviewed under Institute
providing a fast and reliable tool to perform fiber analysis of publication policies. Copyright © 2010, American Concrete Institute. All rights reserved,
cross sections under axial load and biaxial bending, has been including the making of copies unless permission is obtained from the copyright proprietors.
Pertinent discussion including author’s closure, if any, will be published in the May-June
proposed in this study. 2011 ACI Structural Journal if the discussion is received by January 1, 2011.

390 ACI Structural Journal/July-August 2010


M x ⎞ ⎛ M y ⎞ ⎛ 1 – β⎞
⎛ -------- M M
Marco Di Ludovico is an Assistant Professor of structural engineering at the University - + --------- ------------ – 1 = 0 if ⎛ --------x-⎞ > ⎛ --------y-⎞ (4a)
of Naples Federico II, Naples, Italy, where he received his PhD in seismic risk. His ⎝ M x0⎠ ⎝ M y0⎠ ⎝ β ⎠ ⎝ M x0⎠ ⎝ M y0⎠
research interests include nonlinear behavior of reinforced concrete structures, seismic
strengthening of reinforced concrete structures using advanced materials, strengthening
of prestressed concrete bridge members using composites and analysis of reinforced
M x ⎞ ⎛ 1 – β⎞ ⎛ M y ⎞
⎛ -------- M M
- ------------ + --------- – 1 = 0 if ⎛ --------x-⎞ ≤ ⎛ --------y-⎞
concrete members subjected to axial load, and biaxial bending.
(4b)
⎝ M x0⎠ ⎝ β ⎠ ⎝ M y0⎠ ⎝ M x0⎠ ⎝ M y0⎠
Gian Piero Lignola is an Assistant Professor of structural engineering at the
University of Naples Federico II, where he received his PhD in seismic risk. His
research interests include numerical modeling and strengthening of concrete and
masonry structures with advanced materials. where β is a coefficient ranging between 0.5 and 1 depending on
the mechanical and geometrical properties of the cross section.
Andrea Prota is an Assistant Professor of structural engineering at the University of By using a value equal to 1 for α1 and α2, the LC method
Naples Federico II. He received his MS in civil engineering from the University of
Missouri-Rolla, Rolla, MO, and his PhD in structural engineering from the University provides a pyramidal failure surface that is very conservative in
of Naples Federico II. His research interests include the seismic behavior of reinforced most cases. Ghersi and Muratore8 provided simplified and
concrete and masonry structures, the use of advanced materials for new construction,
and the retrofit of existing structures using innovative techniques.
conservative equations to determine uniaxial bending strength
for rectangular cross sections and, in the case of axial load and
Edoardo Cosenza is a Full Professor of structural engineering at the University of biaxial bending, they suggested using the LC method with α1 =
Naples Federico II. His research interests include seismic engineering, steel-concrete
composite structures, and composite materials for construction.
α2 = 1.4 to 1.5, where Mx0 and My0 are computed by using the
simple expressions provided by the authors. Moreover,
Eurocode 29 and Monti and Alessandri10 provide simple
expressions to calibrate α1 = α2 to be used in the LC method as
a function of the normalized axial load v = P/Po to approximate
who proposed to approximate the failure surface by using the the effective moments under biaxial bending.
following expression A different approximate method—the reciprocal load
(RL) method—was also proposed by Bresler.4 In this
α1 α2 method, reported also in the commentary of ACI 318-05,11
Mx ⎞
⎛ -------- M
- + ⎛ --------y-⎞ =1 (1) the failure surface is interpolated by secant planes defined by
⎝ M x0⎠ ⎝ M y0⎠
three points on the failure surface for uniaxial bending

where Mx and My are the effective moments under biaxial 1 1 1 1


bending; Mx0 and My0 are the nominal uniaxial bending --- = ----- + ----- – ----- (5)
P Px Py Po
moments about the x- and y-axis, respectively, under the applied
load P; and parameters α1 and α2 depend on the mechanical and
geometrical properties of the cross section and on the applied where Px and Py are the axial strength under uniaxial
axial load. This expression provides two-dimensional (2D) moments about the x- and y-axis, respectively.
slices of the 3D failure surface in the P-Mx-My space. The The shape of the approximate interaction surface obtained
reliability of this approximate failure surface (Eq. (1)) is by Eq. (5) was confirmed by test results conducted on square
strongly related to the definition of exponents α1 and α2; columns subjected to combined axial load and biaxial
Bresler4 suggested using the same value for both exponents, for bending with constant eccentricity by Ramamurthy.12 On
rectangular or square columns, ranging between 1.15 and 1.55. the other hand, Silva et al.13 showed that such a procedure
An expression for α was also proposed in the 1988 Australian provides a too conservative approximation in most cases and
Standard AS 36005 to better approximate the failure surface is not convenient from a computational point of view
because for every loading case, it requires calculating Px and
Py for given eccentricities ey and ex. Moreover, Wang and
1.7P Hong14 noted that, in the case of RC columns with high-
α = 0.7 + ------------- with 1 ≤ α ≤ 2 (2)
0.6P o strength concrete, the method can lead to unsatisfactory results.
In addition to the LC and RL methods, alternative failure
where P and Po represent the applied axial load under biaxial criteria have also been widely proposed: Sussekind15 used a
eccentricity and the axial compression strength of the cross numerical method to determine a few points on the failure surface
section, respectively. for prismatic RC elements under combined axial load and biaxial
Moreover, Hsu6 proposed to adopt the LC method by bending; Hulse and Mosley16 and Spiegel and Limbrunner17
using a value of 1.5 for both α1 and α2 and to add another provided numerical techniques to compute the capacity of an RC
term to Eq. (1) to take the ratio between the nominal axial member subjected to axial load and biaxial bending; a new
load and the balanced strain axial load directly into account failure criterion was proposed by Silva et al.13 in which it was
assumed that the failure surface can be approximated by a
P – P b ⎞ ⎛ M x ⎞ 1.5 ⎛ M y ⎞ 1.5 given closed convex surface; and Bonet et al.18 provided an
⎛ -----------------
- + --------- + --------- =1 (3) analytical equation based on the definition of the failure surface
⎝ P o – P b⎠ ⎝ M bx⎠ ⎝ M by⎠
by means of two directrix curves and generatrix curves.
Many researchers have proposed design equations and
where Pb is the axial load strength at the balanced strain calculation algorithms to account for the materials’
condition under biaxial bending and Mbx and Mby are the nonlinear behavior. Kawakami et al.19 presented a complete
uniaxial bending moments for balanced strain condition algorithm for the analysis of RC members, including the
about the x- and y-axis, respectively, under the applied axial effect of both mild steel and prestressed reinforcement; but
load P. A variant of the LC method was implemented by their procedure is difficult to implement in a computer
Parme et al.,7 assuming only α1 = α2 = 1 and adding some program. An incremental model for the force-deformation
multiplication factors, as shown in the following expressions behavior of RC cross sections and members under 3D

ACI Structural Journal/July-August 2010 391


under axial load and biaxial bending. For the sake of
simplicity, in the following sections, the procedure will be
illustrated in the case of reinforced concrete cross sections.

Procedure assumptions
The analysis invokes the following assumptions: 1) plane
sections remain plane (classical Bernoulli-Navier hypothesis);
Fig. 1—Constitutive laws implemented in algorithm. 2) the longitudinal stress in a fiber element is dependent only
on the longitudinal strain; 3) perfect bond is assumed at the
interfaces between concrete and steel and/or FRP reinforcing
bars or laminates; and 4) loading and deformation of the
cross section are assumed to take place monotonically.
Cross-section behavior under cyclic loading conditions is
not directly considered, but it could be addressed by
implementing modified material stress-strain relationships.
Linear and nonlinear, also with softening, stress-strain
relationships (constitutive laws) can be considered to
perform the analysis. To date, the three constitutive laws
Fig. 2—(a) Variable axial load with constant eccentricity;
reported in Fig. 1 have been implemented and tested (consti-
and (b) constant axial load with variable eccentricity. tutive law No. 3 is based on the model by Mander et al.26).

Procedure strategy
Two analysis cases, briefly shown in Fig. 2, were considered
to explain the procedure strategy. In the first case (Fig. 2(a)),
an axial load is applied on the cross section with a given
eccentricity; this axial load increases up to failure with
constant eccentricity. In this case, the load path is on a plane
parallel to the P axis and the My /Mx ratio is constant. In the
second case (Fig. 2(b)), the axial load value Pi is constant
and combined with increasing bending moments (for
example, a pier with constant axial load and seismic-type
lateral shear force on top); the failure occurs when the load
path, in a constant Pi plane, reaches the limit surface
(maximum bending moments). In this case, the procedure
Fig. 3—Compression and biaxial bending in cross section. allows the limit values of the eccentricities to be determined.
The procedure described herein refers to the latter case,
whereas the former case can also be solved after plotting the
generalized load or deformation histories, including cyclic entire 3D interaction domain.
loading up to ultimate deformation, was performed by The given axial load Pext remains constant during the load
Bousias et al.20 A method based on the fiber model concept process; and the direction of the horizontal load, or the
and computer graphics was proposed by Sfakianakis21 to application point, remains on the line defined by the centroid
study RC cross sections with regular or irregular shapes; this (or any other reference point) of the gross concrete cross
method requires time-consuming calculations but does not section and a given orientation angle βext (refer to Fig. 3).
include any iteration and, hence, overcomes convergence The value of βext is the orientation angle of the resultant
problems. A new methodology, implemented in a multi-module moment with respect to the horizontal axis x. Flexural
program, was performed by Consolazio et al.22 for the moments computed by the algorithm, including those used to
calculation of moment-curvature diagrams in the case of plot moment-curvature diagrams, are computed with
concrete cross sections of regular or irregular shape under respect to the concrete gross section centroid (or any other
axial load and biaxial bending while De Vivo and Rosati23 reference point).
proposed two algorithms by adopting a secant strategy for
the solution of nonlinear equilibrium equations. Cedolin et al.24 Cross section discretization
proposed a method based on isoparametric discretization of Any regular or irregular concrete cross section should be
the cross section along with a gauss-legendre integration defined by a closed polygonal boundary with straight
scheme to also account for curved boundary. Finally, segments (refer to Fig. 4); linear segments represent the
Furlong et al.25 surveyed the various methods, both equation exterior and interior (if any hole is included) boundaries of
and computer based, proposed in the literature. the cross section. If the cross section has curved boundaries,
it can be approximated using a set of straight segments. The
PROPOSED NUMERICAL PROCEDURE cross section is then meshed into small rectangles, and the
A numerical procedure is presented based on cross section relevant material is assigned to each discrete element.
discretization into rectangular elements. The analytical This procedure calls for particular attention to the order of
procedure can account for regular or irregular cross section the vertices: the vertices of the exterior boundary should be
geometry and material properties, and it can provide numbered counterclockwise, whereas those of the interior
moment-curvature behavior at cross section level to simulate boundaries, if any, should be numbered clockwise (refer to
the load-deformation relationship of structural members Fig. 4).

392 ACI Structural Journal/July-August 2010


Fig. 5—Grid spacing and cross section discretization.

Fig. 4—Vertex numeration: exterior boundary counter


clockwise, interior boundaries clockwise.

The smallest rectangle area enveloping the full cross


section is identified considering maximum and minimum
coordinates of vertices: xmin, xmax and ymin, ymax (refer to
Fig. 5). To divide the cross section into nd × md fiber
elements, a grid with horizontal spacing of xel = (xmax – xmin)/
nd and vertical spacing of yel = (ymax – ymin)/md is defined.
If the centroid of an element (defined by its two coordinates,
abscissa and ordinate) lies inside the boundary of the cross
section and is made by material k, then the element material
k multiplier θk is set to 1. This means it belongs to the cross Fig. 6—Geometry and mechanical properties matrix.
section, and it is active in the calculation; otherwise, θk is set
to 0 (inactive). For this purpose, any centroid is considered values are evaluated in each segment centroid (lumped area).
inside the polygonal closed envelope (defining the cross It is suggested to be consistent with mesh divisions and
section also with some inside voids) if the sum of the angles element size adopted for concrete cross sections.
defined by the centroid and two consecutive vertices of the
envelope is equal to 360 degrees. Iterative procedure description
Once the concrete cross section is discretized into small areas, For a given neutral axis depth and inclination, one step of
each small area is assumed to have a constant strain and stress and the iterative procedure is illustrated as follows. Once the
the integration of the stress field can be performed by adopting the coordinates and material multipliers of each concrete fiber or
midpoint sampling rule, assuming that the integrand function is reinforcement element are known, the perpendicular
linear over the rectangular areas. For the sake of simplicity, this distance of each of them from a given neutral axis can be
simple integration scheme was selected rather than Newton- determined. The neutral axis is characterized by two parameters:
Cotes formulas, Simpson’s rule, or Gauss quadrature, which a point on the plane of the cross section and an angle. It is
could also be more accurate with a smaller mesh division. important to define a direction on the neutral axis because
Similarly to finite element modeling, the accuracy of integration the perpendicular oriented distance of the elements is used to
schemes depends on the resolution of the mesh; therefore, a evaluate the strain in the elements.
general rule valid for any case cannot be provided. For instance, The points at the same perpendicular distance d from the
a mesh division of nd × md = 14 × 14 has been found to be neutral axis have the same strain equal to a given curvature
appropriate in the case of the L-shaped cross section analyzed in in the direction normal to that neutral axis multiplied by d. In
the following sections (convergence with a percent difference the sign convention, the fibers on the same semi-plane of the
smaller than 1%, which is satisfactory in engineering practice). externally applied axial load resultant should have the same
The diameters of the internal reinforcements (that is, steel strain sign as the external load (that is, if compressions are
or FRP bars) are assumed to be negligible when compared to assumed to be positive, then the compression axial load and
the cross section dimensions (meshing procedures such as compressive strains will be positive). Once the strain field is
those used for the concrete cross section are unnecessary). known, the stresses in each fiber can be determined using the
The strain level in the reinforcement is equal to that appropriate constitutive law of the materials concerned.
computed at its centerline and is assumed to be equal to the Particular care must be taken in prestressing elements; the
concrete strain at the same distance from the neutral axis value of strain to account for is the sum of the effective prestrain
(that is, perfect bond). To avoid double counting the area in the tendon (prior to bending loading and after losses have
occupied by reinforcement, particularly significant in cross been accounted for) and the increase in strain caused by
sections with a large quantity of reinforcement, however, the cross section curvature induced by bending loads.
stress in the concrete is subtracted from the stress in the Abscissa and ordinate of the centroid of each element,
reinforcement, both evaluated at the same position and for material multiplier, perpendicular distance from neutral axis,
the same strain. Therefore, the stress in each material is strain and stresses can be stored in a 3D matrix (refer to Fig. 6).
computed based on the relevant constitutive law. The width Each element of the bidimensional nd × md matrix (layer)
of the external reinforcements (that is, FRP laminates) is represents a fiber of the concrete cross section, whereas each
assumed to be negligible when compared to the cross section layer (third coordinate of the matrix) contains specific
dimensions so that their centerlines are considered around information about that fiber (for example, the strain and
the cross section contour and meshed as segments; the strain stress). This matrix can be adopted for spread-sheet

ACI Structural Journal/July-August 2010 393


n rein nd ⋅ md n rein

My = ∫ θ ⋅ σ x dA + ∑ σ j ⋅ A j ⋅ x j
grid
= ∑ θi ⋅ σi ⋅ Ai ⋅ xi + ∑ σj ⋅ Aj ⋅ xj (8)
j=1 i=1 j=1

nd ⋅ md n rein

= ∑ Ni ⋅ xi + ∑ σj ⋅ Aj ⋅ xj
i=1 j=1

and then transported to the centroid of the cross section (or


any other reference point) Mx,g and My,g. Finally the internal
orientation angle is computed as βint = –(My,g/Mx,g).

Convergence criteria
In the case of axial load and bending along the principal
axes of the section, only a nonlinear equation has to be
solved to account for equilibrium. The internal obtained
stress field, integrated over the cross section domain Pint
associated to a given neutral axis depth and curvature, must
be equal to (that is in equilibrium with) external acting force
Pext. This problem is generally nonlinear in the post-elastic
range and can be solved by changing the neutral axis depth
iteratively until equilibrium is found. Usually the convergence
criteria concerns axial load P: |Pint – Pext| ≤ tolerance. The
final result is found if the gap between external axial force
and the internal obtained axial load is less than a given
tolerance value (for example, a percentage of Pext).
In the case of axial load and biaxial bending, such a
procedure is not sufficient; a system of two nonlinear
equations has to be solved simultaneously to account for
equilibrium and external orientation angle βext. The
convergence criterion for the second coupled nonlinear
equation concerns the internal and external orientation
Fig. 7—Flowchart of double iteration procedure. angles: |βint – βext| ≤ tolerance. This problem is nonlinear
with respect to both neutral axis depth and orientation angle;
it involves the simultaneous convergence with two criteria.
It can be solved in a double-nested iterative process: inner—
calculations. Moreover, separate tables can be defined for for a given neutral axis inclination, the neutral axis depth is
each type of reinforcement (internal or external). changed until the internal obtained stress field is in equilibrium
At this stage, the internal axial force Pint can be computed, with Pext; outer—the neutral axis inclination is changed until
adopting the midpoint sampling rule, as the integral of the the internal bending moments has the same orientation
internal stress field over the cross section angle, βint as the external applied bending moment βext.
In other words, the internal obtained stress field, integrated
n rein nd ⋅ md n rein
over the cross section domain, must be in equilibrium with
Pext. for each value of the neutral axis inclination angle. At
P int = ∫ θ ⋅ σ dA + ∑ σj ⋅ Aj = ∑ θi ⋅ σi ⋅ Ai + ∑ σj ⋅ Aj (6)
this stage, the inner iteration process gives a value βint,
grid j=1 i=1 j=1
whereas the outer iteration involves the neutral axis
nd ⋅ md n rein inclination: it is changed until βint approximates βext. A
= ∑ Ni + ∑ σj ⋅ Aj flowchart of the proposed procedure is depicted in Fig. 7.
i=1 j=1 Due to the high nonlinearity of the problem, an optimal
strategy has to be chosen to have a fast convergence and
A constant stress σi is considered inside an active (θi ≠ 0) rapidly solve the nonlinear problem.
fiber, whether concrete or reinforcement, equal to the value The convergence problem must be brought into the standard
at the centroid of the element; Ni is the axial force resultant form f(x) = 0, (that is, Pext – Pint(x) = 0 where x is the neutral
in a meshed element, Ni = θi Aiσi. The integral is approximated axis depth). A method for solving f(x) = 0 has at least two
by a sum over the entire active concrete elements and all the parts: an iteration and a convergence test. The iteration
nrein reinforcement elements, thereby accounting for the generates trial values of the solution and the test indicates
contributions of each material. when to stop iterating.
Similarly, the internal bending moments around origin To solve the axial equilibrium, a linear interpolation
system axes x and y are evaluated method can be adopted.27 This method simply relies on the
fact that, between any pair of points (a,b) that give opposite
n rein nd ⋅ md n rein signs to a continuous function f(x), there must be a zero of
Mx = ∫ θ ⋅ σ y dA + ∑ σ j ⋅ A j ⋅ y j = ∑ θi ⋅ σi ⋅ Ai ⋅ yi + ∑ σj ⋅ Aj ⋅ yj (7) f(x). A closer estimate of the zero is taken by linear interpolating
grid j=1 i=1 j=1 the values of the function in the two bounds of the interval
nd ⋅ md n rein (a,b), adopting a straight line or chord, connecting f(a) and
= f(b) with a slope φ = (f(b) – f(a))/(b – a) and crossing the x-axis
∑ Ni ⋅ yi + ∑ σj ⋅ Aj ⋅ yj
i=1 j=1 in the new estimate of the root xnew = a – (f(a))/φ; this value

394 ACI Structural Journal/July-August 2010


Fig. 8—Limit strain line to evaluate first two trials in linear
interpolation method.

replaces bound a or b if f(xnew) has the same sign as f(a) or


f(b), respectively, and allows iteration with a new slope φnew. Fig. 9—Secant method for axial equilibrium.
The first two bounds of the interval a,b can be estimated
according to two particular positions of the neutral axis Moment-curvature diagram
corresponding to two limit lines tangent to the cross section To evaluate the moment-curvature relationship, the strain
(refer to Fig. 8); the entire cross section is not compressed due in the most compressed element of concrete mesh can be
to the first limit line (Fig. 8(a)), while the entire cross section increased from an initial strain ε0 to an ultimate value εu; ε0
is compressed due to the second limit line (Fig. 8(b)). As the is evaluated under pure compression, and then a fixed-length
proposed method is a bracketing method, the neutral axis increment ε of strain is adopted until ultimate strain in
position is supposed to be in between two bounds where the concrete is attained. If the axial load level is high, then ε0
function changes sign. If f(a) and f(b) have the same sign, the would be high and close to the ultimate strain so that a small
method can be adopted to find a new bound (a, xnew) or (xnew, b), bending capacity and curvature can be expected.
where the function changes its sign and the chord slope is This procedure can account for the debonding of FRP
equal to φ = (f(a) – f(xnew))/(a – xnew) or φ = (f(xnew) – f(b))/ reinforcement or the rupture of prestressed tendons or bars,
(xnew – b). At this stage, to have a faster convergence, the automatically if the constitutive laws are considered as a step
secant method was adopted (refer to Fig. 9). This is a variation function (stress goes to zero when ultimate or debonding strain
of Newton’s method that avoids computing φ at each step and is attained). This leads to automatic exclusion of the contribution
does not break down if φ ≈ 0 in a trial, because it uses the of failed material in the cross section stress integration.
constant slope φ. The curvature is evaluated as the ratio between the strain
To solve the orientation angle problem (|βint – βext| ≤ in a point and the distance of that point from the actual
tolerance), it is instead always possible to adopt bracketing neutral axis. It is also possible to evaluate the curvature
methods directly because the solution domain is smaller than component along preferred reference axes.
180 degrees, in the sense that the neutral axis assumes the Note that the cracking moment can be evaluated by
same position (with opposite direction) when it is rotated by imposing the strain corresponding to tensile strength in the
an angle of 180 degrees. Hence, the solution can be found in a tensile concrete fiber furthermost from the neutral axis.
limited domain. Once it is ensured that the interval contains a Similarly, the yielding moment can be computed by
root, several procedures exist that are guaranteed to converge. imposing the strain corresponding to steel yielding in the
In this section, a strategy to speed up convergence is shown. tensile reinforcement element furthest from the neutral axis.
The first trial value of the neutral axis inclination angle can As an example, the moment-curvature diagrams obtained
be that perpendicular to the orientation angle βext (refer to by considering two different constitutive concrete laws with
Fig. 3). The solution is then supposed to be close to this value reference to an asymmetric L-shaped RC cross section (refer
and a method for locating closer estimates of the root is to to Fig. 10(a)) reinforced by four 0.47 in. (12 mm) diameter
scan the entire domain by small increments and to observe bars (characterized by an elastic-plastic stress-strain curve
the steps in which a change of sign in the function occurs. with yielding strength equal to fy = 40,320 psi [278 MPa])
This signals that the function crosses the root within the and three plies of uniaxial carbon laminates (CFRP) placed
particular step. At this stage, the bisection method can be on the top width (thickness of laminates tf = 0.006 in./ply
adopted; it divides the interval of search by two and always [0.16 mm/ply], Young’s modulus of Ef = 56,565 ksi [390 GPa],
retains that half of the search interval in which the change of ultimate [debonding] strain of εfu = 6.1‰) are reported in
sign has occurred, using midpoint to replace whichever Fig. 10(b). The diagram is computed by considering the
interval limit has the same sign. After each iteration, the previously defined curvature, by assuming an axial load
bounds containing a root decrease by a factor of two. Thus, equal to P = 95.7 kips (426 kN) corresponding to v = P/Po =
the maximum number of iterations n = log2 ψo/τ required to P/(Acon fc′ + As fy) = 0.3 (where Acon is the gross cross section
achieve a given tolerance τ in the solution is known in area, fc′ is the concrete compressive strength, and As is the
advance, where ψo is the size of the initially bracketing total reinforcement area) and an orientation angle βext =
interval (that is, ψo is not greater than 180 degrees). 0 degrees. In particular, Curve A of Fig. 10(b) refers to a
The number of iterations is strictly related to cross section parabolic-rectangular concrete stress-strain relationship
type and materials constitutive laws. It is emphasized, (fc′ = 1926 psi [13.28 MPa] and ultimate axial strain equal to
however, that in the case of asymmetric cross sections, the 3.5‰), whereas Curve B refers to a concrete stress-strain
average number of iterations is usually smaller than 30. relationship characterized by softening (secant modulus of

ACI Structural Journal/July-August 2010 395


same procedure has to be repeated for different values of
external axial load values to plot the 3D interaction domain.
With reference to the cross section reported in Fig. 10(a),
the 2D Mx-My domains for different axial load ratios (ν = P/ P0)
are reported in Fig. 11(a). Finally, Fig. 11(b) shows the
relationship between the axial load ratio versus the ultimate
curvature, χu, in the case of load orientation angle βext = 0
degrees. From this diagram, the ultimate capacity in terms of
curvature can be calculated straightforwardly for each
assigned axial load in the case of biaxial bending.

COMPARISON BETWEEN NUMERICAL AND


Fig. 10—(a) L-shaped cross section; and (b) moment-curvature EXPERIMENTAL RESULTS
diagrams. (Note: Dimensions in mm, 1 mm = 0.0394 in.; The results obtained by using the implemented calculation
1 kNm = 0.74 kip-ft; 1 m = 3.28 ft.) algorithm have been compared with experimental results
obtained on square RC cross sections,6,12 rectangular RC
cross sections,4,12 and a typical prestressed cross section.28
Tables 1 through 3 summarize for square, rectangular, and a
typical prestressed cross section, respectively, both input
data in terms of geometrical cross section dimensions (web
width b and depth h and flange width B and depth s)
geometrical reinforcement ratio ρg; steel yielding strength
fy; concrete compressive strength fc′ ; eccentricities ex and ey;
axial load Pext; and experimental and theoretical results in
terms of ultimate moments about the x-axis, Mx, and the y-axis,
My. In the theoretical calculations, constitutive laws No. 2 and
No. 1 of Fig. 1 were adopted for concrete and steel,
Fig. 11—(a) Two-dimensional Mx-My domains related to cross respectively. In particular, εI = 0.002, εII = 0.0035, and EI =
section of Fig. 10(a); and (b) normalized axial load versus ultimate 0 were assumed in constitutive law No. 2 and EI = 210,000 MPa
curvature. (Note: 1 kNm = 0.74 kip-ft; 1 m = 3.28 ft.) (30,460 ksi), εI = fy /EI, EII = EIII = 0, εII = εIII = 6.75% in
constitutive law No. 1. Linear elastic constitutive model was
adopted for FRP.
the softening branch Ec2 = –1284 ksi [8853 MPa] and ultimate In the last column of Tables 1 through 3, the ratio between
strain εcu = 3.5‰). experimental and theoretical results is also reported.
Although the comparison can be affected by considerable
Interaction diagram plot uncertainty related to the effective material properties and
It is possible to avoid convergence to a given orientation test setup, in every case a good correlation was found. In
angle because ultimate flexural moment is evaluated for particular, for the square members, Table 1 shows that the
different load orientation angles; only axial equilibrium is average percentage theoretical overestimation and under-
needed, assuming that flexural capacity is attained corresponding estimation for each experimental set of tests are smaller than
to the ultimate curvature. The ultimate curvature computation 6%. The average moments ratio MEXP/MTHEOR, computed
involves the following steps: 1) given a neutral axis depth and on all tests analyzed, was the same about the x- and y-axis
orientation angle, the element furthest from the neutral axis is (that is, 1.01), with a standard deviation (SD) of 0.09.
determined for each material and the corresponding ultimate The same good correlation was found with reference to the
value of strain is considered; 2) the ultimate curvature for each rectangular cross sections where the average percentage
material is evaluated dividing ultimate strain by the perpendicular theoretical overestimation and underestimation for each
distance of the corresponding material from the neutral axis; experimental set of tests are smaller than 4%. Also in this
and 3) the ultimate curvature value is taken as equal to the case, the total average ratio between MEXP/MTHEOR was
smallest curvature previously evaluated, such that one material very close to 1 (1.03 for moments about the x-axis, Mx and
(or more at the same time) fails. 0.99 for moments about the y-axis, Mx) with SD ranging
In the case of softening constitutive laws, the problem is between 0.07 and 0.09 (refer to Table 2).
highly nonlinear and the value of curvature corresponding to The correlation with reference to typical prestressed cross
maximum flexural capacity cannot be known a priori sections is presented in Table 3. The original cross section is
because it is smaller than the ultimate curvature. This means made of two different concretes (slab and girder, respectively)
that maximum flexural moment is not achieved when ultimate and two different steels (mild reinforcement and prestressing
strain is attained in one of the materials involved. To overcome strands, respectively) as is common in bridges. Three different
this pitfall, the maximum moment is computed as the configurations were analyzed: S-1 is without external FRP
maximum moment value in the moment-curvature diagram. reinforcement; S-2 is slightly damaged (strands No. 1 and No. 2
Once the axial load Pext is defined, because the axial lost) and reinforced by using two plies of CFRP laminates
equilibrium procedure is repeated for different values of (tf = 0.0129 in. [0.330 mm]); and S-3 is heavily damaged
neutral axis inclination angles instead of the flexural moment (strands No. 1 through 4 lost) and reinforced by using three
resultant orientation angle, the second convergence iterative plies of CFRP laminates (tf = 0.0195 in. [0.495 mm]). Details
procedure is skipped. This procedure allows a slice of the 3D about the experimental program can be found in Di
interaction domain to be designed for a given axial load. The Ludovico et al.28 The comparison between theoretical

396 ACI Structural Journal/July-August 2010


Table 1—Experimental theoretical comparison on square cross sections
MX, kNm MY, kNm Δ = MEXP./MTHEOR. Average SD
ex, ey,
Reference Specimen b, mm ρg, % fy, MPa fc′,MPa mm mm Pext., kN EXP. THEOR EXP. THEOR MX MY MX MY MX MY
U1 102 2.81 503.34 26.93 64 89 42.70 2.71 2.60 3.80 3.64 1.04 1.04
U2 102 2.81 503.34 26.24 76 89 38.70 2.95 2.88 3.44 3.37 1.02 1.02
U3 102 2.81 503.34 26.85 89 89 35.59 3.16 3.15 3.16 3.15 1.00 1.00
1.06 1.05 0.12 0.12
U4 102 2.81 503.34 26.41 51 51 63.61 3.23 3.24 3.23 3.24 1.00 1.00
U5 102 2.81 503.34 25.62 13 140 48.04 0.61 0.47 6.71 5.20 1.30 1.29
Hsu6 U6 102 2.81 503.34 26.86 13 178 27.76 0.35 0.36 4.94 5.07 0.97 0.97
H1 108 4.87 306.83 24.44 76 51 61.83 4.71 4.82 3.14 3.21 0.98 0.98
H2 108 4.87 306.83 26.78 93 57 52.49 4.33 4.93 3.00 3.42 0.88 0.88 0.94 0.94 0.05 0.05
H3 108 4.87 306.83 29.15 64 76 60.50 3.84 3.99 4.61 4.78 0.96 0.96
S1 102 2.75 306.83 22.06 25 38 93.41 2.37 2.55 3.56 3.78 0.93 0.94
0.95 0.95 0.03 0.02
S2 102 2.75 306.83 28.24 25 38 110.32 2.80 2.89 4.20 4.34 0.97 0.97
B1 203 3.88 322.62 29.17 21 79 628.97 13.23 13.15 49.38 48.73 1.01 1.01
B2 203 3.88 322.62 25.75 19 47 771.77 14.98 14.89 36.23 36.15 1.01 1.00
B3 203 3.88 322.62 33.51 51 88 533.79 27.12 25.56 46.97 44.33 1.06 1.06
B4 203 3.88 322.62 31.96 64 110 395.89 25.14 24.11 43.54 41.87 1.04 1.04
Ramanurthy12 1.01 1.01 0.08 0.08
B5 203 3.88 322.62 19.34 36 36 598.29 21.49 25.68 21.49 25.54 .084 0.84
B6 203 3.88 322.62 27.55 65 65 500.42 32.36 32.65 32.36 32.65 0.99 0.99
B7 203 3.88 322.62 29.48 72 72 515.99 37.07 33.92 37.07 33.92 1.09 1.09
B8 203 3.88 322.62 34.13 102 102 369.74 37.57 34.81 37.57 34.81 1.08 1.08
Note: 1 mm = 0.0394 in.; 1 MPa = 145.04 psi; 1 kN = 0.225 kip; 1 kNm = 0.74 kip-ft. Total average 1.03 0.99
Total SD 0.09 0.07

Table 2—Experimental theoretical comparison on rectangular cross sections


Δ = MEXP./
MX, kNm MY, kNm MTHEOR. Average SD
b × h,
Reference Specimen mm x mm ρg, % fy, MPa fc′,MPa ex, mm ey,mm Pext., kN EXP. THEOR EXP. THEOR MX MY MX MY MX MY
B5 152 x 203 2.6 368.88 22.06 76 102 142.3 14.37 14.47 10.81 10.85 0.99 1.00
B6 152 x 203 2.6 368.88 25.51 152 203 75.62 15.37 14.98 11.52 11.30 1.03 1.02
Bresler4 1.02 1.02 0.02 0.04
B7 152 x 203 2.6 368.88 24.13 152 102 93.41 9.49 9.21 14.23 13.78 1.03 1.03
B8 152 x 203 2.6 368.88 24.82 76 203 106.8 21.69 21.26 8.13 7.92 1.02 1.03
D1 152 x 229 4.6 322.62 31.65 25 38 785.1 19.92 20.45 29.04 29.86 0.97 0.97
D2 152 x 229 4.6 322.62 25.44 56 84 400.3 22.31 20.35 31.78 33.84 1.10 0.94
D3 152 x 229 4.6 322.62 24.45 76 114 311.4 23.66 21.07 32.09 35.53 1.12 0.90
Ramamurthy12 1.04 0.96 0.11 0.09
D4 152 x 229 4.6 322.62 25.13 32 32 680.6 22 23.13 23.08 22.00 0.95 1.05
D5 152 x 229 4.6 322.62 30.90 81 81 378.1 30.56 25.55 25.6 30.56 1.20 0.84
D6 152 x 229 4.6 322.62 23.89 79 46 400.3 31.7 34.83 19.77 18.3 0.91 1.08
Note: 1 mm = 0.0394 in.; 1 MPa = 145.04 psi; 1 kN = 0.225 kip; 1 kNm = 0.74 kip-ft. Total average 1.03 0.99
Total SD 0.09 0.07

Table 3—Experimental theoretical comparison on typical prestressed cross section


(reinforced with CFRP laminates)
MX, kNm
Reference Specimen Specimen geometry and material properties ex, mm ey, mm Pext., kN EXP. THEOR. Δ = MEXP./MTHEOR.

S-1 64 60 893 1216 1044 1.16

S-2 38 189 748 1281 1179 1.09


Di Ludovico et al.28

S-3 55 53 601 1123 1104 1.02

Note: 1 mm = 0.0394 in.; 1 MPa = 145.04 psi; 1 kN = 0.225 kip; 1 kNm = 0.74 kip-ft. Total average 1.09
Total SD 0.07

ACI Structural Journal/July-August 2010 397


Fig. 12—(a) Square cross section: two-dimensional Mx-My Fig. 14—Square cross section: (a) comparison between α
domains (ν = 0.2); and (b) percentage of errors between derived from proposed method and simplified expressions;
proposed procedure and simplified expressions in terms of and (b) percentage of errors between proposed procedure
Mx and My. (Note: 1 kNm = 0.74 kip-ft.) and simplified expressions.

Fig. 13—Rectangular cross section: (a) two-dimensional Fig. 15—Rectangular cross section: (a) comparison
Mx-My domains (ν = 0.2); and (b) percentage of errors between α derived from proposed method and simplified
between proposed procedure and simplified expressions in expressions; and (b) percentage of errors between proposed
terms of Mx and My. (Note: 1 kNm = 0.74 kip-ft.) method and simplified expressions.

predictions and experimental results shows an average proposed fiber integration procedure and the commonly used
percentage theoretical underestimation of experimental results simplified expressions provided by Eurocode 8,3 Hsu6
of 9% (refer to Table 3). (Eq. (3)), Parme et al.7 (Eq. (4) with β = 0.578 and β = 0.581,
respectively, according to the graph in Reference 7),
COMPARISON BETWEEN THEORETICAL Bresler,4 and ACI 318-0511 (Eq. (5), RL method) are reported
PROCEDURE AND SIMPLIFIED for the square and the rectangular cross section, respectively.
EXPRESSIONS IN LITERATURE Figures 12(b) and 13(b) show the percentage effective
In this section, a simple case study is investigated to resisting moments Mx and My errors obtained by using
compare the theoretical results provided by the proposed such simplified expressions versus the orientation angle β
theoretical procedure, in terms of effective resisting (–90 degrees ≤ 0 degrees due to the cross section
moments Mx and My under biaxial bending and axial load, symmetry about the x- and y-axes). Such figures clearly
and those obtained by using the main simplified expressions show that the minimum errors are provided by the RL
applied by various authors or reported in the codes. In
method (Eq. (5))4,11 and Eq. (4)7 that supply effective
particular, one square cross section (side b = 11.81 in. [300 mm]
resisting moments very close to those obtained by the
and concrete cover c = 1.57 in. [40 mm]) and one rectangular
proposed fiber algorithm (percentage underestimation less
cross section (width b = 11.81 in. [300 mm], depth h = 19.69 in.
than 5% and 10%, respectively). In the range 10 to 15% is
[500 mm] and concrete cover c = 1.57 in. [40 mm]) were
analyzed. Reinforcement was made of eight 14 mm (0.55 in.) the underestimation of resisting moments computed by using
diameter bars and twelve 16 mm (0.62 in.) diameter bars, Eq. (3), which increases up to 30% by using the equations
respectively, (elastic-plastic stress-strain relationship with provided by Eurocode 8.3 Moreover, the latter method also
yielding strength equal to fy = 56,753 psi [391.3 MPa]). It provides unsafe predictions in the range of 40 degrees ≤ β ≤
was a design example so that design values according to 50 degrees and 25 degrees ≤ β ≤ 35 degrees, in the case of
Eurocode 29 for concrete were adopted (fc′ = αcc ⋅ fck/γc with square or rectangular cross section, respectively.
γc = 1.5 and αcc = 0.85, that is, Italian National Annex). A With reference to the simplified expressions based on the
parabolic-rectangular concrete stress-strain relationship calibration of parameter α1 = α2 = α in the LC method (that
was considered (characteristic cylindrical compressive is, AS 3600,5 according to Eq. (2); Ghersi et al.,8 with α = 1.5;
strength of fck = 3626 psi [25 MPa], design strength fc′ = Eurocode 29; and Monti and Alessandri10). Figures 14(a) and
2055 psi [14.17 MPa], and ultimate axial strain equal to 15(a) show the α values to be used in the LC method to
3.5‰). In Fig. 12(a) and 13(a), the 2D Mx-My domains for obtain the same moments provided by the proposed fiber
an axial load ratio ν = 0.2 (P = 79.0 kips [351.4 kN] and P = algorithm for different axial load ratios ν. In Fig. 14(b) and
137.9 kips [613.8 kN], respectively) obtained by using the 15(b), the percentage difference between α values derived

398 ACI Structural Journal/July-August 2010


by the proposed method and by the existing simplified 4. Bresler, B., “Design Criteria for Reinforced Columns under Axial
expressions are reported. Load and Biaxial Bending,” ACI JOURNAL, Proceedings V. 37, No. 5, Nov.
1960, pp. 481-490.
Figures 14 and 15 show that, especially for 0.1 ≤ ν ≤ 0.4 5. AS 3600-1988, “Concrete Structures,” Standards Association of
(typical normalized axial load in columns of existing structures), Australia, North Sydney, NSW, Australia, 1988, 105 pp.
the most accurate results are provided by Monti and 6. Hsu, C. T. T., “Analysis and Design of Square and Rectangular
Alessandri.10 Rather good predictions are also provided by Columns by Equation of Failure Surface,” ACI Structural Journal, V. 85,
No. 2, Mar.-Apr. 1988, pp. 167-179.
using a fixed value of α = 1.5, even if in some cases slightly 7. Parme, A.; Nieves, J. M.; and Gouwens, A., “Capacity of Reinforced
unsafe resisting moments can be computed. On the other Rectangular Columns Subjected to Biaxial Bending,” ACI JOURNAL,
hand, the α-ν relationships provided by Eurocode 29 and Proceedings V. 63, No. 9, Sept. 1966, pp. 911-923.
AS 36005 are less accurate; in particular, the former appear 8. Ghersi, A., and Muratore, M., “Verifica e Progetto allo Stato Limite
Ultimo di Pilastri in c.a. a Sezione Rettangolare: Un Metodo Semplificato,”
to be too conservative and the latter unsafe in most cases. Ingegneria Sismica, V. XXI, No. 3, Sept.-Dec. 2004, pp. 41-49. (In Italian)
9. EN 1992-1-1, “Eurocode 2: Design of Concrete Structures, Part 1-1:
CONCLUSIONS General Rules and Rules for Buildings,” Dec. 2004, Brussels, Belgium,
An integration procedure was implemented to analyze 274 pp.
10. Monti, G., and Alessandri, S., “Assessment of RC Columns under
cross sections of regular or irregular shapes (for example, Combined Biaxial Bending and Axial Load,” Proceedings of the 2nd
made by concrete) internally or externally reinforced by mild International Congress, fib, Paper 10-8, Naples, Italy, 2006. (CD-ROM)
steel and prestressing tendons as well as by FRP. The 11. ACI Committee 318, “Building Code Requirements for Structural
calculation procedure, based on cross section discretization, Concrete (ACI 318-05) and Commentary (318R-05),” American Concrete
Institute, Farmington Hills, MI, 2005, 430 pp.
allows ultimate moments to be determined under combined
12. Ramamurthy, L. N., “Investigation of the Ultimate Strength of Square
axial load and biaxial bending. The construction of the Mx-My and Rectangular Columns under Biaxially Eccentric Loads,” Reinforced
domain for a given axial load (hence, the complete 3D failure Concrete Columns, SP-13, American Concrete Institute, Farmington Hills,
domain) and moment-curvature diagrams, taking into MI, 1966, pp. 263-298.
account linear and nonlinear material constitutive relationships, 13. Silva, M. A.; Swan, C. C.; Arora, J. S.; and Brasil, R. M. L. R. F.,
“Failure Criterion for RC Members under Biaxial Bending and Axial Load,”
also with softening, can be computed. To give an idea about Journal of Structural Engineering, V. 127, No. 8, 2001, pp. 922-929.
the algorithm processing speed, it is highlighted that an 14. Wang, W., and Hong, H. P., “Appraisal of Reciprocal Load Method
entire 3D failure domain can be evaluated in a few hundred for Reinforced Concrete Columns of Normal and High Strength Concrete,”
iterations, that is, in a few seconds (less than 1 minute), with Journal of Structural Engineering, V. 128, No. 11, 2002, pp. 1480-1486.
a computation time compatible with engineering practice and 15. Sussekind, J. C., “Curso de Concreto,” Editora Globo, Rio de
Janeiro, Brazil, 1989, 376 pp. (in Portuguese)
current hardware computing performance. 16. Hulse, R., and Mosley, W. H., Reinforced Concrete Design by
Reliability of the calculation procedure was assessed by Computer, Macmillan Education Ltd., New York, 1986, 390 pp.
both numerical analyses and experimental versus theoretical 17. Spiegel, L., and Limbrunner, G. F., Reinforced Concrete Design,
comparisons with reference to a large number of experimental Prentice-Hall, Upper Saddle River, NJ, 1988, 516 pp.
18. Bonet, J. L.; Miguel, P. F.; Fernandez, M. A.; and Romero, M. L.,
data available from the scientific literature. All numerical “Analytical Approach to Failure Surfaces in Reinforced Concrete Sections
analyses showed fast convergence of the algorithm and in Subjected to Axial Loads and Biaxial Bending,” Journal of Structural
every case a good correlation emerged. Comparison between Engineering, ASCE, V. 130, No. 12, Dec. 2004, pp. 2006-2015.
the numerical results provided by the theoretical procedure 19. Kawakami, M. T.; Kagaya, M.; and Hirata, M., “Limit States of
Cracking and Ultimate Strength of Arbitrary Concrete Sections under
and the main existing simplified expressions highlighted the Biaxial Loading,” ACI JOURNAL, Proceedings V. 82, No. 2, Mar. 1985,
benefits and limitations of these expressions frequently pp. 203-212.
recalled in design code formulas. Such formulas, either 20. Bousias, S. N.; Panagiotakos, T. B.; and Fardis, M. N., “Modelling of
based on the 2D domain construction or on the definition of RC Members under Cyclic Biaxial Flexure and Axial Force,” Journal of
coefficient α, lead to different degrees of accuracy and are Earthquake Engineering, V. 6, No. 2, 1996, pp. 711-725.
21. Sfakianakis, M. G., “Biaxial Bending with Axial Force of Reinforced,
not conservative at all in some cases. Composite and Repaired Concrete Section of Arbitrary Shape by Fiber
The calculation procedure was implemented in free down- Model and Computer Graphics,” Advanced in Engineering Software, V. 33,
loadable software from the following Web site: http:// 2002, pp. 227-242.
www.reluis.it (area: design, software, biaxial). 22. Consolazio, G. R.; Fung, J.; and Ansley, M., “M-φ-P Diagrams for
Concrete Sections under Biaxial Flexure and Axial Compression,” ACI
Structural Journal, V. 101, No. 1, Jan.-Feb. 2004, pp. 114-123.
ACKNOWLEDGMENTS 23. De Vivo, L., and Rosati, L., “Ultimate Strength Analysis of Reinforced
The analyses were developed within the activities of University Network Concrete Sections Subject to Axial Force and Biaxial Bending,” Computer
of Seismic Engineering Laboratories—ReLUIS for the research program Methods in Applied Mechanics and Engineering, V. 166, 1998, pp. 261-287.
funded by the Department of Civil Protection—Progetto Esecutivo 2005-2008. 24. Cedolin, L.; Cusatis, G.; Eccheli, S.; and Roveda, M., “Capacity of
The authors would like to thank I. Iovinella for his undergraduate thesis on Rectangular Cross Sections under Biaxial Eccentric Loads,” ACI Structural
the research topic presented in the paper and for his collaboration in the Journal, V. 105, No. 2, Mar.-Apr. 2008, pp. 215-224.
development of the related numerical analyses. 25. Furlong, R. W.; Cheng-Tzu, T. H.; and Mirza, S. A.; “Analysis and
Design of Concrete Columns for Biaxial Bending-Overview,” ACI
JOURNAL, Proceedings V. 101, No. 3, May-June 2004, pp. 413-423.
REFERENCES 26. Mander, J. B.; Priestley, M. J. N.; and Park, R., “Theoretical Stress-
1. FEMA 356, “NEHRP Guidelines for the Seismic Rehabilitation of Strain Model for Confined Concrete,” Journal of the Structural Division,
Buildings,” Federal Emergency Management Agency, Washington, DC, 2000. ASCE, V. 114, 1988, pp. 1804-1826.
2. Infrastructure Department, “Guidelines for Constructions,” Supplement 27. Chapra, S. C., and Canale, R. P., Numerical Methods for Engineers,
to Ordinario, N .30 alla G.U. n.29 del 4/2/2008, 428 pp. (in Italian) fourth edition, McGraw-Hill, New York, NY, 2001, 925 pp.
3. EN 1998-1, “Eurocode 8: Design of Structures for Earthquake Resistance, 28. Di Ludovico, M.; Nanni, A.; Prota, A.; and Cosenza, E., “Repair of
Part 1: General Rules, Seismic Actions and Rules for Buildings,” Dec. Bridge Girders with Composites: Experimental and Analytical Validation,”
2003, Brussels, Belgium, 232 pp. ACI Structural Journal, V. 102, No. 5, Sept.-Oct. 2005, pp. 639-648.

ACI Structural Journal/July-August 2010 399

Potrebbero piacerti anche