Engineering Structures 104 (2015) 1–17

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

A new technique for ultimate limit state design of arbitrary shape RC

sections under biaxial bending
R. Vaz Rodrigues
Departamento de Engenharia Civil Arquitectura e Georrecursos, Instituto Superior Técnico, Lisbon, Portugal

a r t i c l e i n f o a b s t r a c t

Article history: The design of reinforced concrete sections of arbitrary shape, namely with variable geometry, holes as well
Received 28 August 2014 as with arbitrary distribution of reinforcing steel bars, is a very common task in civil engineering, rein-
Revised 1 September 2015 forced concrete structures. The design of these sections requires the integration of non-linear stress fields
Accepted 9 September 2015
on complex shapes, because of the non-linear behavior of concrete in compression. In this paper, a novel
Available online 26 September 2015
algorithm is proposed to compute the ultimate strength of reinforced concrete sections under biaxial bend-
ing. The algorithm includes section subdivision into trapezoidal elements using the techniques of polygon
Keywords:
clipping algorithm proposed by Weiler–Atherton. Exact numerical integration for normal strength con-
Cross sections
Reinforced concrete
crete (fck 6 50 MPa) is achieved, for each trapezoid, using the change of variables theorem followed by
Biaxial bending Gauss–Legendre integration. The proposed technique is hereafter referred to as WAGL (Weiler–
Interaction diagram Atherton, Gauss–Legendre). The verification of the proposed algorithm is performed by comparing analyt-
Ultimate bending strength ical results between the WAGL technique and methods proposed by other authors (five examples).
Computer graphics Additionally, the results obtained are also compared with experimental results available in the literature.
The application of the WAGL technique is illustrated with two RC cross-section design examples.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction approximate equation of the failure surface for square and rectan-
gular columns, obtaining good agreement with available test
The design of reinforced concrete cross sections is a common results. The approximate equation of the failure surface is obtained
task in structural engineering and usually involves reinforced con- using the ratio of ‘‘nominal bending moments” about x and y axis
crete buildings, bridges and underground structures. The shape of and the ‘‘nominal bending moments at balanced strain condition”.
cross sections is found to be of various types, which together with The use of computer programs to design reinforced concrete
the bending moments acting simultaneously in both axes and the cross sections is nowadays common. Bentz [6] presents the treat-
axial force turns the design a strongly non-linear problem. ment of reinforced concrete sections of arbitrary shape, including
Simplified methods for approaching the strength of cross sec- arbitrary location of reinforcement bars, using a computer applica-
tions are proposed by Bresler [1], who provides two techniques tion. Papanikolaou [7] shows that the type of algorithm used to cal-
for evaluating the ultimate strength of rectangular columns, the culate the ultimate strength of arbitrary cross sections, along with
load contour method and reciprocal load equation. An application the material constitutive laws, line subdivision and stress integra-
of the load contour method, an approximate formulation for rect- tion scheme are important aspects of the algorithms.
angular, circular or elliptical cross sections, can be found in Euro- Concerning the material constitutive laws for concrete in com-
code 2 (clause 5.8.9) [2]. The CEB/FIP manual on bending and pression, Eurocode 2 defines concrete in compression as parabolic
compression [3] provides design charts for rectangular and circular linear, for design of reinforced concrete cross sections (clause
reinforced concrete cross-sections with various distributions of 3.1.7), therefore this type of relation will be followed in this study.
reinforcement, which are commonly used in current practice. This The use of parabolic linear behavior for concrete was also adopted
analysis shows design charts for some particular distributions of by Rosati et al. [8]. The consideration of fully arbitrary material law
reinforcement along the perimeter of cross sections. Walther and for concrete in compression is evaluated in the studies presented
Houriet [4] provide design charts for hollow reinforced concrete by Papanikolaou [7].
cross sections under biaxial eccentric loads. Hsu [5] proposes an The most critical aspects in terms of efficiency of algorithms are
section subdivision and the stress integration scheme. In order to
calculate the contributions of the compression stress fields, Char-
E-mail address: rui.vaz.rodrigues@tecnico.ulisboa.pt

http://dx.doi.org/10.1016/j.engstruct.2015.09.016
0141-0296/Ó 2015 Elsevier Ltd. All rights reserved.
2 R. Vaz Rodrigues / Engineering Structures 104 (2015) 1–17

Notation

M bending moment euk strain of reinforcing steel at maximum load

N axial force x, y coordinates (domain X)
rc compressive stress in the concrete n, g coordinates (domain T)
fcd design value of concrete compressive strength g, f functions
fck characteristic compressive cylinder strength of concrete Ci,x, Ci,y, Ca/b/c constants
fck,cube characteristic compressive cube strength of concrete As cross sectional area of reinforcement
Ec modulus of elasticity of concrete d effective depth
ec compressive strain in the concrete h overall depth
ec2 compressive strain in the concrete at peak stress d0 distance from top fiber, to center of reinforcing bar
ecu2 ultimate compressive strain in the concrete b cross section width
n exponent value, integer Mpred calculated bending moment
rs stress in reinforcing steel Mtest measured bending moment, at failure
Es modulus of elasticity of reinforcing steel Ptest measured axial force, at failure
es strain of reinforcing steel 1/R curvature
fsd design yield strength of reinforcement a angle; ratio
fy yield strength of reinforcement h angle of inclination of neutral axis
ft tensile strength of reinforcement esup strain at topmost fiber of cross section
k coefficient for considering steel hardening einf strain at bottommost fiber of cross section
eud design strain of reinforcing steel at maximum load
(=0.9 euk)

alampakis and Koumousis [9] divides any section with curvilinear a novel approach, which combines the advantages of numerical
trapezoids and used closed-form solutions to compute the internal integration schemes based on Gauss–Legendre integration, com-
forces, obtaining exact results. Sfakianakis [10] proposes computa- bined with Weiler–Atherton [15] algorithm considered for section
tion of internal forces without section subdivision, using fiber subdivision. This algorithm was developed for application in com-
integration. However, this approach can give approximate results puter graphics, but has proved to be efficient in the division of any
depending on the mesh division. Dias da Silva et al. [11] uses a section in polygons for the computation of internal forces of arbi-
closed-form algorithm applicable to multi-rectangular sections trary reinforced concrete cross sections.
followed by Gauss–Legendre integration. This analysis shows that As presented by Lam et al. [16], the use of spreadsheets can con-
closed-form solutions for computation of the stress resultants are stitute an advantage in terms of its familiarity to the computer user.
preferable in terms of computational efficiency. Rodriguez and The use of extended interpreted programming languages that are
Aristizabal-Ochoa [12] present an algorithm based on section sub- embedded into the main spreadsheet program constitute an addi-
division into polygons and performed closed-form integration per tional advantage. An example of such programming languages is
polygon. Pallarès et al. [13] perform closed-form integration, but Visual Basic for Applications (VBA), which will be used in this study.
without section subdivision. Another interesting integration It should be noted that classic languages (e.g. C) can also be embed-
scheme using boundary integrals is proposed by De Vivo and ded in spreadsheets using a dynamic link library implementation
Rosati [14]. (DLL), providing substantially better computational performance.
Table 1 presents stress integration techniques available in
selected previous studies, compared with the suggested methodol- 2. Material properties
ogy. It can be seen that most previous studies use analytical
solutions (closed-form functions) that have the main advantage of 2.1. Stress strain relations of concrete for the design of cross sections
providing exact results, however they have the inconvenient of
being restricted to specific stress–strain material law or section Eurocode 2 [2], clause 3.1.7, gives the following stress–strain
shapes. A second option can be considered using stress integration relations for the design of cross sections and which are used in
schemes based on fiber integration, but they are usually associated the current study:
with the inconvenient of providing approximate solutions, depend- 8

n 
>
< rc ¼ f cd 1  1  ec2
ec
ing on fiber mesh density. A third option, the use of a suitable > for 0 6 ec 6 ec2
numerical integration (Green/Gauss) scheme has been successfully
rc ¼ f cd for ec2 6 ec 6 ecu2 ð1Þ
>
>
used as integration scheme, consisting of an excellent approach :
both in terms of accuracy and execution times. This paper presents rc ¼ 0 for ec 6 0

Table 1
Stress integration schemes of selected previous studies.

Stress integration schemes Authors

Closed-form per trapezoid/polygon Rosati et al. [8], Charalampakis et al. [9], Rodriguez and Aristizabal-Ochoa [12],
Pallarès et al. [13]
Closed-form, multi-rectangular Dias da Silva et al. [11]
Fiber integration Sfakianakis [10]
Numerical Green/Gauss with adaptive strain-mapping Papanikolaou [7]
(Section subdivision using Weiler–Atherton algorithm) numerical Gauss– Suggested methodology (WAGL)
Legendre integration
 R. Vaz Rodrigues / Engineering Structures 104 (2015) 1–17 3

σc σs

fcd k fsd
fsd

εc2 εcu2 εc fsd ε ud εc

Es
(a) (b)

Fig. 1. Unixial constitutive stress–strain relationships considered in present study for: (a) concrete under compression and (b) steel reinforcement under compression/
tension.

Table 2
Strength and strain values of concrete under compression according to Eurocode 2 (2004).

fck (MPa) 12 16 20 25 30 35 40 45 50 55 60 70 80 90
ec2 (‰) 2.0 2.2 2.3 2.4 2.5 2.6
ecu2 (‰) 3.5 3.1 2.9 2.7 2.6 2.6
n 2.0 1.75 1.60 1.45 1.40 1.40

8
The graphical representation of these equations is shown in >
>
rs ¼ Es es for 0 6 es 6 f sd =Es
>
>
Fig. 1(a). The values for compressive strain at reaching maximum < rs ¼ f sd þ ðk1Þ f sd ðes  f sd =Es Þ for f sd =Es 6 es 6 eud
e f =Es
strength ec2, ultimate strain in the concrete ecu2 and exponent value ud sd
ð5Þ
n are given as a function of the characteristic compressive strength
>
>
> rs ¼ Es es for  f sd =Es 6 es 6 0
>
:
fck, as shown in Table 2. Note that fcd is considered as fcd = a fck/cc, rs ¼ f sd þ eðk1 Þ f sd
f =Es
ðes þ f sd =Es Þ for  eud 6 es 6 f sd =Es
ud sd
according to Eurocode 2 clause 3.1.6.
Besides the tabulated values for the strains ec2, ecu2 and expo- Accordingly to Eurocode 2, the characteristic value of the yield-
nent n, Eurocode 2 provides analytical expressions for these values ing strength ranges between 400 MPa and 600 MPa. In addition,
as a function of the characteristic compressive strength fck, only Eurocode 2 states that the constant k and the strain at maximum
applicable for high strength concrete (fck P 50 MPa): stress euk depend on the ductility class, as shown in Table 3. How-
ever, one can consider k = 1, corresponding to horizontal yield pla-
ec2 ð‰Þ ¼ 2:0 þ 0:085ðf ck  50Þ0:53 ð2Þ teau, in any case. Note that the design strain at maximum stress
corresponds to eud = 0.9euk and that the design yield strength corre-
ecu2 ð‰Þ ¼ 2:6 þ 35½ð90  f ck Þ=1004 ð3Þ sponds to fsd = fy/1.15. The graphical representation of these equa-
tions is shown in Fig. 1(b).
4
n ¼ 1:4 þ 23:4½ð90  f ck Þ=100 ð4Þ
As mentioned, the concrete model adopted in the paper is the 3. Computation of internal forces in a trapezoidal element
model considered in Eurocode 2, clause 3.1.7 (stress–strain rela-
tions for the design of cross-sections), and is the model for uncon- 3.1. Axial force due to concrete in compression
fined concrete. The effect of confinement results in higher
compressive strengths and higher strains which can be considered The calculation of resultant compression force N is expressed by
using values of fck,c, ec2,c and ecu2,c (clause 3.1.9 of Eurocode 2). the integral in integration domain W, as indicated in Fig. 2 and
expressed by the following equation:
2.2. Stress strain relations of reinforcement steel Z
N¼ rc ðx; yÞ dx dy ð6Þ
The behavior of steel in compression or tension is assumed to be W
defined by a bilinear relation, with k = ft/fy, the ratio between the In a local coordinate system (xlocal, ylocal), assuming that neutral
tensile strength and the yielding strength, as shown in Eq. (5). axis is coincident with xlocal axis (Fig. 2(a)), the strains are
expressed as:
Table 3
Values of k and strain at maximum stress euk. 1
ec ðx; yÞ ¼ y ð7Þ
Class k euk (‰) R
A P1.05 P2.5 Replacing this result for the strain distribution, the concrete
B P1.08 P5.0 compressive stresses can be obtained as a function of the local
C P1.15 P7.5 coordinate y and of the curvature 1/R. Hence, two ranges can be
<1.35
identified:
4 R. Vaz Rodrigues / Engineering Structures 104 (2015) 1–17

Strain profile

εsup

Neutral axis
My (>0) ylocal
(a) x local

(x0, y0 ) εinf
y

x Mx (>0) N (>0) Compression

N (<0) Tension
(b)

(-1;1) (1;1)
(x4; y4) (x3; y3)
T
W
y (x1; y1)
(x2; y2) (-1;-1) (1;-1)
(Local)
x

Fig. 2. (a) Global and Local Coordinate systems, neutral axis inclination h and sign conventions. (b) Coordinate transformation g(n, g) from T domain (n, g), to W domain (x, y).

8 
  2 3
> 1y n @g 1 ðn;gÞ @g 1 ðn;gÞ
< rc ¼ f ec2
cd 1  1  ec2 for 0 6 y 6 1=R @g
R @n
rc ðyÞ ¼ ð8Þ D gðn; gÞ ¼ 4 @g 5 ð12Þ
2 ðn;gÞ @g 2 ðn;gÞ
>
:r ¼ f ec2
c cd for 1=R
6 y 6 e1=R
cu2 @n @g

The derivatives shown in the expression above can be explicitly

In order to perform integration of the compression stresses
written as:
along the domain W, defined by an arbitrary trapezoid, it is useful 8




to use the change of variables theorem. Therefore, the function > @g 1 1þg
x1 þ 14 g x2 þ 1þ4 g x3 þ 1 g
> @n ¼
>
> 4 4
x4
g, used for variable transformation (mapping), is defined for the >
> 1þn 1n 1þn 1n
>
< @g 1
¼ 4 x1 þ 4 x2 þ 4 x 3 þ 4 x 4
corresponding points between domain T and domain W, as @g




 ð13Þ
indicated in Fig. 2(b). >
> @g 1þg
y1 þ 14 g y2 þ 1þ4 g y3 þ 1 g
> @n2 ¼
> y4
Given a point of coordinates n and g, defined in the coordinate >
>
4 4
>      
: @g2 ¼ 1þn y þ 1n y þ 1þn y þ 1n y  
system of domain T, the corresponding point (x, y) in the domain W @g 4 1 4 2 4 3 4 4
will be defined by the following function g (R2 ? R2):
8 After simplifying terms and grouping the terms of variables n
> X
4
and g, the following expression is obtained for the determinant
>
> g 1 ðf; gÞ ¼
>
<
Ni ðn; gÞ xi
of the Jacobian of function g:
ðx; yÞ ¼ gðn; gÞ ¼
i¼1
ð9Þ
>
>
> X
4 j det½D gðn; gÞj ¼ jðC 2x C 1y  C 1x C 2y Þn þ ðC 1x C 3y  C 3x C 1y Þg
: g 2 ðf; gÞ ¼
> Ni ðn; gÞ yi
i¼1
þ ðC 2x C 3y  C 3x C 2y Þj ð14Þ

The functions Ni, with i = 1–4, are defined as follows: Next, the constants C1,x, C1,y, C2,x, C2,y, C3,x, C3,y and C4,y are defined
8 as a combination of trapezoidal element coordinates, in a local
> N1 ¼ ð1  g  n þ gnÞ=4
>
> coordinate system (x, y):
<N ¼ ð1  g þ n  gnÞ=4 8
2
ð10Þ >
> C 1;x ¼ ðx1 x2 þx 3 x4 Þ
C 1;y ¼ ðy1 y2 þy3 y4 Þ
>
>
> N 3 ¼ ð1 þ g þ n þ gnÞ=4 >
>
>
4 4
: >
> ðx1 þx2 þx3 x4 Þ
C 2;y ¼ ðy1 þy24þy3 y4 Þ
N4 ¼ ð1 þ g  n  gnÞ=4 > C 2;x ¼
> 4
>
<
C 3;x ¼ ðx1 x24þx3 þx4 Þ C 3;y ¼ ðy1 y24þy3 þy4 Þ
As already mentioned, in order to perform integration of the com- ð15Þ
>
> C 4;x ¼ ðx1 þx2 þx3 þx4 Þ C 4;y ¼ ðy1 þy2 þy3 þy4 Þ
pressive stresses in the trapezoid concrete element, the function g >
>
>
>
4 4
has been used for variable change, resulting in the following expres- >
> C a ¼ ðC 2x C 1y  C 1x C 2y Þ C b ¼ ðC 1x C 3y  C 3x C 1y Þ
>
>
sion for integration on domain T (1 6 n 6 1 and 1 6 g 6 1): :
C c ¼ ðC 2x C 3y  C 3x C 2y Þ
Z Z !
X
4
Using the (above defined) constants Ca, Cb and Cc, the
N¼ rc ðyÞdxdy ¼ rc Ni ðn; gÞyi jdet½Dgðn; gÞjdndg ð11Þ
X T i¼1
determinant of the Jacobian of function g can thus be written in
the following form:
where D g(n, g) is the Jacobian of the function g, as shown in the
j det½D gðn; gÞj ¼ jC a n þ C b g þ C c j ð16Þ
following equation:
 R. Vaz Rodrigues / Engineering Structures 104 (2015) 1–17 5

Z þ1 Z þ1
And Eq. (11) can now be written as:
M¼ f cd ½C 3;y g þ C 2;y n þ C 1;y gn
X
4 1 1
"
Ni ðn; gÞyi ¼ C 3;y g þ C 2;y n þ C 1;y gn þ C 4;y ð17Þ  n #
1
½C 3;y g þ C 2;y n þ C 1;y gn þ C 4;y 
þ C 4;y  1  1  R
i¼1
ec2
Therefore, the resultant axial force in a trapezoidal element,
corresponding to the first range of concrete’s constitutive law
 j½C A n þ C B g þ C C jdndg ð25Þ
Eq. (8), can be written as: In order to numerically evaluate the double integral,
Z Z "  n # Gauss–Legendre quadrature is used. Points defined with 3  3 grid
þ1 þ1 1
½C 3;y g þ C 2;y n þ C 1;y gn þ C 4;y  are necessary, which gives exact results for polynomial expressions
N¼ f cd 1 1 R
1 1 ec2 up to degree of five. Therefore, the final expression of the resultant
j½C A n þ C B g þ C C jdndg ð18Þ bending moment in a trapezoidal element is given by:
rffiffiffi rffiffiffi! rffiffiffi ! rffiffiffiffiffi rffiffiffi!
In order to numerically evaluate the double integral, 55 3 3 58 3 55 3 3
M¼ h  ; þ h  ;0 þ h  ;
Gauss–Legendre quadrature is used with 2  2 points, which gives 99 5 5 99 5 99 5 5
an exact result for polynomial expressions up to degree of three. rffiffiffi! rffiffiffi!
It should be noted that exact results can only be obtained for 85 3 88 85 3
þ h 0;  þ hð0; 0Þ þ h 0;
concrete strength up to fck = 50 MPa. For high-strength concrete 99 5 99 99 5
(fck > 50 MPa), the first branch of the concrete stress–strain rela- rffiffiffi rffiffiffi! rffiffiffi ! rffiffiffi rffiffiffi!
55 3 3 58 3 55 3 3
tionship ceases to be polynomial (non-integer exponent value) þ h ; þ h ;0 þ h ; ð26Þ
99 5 5 99 5 99 5 5
and hence the employed Gauss–Legendre integration yields
approximate results.
where h(n, g) is given by:
Therefore, the final expression of the axial resultant force in a
trapezoidal element is: hðn; gÞ ¼ f cd ½C 3;y g þ C 2;y n þ C 1;y gn
        "  n #
1 1 1 1 1 1 1 1
N ¼ f pffiffiffi ; pffiffiffi þ f  pffiffiffi ; pffiffiffi þ f pffiffiffi ;  pffiffiffi þ f  pffiffiffi ;  pffiffiffi
1
½C 3;y g þ C 2;y n þ C 1;y gn þ C 4;y 
þ C 4;y  1  1  R
3 3 3 3 3 3 3 3 ec2
ð19Þ
 j½C A n þ C B g þ C C j ð27Þ
"  n #
1
½C 3;y g þ C 2;y n þ C 1;y gn þ C 4;y  Similarly, the explicit expression for the resulting bending
With f ðn; gÞ ¼ f cd 1 1 R
ec2 moment, corresponding to the second range of Eq. (8), results in:
Z Z !
j½C A n þ C B g þ C C j ð20Þ X
4
M¼ yrc ðyÞ dx dy ¼ Ni ðn; gÞyi f cd j det½D gðn; gÞjdndg
Similarly, the explicit expression for the resultant axial force in X T i¼1
a trapezoidal element, corresponding to the second range of ð28Þ
concrete’s constitutive law of Eq. (8), results in:
Z Z Using the results from Eqs. (16) and (17), the following
N¼ rc ðyÞ dxdy ¼ f cd j det½D gðn; gÞ jdndg expression is obtained:

Z
X
Z
T Z 1 Z 1
1 1
¼ f cd j½C A n þ C B g þ C C j dgdn ð21Þ M¼ ½C 3;y g þ C 2;y n þ C 1;y gn þ C 4;y f cd j½C A n þ C B g þ C C j dgdn
1 1
1 1
ð29Þ
The numerical evaluation of this expression is obtained using
2  2 integration points and using Eq. (19), but taking f(n, g) as: The numerical evaluation of this expression is obtained using
2  2 integration points and using Eq. (19) (with M instead of N),
f ðn; gÞ ¼ f cd j½C A n þ C B g þ C C j ð22Þ but taking f(n, g) as:

f ðn; gÞ ¼ ½C 3;y g þ C 2;y n þ C 1;y gn þ C 4;y f cd j½C A n þ C B g þ C C j ð30Þ

3.2. Bending moment due to concrete in compression
The position of the resultant axial force is yR = M/N, with M and
The calculation of the resultant bending moment M, with N obtained from Eqs. (25) and (18) for the first range or
respect to local axis x, is expressed by the integral on domain Eqs. (29) and (21) for the second range.
W, as shown in Fig. 2(a) and expressed in the following equation:
Z 4. Subdivision of cross section into trapezoidal elements
M¼ yrc ðx; yÞ dxdy ð23Þ
W
The subdivision of cross section into trapezoidal elements is
In order to perform integration of the compressive stresses in based on the Weiler–Atherton algorithm [15]. This algorithm is
the concrete trapezoidal element, the function g for variable based on a two-dimensional polygon clipper originally proposed
transformation is used, resulting in the following expression for in 1977 by Kevin Weiler and Peter Atherton, in the scope of com-
integration on domain T (1 6 n 6 1 and 1 6 g 6 1): puter graphics, with the aim of performing hidden surface removal
Z Z ! ! using polygon area sorting. As originally mentioned in [15], ‘‘the
X
4 X
4
output of the method is in the form of polygons, making it useful
M¼ yrc ðyÞ dxdy ¼ Ni ðn; gÞyi rc N i ðn; gÞyi
X T i¼1 i¼1
in a variety of situations including the usual display applications.”
One of the advantages of the Weiler–Atherton algorithm is that
j det½D gðn; gÞjdndg ð24Þ
it preserves the polygon data structure, both in output and input
Using the results from Eqs. (16) and (17), the explicit expression data, which allows recursive use of the algorithm when additional
for the bending moment results in: subdivision of the cross-section is required.
6 R. Vaz Rodrigues / Engineering Structures 104 (2015) 1–17

In this paper, the algorithm is used to perform polygon clipping Table 4

of a base polygon after a polygonal window, as shown in Fig. 3. Lists X, Y, Z required for implementing the Weiler–Atherton algorithm.

The basic steps of the algorithm are the following: X {P1, P2, B, P3, P4, P5, P6, P7, P8, P9, A, P10}
Y {W1, W2, W3, B, A, W4}
(1) Calculate all the intersections between the base polygon and Z {A}

the clipping window (points A and B in Fig. 3).

(2) Compile a list (X, see Table 4) containing all vertices of the
base polygon and the intersections points calculated at step the four equations just mentioned and multiplying the value of
(1). axial force and bending moment by (1). The contribution of each
(3) Compile a list (Y, see Table 4) containing all vertices of the reinforcing bar is direct: after defining the strains on top and bot-
clipping window and the intersection points calculated at tom fibers, the curvature becomes known, and therefore the contri-
step (1). butions of all reinforcing steel bars are:
(4) Compile a list (Z, as indicated in Table 4) containing only the X
n
1
intersection points that ‘‘step in” (when ‘‘walking” along the Ns ¼ yi rsi Asi ð31Þ
R
base polygon) the clipping window. Please note that is i¼1

required to define a direction associated with the base poly-

gon. In the case of this example (see Fig. 3), counter clock- X
n
1
Ms ¼ y2i rsi Asi ð32Þ
wise direction is chosen, so that vertex A is in this case the i¼1
R
only vertex that ‘‘steps in” the clipping window.
(5) Select a vertex from list Z (A in the current example) and travel The interaction diagram point (Ntot, Mtot) is given by:
along the list X (starting from this vertex). When an intersec- X
n
tion vertex is found on list X (B in this example), one should Ntot ¼ Nci þ Ns ð33Þ
change the ‘‘walking path” to list Y. Again, when an intersec- i

tion vertex is found on list Y, one should return to the list X.

The procedure arrives to an end when an intersection vertex
X
n
Mtot ¼ Mci þ M s ð34Þ
that has already been selected is found (A, in list Y). The output i
of this example would then be: {A, P10, P1, P2, B}.
The axial force contribution Nci of trapezoid i is obtained from
Using the exemplified algorithm, it is possible to divide any Eqs. (18) and (21) and moment contribution Mci using Eqs. (25)
type of polygon into a group of four (or three) point polygons that and (29). One should note that the point where the resultants Mtot
can be treated as indicated in the previous section. and Ntot are calculated could be at any point of the section. In order
to allow comparison with the work of other researchers, the gross
section’s geometric centroid will be used in this study.
5. Computation of internal forces in complete section
5.1. Possible range of strain distributions
For each imposed strain on top and bottom fibers, the corre-
The interaction diagram between bending moments and axial
sponding part of the section that is under compression strains is
force can be obtained by integration of the stresses that result from
directly defined. The calculation of a complete section is afterwards
the strain states. In order to compute the resultant internal forces
performed by treatment of cross section contour, defined by a
(moment and axial force) associated with a particular strain state,
polygon. By applying the Weiler Atherton algorithm, any polygon
it is necessary to define limiting values of strains in steel and con-
is divided into a group trapezoidal elements, or triangular ele-
crete. Eurocode 2 [2] states that the compressive strain in concrete
ments. Note that the equations developed for trapezoidal elements
is in general limited to ecu2. However, if the section is under uni-
are also applicable for triangular elements; two points are in this
form compression, the strain value should be limited to ec2. With
case coincident. The calculation of the moment and the axial force
respect to reinforcement steel, strain values are limited to eud.
contributions, for each individual trapezoid situated inside the
The implementation of such principles can be formulated in terms
compression zone, can be performed using the Eqs. (18), (21),
of a strain path, defined by means of a diagram where the strains at
(25) and (29). With respect to openings, the contributions of the
the top fiber are plotted against the strains at the bottom fiber, as
axial force and the moment for each opening can be obtained using
exemplified in Fig. 4(a). The proposed technique allows the defini-
tion of the complete failure surface of R/C cross sections, including
W1 W2 the points at (Mx–My) plane with N = 0, using a non-iterative proce-
P1 P2 dure. The overall depth d and effective depth h are represented in
Vertex A : Clipping Fig. 4(b). The strain path key values are listed in Table 5 for an arbi-
“Step-in” vertex Window trary section.
P10
A B
W4 W3 6. Computation of failure surface and moment interaction
P9 P3 curves
P8

In order to calculate a point at the failure surface, it is required

to calculate the axial force N, and moments Mx and My, which are
P4 represented in Fig. 2(a). The axial force and bending moments
P7 Base Polygon P6 result from stress integration, for each inclination of neutral axis
and strain state defined by imposed strains at top and bottom
P5
fibers. The axial force and bending moments are, in a first stage,
Fig. 3. Application of Weiler–Atherton algorithm: Base polygon and clipping calculated about the neutral axis, and are afterwards converted
window. to the center of geometries.
 R. Vaz Rodrigues / Engineering Structures 104 (2015) 1–17 7

d’

d
h

(a) (b)
Fig. 4. Strain path for cross sections: (a) strain path for ecu = 2.0‰, ecu2 = 3.5‰, eud = 22.5‰, h = 1.90 m, d = 1.80 m and d0 = 0.05 m; (b) definition of overall depth h, effective
depth d and distance d0 in an arbitrary cross section.

My ¼ M x;local sin h þ M y;local cos h þ Ny0;local ð39Þ

Table 5 Note that, in this paper, the origin of global axis x and y is
Definition of admissible strain states in cross section. located at the geometric centroid of the section.
Strain state esup einf Fig. 5 illustrates the construction of the failure surface for a rect-
angular cross section with dimensions b = 25 cm  h = 45 cm and
1 ec2 ec2
2 ecu2 0 total amount of reinforcement 8  2.01 cm2. For a given fixed neu-


3 ecu2 ecu2 þ fEsds hd
h d tral axis inclination, the cross section is subjected to a series of
hd
  strain states (Table 5) that generate a non planar curve (Mx, My,
4 ecu2 ecu2 þ e h d
ud hd hd
N). It can be observed that the values of the bending moments
5 eud eud
6 d
0
ecu2 þ eud hdh ecu2 (Mx and My) depend of the strain state of the section, and especially
hd0 0

d0 ecu2 of the geometry of the section. As a consequence the ratio (Mx/My)

7
hd
0 ecu2 þ Es hd0
f sd h
is not constant along this curve. The computation is performed
8 0 ecu2
using neutral axis angle of inclination increments Dh = 3° and
800 strain states for each neutral axis angle of inclination, which
results in a total of 48 800 points. Note that the neutral axis angle
The axial force and bending moments (about the neutral axis)
of inclination ranges between h = 0° and h = 180° (covering all pos-
result from stress integration, for each inclination of neutral axis
sible strain states, according with strains limits indicated in
and strain state defined by imposed strains at top and bottom
Table 5) which results in a total of 48 800 points for the complete
fibers. The computation of axial force and bending moments about
surface.
the neutral axis (Mx,local) is done using Eqs. (33) and (34).
Additionally, it is required to evaluate the following integral in
order to compute the moment about the local y axis. The integra- 7. Examples and verification
tion of concrete stresses is expressed in the following equation:
Z 7.1. Example 1: Rectangular cross section
M y;local ¼ xrc ðx; yÞ dx dy ð35Þ
w
The first example shows the response of a rectangular cross sec-
For the first range of Eq. (8), this integral can be evaluated using tion with dimensions b = 0.30 m  h = 0.70 m, with total reinforce-
Eq. (26), where function h(n, g) is given by: ment As,tot = 4  10 cm2, placed at the corners of the cross section,
with cover equal to 5 cm, concrete strength fcd = 20 MPa and steel
hðn; gÞ ¼ f cd ½C 3;x g þ C 2;x n þ C 1;x gn
" yielding strength fsd = 500 MPa. Analysis is performed using neutral
 n #
1
½C 3;y g þ C 2;y n þ C 1;y gn þ C 4;y  axis angle of inclination increments Dh = 5° and 800 strain states for
þ C 4;x  1  1  R
ec2 each neutral axis angle of inclination. The moment interaction
curves are shown in Fig. 6 for (compression) axial force values of
 j½C A n þ C B g þ C C j ð36Þ Nsd = 5000 kN, 1000 kN, 0 kN and (tensile) axial force values of
Similarly, the explicit expression for the resulting bending Nsd = 750 kN and 1500 kN. The results of the present study com-
moment, corresponding to the second range of concrete’s constitu- pare very well with those obtained by Papanikolaou [7] using stress
tive law of Eq. (8), is computed using Eq. (19) (with M instead of N) integration by Green path integrals with adaptive strain-mapped
taking f(n, g) as: Gaussian sampling.

f ðn; gÞ ¼ ½C 3;x g þ C 2;x n þ C 1;x gn þ C 4;x f cd j½C A n þ C B g þ C C j ð37Þ

7.2. Example 2: U-shaped cross section
Finally, the bending moments with respect to global axis x and y
are obtained by the following equations: The second example shows the response of a U-shaped cross
section, analyzed in a study presented by Rosati et al. [8], with
M x ¼ M x;local cos h  My;local sin h þ Nx0;local ð38Þ
dimensions b = 0.80 m  h = 0.60 m and wall thickness of 20 cm,
8 R. Vaz Rodrigues / Engineering Structures 104 (2015) 1–17

Table 6
Comparison between numerical results and experimental data from Ramamurthy [18], and Pallarès et al. [19].

Test fcm,cube (MPa) a (°) fy (MPa) Ptest (kN) Mtest (kN m) Bars As (mm2) b (mm) d (mm) d0 (mm) Mpred (kN m) Mpred/Mtest
Ramamurthy – series A
A-1 52.1 15 291.9 564.9 51.7 8#4 1032 203.2 203.2 31.75 54.74 1.06
A-2 57.2 15 291.9 395.9 52.1 8#4 1032 203.2 203.2 31.75 50.34 0.97
A-3 55.2 15 291.9 378.1 49.7 8#4 1032 203.2 203.2 31.75 49.00 0.99
A-4 47.9 15 291.9 283.6 44.7 8#4 1032 203.2 203.2 31.75 42.80 0.96
A-5 42.7 15 291.9 235.8 43.4 8#4 1032 203.2 203.2 31.75 39.30 0.91
A-6 40.3 14.5 291.9 171.9 36.0 8#4 1032 203.2 203.2 31.75 35.66 0.99
A-7 44.5 11.3 291.9 146.8 38.0 8#4 1032 203.2 203.2 31.75 35.11 0.92
A-8 31.9 20 291.9 476.0 38.6 8#4 1032 203.2 203.2 31.75 40.53 1.05
A-9 46.2 20 291.9 280.2 45.5 8#4 1032 203.2 203.2 31.75 41.80 0.92
A-10 55.4 30 291.9 462.6 47.5 8#4 1032 203.2 203.2 31.75 51.23 1.08
A-11 57.4 30 291.9 264.7 38.8 8#4 1032 203.2 203.2 31.75 43.34 1.12
A-12 49.6 30 291.9 170.1 34.6 8#4 1032 203.2 203.2 31.75 36.94 1.07
A-13 20.9 33.7 291.9 164.6 30.1 8#4 1032 203.2 203.2 31.75 28.32 0.94
A-14 29.6 26.57 291.9 160.1 36.4 8#4 1032 203.2 203.2 31.75 31.94 0.88
A-15 23.7 45 291.9 266.9 33.9 8#4 1032 203.2 203.2 31.75 30.91 0.91
Ramamurthy – series B
B-1 32.4 15 322.6 629.0 51.1 8#5 1600 203.2 203.2 33.3 49.97 0.98
B-2 28.6 22.5 322.6 771.8 39.2 8#5 1600 203.2 203.2 33.3 40.41 1.03
B-3 37.2 30 322.6 533.8 54.2 8#5 1600 203.2 203.2 33.3 51.16 0.94
B-4 35.5 30 322.6 395.9 50.3 8#5 1600 203.2 203.2 33.3 49.31 0.98
B-5 21.7 45 322.6 598.3 30.4 8#5 1600 203.2 203.2 33.3 34.23 1.13
B-6 30.6 45 322.6 500.4 45.8 8#5 1600 203.2 203.2 33.3 44.35 0.97
B-7 32.8 45 322.6 516.0 52.4 8#5 1600 203.2 203.2 33.3 45.98 0.88
B-8 37.9 45 322.6 369.8 53.1 8#5 1600 203.2 203.2 33.3 48.87 0.92
Ramamurthy – series C
C-1 34.6 15 275.8 464.8 17.7 8#4 1032 152.4 152.4 31.75 18.90 1.07
C-1(a) 41.2 15 275.8 569.4 21.7 8#4 1032 152.4 152.4 31.75 20.53 0.95
C-2 37.7 15 275.8 400.3 18.3 8#4 1032 152.4 152.4 31.75 21.51 1.18
C-2(a) 48.3 15 275.8 489.3 22.4 8#4 1032 152.4 152.4 31.75 25.34 1.13
C-3 32.2 30 275.8 460.4 17.5 8#4 1032 152.4 152.4 31.75 16.92 0.96
C-4 26.8 30 275.8 378.1 17.3 8#4 1032 152.4 152.4 31.75 15.47 0.90
C-5 27.6 45 275.8 506.0 15.4 8#4 1032 152.4 152.4 31.75 13.05 0.85
C-6 34.5 45 275.8 350.3 21.4 8#4 1032 152.4 152.4 31.75 18.73 0.88
Ramamurthy – series R
R-138 25.9 45 275.8 138.3 14.9 8#3 568 152.4 152.4 30.16 12.30 0.83
R-238 34.5 45 275.8 160.1 17.3 8#3 568 152.4 152.4 30.16 14.39 0.83
R-338 36.2 33.7 275.8 118.8 16.3 8#3 568 152.4 152.4 30.16 13.89 0.85
R-438 29.6 33.7 275.8 71.2 13.0 8#3 568 152.4 152.4 30.16 11.69 0.90
Ramamurthy – series D
D-1 35.2 33.7 322.6 785.1 35.9 8#5 1600 152.4 228.6 33.3 42.78 1.19
D-2 28.3 33.7 322.6 400.3 40.7 8#5 1600 152.4 228.6 33.3 43.98 1.08
D-3 27.2 33.7 322.6 311.4 42.7 8#5 1600 152.4 228.6 33.3 43.73 1.02
D-4 27.9 45 322.6 680.6 31.1 8#5 1600 152.4 228.6 33.3 33.13 1.06
D-5 34.3 45 322.6 378.1 43.2 8#5 1600 152.4 228.6 33.3 42.31 0.98
D-6 26.5 60 322.6 400.3 36.6 8#5 1600 152.4 228.6 33.3 31.17 0.85
Ramamurthy – series E
E-1 26.1 26.57 322.6 464.8 59.0 8#5 1600 152.4 304.8 33.3 74.79 1.27
E-2 23.8 26.57 322.6 311.4 53.1 8#5 1600 152.4 304.8 33.3 70.87 1.34
E-3 30.8 45 322.6 435.9 53.1 8#5 1600 152.4 304.8 33.3 63.83 1.20
E-4 27.6 60 322.6 542.7 41.4 8#5 1600 152.4 304.8 33.3 45.83 1.11
Ramamurthy – series F
F-1 32.6 15 291.9 600.5 34.3 8#4 1032 152.4 228.6 31.75 40.16 1.17
F-2 39.9 33.7 291.9 533.8 30.5 8#4 1032 152.4 228.6 31.75 42.86 1.41
F-3 20.0 33.7 291.9 384.8 22.0 8#4 1032 152.4 228.6 31.75 27.21 1.24
F-4 28.1 33.7 291.9 266.9 24.4 8#4 1032 152.4 228.6 31.75 34.07 1.40
F-5 29.6 45 291.9 466.0 26.6 8#4 1032 152.4 228.6 31.75 31.18 1.17
Ramamurthy – series G
G-1 36.6 15 291.9 827.4 63.0 8#4 1032 152.4 304.8 31.75 74.71 1.18
G-2 28.1 15 291.9 418.1 53.1 8#4 1032 152.4 304.8 31.75 64.35 1.21
G-3 24.2 26.57 291.9 507.1 38.6 8#4 1032 152.4 304.8 31.75 54.87 1.42
G-4 41.4 26.57 291.9 333.6 50.8 8#4 1032 152.4 304.8 31.75 65.47 1.29
G-5 34.5 45 291.9 584.9 35.7 8#4 1032 152.4 304.8 31.75 55.95 1.57

Pallarès
10_05_3 94.1 14.04 558 473.9 35.0 4/10 314 100 200 20 40.15 1.15
10_05_4 94.1 14.04 558 175.2 25.8 4/10 314 100 200 20 28.11 1.09
10_1_4 95.1 26.56 558 166.1 23.0 4/10 314 100 200 20 25.84 1.12
 R. Vaz Rodrigues / Engineering Structures 104 (2015) 1–17 9

Table 6 (continued)

Test fcm,cube (MPa) a (°) fy (MPa) Ptest (kN) Mtest (kN m) Bars As (mm2) b (mm) d (mm) d0 (mm) Mpred (kN m) Mpred/Mtest
10_2_2 92.2 45 558 899.9 22.8 4/10 314 100 200 20 26.74 1.17
10_2_3 94.1 45 558 436.5 22.1 4/10 314 100 200 20 27.94 1.27
10_2_4 94.1 45 558 142.0 14.4 4/10 314 100 200 20 20.64 1.44
Average 1.07
Standard
deviation
0.17

Secon 25 x 45 cm, cover 5 cm Curve for fixed

N [kN] neutral axis inclination
fc = 47.6 MPa
EC2 parabolic linear 7 000
fs = 530 MPa.
N [kN] As,tot = 8 x 2.01 cm2 6 000

5 000

4 000

3 000

2 000

1 000

0.000

-1 000
My Mx [kNm] -500 -400 -300 -200 -100 0.00 100 200 300 400 500
[kNm]
Mx [kNm]
(a) (b)

7 000 Curve for fixed

N neutral axis inclination Curve for fixed
[kN] 6 000 neutral axis inclination
-300
5 000 My
[kNm] -200
4 000
-100
3 000
0.00
2 000
100
1 000
200
0.000

-1 000 300
-300 -200 -100 0.0 100 200 300 -500 -400 -300 -200 -100 0.00 100 200 300 400 500

My [kNm] Mx [kNm]
(c) (d)
Fig. 5. Computational procedure for obtaining the failure surface of rectangular cross section with dimensions b = 25 cm  h = 45 cm and total amount of reinforcement
8  2.01 cm2: (a) tridimensional view, (b) projected view on the N–Mx plane, (c) projected view on the N–My plane and (d) projected view on the My–Mx plane.

with total reinforcement As,tot = 24  3.14 cm2, cover equal to 3 cm, curves are shown in Fig. 7 for (compression) axial force values of
concrete strength fcd = 0.85  20.75/1.6 MPa and steel yielding Nsd = 4037 kN, 1439 kN, 140 kN and (tensile) axial force values of
strength fsd = 375/1.15 MPa. Analysis is performed using neutral Nsd = 1809 kN. The results of the present study also compare well
axis angle of inclination increments Dh = 6° and 800 strain states with those obtained by Rosati et al. [8] using a method with secant
for each neutral axis angle of inclination. The moment interaction elastic iterations.
10 R. Vaz Rodrigues / Engineering Structures 104 (2015) 1–17

350
Secon 30 x 70 cm,cover 5 cm Present N=5000 kN
fcd = 20 MPa, EC2 parabolic linear Present N=-750 kN
300 fsd = 500 MPa, As, tot = 40 cm2
Present N = -1500 kN
Present N = 0 kN
250 Papanikolau
Papanikolaou[7]N = 1000 kN
Present N = 1000 kN
Mxy

My[kNm]
200

150

100

50

0
0 100 200 300 400 500 600 700 800 900 1000
Mx [kN m]

Fig. 6. Comparison of moment interaction curves for rectangular section, between present method and Papanikolaou [7].

Present N=4037 kN Present N=1439 Present N=140 kN Present N= -1809 kN

Rosati et al. (N = 4037 kN) Rosati et al. (N = 1439 kN) Rosati et al. (N = 140 kN) Rosati et al. (N = - 1809 kN)

U-shaped Secon,cover 3 cm
fcd = 0.85 x 20.75/1.6 MPa
EC2 parabolic linear
800 fsd = 375 / 1.15 MPa.
Bar diameter 20 mm

300
My[kN]

-200

-700

-1200
-800 -600 -400 -200 0 200 400 600 800
Mx [kNm]

Fig. 7. Comparison of moment interaction curves for U-shaped section, between present method and Rosati et al. [8].

Papanikolaou [7] N= -2 657 kN Present N= -2 657 kN Papanikolaou [7] N = 6 182 kN

Present N = 6 182 kN Papanikolaou [7] N = 879 kN Present N = 879 kN
2000
Gshaped Secon,cover 3 cm
fcd = 0.85 x 20.75/1.6 MPa, EC2 parabolic linear
1500 fsd = 375 / 1.15 MPa. Bar diameter 24 mm

1000

500
My [kNm]

-500

-1000

-1500

-2000
-1000 -800 -600 -400 -200 0 200 400 600 800 1000

M x [kNm]

Fig. 8. Comparison of moment interaction curves for G-shaped section, between present method and Papanikolaou [7].
 R. Vaz Rodrigues / Engineering Structures 104 (2015) 1–17 11

7000
Rodriguez et al. 1999 [12] Hollow Circular Column
Present Reseach fcd = 27.58 MPa
6000
fsd = 413.69 MPa.
Bar diameter 7/8''
5000

4000

3000

N [kN]
2000

1000

-1000

-2000
0 100 200 300 400 500 600
M [kNm]

Fig. 9. Comparison of moment–axial force interaction curve for hollow circular section, between present method and Rodriguez et al. [12].

Papanikolaou [7] N= -5676 kN Papanikolaou [7] N = 34 129 kN Papanikolaou [7] N = 10 246 kN

Present N=10 246 kN Present N=34 129 kN Present N= -5676 kN

20000
MulcellSecon,cover 3 cm
fcd = 0.85 x 20.75 /1.6 MPa
15000 fsd = 375 / 1.15 MPa.
Bar diameter 32 mm

10000

5000
My [kNm]

-5000

-10000

-15000

-20000
-15000 -10000 -5000 0 5000 10000 15000
Mx [kNm]

Fig. 10. Comparison of moment interaction curves for multicell section, between present method and Papanikolaou [7].

7.3. Example 3: G-shaped cross section Aristizabal-Ochoa [12], with external diameter £ext = 609.6 mm
and wall thickness of 254 mm, total reinforcement
The third example shows the response of a G-shaped cross sec- As,tot = 8  3.87 cm2, placed at radius of 244 mm, concrete strength
tion analyzed in a study presented by Papanikolaou [7], with dimen- fcd = 27.6 MPa and steel yielding strength fsd = 413.7 MPa. The
sions b = 1.00 m  h = 0.60 m and wall thickness of 20 cm, with total moment–axial force interaction curve is shown in Fig. 9. A polygon
reinforcement As,tot = 24  4.52 cm2, cover equal to 3 cm, concrete of 16 points is used to approximate the exterior and interior
strength fcd = 0.85  20.75/1.6 MPa and steel yielding strength circular boundaries. The results of the present study compare well
fsd = 375/1.15 MPa. Analysis is performed using neutral axis angle with those obtained by Rodriguez and Aristizabal-Ochoa [12].
of inclination increments Dh = 6° and 800 strain states for each neu-
tral axis angle of inclination. The moment interaction curves are
shown in Fig. 8 for (compression) axial force values of Nsd = 6182 kN, 7.5. Example 5: Multicell cross section
879 kN and (tensile) axial force values of Nsd = 2657 kN. The
results of the present study are almost coincident with those The fifth example shows the response of a multicell cross sec-
obtained by Papanikolaou [7] using stress integration by Green path tion analyzed in a study presented by Papanikolaou [7], with
integrals with adaptive strain-mapped Gaussian sampling. It should
also be noted that small differences may occur when taking into Table 7
account the ‘‘bar holes” as if it was of concrete. Execution times for the full interaction surface (48 800 points).

Section A B C D E F
7.4. Example 4: Circular hollow section Time (s) 145 279 307 295 119 220

A: Rectangular section (Fig. 6); B: U-shaped section (Fig. 7); C: G-shaped section
The fourth example shows the response of a circular hollow (Fig. 8); D: multicell section (Fig. 10); E: design example 1 (Fig. 16); F: design
section analyzed in a study presented by Rodriguez and example 2 (Fig. 17).
12 R. Vaz Rodrigues / Engineering Structures 104 (2015) 1–17

Q 1 = 96.1 kN Q 2 = 18.7 kN

R1 R2

0,20 3,00 3,00 m 2,00 0,20

-20
0
40
80
120
160
200 M [kN·m]
170.9 kN m (Measured at failure)

Fig. 11. Bending moments at failure of beam. Failure caused by tensile failure of reinforcing steel.

dimensions b = 2.50 m  h = 1.40 m and wall thickness of 30 cm, using stress integration by Green path integrals with adaptive
with total reinforcement As,tot = 52  8.04 cm2, cover equal to strain-mapped Gaussian sampling.
3 cm, concrete strength fcd = 0.85  20.75/1.6 MPa and steel From the comparisons presented in this section, it is concluded
yielding strength fsd = 375/1.15 MPa. Analysis is performed using that the methodology proposed in the present study (WAGL)
neutral axis angle of inclination increments Dh = 6° and 800 strain provides accurate results for symmetric and non-symmetric
states for each neutral axis angle of inclination. The moment sections of complex shapes, including consideration of hollow cross
interaction curves are shown in Fig. 10 for (compression) axial sections.
force values of Nsd = 34 129 kN, 10 246 kN and (tensile) axial force Execution times for obtaining the full interaction surface with
values of Nsd = 5676 kN. The results of the present study are 48 800 points are indicated in Table 7, obtained using a system
almost coincident with those obtained by Papanikolaou [7] with processor Intel Core i5 CPU M460 @ 2.53 GHz and installed

4 Ø16

8Ø 12.5 2Ø 10
M 0.020
0.025
0.20 m
0.203

0.45 m α d
0.404 m
2Ø 10
0.10 m
b
0.203
4 Ø16

0.25 m
(a) (b) (c)
Fig. 12. Cross section and reinforcement layout of evaluated tests: (a) Vaz Rodrigues et al. [17]; (b) Ramamurthy [18], A-series; (c) Pallarès et al. [19].
 R. Vaz Rodrigues / Engineering Structures 104 (2015) 1–17 13

7000
M
6000

5000

4000

N [kN]
3000

2000

1000

-1000 Experimental point with

Vaz Rodrigues et al. 2010, N=0; M = 170.9 kNm

-2000
-500 -400 -300 -200 -100 0 100 200 300 400 500
M [kN·m]

Fig. 13. Interaction M–N diagrams for rectangular section, for various inclinations of neutral axis h.

memory RAM 4.00 GB, and implemented using Visual Basic for indicated in Fig. 14 – a representation of moment interaction curves
Applications (VBA), as mentioned before. of the cross section for N = 0 kN.
Secondly, the results from the WAGL technique are compared
with the results provided by Ramamurthy [18]. The cross sections
8. Comparison with experimental data tested by this author are illustrated in Fig. 12(b) and the test results
are indicated in Table 6. Note that Ramamurthy evaluated the
Firstly, the results from the WAGL technique are compared with concrete compressive strength in cubes. Therefore the values to
the experimental values for simple bending. The simple cross adopt for cylinders are considered to be fck = 0.85 fck,cube. The angle
section tested by Vaz Rodrigues et al. [17] is analyzed. The beam a is defined in Fig. 12(b), along with the column width and overall
was tested in two point bending with material properties for depth. The total area of reinforcing steel, the number of provided
compressive concrete strength of fc = 47.6 MPa and reinforcing steel bars and distance to concrete surface d0 are indicated in Table 6,
yield strength fs = 530 MPa. The experimental bending strength of along with the theoretical bending strengths, showing good agree-
the beam is of 170.9 kN m. The spans and bending moment diagram ment with measured values. This fact also results from Fig. 15,
at failure of the beam are shown in Fig. 11. The cross section of the which shows the agreement for each series. It can also be observed
beam is shown in Fig. 12(a). The evaluation of the theoretical ulti- that for the columns of Series E, F, G the agreement is less effective.
mate strength of the beam can be done after the interaction diagram However this lack of agreement was also confirmed by other
(M–N) using N = 0 kN for input. The interaction diagram is shown in authors, as in Pallarès et al. [13].
Fig. 13, corresponding to the curve indicated with h = 0°. The theo- Thirdly, the results from the WAGL technique are also com-
retical ultimate bending strength is 164 kN m using the procedures pared with the experimental results provided by Pallarès et al.
described in this paper. The steel hardening is not considered in the [19]. The results of Table 6 also confirm good agreement with
analysis, therefore k = 1.00. The analysis of the cross section is also the experimental results provided by these authors.
performed using a different angle of inclination of neutral axis h. The average ratio (for all 61 tests) of the bending moment
The experimental point is represented in the failure surface, as obtained through the WAGL technique and the experimental

Present N = 0 Present N = -600 kN Present N = 600 kN Present N=1 200 kN

180
Secon 25 x 45 cm,cover 3.30 cm
160 fc = 47,6 MPa, EC2 parabolic linear
fy = 530 MPa, As, tot = 8 x 2,01 cm2

140

120
My [kNm]

100

80

60

40 Experimental point
Vaz Rodrigues et al. 2010
(Mx = 170.9 kNm)
20

0
0 50 100 150 200 250 300 350 400
Mx [kNm]

Fig. 14. Moment interaction curves for rectangular section and comparison with experimental results.
14 R. Vaz Rodrigues / Engineering Structures 104 (2015) 1–17

80.0 80.0 80.0

70.0
20.3 x 20.3 cm 70.0
20.3 x 20.3 cm 70.0
15.2 x 15.4 cm
8 #4 8 #5 8 #4
60.0 fy = 292 MPa 60.0 fy = 322 MPa 60.0 fy = 276 MPa

M test [kNm]
M test [kNm]

M test [kNm]
50.0 50.0 50.0

40.0 40.0 40.0

30.0 30.0 30.0

20.0 20.0 20.0

10.0 10.0 10.0

0.0 0.0 0.0

0.00 20.00 40.00 60.00 80.00 0.00 20.00 40.00 60.00 80.00 0.00 20.00 40.00 60.00 80.00
M pred [kNm] M pred [kNm] M pred [kNm]
(a) Ramamurthy – series A (b) Ramamurthy – series B (c) Ramamurthy – series C

80.0 80.0 80.0

15.2 x 15.4 cm 15.2 x 22.9 cm 15.2 x 30.5 cm
70.0 70.0 70.0
8 #3 8 #5 8 #5
60.0 fy = 276 MPa 60.0 fy = 322 MPa 60.0 fy = 322MPa

M test [kNm]
M test [kNm]

M test [kNm]

50.0 50.0 50.0

40.0 40.0 40.0

30.0 30.0 30.0

20.0 20.0 20.0

10.0 10.0 10.0

0.0 0.0 0.0

0.00 20.00 40.00 60.00 80.00 0.00 20.00 40.00 60.00 80.00 0.00 20.00 40.00 60.00 80.00
M pred [kNm] M pred [kNm] M pred [kNm]

(d) Ramamurthy – series R (e) Ramamurthy – series D (f) Ramamurthy – series E

80.0 80.0 80.0

15.2 x 22.9 cm 15.2 x 30.5 cm 10 x 20 cm
70.0 70.0 70.0
8 #4 8 #4 4 Ø10
60.0 f y = 292 MPa 60.0 fy = 292 MPa 60.0 fy = 558 MPa
M test [kNm]

M test [kNm]

M test [kNm]

50.0 50.0 50.0

40.0 40.0 40.0

30.0 30.0 30.0
20.0 20.0 20.0
10.0 10.0 10.0
0.0 0.0 0.0
0.00 20.00 40.00 60.00 80.00 0.00 20.00 40.00 60.00 80.00 0.00 20.00 40.00 60.00 80.00
M pred [kNm] M pred [kNm] M pred [kNm]
(g) Ramamurthy – series F (h) Ramamurthy – series G (i) Pallarès
Fig. 15. Comparison between numerical results and experimental data (square and rectangular cross sections) from Ramamurthy [18] and Pallarès et al. [19].

bending moment is 1.07, as indicated in Table 6, which confirm force is Nd = 5000 kN. The material properties at design level are
good agreement. the compressive strength fcd = 14.57 MPa and the yield strength
fsd = 435 MPa (horizontal yield plateau considered). From the
analysis of Fig. 16, it can be seen that the total amount steel,
9. Design examples
105.6 cm2, is adequate to provide the required bending strength.
Finally, the detailing of the bending reinforcement is shown in
The application to design is illustrated by means of two exam-
Fig. 16 and the element itself during the construction of the
ples. The first example consists if a rectangular cross section under
column.
biaxial bending moments and second example consists of a hollow
cross section. In both examples the designer can plot a series of
curves, each curve for a given amount of bending reinforcement. 9.2. Hollow cross section under biaxial loading

9.1. Rectangular cross section under biaxial loading Consider in Fig. 17 a square box section (2.0  2.0 m) with wall
thickness of 0.30 m. The applied bending moments are
Consider the reinforced concrete column of a powerhouse, Msd,x = 10 400 kN m and Msd,y = 10 400 kN m, combined with axial
shown in Fig. 16, with overall depth of 2.20 m and width of compression force of Nd = 10 000 kN. The principal bending
0.80 m. Assume that applied bending moments are moment is therefore Msd = 14 707 kN m, applied on an angle of
Mxd = 6000 kN m, Mysd = 1700 kN m and that the axial compression 45° with respect to the major bending axis. Material properties
 R. Vaz Rodrigues / Engineering Structures 104 (2015) 1–17 15

4000
As,tot = 105.6 cm2 5Ø25
Nsd = 5 000 kN
95.0cm2 (each face)
84.5 cm2
3000
73.9 cm2
63.6 cm2

2000
52.8 cm2 9Ø20
42,2 cm2 2.2 (each
31.7 cm2
face)
21.2 cm2
10.6 cm2
1000
0.0 cm2
My [kNm]

0
0.80

-1000

-2000

-3000

-4000
-10000 -8000 -6000 -4000 -2000 0 2000 4000 6000 8000 10000

Mx [kNm]

Fig. 16. Design example for rectangular cross section (Msd,x = 6000 kN m; Msd,y = 1700 kN m; Nsd = 5000 kN): moment interaction Mx–My curves for different amounts of
reinforcement, cross section details, and column during construction.

20000
As,tot = 105.6 cm2
Nsd = 10 000 kN 95.0 cm2
15000 84.5 cm2
73.9 cm2
63.6 cm2
10000
52.8 cm2
42,2 cm2
31.7 cm2
5000
21.2 cm2
My [kNm]

10.6 cm2
0 0.0 cm2

-5000

-10000

-15000
without strain hardening (k = 1.00)
with strain hardening (k = 1.35)
-20000
-20000 -15000 -10000 -5000 0 5000 10000 15000 20000
Mx [kN·m]

4x13Ø20 (ext)
4x11Ø20 (int) Mx
2.00
M

My
45°

0.30 m

2.00

Fig. 17. Design example for hollow cross section (Msd,x = 10 400 kN m; Msd,y = 10 400 kN m; Nd = 10 000 kN): moment interaction Mx–My curves for different amounts of
reinforcement and cross section details.
16 R. Vaz Rodrigues / Engineering Structures 104 (2015) 1–17

Strain [‰] Stress [MPa]

-2.00 20 MPa
(1.0;2.5)

fiber

(2.0;1.50)

(0.25;0.75)
-0.40
(0.50;0.50) 7.2 MPa
yLocal
neutral axis
xLocal

Fig. 18. Example used for comparing the WAGL technique with the fiber element approach.

 The WAGL technique uses a polygon clipping algorithm, origi-

nally developed in the scope of computer graphics, by Weiler
and Atherton [15]. This was found to be an efficient formulation
0.50 2500 for dividing any cross section into trapezoidal elements.
Number of operaons of  The results from the WAGL technique (using the material prop-
0.45
Fiber element approach erties of concrete and steel from Eurocode 2), show a good
0.40 2000 agreement with the available experimental data.
 The WAGL technique provides a method without need of itera-
Number of operaons

0.35
tion, achieved by evaluating a finite number of algebraic expres-
0.30 Error 1500 sions that are derived in this paper, which constitutes an
Error [‰]

0.25 indisputable advantage. The findings of this paper confirm

those identified by Bonet et al. [20], i.e., integration methods
0.20 1000
based on the Gauss–Legendre quadrature (with decomposition
0.15 of the compression zone into ‘‘thick” layers) are very effective
Number of operaons of
for computing interaction diagrams of concrete section. An
0.10 WAGL technique 500
example is included in Appendix A, comparing results of the
0.05 WAGL technique, with the results of fiber element approach,
which confirms these findings.
0.00 0
0 100 200 300 400 500  The WAGL technique provides an efficient method in the
Number of fibers (along y) laborious task of decomposing the section into trapezoidal
elements. As a consequence, high computational efficiency is
Fig. 19. Comparison between the WAGL technique and the fiber element approach.
obtained, which is considered to improve both flexibility and
productivity in the engineering task of reinforced concrete cross
section design.
at design level, are the compressive strength fcd = 16 MPa and the
yield strength fsd = 400 MPa (horizontal yield plateau considered).
From the analysis of Fig. 17, it can be seen that the total amount Appendix A. Comparison between the WAGL technique and the
steel, 301.4 cm2, resulting from 4  13 + 4  11, 20 mm diameter fiber element approach
bars, is adequate to provide the required inclined bending strength.
The effect of reinforcing steel strain hardening is also analyzed Given the concrete section shown in Fig. 18, the resultant axial
(with k = 1.15) showing a very marginal influence at this level of force N under the indicated strains is calculated considering the
axial compression. proposed technique (WAGL) and the fiber element approach. The
exact value for the axial force NWAGL is (24 822.92 kN). The integral
10. Conclusions is evaluated using 51 operations of sum and 46 multiplications,
which results in 97 operations. The same integral can be evaluated
A new procedure (WAGL) is developed in this study to evaluate using the fiber method approach; in this case the accuracy of the
the bending strength of polygonal sections of arbitrary shape, result depends on the mesh division. The error of the fiber method
including holes, and arbitrary locations of reinforcing bars. The is therefore defined as |NWAGL  NFiber|/NWAGL, and is represented in
following conclusions are achieved: Fig. 19. It can be seen that the fiber method approach converge to
the exact value. The required operations for the fiber method are
 The proposed section subdivision and stress integration tech- also indicated in Fig. 19. By comparing both techniques, it can be
nique (WAGL), based on the change of variables theorem, is seen that, for the same number of operations (97), the WAGL tech-
an effective formulation to provide exact results for the biaxial nique leads to the exact value of the integral, whereas the fiber
bending strength (combined with axial force) of normal method approach leads to a relative error of about 0.5‰, in the pre-
strength (fck 6 50 MPa) reinforced concrete cross sections. sent numerical example.
 R. Vaz Rodrigues / Engineering Structures 104 (2015) 1–17 17

References [12] Rodriguez JA, Aristizabal-Ochoa JD. Biaxial interaction diagrams for short RC
columns of any cross section. J Struct Eng 1999;125(6):672–83.
[13] Pallarès L, Miguel PF, Fernández-Prada MA. A numerical method to design
[1] Bresler B. Design criteria for reinforced concrete columns under axial load and
reinforced concrete sections subjected to axial forces and biaxial bending
biaxial loading. J Am Concr Inst 1960;57(5):481–90.
based on ultimate strain limits. Eng Struct 2009;31:3065–71.
[2] CEN. Eurocode 2: design of concrete structures – Part 1-1: General rules and
[14] De Vivo L, Rosati L. Ultimate strength analysis of reinforced concrete sections
rules for buildings; 2004.
subject to axial force and biaxial bending. Comput Methods Appl Mech Eng
[3] CEB/FIP, manual CEB/FIP on bending and compression. Bulletin no. 141,
1980;166:261–87.
Construction Press; 1982.
[15] Weiler K, Atherton P. Hidden surface removal using polygon area sorting.
[4] Walther R, Houriet B. Design charts for reinforced concrete sections, hollow
Comput Graph 1977;11(2):214–22.
sections. Ecole Polytechnique Fédérale de Lausanne; 1980.
[16] Lam N, Wilson J, Lumantarna E. Force–deformation behavior modeling of
[5] Hsu C-TT. Analysis and design of square and rectangular columns by equation
cracked reinforced concrete by EXCEL spreadsheets. Comput Concr 2011;8
of failure surface. ACI Struct J 1988;85(20):167–78.
(1):43–57.
[6] Bentz EC. Sectional analysis of reinforced concrete members [PhD thesis].
[17] Vaz Rodrigues R, Muttoni A, Fernández Ruiz M. Influence of shear on rotation
University of Toronto; 2000.
capacity of reinforced concrete members without shear reinforcement. ACI
[7] Papanikolaou VK. Analysis of arbitrary composite sections in biaxial bending
Struct J 2010;107(50):516–25.
and axial load. Comput Struct 2012;98(99):33–54.
[18] Ramamurthy LN. Investigation of the ultimate strength of square and
[8] Rosati L, Marno F, Serpieri R. Enhanced solution strategies for the ultimate
rectangular columns under biaxial eccentric loads. American Concrete
strength analysis of composite steel–concrete sections subject to axial force
Institute; 1966. p. 263–98 [Publ. 13].
and biaxial bending. Comput Methods Appl Mech Eng 2008;197:1033–55.
[19] Pallarès L, Bonet JL, Miguel PF, Fernández Prada MA. Experimental research on
[9] Charalampakis AE, Koumousis VK. Ultimate strength analysis of composite
high strength concrete slender columns subjected to compression and biaxial
sections under biaxial bending and axial load. Eng Struct 2008;39:923–36.
bending forces. Eng Struct 2008;30:1879–94.
[10] Sfakianakis MG. Biaxial bending with axial force of reinforced, composite and
[20] Bonet JL, Barros MHFM, Romero ML. Comparative study of analytical and
repaired concrete sections of arbitrary shape by fiber model and computer
numerical algorithms for designing reinforced concrete sections under biaxial
graphics. Adv Eng Softw 2002;33:227–42.
bending. Comput Struct 2006;84:2184–93.
[11] Dias da Silva V, Barros MHFM, Júlio ENBS, Ferreira CC. Closed form ultimate
strength of multi-rectangle reinforced concrete sections under axial load and
biaxial bending. Comput Concr 2009;6(6):505–21.