You are on page 1of 11

Engineering Structures 30 (2008) 3003–3013

Contents lists available at ScienceDirect

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

Proportioning of steel beam–column members based on RSD


optimization methodology
Luisa María Gil-Martín a , Mark Aschheim b , Enrique Hernández-Montes a,∗
a University of Granada, Campus de Fuentenueva, 18072 Granada, Spain
b Santa Clara University, 500 El Camino Real, Santa Clara, CA, United States

article info a b s t r a c t

Article history: Due to the cost and energy embodied in steel production, a reduction in the cross-sectional area of a steel
Received 11 February 2008 structural member could imply a significant saving of money and reduction in greenhouse gas emissions.
Received in revised form Hernandez-Montes et al. [Hernández-Montes E, Aschheim M, Gil-Martin LM. The impact of optimal
2 April 2008
longitudinal reinforcement on the curvature ductility capacity of reinforced concrete column sections.
Accepted 3 April 2008
Available online 20 May 2008
Mag Concrete Res 2004;56(9):499–512; Hernández-Montes E, Gil-Martín LM, Aschheim M. The design of
concrete members subjected to uniaxial bending and compression using reinforcement sizing diagrams.
Keywords:
ACI Struct J 2005;102(1):150–8] proposed an analytical approach for reducing the amount of longitudinal
Steel structures reinforcement in reinforced concrete members subjected to the action of axial force and bending moment
Structural optimization acting about a principal axis of the cross-section. The approach, which makes use of Reinforcement
Sizing Diagrams (RSD), makes use of a graphical representation of all possible reinforcement solutions
for a particular concrete cross-section subjected to a combined loading consisting of bending moment
and axial load (M, N). The common symmetric solution is recognized as just one of the infinite number
of possible solutions. The RSD methodology used in that approach is extended in the present paper to
steel sections having at least one axis of symmetry, subjected to axial force, bending moment about the
strong axis, and shear acting in the plane of the bending moment. Special considerations are introduced to
address the instabilities associated with slender steel elements. The methodology is developed following
the Eurocode 3 provisions for compact steel members.
© 2008 Elsevier Ltd. All rights reserved.

1. Introduction shear force, V , acting in the plane of the moment, and an axial force,
N, considered to be applied at the centroid of the rectangular web of
Structural members may be present as individual members the section, and directed along the longitudinal axis of the member
or as a part of a sub-assembly of a more complex structural (see Fig. 1). Beam–columns subjected to torsion or biaxial bending
framework. Members of indeterminate structures interact with are not considered in the present work. Furthermore, this study
one another, with respect to redistribution of loads and the is restricted to compact sections (i.e. Class 1 sections per EC3 [3])
development of instabilities. Beam–column members, which that are symmetric about the minor principal axis of the section
are subjected simultaneously to bending and axial force, occur (Fig. 1). The section is proportioned to provide sufficient strength
to resist these actions (M, N, V ) and sufficient stiffness to prevent
frequently in typical steel structures. For this reason the behaviour
premature buckling. An additional constraint, relating to the stiff-
and design of such members has been central to steel design.
ness necessary to limit deflections to acceptable levels, may also be
Beams constitute a special case, where the axial load (N) is
imposed.
negligible or zero, and are considered to be a subset of the more
It is common in concrete structures to have longitudinal
general beam–column design problem.
reinforcement arranged symmetrically in the cross-section. In
The present paper considers the optimal design of beam–column fact, the design of longitudinal reinforcement is often made
members in which external loading causes an in-plane bending with the assistance of N–M interaction diagrams, which generally
moment, M, acting about the strong axis of the cross-section, a are presented only for symmetric reinforcement. However, it
is evident from RSD design approaches that in some cases it
is feasible and economically advantageous to use asymmetric
∗ Corresponding author. Tel.: +34 958 249965; fax: +34 958 249959. reinforcement distributions. Fig. 2 illustrates this for an example
E-mail addresses: mlgil@ugr.es (L.M. Gil-Martín), maschheim@scu.edu described in Hernandez-Montes et al. [1,2]. RSD methodology
(M. Aschheim), emontes@ugr.es (E. Hernández-Montes). consists in the consideration of all the possible solutions, for
0141-0296/$ – see front matter © 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2008.04.004
3004 L.M. Gil-Martín et al. / Engineering Structures 30 (2008) 3003–3013

Fig. 1. Nomenclature.

a design problem, through the representation called RSD, as


illustrated for a reinforced concrete section in Fig. 2.
In steel structures, doubly symmetric cross-sections are fre-
quently used, although singly symmetric cross-sections are being
used for columns to achieve economy in large warehouses. This ar-
ticle extends the RSD technique to steel construction to provide a
more precise and useful design procedure, and demonstrates that
reductions in the area of steel necessary to resist a given combi-
nation of axial load and moment (N, M) may be achieved with an
RSD-type optimization approach. This new approach allows the
engineer to choose among all the possible solutions, considering
minimum weight, availability of steel shapes, simplicity on the job
site, and so on. Fig. 2. Example of reinforcement solution for a rectangular cross-section with
top and bottom reinforcement (A0s and As respectively). (a) Reinforcement Sizing
Fig. 1 shows the basic nomenclature for the cross-section. The
Diagram for flexure according to ACI-318, for φNn = 2500 kN and φMn =
web, having thickness tw and height, d, connects to flanges having 1000 kN m. (b) Strain diagrams for symmetric reinforcement (point A) and optimal
areas A1 at the top and A2 at the bottom. The moment, M, is reinforcement (point B).
considered positive when the top flange in the cross-section is
in compression and the axial force, N, is considered positive in
tension. To simplify the equations, the fillets in rolled sections
and throat thickness in welded sections have been ignored in the
derivations. The restriction that all sections are compact (Class 1
sections per EC3 [3]) implies that the full plastic capacity of the
cross-section can be developed; that is, local buckling will not
occur.
Fig. 3 presents the flow chart for the design procedure described
in this paper.
As was observed with reinforced concrete, an infinite number
of solutions exist for the design of a steel cross-section subjected
to combined loads N and M. These solutions can be presented
using graphics similar to those used in the reinforced concrete RSD
representation (Fig. 2).
The extension to steel members requires consideration of
buckling modes. The buckling modes considered herein are
identified in Table 1. For members in compression, two types of
bucklings are considered: flexural buckling (i.e. Euler buckling
for slender members and inelastic buckling for stocky members)
and torsional-flexural buckling. In the former, the cross-section
does not rotate about the longitudinal axis; in the latter, rotation
occurs. For members bending about the strong axis, the instability
phenomena considered is lateral-torsional buckling.
The EC3 [3] approach for checking buckling of members
subjected to axial compression and bending known as the General
Fig. 3. Flow chart of the proposed method.
Method was adopted for the developments described in this article.
Specifically, the conditions considered are single-span members avoided by enforcing constraints on web dimensions in the design
of constant cross-section under axial compression, with end-fork process outlined herein.
conditions (i.e. pin supported without warping restraint).
Other potential buckling modes involve buckling of the web due 2. Constraints on web dimensions
to shear and local buckling of plates loaded – totally or partially – in
compression. The restriction that all sections are compact implies The analytical approach developed herein begins with an
that the full plastic capacity of the cross-section can be developed; assumed web thickness, tw . Given tw , the height of the web, d, is
that is, local buckling will not occur. Shear buckling of the web is limited as follows:
L.M. Gil-Martín et al. / Engineering Structures 30 (2008) 3003–3013 3005

Table 1
Buckling modes considered
Members in compression (M = 0, - · - · - buckling axis) Members in bending about the Members subjected to shear Plates under compression
strong axis
Lateral buckling (N = 0, - · - · - buckling axis)
Flexural buckling Torsional-flexural Lateral-torsional buckling Buckling shear Local buckling
buckling

Avoided using a web thick In order to force Class 1


enough (cross-section can form a
plastic hinge with the capacity
required from plastic analysis
without reduction of the
resistance) the
width-to-thickness ratios of
plates in cross-section
(compressed flange and web)
are limited

Members in bending and axial compression: Interaction formulation


Item 6.3.1.3 Item 6.3.1.4 Item 6.3.2 Section 5 Table 5.2
Eurocode 3. Part 1-1. Eurocode 3. Part 1-1. Eurocode 3. Part 1-1. Eurocode 3. Part 1-5. Eurocode 3. Part 1-1.

1. Shear strength: To provide sufficient shear strength, EC3 [3] expressions that may be used to determine the flange areas, A1 and
provisions for shear strength require that: A2 :

V d
d≥ √ = dmin . (1) A1 · fyf · d + M − N · =0
tw · fyw / 3 2
(3)
d
2. Shear buckling: To prevent shear buckling, the slenderness of A2 · fyf · d + M + N · = 0.
2
the web has to be limited. EC3 [4] (Section 5.1(2) of part 1-5) Once preliminary values of A1 and A2 are obtained using Eqs. (3),
provisions require, where webs are not stiffened, that then the width of the flanges may be determined as follows.
d/tw 0.83 Depending on the ratio of axial load to moment, one or both
λ̄w = √ < flanges may be in compression. The aspect ratio (or equivalently,
37.4 · ε · kτ η
flange slenderness, b/tw ) of the compression flange(s) is limited to
or equivalently, prevent local buckling of the flange in the plane of the web. EC3 [4]
expressions for local buckling of the flange (Section 8 of Eurocode
0.83
< · 37.4 · ε · kτ · tw = d1 3, Parts 1–5) are used to obtain an upper limit for the thickness of
p
d (2)
η the compression flange in relation to the thickness of the web and
where kτ = the shear buckling coefficient for the web, η = the the height of the web, given that A1 = bf c · tf c :
strain hardening coefficient and ε is given by the following s
d E d · tw k2 · E2 · tw
3
expression ≤k ⇒ tf c ≤ (4)
s tw fyf A1 fyf2 · bf c · d
235
ε= . where E = 210,000 N/mm2 and k = 0.3 for compression flange in
fy (N/mm2 ) Class 1.
An accepted value of η, equal to 1.2, is used herein. Assuming Once the thickness of the top flange, tf c , and bottom flange, tf t ,
that there are web stiffeners only at supports results in kτ = are established, the corresponding widths are easily determined,
given the required areas A1 and A2 from Eqs. (3).
5.34.
A1 A2
Eurocode 3 limits the height of the web for Class 1 members; the bf c = and bf t = . (5)
tf c tf t
maximum height of the web, d2 , is discussed in Table 5.2 of EC3 [3].
Thus, the height of the web, d, is limited to dmin ≤ d ≤ dmax In the procedure described below, tf c is set equal to tf t , thus
where dmax = min(d1 , d2 ), d1 is given by Eq. (2), d2 is given by Table eliminating one variable.
Since the cross-section is compact (Class 1), the width of the
5.2 of EC3 [3] and dmin is given by Eq. (1).
compression flange is limited to
bf c ≤ 18 · ε · tf c + tw (6)
3. Preliminary proportioning of flanges
as described in Table 5.2 of Part 1-1 of EC3 [3].
In the remainder of the paper, the flanges are assumed to have
This section describes simple approximations that are used to identical thickness (tf c = tf t = tf ) and values of bf c and bf t obtained
establish preliminary proportions of the flanges. Equilibrium of from Eqs. (5) are rounded up to the nearest integer (in millimeters).
forces acting on the cross-section is established for preliminary Cross-sections whose neutral plastic fiber (NPF) is located
proportioning throughout the method assuming that axial load is outside the web when subjected to conditions of pure flexure
applied at the centroid of the rectangular web (Fig. 1). In order (i.e., for N = 0, the neutral plastic fiber, which divides the cross-
to determine the preliminary proportions of the flanges, as a first section area into two equal areas, is located in a flange) are not
approximation, the flexural resistance of the web is ignored and considered further. Then, with reference to Fig. 1,
the forces carried by the top and bottom flanges are assumed to act
bf c · tf + tw · (d + 2 tf ) − bf t · tf
at the ends of the web. Thus, assuming that the flanges are yielding, tt < < tf + d. (7)
2 · tw
summation of moments about either ends of the web results in
3006 L.M. Gil-Martín et al. / Engineering Structures 30 (2008) 3003–3013

4. Treatment of instabilities at member level where

zg = za − z0
4.1. Lateral buckling
(y2 + z2 )
Z
zj = z0 − 0.5 z dA
The structural member can freely deflect about both principal A Iy
axes of the cross-section. As in EC3 [3], axis y–y is the strong one
and z0 is the distance to the shear center from the centroid of
and axis z–z is the weak one.
the section, za is the location of the point of load application with
The design buckling resistance of a compression member is
respect to the centroid of the section, k and kw are effective length
established in EC3 [3] as:
factors, C1, C2 and C3 are factors depending on the loading and end
χ · A · fy restraint, It is the torsion constant, Iw is the warping constant, L is
Nb,Rd = (8) the length of the member between lateral restraints and Iz is the
γM 1
moment of inertia along weak axis z–z.
where A is the cross-section area, fy is the yield strength, χ is the
reduction factor for pure axial compression corresponding to the
4.3. Combined lateral and lateral-torsional buckling
relevant buckling mode and γM1 = 1.0.
Depending on the restraint conditions, the relevant mode of
To verify the resistance to combined lateral and lateral-
buckling for doubly symmetric cross-sections may correspond to
torsional buckling for a prismatic member in bending and axial
weak axis buckling, for which
compression the General Method proposed by EC3 [3] is used, as
π2 EIz described in this section.
Ncr = Ncr,z = .
`2 The interaction formulae for instability effects for structural
components subject to compression and uniaxial bending in the
In the case of a singly symmetric section, the relevant buckling
plane is given by:
mode corresponds to torsional-flexural buckling, obtained from
the non-dimensional slenderness defined as (Kaim [6]): NEd My,Ed
+ ≤1 (11)
s
A fy
χ · NRk /γM1 χLT · My,Rk /γM1
λ=
Ncr,FT where γM1 is the partial safety factor for the building (γM1 =
1.0), NEd is the design normal force, My,Ed is the design bending
where moment about the y–y axis, NRk is the characteristic resistance
 v 
!2 ! to normal force of the critical cross-section, equal to Npl (full
z02
u
Ncr,z Ncr,T u Ncr,T Ncr,T 
Ncr,FT =  1 +

− t
1− +4 plastic axial force) for Class 1 members, My,Rk is the characteristic
i20

2

z0 Ncr,z Ncr,z Ncr,z moment resistance of the critical cross-section about y–y axis,
2 1− i2
0
equal to Mpl,y (full plastic bending moment about the strong axis)
π2 E Iw for Class 1 members, χ is the relevant reduction factor for pure axial
!
1
with Ncr,T = G It + and i2o = i2y + i2z + zo2
i20 L2 compression, and χLT is the reduction factor for lateral-torsional
buckling, defined previously in Sections 4.1 and 4.2, respectively.
where variables iy and iz are the radii of gyration about strong and
weak axes, respectively, and zo is the distance between the centroid
5. Cross-section strength
and the shear center of the section. For the determination of χ in
Eq. (8) the buckling curve associated with weak axis bending, given
Since only Class 1 sections are considered in this paper, the
in Section 6.3.1.4 of EC3 [3], may be used.
plastic capacity is of interest. Where shear force and axial force are
present, allowance should be made for the effect of both shear and
4.2. Lateral-torsional buckling axial force on the resistance moment. However, EC3 [3] establishes
that this effect can be neglected where the shear force is less than
Design of buckling resistance for a bending member under half the plastic shear capacity (Vpl,Rd ). For larger shear forces, the
lateral-torsional buckling is defined in general in Section 6.3.2 of design resistance of the cross-section to combinations of moment
EC3 [3] as: and axial force should be calculated using a reduced yield strength
χLT · Wpl,y · fy for the shear area, (1 − ρ)fyw , where
Mb,Rd = (9)
γM1 !2
2 · VEd fyw
where Wpl,y is the plastic section modulus for bending about
ρ= −1 and Vpl,Rd = (d · tw ) · √ (12)
Vpl,Rd 3
y–y axis, fy is the yield strength, χLT is the reduction factor for
lateral-torsional buckling and γM1 = 1.0. For a cross-section of where an axial force is present, allowance should be made for its
a statically determinant beam, the value of χLT depends on the effect on the plastic moment resistance (EC3 [3]). According to
relationship between the plastic moment and the critical moment EC3 [3], for doubly symmetric I- and H-sections allowance need not
for elastic lateral-torsional buckling, Mcr (or squared dimensionless to be made for the effect of the axial force on the plastic resistance
slenderness ratio). moment about the y–y axis when both the following criteria are
For a cross-section symmetric about the minor axis z–z, Mcr is satisfied:
given as (annex F EC3 [5]): 0.5 · d · tw · fy
NEd ≤ 0.25 · Npl,Rd and NEd ≤ . (13)
π2 E · Iz (kL)2 G · It γM0
(" 2
k Iw
Mcr = C1 +
(kL)
2 kw 2 Iz π E · Iz Otherwise it is necessary to account for the effect of axial force on
#1/2  flexural strength. Cases, in which the axial force causes a reduction

+ C2 · zg − C3 · zj
2 
− C2 · zg − C3 · zj

(10) in flexural strength, as illustrated in Fig. 4, are considered in this
 section.
L.M. Gil-Martín et al. / Engineering Structures 30 (2008) 3003–3013 3007

Fig. 4. Different cases where axial force causes a reduction in flexural strength.

The reduced plastic moment capacity, taking into account the


influence of the axial load N on the plastic bending strength, is (see
Fig. 5):
tf tf
    
MRd = fyf bf c · tf · d + tf + − PNA + bf t · tf · PNA −
2 2
 1
 
+ , fyw · ρ · tw · d + tf − PNA − y · · d + tf − PNA + y

2
 1
 
+ fyw · ρ · tw · PNA − tf − y · · PNA − tf + y .

(17)
2
Case b: PNA + |y| ≤ d + tf and PNA − |y| < tf
Values of y1 and δt indicated in Fig. 4(b) can be obtain from the
Fig. 5. Example corresponding to pure flexure (triangular moment diagram). system of two equations:

N − fyw · ρ · tw · PNA − tf + y1 − fyf · δt · bf t = 0



To account for the effect of axial force on flexural strength, an
 !
auxiliary parameter y is defined (Chen [7]) as (see Fig. 4(a)): y21 PNA − tf 2
fyw · ρ · tw · − − fyf · δt · bf t
2 2 (18)
N
y= . (14) δt
2 · tw · fyw · (1 − ρ)
 
· PNA − tf + = 0.
2
This parameter represents the portion of the web in the vicinity of
the plastic neutral axis that is considered to resist the axial force, N, Now, the α parameter, defined in Table 5.2 of EC3 [3], is given as a
function of N by
where fyw is the specified minimum yield strength for the steel of
the web, tw is the thickness of the web and ρ is the factor defined tf + d − PNA − y1
in Eq. (12). Thus, determination of the flexural strength depends If N ≥ 0 (tension) ⇒ α =
d (19)
on where the neutral axis is relative to the flanges. Four cases are If N < 0 (compression) ⇒ α = 1.
considered:
Then, the maximum value of web height for a Class 1 section is
Case a: PNA + |y| ≤ d + tf and PNA − |y| ≥ tf
In this case the portion of the web in compression, α (Table 5.2 396 ε
 
If α > 0.5 ⇒ d ≤ · tw 

 
α

EC3 [3]) can be obtained from: For N ≥ 0 ⇒ 13 · − 1 .
36 ε (20)
If α ≤ 0.5 ⇒ d ≤ · tw

 

tf + d − PNA − y α

α= . (15)
d For N < 0 ⇒ d ≤ 33ε · tw
The maximum web height for a Class 1 section depends on α as and the plastic moment capacity accounting for N can be obtained
follows: as (Fig. 4(b))
396 ε tf
  
If α > 0.5 ⇒ d ≤ · tw . − PNA + bf t · tf − δt

MRd = fyf bf c · tf · d + tf +
13 · α − 1 2
(16)
36 ε 1
 
If α ≤ 0.5 ⇒ d ≤ · tw . × PNA − · ttf − δt + fyw · ρ · tw

α 2
3008 L.M. Gil-Martín et al. / Engineering Structures 30 (2008) 3003–3013

 1
 
6. Design refinement
.

× d + tf − PNA − y1 · · d + tf − PNA + y1 (21)
2
In the preliminary design we had assumed that the axial force
Case c: PNA + |y| > d + tf and PNA − |y| > tf
was applied at the center of the web. The section dimensions
Values of y2 and δc in Fig. 4(c) can be obtain from the system: may now be refined to recognize that the axial force is applied
at the center of gravity of the cross-section. The center of gravity
N − fyw · ρ · tw · d + tf − PNA + y2 − fyf · δc · bf c = 0

is located a distance h from the lowest fiber of the cross-section
2 !
y22 d + tf − PNA (Fig. 1).
fyw · ρ · tw · − We may define e as the distance between the center of gravity
2 2 (22)
and the initial point of application of axial force (at the center of
δc
 
− fyf · δc · bf c · d + tf − PNA + = 0. the web). Then
2
d
 
Then α (Table 5.2 EC3 [3]) is e=h− + tf . (30)
2
If N > 0 (tension) ⇒ α = 0 Therefore, the bending moment at the center of gravity of the
tf + d − PNA − y2 (23) cross-section, Mg , is
If N ≤ 0 (compression) ⇒ α = .
d Mg = M + N · e. (31)
The maximum value of web height of a Class 1 section is: To provide a cross-section of minimum cross-sectional area
that satisfies the preceding constraints relating to strength and
396 ε
 
If α > 0.5 ⇒ d ≤

 · tw 
 stability, the following procedure is adopted:
13 · α − 1

For N ≤ 0 ⇒
36 ε
.
+ χLT ·Myy,,RkEd/γM1 < 0.95 the section provides
MRd NEd M
(24) 1. If ≥ 1 and
If α ≤ 0.5 ⇒ d ≤ · tw Mg χ·NRk /γM1

 

α

excess capacity. To reduce the cross-sectional area, the widths
For N > 0 ⇒ no limit of both flanges are incrementally reduced until:
MRd NEd My,Ed
and the reduced flexural strength accounting for N is (Fig. 4(c)) ≥ 1 and 0.95 ≤ + ≤ 1.
Mg χ · NRk /γM1 χLT · My,Rk /γM1
tf + δc
"  !
MRd = fyf bf c · tf − δc · d + tf + < 1 or > 1 the section has
 MRd NEd My,Ed
− PNA 2. If Mg χ·NRk /γM1 + χLT ·My,Rk /γM1
2
# inadequate strength or is governed by instability. To provide
1

+ bf t · tf · PNA − · tf sufficient strength, the flange areas must be increased. The axial
2 force and bending moment determine the approach to increase
flange area, as follows.
 1
 
+ fyw · ρ · tw · PNA − tf − y2 · · PNA − tf + y2 . (25)

2 (a) If M = 0 or N = 0, the section is symmetric (Eqs. (3))
and areas of both flanges are increased incrementally by the
Case d: PNA + |y| > d + tf and PNA − |y| < tf same amount.
Values of δt and δc in Fig. 4(d) can be obtained from the system (b) If M 6= 0 and N 6= 0, then one of the flange areas is
of equations: increased as applicable in order to reduce the eccentricity
given in Eq. (30) and the corresponding moment given in
N − fyw · ρ · tw · d − fyf · δc · bf c + δt · bf t = 0 Eq. (31). If the axial force is tensile (N > 0), the area of


1 2  the top flange (which is in compression) is increased. If


fyw · ρ · tw · d + tf − PNA − (PNA − tf ) + fyf · δc · bf c
2
(26) the axial force is compressive (N < 0), the area of the
2

δc
 
δt
 bottom flange (which is in tension) is increased. Since the
× d + tf − PNA + − fyf · δt · bf t · PNA − tf + = 0. objective is to minimize the cross-sectional area, it may be
2 2
possible to reduce the area of one flange when the other
Then, α is given as a function of N in Table 5.2 EC3 [3] as one is increased. Consequently, there is a large family of
solutions, in which a doubly symmetric cross-section is only
If N > 0 (tension) ⇒ α = 0. one possible solution.
(27)
If N < 0 (compression) ⇒ α = 1. The iterative refinement of the flange areas can stop when:

The maximum web height of a Class 1 section is MRd NEd My,Ed


≥ 1 and 0.95 ≤ + ≤ 1.
Mg χ · NRk /γM1 χLT · My,Rk /γM1
For N > 0 ⇒ d ≤ 33 · ε · tw .
For each value of d a large number of different possible solutions
For N = 0 ⇒ d ≤ 72 · ε · tw . (28) exist. Thus, the designer may select the most economical cross-
For N > 0 ⇒ no limit. section, considering total area and perhaps other factors such
as the discrete sizes of component materials and the ease of
The reduced flexural strength accounting for N is (Fig. 4(d)) fabrication.
In the event that several different solutions are found that have
tf + δc
"  !
MRd = fyf bf c · tf − δc · d + tf + the same minimum area, we select as optimal the solution having

− PNA
2 the minimum value of the interaction factor (Eq. (11)), defined as:
#
1

+ bf t tf − δt · PNA −

tf − δt

. (29) NEd My,Ed
2 ζ= + . (32)
χ · NRk /γM1 χLT · My,Rk /γM1
L.M. Gil-Martín et al. / Engineering Structures 30 (2008) 3003–3013 3009

Fig. 6. Flange area solutions: (a) several flange thicknesses; (b) flange thickness of
19 mm.

7. Examples

7.1. Strong axis bending

The first example is a simply supported beam with end-fork


conditions (i.e. pin supported and free (unrestrained) warping)
and subject to a triangular bending moment diagram (ψ = 0), as
shown in Fig. 5. The span is 4 m and the peak bending moment is
M = 407 KN m. This moment corresponds to the lateral-torsional
buckling moment of H-shaped HEB300 made from Grade 235 steel
Fig. 7. Cross-section area and width of flanges; bending moment.
(fy = 235 N/mm2 ), for which the total cross-sectional area is
14,910 mm2 and section dimensions are bf c = bf t = 300 mm, tw = solution for the standard HEB300 shape is identified in Fig. 6(b), and
11 mm, tf c = tf t = 19 mm, and d = 262 mm. clearly is just one of the many possible solutions.
Design solutions were obtained for different web depths: d. The total cross-sectional area for the preceding solutions is
Fig. 6(a) shows values of flange areas for a doubly symmetric cross- plotted in Fig. 7(a) and (b). Fig. 7(a) presents results for flange
section obtained from Eqs. (3) with A1 = A2 as a continuous curve. thicknesses, tf , between 3 and 22 mm and Fig. 7(b) presents the
Individual points on the plot of Fig. 6(a) correspond to refined subset of results for tf equal to 19 mm. The minimum cross-
values, obtained using the method of Section 6. sectional area is identified by Point A in these figures. In Fig. 7(a)
During the refinement process, the flange dimensions are the minimum cross-sectional area is 10,648 mm2 and corresponds
iteratively adjusted to satisfy all considerations as described to d = 550 mm, tf c = tf t = 11 mm, and bf c = bf t = 209 mm.
previously (i.e. Class 1 cross-section, strength and instabilities). The In Fig. 7(b), the minimum cross-sectional area is 11,701 mm2 and
refinements were completed for flange thicknesses in increments corresponds to tf c = tf t = 19 mm, d = 511 mm and bf c =
of 2 mm and web heights in increments of 5 mm. For each bf t = 160 mm. The point corresponding to HEB300 is identified in
combination of flange thickness and web height, the required Fig. 7(b), where a significant saving of cross-sectional area (about
21.5 %) is readily apparent. Fig. 7(c) compares flange widths for the
flange area was determined, and this flange area corresponds to a
subset of solutions for which tf c = tf t = 19 mm.
particular value of flange width. For the special case of this example
For members governed by strength rather than instability, it
(N = 0) doubly symmetric sections are optimal solutions, bf =
is obvious that a slender section (such as an I shape) is more
bf c = bf t . Values of flange thickness, tf = tf c = tf t varied from 3 mm
economical than a stocky section, such as an H shape. The
to 2 · tw (with tw = 11 mm) and web heights, d, varied from 10 mm numerical results indicate that for this case, even where potential
to dmax . instabilities are considered, the slender shape is more economical
A subset of these results, corresponding to flange thickness than a stocky one.
tf = 19 mm, is examined in Fig. 6(b). One may appreciate
that the initial approximation from Eqs. (3) (given by the solid 7.2. Pure compression
line) overestimated the dimensions of the flange because the
contribution of the web was ignored. During the refinement The simply supported element considered in the previous
process this error was eliminated. The point corresponding to the example is considered here subject to a concentric compressive
3010 L.M. Gil-Martín et al. / Engineering Structures 30 (2008) 3003–3013

Fig. 9. Interaction equation corresponding to IPE500, for fy = 235 N/mm2 , lb =


4 m, and ψ = 0.

was satisfied, taking into account various limitations for Class 1


sections.

7.3. Simultaneous compression and bending moment about strong


axis

A standard I-shaped IPE500 (fy = 235 N/mm2 ) shape


was adopted as a benchmark for applying this optimization
methodology to the case of simultaneous axial compression and
strong axis bending. The length of the element is 4 m and it
is considered to be simply supported with end-fork conditions
(warping unrestrained). The bending moment diagram has a
triangular shape (ψ = 0). Lateral and lateral-torsional bucklings
are considered.
The General bending moment-axial force interaction diagram
for an IPE500 shape (Eq. (11)) is plotted in Fig. 9. Two combinations
of load will be studied, represented by points D and H in Fig. 9.
The axial compression force and bending moment corresponding
to these points are:

Point D: N = −483 kN & My = 288 kN m


Point H: N = −1130 kN & My = 112 kN m.

The axial force is negative because it is compressive, while the


moment is positive because it causes compression in the upper
Fig. 8. Pure compression (tf = 19 mm).
flange.
Fig. 10(a) shows, for each value of the height of the web, top and
axial force, N, equal to −2.73 × 106 N. This load corresponds
bottom flange areas obtained from Eqs. (3) (A1 and A2 ) and the total
to the buckling capacity of the H-shaped HEB300 (bf c = bf t =
area of the cross-section (A1 + A2 + d · tw ), for load corresponding
300 mm, tw = 11 mm, tf c = tf t = 19 mm, d = 262 mm) having
to Point D in Fig. 9. In this example, the thickness of web and
Grade 235 steel (fy = 235 N/mm2 ) and lb = 4 m.
flanges of the IPE500 shape were adopted (tw = 10.2 mm and
Since M = 0, flanges’ areas determined from Eqs. (3) are not
tf c = tf t = 16 mm). The web depth, d, is considered in increments
influenced by d, the height of the web. So, initial values of flange
dimensions of flanges (Eqs. (3)) represent flange areas (A1 = A2 of 5 mm.
in Eqs. (3)) that are invariant with the height of the web, and thus Fig. 10(b) plots flange areas and total cross-section area as a
plot as a horizontal line in Fig. 8(a). Points corresponding to refined function of the web height, d. The results reflect the refinements
solutions for the subset of results corresponding to tf c = tf t = described in Section 6, addressing requirements for Class 1
19 mm, are indicated in Fig. 8(a). sections, strength and instabilities including interaction. The
Fig. 8(b) shows the variation of total cross-sectional area as minimum cross-section area corresponds to Point A (10,973 mm2 ),
a function of the depth of web, d. Points in this figure are for which the dimensions are d = 395 mm, bf c = 220 mm and
identified corresponding to minimum area (Point A, corresponding bf t = 214 mm. Point B of Fig. 10(b) identifies a cross-section
to 13,300 mm2 ) and the standard HEB300 section. In this example approximately equivalent to IPE500 (11,034 mm2 ) for which the
the cross-sectional area can be reduced by about 10%. Fig. 8(c) dimensions are d = 470 mm, bf c = 217 mm, and bf t = 173 mm.
plots the flange width as a function of d. Point A corresponds to Fig. 10(c) plots the flange width as a function of d after the
bf c = bf t = 306 mm, d = 152 mm. Dimensions of cross-section, refinements of Section 6.
obtained from Eqs. (3) for thickness of flanges equal to 19 mm, are Fig. 10(d) plots other possible solutions for point D in Fig. 9
inadequate because of buckling (ζ > 1) while in this example the under the additional constraint that the cross-section be doubly
refinement process results in ζ ≤ 1 for all sections (the element symmetric. In this case, the minimum cross-section area is
will not buckle under the applied load). In this case the refinement 10,983 mm2 and occurs for d = 396 mm and bf c = bf t = 217 mm.
process required an increase in both flange areas until Eq. (11) In this figure, a point corresponding to IPE500 is also shown.
L.M. Gil-Martín et al. / Engineering Structures 30 (2008) 3003–3013 3011

Table 2(a)
Summary of results for Point D in Fig. 9
d bf c bf t Area Class cross-section Verification of resistance (EC3)
(mm) (mm) (mm) (mm2 ) [C1: Class 1; C2: Class 2; C3: Class 3; C4: Class 4]

IPE500 468 200 200 11,173.6 Compression flange: 200−210.2 = 94.9 ≤ 9 · 16 = 144 ⇒ C1 NEd = 4.83 × 105 N ≤ Min{0.25 · Npl = 6.56 × 105 ,
Web: 0.5 · d · tw · fy = 5.61 × 105 } = 5.61 × 105 N,
486
N (M = 0): 10 .2 = 45.9 > 42 ⇒ C4 ⇒ ρ = 0.9 (the effect of axial force on the plastic moment of
486
M (N = 0): 10 .2 = 45.9 ≤ 72 ⇒ C1
resistance does not need to be taken into account)
486 396 My,Ed = 2.88 × 108 N mm ≤ My,Rd = 4.95 × 108 N mm
N + M : α = 0.72 ⇒ 10 .2 = 45.9 ≤ 13·α−1 = 47.74 ⇒ C1
Singly 395 220 214 10,973.0 Compression flange: Case a (in Section 5):
220−10.2
symmetric 2
= 104.9 ≤ 9 · 16 = 144 ⇒ C1 My,Ed = 2.88 × 108 N mm ≤ MN,y,Rd
cross-section Web: = 4.045 × 108 N mm
395
N (M = 0): 10 .2 = 38.73 < 42 ⇒ C3
395
M (N = 0): 10 .2 = 38.73 < 72 ⇒ C1
395 396
N + M : α = 0.751 ⇒ 10 .2 = 38.73 < 13·α−1 = 45.19 ⇒
C1

Doubly 396 217 217 10,983.2 Compression flange: NEd = 4.83 × 105 N
217−10.2
symmetric 2
= 103.4 ≤ 9 · 16 = 144 ⇒ C1 > Min{0.25 · Npl = 6.45 × 105 , 0.5 · d · tw · fy
cross-section Web: = 4.75 × 105 } = 4.75 × 105 N,
396
N (M = 0): 10 .2 = 38.82 ≤ 42 ⇒ C3 (the effect of axial force on the plastic moment of
396
M (N = 0): 10 .2 = 38.82 ≤ 72 ⇒ C1
resistance need to be taken into account). Section 6.2.9
N + M : α = 0.75 ⇒ 10396 396 of EC3 [3].
.2 = 38.82 ≤ 13·α−1 = 45.26 ⇒ C1 A−2bf tf 5
a= = 0.37 ≤ 0.5; n = 4.83×106 = 0.19;
A 2.58×10
My,Ed = 2.88 × 108 N mm
1−n
≤ MN,y,Rd = My,Rd 1− 0.5a
= 4.27 × 108 N mm

(a) Top and bottom flange areas and the total area of the cross-section (b) Solutions after refinement procedure for load corresponding to Point
obtained from Eqs. (3), for load corresponding to Point D in Fig. 9. D in Fig. 9. Point A corresponds to minimum cross-section area and
point B is closer to height of web of IPE500.

(c) Solutions after refinement procedure. Flange widths (d) Solutions after refinement procedure imposing double symmetry. Point
corresponding to Points A (minimum cross-section area) and B A1 corresponds to minimum cross-section area.
(closer to IPE500).

Fig. 10. Solutions from RSD methodology for point D in Fig. 9.

The saved cross-area is not significant. Details are summarized in area for initial possible solutions from Eqs. (3) for the specific
Table 2(a). case of flange thickness, tf equal to 16 mm and web thickness, tw
Fig. 11 plots solutions for loading corresponding to Point H equal to 10.2 mm. These values correspond to those of the IPE500
in Fig. 9. Fig. 11(a) shows flange areas and total cross-sectional shape. In order to avoid numerical problems, the minimum value
3012 L.M. Gil-Martín et al. / Engineering Structures 30 (2008) 3003–3013

(a) Solutions from Eqs. (3) with minimum width of flange equal to 2 · tw (b) Solutions after refinement procedure. Point A (minimum cross-section area)
(with tw = 10.2 mm) for load corresponding to point H in Fig. 9. corresponds to d = 290 mm, bf c = 253 mm, bf t = 197 mm. Point B (closer to
height of web of IPE500) corresponds to d = 470 mm, bf c = 210 mm,
bf t = 177 mm.

(c) Solutions after refinement procedure (widths of flanges for Point A (d) Solutions after refinement procedure imposing double symmetry.
corresponds to minimum cross-section area and Point B is closer to Point A1 (minimum cross-section area) corresponds to
IPE500). d = 296 mm, bf = 228 mm. Note that the section closer to IPE500
(d = 467 mm, bf = 200 mm) is out of range because it does not satisfy
Class 1 requirements for compact sections.

Fig. 11. Solutions from RSD methodology for Point H in Fig. 9.

of the flange width was fixed as 2 · tw . Consequently, T-shaped 8. Conclusion


sections are not considered here. In Fig. 11, the range of web
height over which possible solutions exist is indicated. One may Design generally implies the management of several unknowns
appreciate that solutions exist for values of d ranging between 185 in the pursuit of a practical solution. In many cases, design
and 335 mm and between 440 and 520 mm, but for 335 < d < problems are not sufficiently constrained to have a single solution;
440 mm the section does not satisfy the requirements for Class 1 an infinite number of theoretical solutions exist where the
number of unknowns is greater than the number of equations.
sections and thus no solutions could be obtained in this range.
Minimization is always a way to obtain a particular solution,
Fig. 11(b) shows a subset of the solutions obtained after but often the solution obtained using a minimization procedure
refinement, for tf = 16 mm and tw = 10.2 mm. The minimum is not feasible for construction. The optimization process based
area solution is identified by Point A (10,158 mm2 ), for which d = on RSD proposed in this paper is a way of representing the
290 mm, bf c = 253 mm and bf t = 197 mm. Point B corresponds infinite solutions and allows engineers to choose constructable
approximately to the IPE500 shape (10,986 mm2 ), for which d = solutions that require less material than that would be determined
470 mm, bf c = 210 mm, and bf t = 177 mm. Fig. 11(c) plots the using conventional design approaches. Since multiple possible
flange widths as a function of web depth, d, for the refined solutions solutions are portrayed on the RSD representation, the engineer
that are plotted in Fig. 11(b). may consider factors other than mathematically pure optimization
Fig. 11(d) plots additional solutions under the constraint that when deciding upon a particular design solution, such as
the section is doubly symmetric, for the loading corresponding to availability of steel shapes and price differentials, simplicity of the
Point H of Fig. 9. These solutions are different from those plotted job site, and labour. Similar advantages have already been realized
in reinforced concrete design, in which significant savings in steel
previously due to the fact that symmetric constraint was not
reinforcement are possible where non-symmetric solutions may
considered before. In this case the minimum cross-section area
be used.
is 10,315 mm2 and occurs for d = 296 mm and bf c = bf t =
The assumptions required to implement the EC3 methodology
228 mm. In Fig. 11(d), the point corresponding to the IPE500 shape in this paper can be modified to suit the requirements of other
is located outside of the range of possible solutions because for this codes without the methodology losing validity. As well, the
load combination, the IPE500 shape is not in Class 1. Details are methodology can be extended sections in classes other than
summarized in Table 2(b). Class 1.
L.M. Gil-Martín et al. / Engineering Structures 30 (2008) 3003–3013 3013

Table 2(b)
Summary of results for Point H in Fig. 9
d bf c bf t Area Class cross-section Verification of resistance (EC3)
(mm) (mm) (mm) (mm2 ) [C1: Class 1; C2: Class 2; C3: Class 3; C4: Class 4]

IPE500 468 200 200 11,173.6 Compression flange: 200−210.2 = 94.9 ≤ 9 · 16 = 144 ⇒ C1 NEd = 1.13 × 106 N > Min{0.25 · Npl = 6.56 × 105 ,
Web: 0.5 · d · tw · fy = 5.61 × 105 } = 5.61 × 105 N
486
N(M = 0): 10 .2 = 45.9 > 42 ⇒ C4 ⇒ ρ = 0.9 (the effect of axial force on the plastic moment of
486
M (N = 0): 10 .2 = 45.9 ≤ 72 ⇒ C1
resistance need to be taken into account). Section 6.2.9
N + M : α = 1 ⇒ 10 486 456 of EC3 [3].
.2 = 45.9 > 13·α−1 = 38; A−2bf tf 6
ψ = 0.226; 45.9 ≤ 56.4 ⇒ C3 a= = 0.43 ≤ 0.5; n = 1.13×106 = 0.43;
A 2.63×10
My,Ed = 1.12 × 108 N mm ≤ MN,y,Rd
1−n
= My,Rd 1− 0.5a
= 3.6 × 108 N mm
Singly 290 253 197 10,158.0 Compression flange: Case d (in Section 5):
253−10.2
symmetric 2
= 121.4 ≤ 9 · 16 = 144 ⇒ C1 My,Ed = 1.12 × 108 N mm ≤ MN,y,Rd
cross-section Web: = 2.00 × 108 N mm
290
N(M = 0): 10 .2 = 28.43 < 33 ⇒ C1
290
M (N = 0): 10 .2 = 28.43 < 72 ⇒ C1
290 396
N + M : α = 1 ⇒ 10 .2 = 28.43 ≤ 13·α−1 = 33 ⇒ C1
Doubly 296 228 228 10,315.2 Compression flange: NEd = 1.13 × 106 N > Min{0.25 · Npl
228−10.2
symmetric 2
= 108.9 ≤ 9 · 16 = 144 ⇒ C1 = 6.06 × 105 , 0.5 · d · tw · fy = 3.57 × 105 } = 3.57 × 105 N
cross-section Web: (the effect of axial force on the plastic moment of
296
N(M = 0): 10 .2 = 29.01 ≤ 33 ⇒ C1 resistance need to be taken into account). Section 6.2.9
296
M (N = 0): 10 .2 = 29.01 ≤ 72 ⇒ C1
of EC3 [3].
A−2bf tf 6
N + M : α = 1 ⇒ 10 298 396
.2 = 29.01 ≤ 13·α−1 = 33 ⇒ C1 a= = 0.29 ≤ 0.5; n = 1.13×106 = 0.47;
A 2.42×10
My,Ed = 1.12 × 108 N mm ≤ MN,y,Rd
1−n
= My,Rd 1− 0.5a
= 1.98 × 108 N mm

References [4] EC3. European Committee for Standarization. prEN1993-1-5. Eurocode 3.


Design of steel structures. Part 1.5: Plated structural elements. Brussels;
[1] Hernández-Montes E, Aschheim M, Gil-Martin LM. The impact of optimal 2004.
longitudinal reinforcement on the curvature ductility capacity of reinforced [5] EC3. Eurocódigo 3. Proyecto de Estructuras de acero. UNE-ENV 1993-1-1. Parte
concrete column sections. Mag Concrete Res 2004;56(9):499–512. 1-1: Reglas generales y para edificación. AENOR; 1996.
[2] Hernández-Montes E, Gil-Martín LM, Aschheim M. The design of concrete [6] Kaim P. Spatial buckling behavior of steel members under bending and
members subjected to uniaxial bending and compression using reinforcement compression. Ph.D. thesis. Institute for Steel, Timber and Shell Structures, Graz
sizing diagrams. ACI Struct J 2005;102(1):150–8. University of Technology, H.12; 2004.
[3] EC3. European Committee for Standarization. ENV1993-1-1. Eurocode 3: Design [7] Chen WF, Shoal I. Plastic design and second-order analysis of steel frames.
of steel structures. Part 1.1: General rules and rules for buildings. Brussels; 2005. Springer-Verlag. New-York Inc.; 1995.