Optimum RC Column Reinforcement Considering Multiple Load Combinations

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Struct Multidisc Optim (2009) 39:153–170

DOI 10.1007/s00158-008-0318-4

INDUSTRIAL APPLICATION

Optimum RC column reinforcement considering


multiple load combinations
Ho Jung Lee · Mark Aschheim ·
Enrique Hernández-Montes ·
Luisa María Gil-Martín

Received: 29 October 2007 / Revised: 8 August 2008 / Accepted: 7 September 2008 / Published online: 17 October 2008
© Springer-Verlag 2008

Abstract A general method is proposed to determine Notation


optimal reinforcement distributions for rectangular
reinforced concrete column sections subjected to multi- Ag gross area of the cross section
ple combinations of axial load and moment. The design As area of bottom reinforcement
problem is formulated as a general constrained nonlin- A′s area of top reinforcement
ear optimization problem and is solved both mathemat- b width of the cross section
ically and graphically. Ultimate strength and strength C a value at the coordinate axes, As and A′s
reduction factors are evaluated using ACI code provi- c depth of the neutral axis
sions. The method provides a simple approach to deter- Cc compressive force carried by concrete
mine optimal reinforcement under multiple loading Cs compressive force carried by top reinforcement
combinations. The use of optimal reinforcement rather d distance from the centroid of the bottom rein-
than conventional (symmetric) distributions of rein- forcement to the upper fiber
forcement can lower construction costs and environ- dt distance from the centroid of the bottom rein-
mental impacts. forcement to the upper fiber
d′ distance from the centroid of the top reinforce-
Keywords Concrete column · Nonlinear constraints · ment to the upper fiber
Optimization · Multiple load combinations e eccentricity of axial load
fc′ specified compressive strength of concrete
fs tensile stress in the bottom reinforcement
fs′ compressive stress in the top reinforcement
fy specified yield strength of the reinforcement
H. J. Lee (B)
SC Solutions, Inc., 1261 Oakmead Parkway, h height of the cross section
Sunnyvale, CA, USA M moment
e-mail: hlee@scsolutions.com Mu factored moment
Mn nominal moment
M. Aschheim
Santa Clara University, 500 El Camino Real, P axial load
Santa Clara, CA, USA Pu factored axial load
e-mail: maschheim@scu.edu Pn nominal axial load
Ts tensile force carried by bottom reinforcement
E. Hernández-Montes · L. M. Gil-Martín
University of Granada, Campus de Fuentenueva, β1 rectangular stress block coefficient, defined in
18072 Granada, Spain Section 10.2.7.3 of ACI 318
E. Hernández-Montes γ ratio of distance between centroids of outer rows
e-mail: emontes@ugr.es of bars and cross section thickness
L. M. Gil-Martín εc strain at the upper fiber
e-mail: mlgil@ugr.es εs tensile strain in extreme bottom reinforcement
154 H.J. Lee et al.

εt tensile strain in extreme bottom reinforcement tigated. They showed that the discretized continuum-
εs′ compressive strain in top reinforcement type optimality criteria-based methods are particularly
φ strength reduction factor efficient for the design of large RC frames.
Despite these theoretical advances, in practice, de-
sign typically precedes using an iterative design ap-
1 Introduction proach, generally emphasizing analysis to establish that
the ultimate strength of a section with known rein-
Uniaxial P–M interaction diagrams have become forcement is adequate, rather than solving directly for
widely used for determining column reinforcement the reinforcement required to provide a section with
since their original presentation by Whitney and Cohen adequate strength. Hernández-Montes et al. (2005)
(1956). P–M interaction diagrams allow the amount of reformulated the sectional design problem to deter-
reinforcement to be determined to provide sufficient mine the required reinforcement directly as a func-
strength to resist one or more combination of axial load tion of the neutral axis location, for given axial load
(P) and moment (M). A symmetric pattern of rein- and moment. Consideration of all possible neutral axis
forcement is typically used to generate the P–M inter- locations allows the solution having the least amount of
action diagrams, and is most suitable where symmetric total reinforcement to be selected; the results are easily
values of moment (+M and −M) occur in combination visualized by plotting the amounts of reinforcement as
with any particular axial load. However, some column- a function of the neutral axis depth on a “Reinforce-
type members may be subjected to asymmetric load- ment Sizing Diagram.” Such a solution is valid for the
ing, such as edge and corner columns of buildings, prescribed combination of axial load and moment, but
bridge columns having C-bent configurations, the walls may not satisfy other combinations of axial load and
of box culverts, and piers supporting retaining walls. moment that may occur under other load combinations.
For such members, optimal reinforcement can signifi- The present paper extends the previous study by
cantly reduce the amount of longitudinal reinforcement Hernández-Montes et al. (2005) to consider optimal
(Hernández-Montes et al. 2005). reinforcement under multiple combinations of axial
Karihaloo and his coworkers have given significant load and moment. A general approach to determining
attention to the optimal design of reinforced concrete optimal reinforcement for multiple combinations of
structures. Kanagasundaram and Karihaloo (1990, axial load and moment is proposed. ACI code provi-
1991a, b) used sequential linear programming and sions (2005) are used to determine ultimate strengths.
sequential convex programming to address minimum- Sources of nonlinearity in the constraints of the opti-
cost design of multi-span beams and columns consid- mization problem include the strength reduction factor,
ering stability, strength, serviceability, durability and the stress–strain relationship of reinforcement, and the
fire resistance. Adamu et al. (1994) used an indi- concrete compressive stress block as defined in the ACI
rect minimization technique based on continuum-type code provisions.
optimality criteria for the design of a reinforced con- The problem is solved using a simple graphical
crete propped cantilever beam. Necessary conditions approach that makes use of “Load Combination
for cost minimality were derived by applying the cal- Reinforcement Diagrams” and using a formal con-
culus of variations to an augmented Lagrangian, con- strained nonlinear optimization formulation which can
sidering maximum deflection, flexure, and shear as be solved using functions available in MATLAB
constraints. Adamu and Karihaloo (1995c) applied this (2006). Both solution approaches are shown to be
method to reinforced concrete beams with prismatic effective. Savings in reinforcement can be significant,
segments, because continuous variation of the cross- especially for the column subjected to multiple loading
section over the domain is not practical. Adamu and combinations.
Karihaloo (1994a, b) also developed a procedure for
optimal design based on discretized continuum-type
optimality criteria, since direct and indirect methods 2 Design assumptions in the ACI code provisions
based on continuum-type optimality criteria are not
efficient for large, complex, structural systems. Adamu Assumptions used in the analysis of a rectangular cross
and Karihaloo (1995a, b) then applied the discretized section under combined flexure and axial load are illus-
continuum-type optimality criteria-based method to trated in Fig. 1. The required cross-sectional areas of
the minimum cost design of RC frames. Convention- top and bottom reinforcement are determined to pro-
ally reinforced columns under both uniaxial bending vide adequate ultimate strength in accordance with
actions and under biaxial bending actions were inves- current ACI 318 design code provisions. Positive axial
RC column reinforcement considering multiple load combinations 155

Fig. 1 Strain and stress εc=0.003


diagrams for ultimate fc = 0.85—f’c
strength analysis per As ' d' ε ′s β 1c fs Cs
ACI 318 c β 1c Cc
h 2
2 M
h d P

As εs fs Ts
Stress Free body
b Strain
diagram
(a) (b) (c) (d)

load and moment are illustrated, where axial load is compressive force carried by the concrete, Cc , is deter-
positive in compression and moment is positive if the mined as
top fiber is compressed relative to the bottom fiber.
Cc = 0.85 fc′ (β1c ) (b )
¡ ¢
These terms and sign conventions are consistent with (1)
the conventional analysis for the sectional design of a
reinforced concrete member. where b = the width of the cross section. Since it was
According to the provisions of ACI 318, the ulti- assumed that top reinforcement is within the compres-
mate strength of the cross section is determined for sion zone (β 1 c > d′ ), the stress assigned to the rectangu-
an extreme fiber compressive strain of 0.003 using lar concrete block should not be counted in determining
Bernoulli’s hypothesis that plane sections remain plane. the force carried by the top steel, Cs . Hence,
The concrete contribution in compression is simplified
Cs = fs′ − 0.85 fc′ A′s
¡ ¢¡ ¢
using the Whitney stress block, for which a stress of (2)
0.85 fc′ is present over the width of the section and over
a depth of β 1 c from the top of the section where fc′ is where A′s = the cross-sectional area of top reinforce-
concrete compressive strength, β 1 is a factor relating ment located at a distance d′ from the top of the sec-
the depth of equivalent rectangular compressive stress tion. Based on the strain, εs′ , obtained by using similar
block to neutral axis depth, c. The strains εs and εs′ rep- triangles, the top reinforcement stress, fs′ , can be deter-
resent the strains in the bottom and top reinforcement, mined as
respectively, and are considered positive for tension
fs′ = (Es ) εs′ = (Es ) 0.003 1−d′ c if 0 ≤ εs′ ≤ f y Es
¡ ¢ ¡ ¡ ± ¢¢ ±
in the bottom reinforcement and compression in the
if εs′ ≥ f y Es
±
top reinforcement. These strains can be determined = fy
by similar triangles as a function of the neutral axis (3)
depth, c (Fig. 1b). The steel reinforcement is considered
to have elastic–plastic behavior. In the elastic range, where f y = the specified yield strength of the reinforce-
the reinforcement stress, fs or fs′ is determined by the ment. The stresses fc′ and fs′ are considered positive in
product of the elastic modulus, Es , and the reinforce- compression, as are the corresponding stress resultants
ment strain. This assumption allows the reinforcement Cc and Cs . Since β 1 c was assumed to be greater than
stresses to be determined as a function of the neutral d′ , εs′ becomes positive for the example selected in this
axis depth, c. In the plastic range, the reinforcement study.
stress is considered to be equal to the specified yield The force carried by the bottom steel, Ts , can be
strength of the reinforcement. determined as
The internal stress resultants Cc , Cs , and Ts can
be determined as the product of the stresses and the Ts = ( fs ) (As ) (4)
corresponding areas. Since this study focuses on the
combined effect of flexure and axial load, it is assumed where As = the cross-sectional area of bottom rein-
that the compression zone, β 1 c, is greater than d′ and forcement located at a distance d from the top of
less than d. Based on the Whitney stress block, the the section. Based on the strain, εs , obtained by using
156 H.J. Lee et al.

similar triangles, the bottom reinforcement stress, fs , Where the net tensile strain exceeds 0.005, the section is
can be determined as classified as tension-controlled. In this zone, the bottom
¡ ¡ ± ¢¢ ± reinforcement yields and a higher strength reduction
fs = (Es ) (εs ) = (Es ) 0.003 d c−1 if 0 ≤ εs ≤ f y Es factor, equal to 0.9, is specified. For the intermediate
= fy
±
if εs ≥ f y Es strains, the strength reduction factor is determined by
linear interpolation.
(5)

The bottom reinforcement stress, fs , is considered 3 Graphical solutions for individual load combinations
to be positive in tension, as is the corresponding stress
resultant, Ts . Since c was assumed to be less than d, εs Before presenting a general nonlinear optimization for-
becomes positive for the example selected in this study. mulation of the column design, a graphical approach
In a well designed beam, the internal stress resul- is illustrated for the purpose of showing the charac-
tants, Cc , Cs , and Ts equilibrate the applied axial load, teristics of the column design problem using the ACI
P, and moment, M. code provisions. To illustrate a graphical approach, the
In ultimate± strength design,±a design solution is to en- design of a column section shown in Fig. 3 is selected
sure that Pu φ ≤ Pn and Mu φ ≤ Mn where the sub- for the optimal design study under multiple load com-
script u represents a factored load effect, the subscript binations. The cross section dimensions are 457 mm ×
n represents a nominal strength, and φ is a strength 457 mm (18 in. × 18 in.) The reinforcing position is
reduction factor. A unified strength reduction factor, φ, assumed to be given by γ = 0.75. The concrete com-
is specified in the ACI 318 (2002, 2005) code provisions. pressive strength is fc′ = 27.6 MPa (4 ksi) and the spec-
As shown in Fig. 2, the strength reduction factor varies ified yield strength of reinforcement is f y = 414 MPa
with the net tensile strain of the bottom reinforcement. (60 ksi).
Where the net tensile strain is less than 0.002, the The load combinations considered for the optimal
section is classified as compression-controlled. In this design of the column are as follows:
zone, the stress of the bottom reinforcement is in the ± ±
elastic range and thus a lower strength reduction factor 1. Load case 1: Pu ± Ag fc′ = 0.1, Mu ± Ag fc′ h = 0.2
of 0.65 is specified for a rectangular (non-spiral) section. 2. Load case 2: Pu Ag fc′ = 0.3, Mu Ag fc′ h = 0.23

Fig. 2 Variation of φ with (B)


φ (A)
net tensile strain εt and c / dt
for Grade 60 reinforcement
and for prestressing steel φ = 0.7 + (ε t _ 0.002)(50)
(ACI 318 2002 and 2005)

0.90

250
Spiral φ = 0.65 + (ε t _ 0.002)( )
0.70 3

Other
0.65

Compression Transition Tension


controlled controlled

ε t = 0.002 ε t = 0.005
c / d t = 0.6 c / d t = 0.375
c
Interpolation on :
dt
1 5 1 5
Spiral: φ = 0.70 + 0.15( − ) Other: φ = 0.65 + 0.20( − )
c / dt 3 c / dt 3
RC column reinforcement considering multiple load combinations 157

b = h = 457 mm (18 in.), γ = 0.75 Sizing Diagram. The reinforcement sizing diagram can
be obtained by formulating equilibrium equations sepa-
h rately as the sum of moments about the centroid of the
top reinforcement and the sum of moments about the
h centroid of the bottom reinforcement. By doing this,
the following uncoupled equations for As and A′s are
obtained.
As ³ ´
Mu Pu
¡h ¢ β1 c
φ
− φ 2
− d′ − C c d′ − 2
As' As ≥ (6)
b fs (d − d′ )

³ ´
Mu Pu β1 c
d − h2 − Cc d −
¡ ¢
φ
− φ 2
A′s ≥ (7)
fs′ (d − d′ )

e Pu Mu=Pue
Before analyzing the optimal design solution under
multiple load combinations, the reinforcement sizing
diagrams for each load combination were obtained, to
provide a baseline comparison to the design solution
obtained for multiple loading combinations.
¡
Fig. 3 Section dimensions of the column selected for the optimal For Load¢ Combination 1 Pu /Ag fc′ = 0.1, Mu /
design study Ag fc′ h = 0.2 , reinforcement sizing diagram variations
with the strength reduction factor specification are
shown in Fig. 4. The symmetric design does not produce
±
3. Load case 3: Pu ± Ag fc′ = 0.35, M±
± ′ the optimal design solution for either of the strength
u Ag fc h = 0.27
′ ′
4. Load case 4: Pu Ag fc = 0.4, Mu Ag fc h = 0.15 reduction factor specifications. In this particular case,
the optimal design obtained using the ACI code speci-
where Ag is the gross section area of the column sec- fication for the strength reduction factor is close to the
tion. In the column design of civil structures, it is symmetric design, while the optimal design obtained
assumed that the column is subjected to one load using an invariant strength reduction factor differs
combination at a time. Therefore, the adequacy of the significantly from the symmetric design solution. The
column should be established independently for each strength reduction factor specified in the ACI code
load combination. provisions tends to promote tension-controlled designs,
To investigate the influence of the strength reduction which are considered more desirable because of their
factor on design, two specifications for the strength re- enhanced ductility. The required reinforcement amount
duction factor were used. One is the strength reduction is sensitive to the strength reduction factor and the
factor specification in the Chapter 9 of the ACI 318 value c/d. Since the reinforcement sizing diaphragm is
(2002 and 2005) code provisions. The second assumes represented explicitly terms of c/d, a designer can es-
the strength reduction factor is equal to unity, in ef- tablish the range of solutions (in terms of c/d) for which
fect± resulting in design
± for specified nominal strengths the design is controlled by tension
¡ or compression.
′ ′ ′ ′
Pn ± Ag fc and Mn ±Ag fc h rather than ultimate strengths For
¢ Load Combination 2 Pu /Ag fc = 0.3,Mu /Ag fc h =
Pu Ag fc′ and Mu Ag fc′ h. 0.23 , reinforcement sizing diagrams are shown for the
Hernández-Montes et al. (2005) proposed a new two strength reduction factor specifications in Fig. 5.
approach to design in which the reinforcement areas, For each strength reduction factor specification, the
As and A′s , required to provide nominal strengths Pn symmetric design is not an optimal solution. In this
and Mn are determined as a function of the neutral case, the invariant strength reduction factor produces
axis depth, c. For each value of c, the two equations an optimal solution close to the symmetric design, while
of equilibrium can be solved directly for the two steel the ACI code specification produces an optimal design
areas required for a given Pn and Mn . The solution solution that differs significantly from the symmetric
corresponding to a neutral axis depth that minimizes design solution. Again, the ACI specification for the
the total reinforcement area for a given axial load strength reduction factor tends to produce an optimal
and moment can be identified using the Reinforcement solution that is tension-controlled, while the symmetric
158 H.J. Lee et al.

Fig. 4 Reinforcement sizing Reinforcement Sizing Diagram - Load Case 1


diagram for the load case (Pu/Agf'c=0.1, Mu/Agf'ch=0.2)
of Pu /Ag fc′ = 0.1 and 20000 31
Mu /Ag fc′ h = 0.2
As
As'
As+As'
15000 Optimal solution under multiple load cases 23.25

Steel Area, mm 2

Steel Area, in 2
10000 Optimal 15.5
solution

Symmetric
5000 solution 7.75

Tension- Compression-
controlled (A) (B)
0 controlled 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Normalized Neutral Axis Depth, c/d
(a) Strength reduction factor based on the ACI code provisions

Reinforcement Sizing Diagram - Load Case 1


(Pu/Agf'c=0.1, Mu/Agf'ch=0.2)
20000 31
As
As'
As+As'
15000 Optimal solution under multiple load cases 23.25
2
Steel Area, mm

Steel Area, in 2
10000 15.5

Optimal
solution
Symmetric
5000 7.75
solution

Tension- (B) Compression-


controlled (A) controlled
0 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Normalized Neutral Axis Depth, c/d

(b) Invariant strength reduction factor (φ=1.0)

¡
solution produces a compression-controlled condition For Load ¢Combination 3 Pu /Ag fc′ = 0.35, Mu /
for the ACI specification of the strength reduction Ag fc′ h = 0.27 , reinforcement sizing diagrams are pro-
factor. Hence, the symmetric design does not always vided in Fig. 6 for the two specifications of strength
produce a favorable failure mode. A designer can ef- reduction factor. Design patterns observed for Load
ficiently use reinforcement sizing diagrams to obtain Combination 3 are similar to those of Load Combi-
a desirable design solution by considering the load nation 2. In this case, an invariant strength reduction
combination and desired mode of failure together with factor produces an optimal solution very close to the
the strength reduction factor specification. symmetric design, while the ACI strength reduction
RC column reinforcement considering multiple load combinations 159

Fig. 5 Reinforcement sizing Reinforcement Sizing Diagram - Load Case 2


diagram for the load case
(Pu/Agf'c=0.3, Mu/Agf'ch=0.23)
of Pu /Ag fc′ = 0.3 and 20000 31
Mu /Ag fc′ h = 0.23 As
As '
As +A s '
Optimal solution under multiple load cases
15000 23. 25

Steel Area, mm2

Steel Area, in 2
Optimal
10000 solution 15. 5
Symmetric
solution

5000 7.75

Tension- Compression-
(A) (B) controlled
0 controlled 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Normalized Neutral Axis Depth, c/d
(a) Strength reduction factor based on the ACI code provisions

Reinforcement Sizing Diagram - Load Case 2


(Pu/Agf'c=0.3, Mu/Agf'ch=0.23)
20000 31
As
As '
As +A s '
15000 Optimal solution under multiple load cases 23.25
2
Steel Area, mm

2
Steel Area, in
10000 15.5
Optimal
Symmetric solution
solution

5000 7.75

Tension- Compression-
controlled (A) (B) controlled
0 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Normalized Neutral Axis Depth, c/d

(b) Invariant strength reduction factor (φ=1.0)

factor produces an optimal design solution that dif- reduction factor produces a compression-controlled
fers significantly from the symmetric design solution. design in this case. ¡
As in the previous cases, the ACI strength reduction For Load ¢ Combination 4 Pu /Ag fc′ = 0.4, Mu /
factor produces an optimal solution that is considered Ag fc′ h = 0.15 , reinforcement sizing diagrams are
a tension-controlled design. The symmetric design so- shown in Fig. 7 for each strength reduction factor spec-
lution produces a compression-controlled condition. ification. In this case, the invariant strength reduction
Thus, based on ductility considerations, the optimal factor produces an optimal design solution close to the
design solution is more desirable than the symmetric symmetric design, for which the entire section is in com-
design because it is more ductile. The optimal de- pression. In contrast, the optimal design obtained using
sign solution determined using an invariant strength the ACI code specification for φ produces an optimal
160 H.J. Lee et al.

Fig. 6 Reinforcement sizing Reinforcement Sizing Diagram - Load Case 3


diagram for the load case (Pu/Agf'c=0.35, Mu/Agf'ch=0.27)
of Pu /Ag fc′ = 0.35 and 20000 31
Mu /Ag fc′ h = 0.27 As
As'
As+As'
Optimal solution under multiple load cases
15000 23.25

Steel Area, mm 2

Steel Area, in 2
Optimal
solution Symmetric
10000 solution 15.5

5000 7.75

Tension- Compression-
controlled
(A) (B) controlled
0 0
0.2 0.3 0.4 0.5 0.6 0.7 0.8
Normalized Neutral Axis Depth, c/d
(a) Strength reduction factor based on the ACI code provisions

Reinforcement Sizing Diagram - Load Case 3


(Pu/Agf'c=0.35, Mu/Agf'ch=0.27)
20000 31
As
As'
As+As'
15000 Optimal solution under multiple load cases 23.25
2
Steel Area, mm

Steel Area, in 2
Optimal
10000 solution 15.5

Symmetric
5000 solution 7.75

Tension- Compression-
controlled
(A) (B) controlled
0 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Normalized Neutral Axis Depth, c/d

(b) Invariant strength reduction factor (φ=1.0)

design solution for which the bottom reinforcement, optimal solution may deviate from an optimal solution
if present, would be in tension; however, the required obtained for any individual load combination. This sec-
area of bottom reinforcement is zero. tion presents a graphical implementation of a linear
programming technique to determine the optimal rein-
forcement for a section required to resist two or more
4 Graphical solutions for multiple load combinations load combinations. The optimal solution must provide
adequate strength to resist any of the load combinations
The preceding section identified optimal solutions for taken individually. Both the ACI 318 strength reduc-
each load combination considered individually. Where tion factor and an invariant strength reduction fac-
multiple load combinations must be considered, the tor of 1.0 are considered. The four load combinations
RC column reinforcement considering multiple load combinations 161

Fig. 7 Reinforcement sizing Reinforcement Sizing Diagram - Load Case 4


diagram for the load case
(Pu/Agf'c=0.4, Mu/Agf'ch=0.15)
of Pu /Ag fc′ = 0.4 and 16000 24.8
Mu /Ag fc′ h = 0.15
As
As '
As +A s '
Optimal solution under multiple load cases
12000 18.6

Steel Area, mm 2

Steel Area, in 2
8000 Symmetric 12.4
solution
Optimal
solution
4000 6.2

Tension-
controlled (A) (B) Compression-
controlled
0 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Normalized Neutral Axis Depth, c/d
(a) Strength reduction factor based on the ACI code provisions

Reinforcement Sizing Diagram - Load Case 4


(Pu/Agf'c=0.4, Mu/Agf'ch=0.15)
16000 24.8
As
As '
As +A s '
12000 Optimal solution under multiple load cases 18.6
2

Steel Area, in 2
Steel Area, mm

Tension- Compression-
controlled
(A) (B)
8000 controlled 12.4

Optimal
4000 6.2
solution

Symmetric
solution
0 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Normalized Neutral Axis Depth, c/d

(b) Invariant strength reduction factor ( φ=1.0)

considered in previous section constitute the multiple axis depth, c; alternatively, the equilibrium equations
load combinations. can be solved to yield A′s , as a function of As (by
The optimal design solution for multiple load com- removing c). The LCRD allows a design engineer to
binations can be obtained graphically using a new di- easily determine an optimal design solution from the
agram. This diagram, termed a “Load Combination feasible solutions plotted on a LCRD.
Reinforcement Diagram” (LCRD) plots reinforcement The feasible solution region based on the ACI 318
solutions in a two-dimensional space defined by the strength reduction factor is indicated on the load com-
coordinates As and A′s . Each solution can be viewed bination reinforcement diagram in Fig. 8. The feasible
as having been obtained for a different value of neutral solution region is determined by the boundaries of
162 H.J. Lee et al.

Fig. 8 Feasible design Load Case 1 (Pu/Agf'c=0.1, Mu/Agf'ch=0.2)


solution region on the Load Load Case 2 (Pu/Agf'c=0.3, Mu/Agf'ch=0.23)
Feasible solution region Load Case 3 (Pu/Agf'c=0.35, Mu/Agf'ch=0.27)
Combination Reinforcement As (in2) Load Case 4 (Pu/Agf'c=0.4, Mu/Agf'ch=0.15)
Diagram and an optimal
0 3.1 6.2 9.3 12.4 15.5 18.6 21.7
design solution and a 14000 21.7
symmetric design solution
under multiple load
As = A's
combinations with the 12000 18.6
strength reduction factor in
the ACI code provisions
10000 15.5
Optimal design
solution
8000 Symmetric 12.4
As' (mm )

As' (in )
2

2
design solution

6000 9.3

4000 6.2

2000 3.1

0 0
0 2000 4000 6000 8000 10000 12000 14000
As (mm 2)

Loading Combinations 1 and 3. In this case, a design shown in Fig. 8. The line As + A′s = C, where C = a
solution within this feasible solution region automati- constant, has intercepts (0, C) and (C, 0). Thus, the
cally satisfies Load Combinations 2 and 4. An optimal optimal solution can be found by increasing C from 0
design solution and a symmetric design solution are also to the point that the line intersects the feasible solution

Fig. 9 Feasible design Load Case 1 (Pu/Agf'c=0.1, Mu/Agf'ch=0.2)


solution region on the Load Load Case 2 (Pu/Agf'c=0.3, Mu/Agf'ch=0.23)
Feasible solution region Load Case 3 (Pu/Agf'c=0.35, Mu/Agf'ch=0.27)
Combination Reinforcement As (in 2) Load Case 4 (Pu/Agf'c=0.4, Mu/Agf'ch=0.15)
Diagram and an optimal 0 1.55 3.1 4.65 6.2 7.75 9.3 10.85
design solution and a 7000 10.85
symmetric design solution
under multiple load
combinations with the As = A's
6000 9.3
invariant strength reduction
factor of 1.0
5000 7.75

4000 6.2
As' (mm )

As' (in )
2

Optimal solution close to the


symmetric design solution
3000 4.65

2000 3.1

1000 1.55

0 0
0 1000 2000 3000 4000 5000 6000 7000
As (mm 2)
RC column reinforcement considering multiple load combinations 163

region. In this case, the optimal design solution occurs the origin. The line represents symmetric reinforce-
at the common point of the boundaries of Load- ment, and the symmetric reinforcement solution is ob-
ing Combinations 1 and 3. In this case, the opti- tained where this diagonal line intersects the feasible
mal reinforcement solution requires As = 3,245 mm2 solution region. In this case, the minimum symmetric
(5.03 in.2 ), A′s = 6,052 mm2 (9.38 in.2 ), and a total re- reinforcement solution is determined only by Load-
inforcement area of As + A′s = 9,297 mm2 (14.41 in.2 ). ing Combination 3. The symmetric reinforcement so-
The line As = A′s is a diagonal line emanating from lution requires As + A′s = 6,935 mm2 (10.75 in.2 ), with

Fig. 10 P–M interaction P-M Interaction Diagram


diagrams of the optimally 1.0
designed section under
multiple load combinations 0.8

0.6 φPn,max

0.4

0.2
φPn/(Agf'c)

0.0
-0.30 -0.25 -0.20 -0.15 -0.10 -0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30
-0.2

-0.4

-0.6
Load Case 1 (Pu/Agf'c=0.1, Mu/Agf'ch=0.2)
Load Case 2 (Pu/Agf'c=0.3, Mu/Agf'ch=0.23)
-0.8 Load Case 3 (Pu/Agf'c=0.35, Mu/Agf'ch=0.27)
Load Case 4 (Pu/Agf'c=0.4, Mu/Agf'ch=0.15)
-1.0
φ Mn/(Agf'ch)

(a) Strength reduction factor based on the ACI code provisions


P-M Interaction Diagram
1.4

1.2

1.0

0.8 Pn,max

0.6
Pn/(Agf'c)

0.4

0.2

0.0
-0.30 -0.20 -0.10 0.00 0.10 0.20 0.30
-0.2

-0.4
Load Case 1 (Pu/Agf'c=0.1, Mu/Agf'ch=0.2)

-0.6 Load Case 2 (Pu/Agf'c=0.3, Mu/Agf'ch=0.23)


Load Case 3 (Pu/Agf'c=0.35, Mu/Agf'ch=0.27)
-0.8 Load Case 4 (Pu/Agf'c=0.4, Mu/Agf'ch=0.15)

Mn/(Agf'ch)

(b) Invariant strength reduction factor (φ =1.0)


164 H.J. Lee et al.

total reinforcement area of As + A′s = 13, 870 mm2 solution was determined by only Load Combination 3.
(21.50 in.2 ). The symmetric design solution requires Hence, Load Combination 3 plots on the boundary
49.2% more reinforcement than the optimal design of the P − M interaction diagram. The conventional
solution. design approach might lead to a condition with all load
The feasible solution region obtained for an invari- combinations plotting within the P − M interaction di-
ant strength reduction factor (equal to 1.0) is shown on agram, or at most one load combination plotting at the
the load combination reinforcement diagram in Fig. 9. boundary of the P − M interaction diagram, in general.
The boundary curves of each loading case are smoother Load combination reinforcement diagrams can help
than those based on the ACI 318 strength reduction fac- to visually establish the optimal solution under multiple
tor because the nonlinearity due to the strength reduc- load combinations, while hand calculations and approx-
tion factor is removed in this case. The feasible solution imations required in conventional iterative design
region is determined by the boundaries of Load Com- approaches may produce a more conservative design,
binations 1 and 3. A design solution within this feasible especially for sections subjected to multiple load com-
solution region automatically satisfies Load Combina- binations. In order to make the procedure amenable to
tions 2 and 4. An optimal design solution is indicated algorithmic implementation, a mathematical formula-
in Fig. 9. The optimal design solution occurs at the tion for the preceding optimal design problem is devel-
boundary of Loading Combination 3. The boundary of oped in the following.
Load Combination 1 is not involved in determining the
optimal design solution. For this case, the optimal rein-
forcement solution requires As = 3,206 mm2 (4.97 in.2 ),
A′s = 3,129 mm2 (4.85 in.2 ) and a total reinforcement 5 Mathematical formulation for optimal design
area of As + A′s = 6,335 mm2 (9.82 in.2 ). The sym- under multiple load combinations
metric design solution is very close to the optimal
design solution. The symmetric reinforcement solution A conventional approach to the design of a reinforced
requires As = A′s = 3, 174 mm2 (4.92 in.2 ) and a total concrete section is to apply equilibrium equations to
reinforcement area of As = A′s = 6,348 mm2 (9.84 in.2 ). evaluate the nominal strengths Pn and Mn for a given
These optimal design solutions, considering multiple neutral axis depth, material properties, and reinforce-
loading combinations, are indicated on the reinforce- ment areas As and A′s . By considering different neutral
ment sizing diagrams determined for each single load axis depths, the entire P − M interaction surface is
case in Figs. 4–7. As expected, the optimal design so- constructed. Because this approach requires a prede-
lution under multiple load combinations using the ACI termined pattern of reinforcement (ratios or amounts
strength reduction factor lies on the boundary of Load of As and A′s ) it is not generally possible to find the
Combinations 1 and 3 and lies above the boundary optimal amounts (and ratios) of As and A′s to minimize
of Load Combinations 2 and 4. The optimal design the total reinforcement of As + A′s .
solution under multiple load combinations using the The mathematical formulation to determine optimal
invariant strength reduction factor lies on the boundary reinforcement for the design of a reinforced concrete
of Load Combination 3 and lies above the boundary of section under multiple load combinations is developed
the other load combinations. In these cases, the optimal as follows. Optimization techniques are used to find a
design solution under multiple load combinations does set of design parameters, x = {x (1) , x (2) , . . . , x (n)},
not coincide with the optimal design solution under any that can be defined as optimal. A general constrained
single load combination and requires significantly more minimization problem can be stated as
reinforcement than is required for some of the single
load combination cases. Minimize f (x) (8)
A P − M interaction diagram of the section designed x
using the optimal solution for multiple load combina-
tions is shown in Fig. 10. For the ACI strength reduction subject to equality constraints, Ceq(x) = 0, inequality
factor, the optimal solution was determined by the constraints, C(x) ≤ 0, and bounds a ≤ b ≤ b, where x
boundary of Load Combinations 1 and 3. Hence, Load is the vector of length n design parameters, f (x) is the
Combinations 1 and 3 plot on the boundary of the P − objective function, which returns a scalar value, and the
M interaction diagram. As expected, Load Combina- vector functions Ceq(x) and C(x) contain the values
tions 2 and 4 plot within the P − M interaction diagram. of the equality and inequality constraints evaluated at
For the invariant strength reduction factor, the optimal x and return a vector of lengths i and j, respectively.
RC column reinforcement considering multiple load combinations 165

The vectors a and b are lower bound and upper bound constraints and lower bound and upper bound values
values of the design parameters x, respectively. can be stated as follows for each domain:
When specialized to the present problem,
1. Domain I
• The objective function is simply As + A′s , and thus
Pui 0.9 − 0.85 fc′ β1 b x (i + 2) + f y x (1)
±
is linear.
− 0.003Es − 0.85 fc′
¡
• Due to the asymmetry of the interaction± diagram
with respect to P and M, for which Mu φ ≤ Mn at
− 0.003Es d′ x (i + 2) x (2) = 0
± ¢
(9)
a given Pn = Pu / φ (but the converse is not true),
equilibrium constraints are established as follows:
±
Pn = Pu / φ (an equality constraint) and Mu φ ≤ Mui 0.9−0.85 fc′ β1 b x (i + 2) h 2−β1 x(i + 2) 2
± ¡ ± ± ¢
Mn (an inequality constraint).
+ f y d − h 2 x (1) − 0.003Es −0.85 fc′
¡ ± ¢ ¡

The stress in the reinforcement and the strength − 0.003Es d′ x (i+2) h 2−d′ x (2) ≤ 0
± ¢¡ ± ¢
(10)
reduction factor are nonlinear functions of the neutral
axis depth, c. The location of the internal concrete
stress resultant Cc also varies with the neutral axis x (1) ≥ 0, x (2) ≥ 0, d′ β1 ≤ x (i+2) ≤ 3d 8
± ±
(11)
depth, c. Therefore, the constraints are nonlinear. Due
to the nonlinear constraints, the problem is a nonlinear 2. Domain II
optimization problem and a solution may be obtained
Pui 0.233 + 0.25d x (i + 2) −0.85 fc′ β1 b x (i + 2)
±¡ ± ¢
by considering solution intervals over the neutral axis
+ f y x (1) − 0.003Es − 0.85 fc′
¡
depth c. Four intervals, or domains, are considered:
− 0.003Es d′ x(i + 2) x(2) = 0
± ± ± ¢
1) Domain I: d′ β1 ≤ c< 3d 8. In this domain, the
φ = 0.9 and the ± bottom reinforcement
±¡ is yielding. ¢ (12)
2) Domain II: 3d 8 ≤ c < .003Es d′ 0.003Es − f y .
In this domain, the φ = 0.233 + 0.25d/c(based on
Mui
±¡ ± ¢
the φ = 0.65 + (εt − 0.002) (250 / 3) in Fig. 2), and 0.233 + 0.25d x (i + 2)
the bottom reinforcement is yielding.
¡ ¢ − 0.85 fc′ β1 b x (i + 2) h 2 − β1 x (i + 2) 2
¡ ± ± ¢
3) Domain III: 0.003Es d′ / 0.003Es − f y ≤ c <
+ f y d − h 2 x (1) − 0.003Es − 0.85 fc′
¡ ± ¢ ¡
0.003Es d/( f y + 0.003Es ). In this domain, φ =
0.233 + 0.25d / c, and both the top and bottom − 0.003Es d′ x (i + 2) h 2 − d′ x (2) ≤ 0 (13)
± ¢¡ ± ¢
reinforcement are yielding.
±¡ ¢
4) Domain IV: 0.003Es d f y + 0.003Es ≤ c ≤ d. In ±
this domain, φ = 0.65, and the top reinforcement is x (1) ≥ 0, x (2) ≥ 0, 3d 8 ≤ x (i + 2)
yielding.
≤ 0.003Es d′ 0.003Es − f y
±¡ ¢
(14)

The domain with c < d /β1 and the domain with c > d 3. Domain III
are excluded in this study. These domains represent
Pui 0.233+0.25d x (i+2) −0.85 fc′ β1 b x (i+2)
±¡ ± ¢
the column under the compression or tension dominant
loading condition. For such condition, column design
+ f y x (1) − f y − 0.85 fc′ x (2) = 0
¡ ¢
(15)
problems are trivial. Therefore, the optimization for-
mulation presented in this study does not include such
condition for simplicity, although the formulation can Mui
±¡ ± ¢
0.233 + 0.25d x (i + 2)
be extended to consider such conditions.
− 0.85 fc′ β1 b x (i + 2) h 2 − β1 x (i + 2) 2
¡ ± ± ¢
A set of design parameters in the optimization
problem can be defined as x = {As , A′s , c1 , c2 ,. . . ,
¡ ± ¢
+ f y d−h 2 x (1)
cm }, where x(1) = As , x(2) = A′s , x(3) = c1 , x(m + 2) =
− f y −0.85 fc′ h 2−d′ x (2) ≤ 0
¡ ¢¡ ± ¢
(16)
cm , ci = the neutral axis depth under the i-th load com-
bination, and m = the number of multiple load com-
bination. The objective function f (x) can be stated as x (1) ≥ 0, x (2) ≥ 0, 0.003Es d′ 0.003Es − f y
±¡ ¢
x(1) + x(2) and our goal is to minimize f (x). For the i-th ±¡ ¢
load condition, Pui and Mui , the equality and inequality ≤ x (i + 2) ≤ 0.003Es d f y 0.003Es (17)
166 H.J. Lee et al.

4. Domain IV Pu4 0.65 − 0.85 fc′ β1 b x (6)


±

Pui 0.65 − 0.85 fc′ β1 b x (i + 2)


± ¡ ± ¢
+ 0.003Es d x (6) − 0.003Es x (1)
¡ ± ¢
+ 0.003Es d x (i + 2) − 0.003Es x (1) − f y − 0.85 fc′ x (2) = 0
¡ ¢
(24)
− f y − 0.85 fc′ x (2) = 0
¡ ¢
(18)
4. Inequality constraints

Mui 0.65
±
Mu1 0.9 − 0.85 fc′ β1 b x (3) h 2 − β1 x (3) 2
± ¡ ± ± ¢

− 0.85 fc′ β1 b x (i + 2) h 2 − β1 x (i + 2) 2
¡ ± ± ¢
+ f y d − h 2 x (1) − 0.003Es − 0.85 fc′
¡ ± ¢ ¡
¡ ± ¢¡ ± ¢
+ 0.003Es d x (i + 2) − 0.003Es d−h 2 x (1) − 0.003Es d′ x (3) h 2 − d′ x (2) ≤ 0
± ¢¡ ± ¢
(25)
− f y − 0.85 fc′ h 2 − d′ x (2) ≤ 0
¡ ¢¡ ± ¢
(19)
Mu2
±¡ ± ¢
0.233 + 0.25d x (4)
±¡ ¢
z (1) ≥ 0, x (2) ≥ 0, 0.003Es d f y + 0.003Es − 0.85 fc′ β1 b x (4) h 2 − β1 x (4) 2
¡ ± ± ¢

≤ x (i + 2) ≤ d (20) + f y d − h 2 x (1) − 0.003Es − 0.85 fc′


¡ ± ¢ ¡

All four domains must be considered for each load − 0.003Es d′ x (4) h 2 − d′ x (2) ≤ 0
± ¢¡ ± ¢
(26)
combination. Hence, an optimal design solution must
be found among the 4m combinations of domains under
Mu3
±¡ ± ¢
multiple loading combinations. Note that the optimal 0.233 + 0.25d x (5)
solution for multiple load combinations need not cor- − 0.85 fc′ β1 b x (5) h 2 − β1 x (5) 2
¡ ± ± ¢
respond to a neutral axis depth for which the optimal ¡ ± ¢
solution is found for any one load combination indi- + f y d − h 2 x (1)
vidually. By using four loops, the minimum amount − f y − 0.85 fc′ h 2 − d′ x (2) ≤ 0
¡ ¢¡ ± ¢
(27)
of As + A′s is searched among the 4m combinations
of domains. To illustrate further, one possible domain
Mu4 0.65 − 0.85 fc′ β1 b x (6) h 2 − β1 x (6) 2
± ¡ ± ± ¢
combination among the 4m combinations of domains is
the combination with Load Combination 1 in Domain I, ¡ ± ¢¡ ± ¢
+ 0.003Es d x (6) − 0.003Es d − h 2 x (1)
Load Combination 2 in Domain II, Load Combination
− f y − 0.85 fc′ h 2 − d′ x (2) ≤ 0
¡ ¢¡ ± ¢
3 in Domain III, and Load Combination 4 in Domain (28)
IV. For this particular combination, the optimization
problem can be stated as 5. Lower and upper bounds

1. Design parameters x = {As , A′s , c1 , c2 , c3 , c4 }, x (1) ≥ 0, x (2) ≥ 0, d′ β1 ≤ x (3) ≤ 3d 8


± ±
(29)
x(1) = As , x(2) = A′s , x(3) = c1 , x(4) = c2 , x(5) = c3 ,
x(6) = c4
3d 8 ≤ x (4) ≤ 0.003Es d′ 0.003Es − f y
± ±¡ ¢
2. Objective function f (x)= x(1) + x(2) (30)
3. Equality constraints
0.003Es d′
±¡ ¢
Pu1 0.9 − 0.85 fc′ β1 b x (3) + f y x (1) 0.003Es − f y ≤ x (5)
±
±¡ ¢
− 0.003Es − 0.85 fc′ − 0.003Es d′ x (3) x (2) = 0 ≤ 0.003Es d f y + 0.003Es (31)
¡ ± ¢

(21) ±¡ ¢
0.003Es d f y + 0.003Es ≤ x (6) ≤ d (32)

Pu2 0.233 + 0.25d x (4) − 0.85 fc′ β1 b x (4)


±¡ ± ¢

The optimal solution under this particular domain


+ f y x (1)− 0.003Es −0.85 fc′
¡
combination is compared with the solutions under the
− 0.003Es d′ x (4) x (2) = 0
± ¢
(22) other possible domain combinations. In other words,
the minimum total reinforcement area, As + A′s , for
each solution is determined for the 4m combinations of
Pu3 0.233 + 0.25d x (5) − 0.85 fc′ β1 b x (5)
±¡ ± ¢
domains, and the smallest of these minima we retained
+ f y x (1) − f y − 0.85 fc′ x (2) = 0
¡ ¢
(23) as the optimal solution.
RC column reinforcement considering multiple load combinations 167

12000 Table 2 Output history information for the case with a starting
point x0 = [2,000, 2,000, 108.63, 108.63, 167.08, 318.41]
Objective function value (mm2)

10000 Iteration F-count f (x) Maximum constraint Directional


violation derivative
8000 0 7 4,000 3.47E+08
1 14 9,051.3 1.65E+08 5050
6000 2 21 9,508.99 2.04E+07 458
3 28 9,551.21 4.35E+05 42.2
4000 4 35 9,539.13 2.22E+03 −12.1
5 42 9,265.7 2.06E+06 −273
2000 Starting point of x(1)=0.0 and x(2)=0.0 6 49 9,296.96 3.13E+03 31.3
Starting point of x(1)=2000 and x(2)=2000 7 56 9,296.97 1.16E−02 0.00549
0 8 63 9,296.97 1.19E−07 1.06E−08
0 1 2 3 4 5 6 7 8 9
F-count = cumulative number of function evaluations, f (x) =
Iteration number function value at current point
(a) Convergence plot of objective function

7.E+08 optimization) of the function fmincon was used in this


Starting point of x(1)=0.0 and x(2)=0.0 study. This algorithm uses a sequential quadratic pro-
Maximum constraint violation

6.E+08
Starting point of x(1)=2000 and x(2)=2000 gramming method, wherein a quadratic programming
5.E+08

4.E+08 Reinforcement Amount


7000
Design parameter value (mm )
2

3.E+08
6000

2.E+08
5000

1.E+08 4000

0.E+00 3000
0 1 2 3 4 5 6 7 8 9
2000
Iteration number
(b) Convergence plot of constraint function 1000 x(1)
x(2)
Fig. 11 Convergence plots of objective and constraint functions 0
against iteration numbers 0 1 2 3 4 5 6 7 8 9

Iteration number
(a) Design parameters related to reinforcement amount
The solution of this constrained nonlinear optimiza-
tion can be obtained easily using the MATLAB func- Neutral Axis Location
400
tion fmincon. The standard algorithm (medium-scale x(3)
Design parameter value (mm)

x(4)
x(5)
300 x(6)
Table 1 Output history information for the case with a starting
point x0 = [0, 0, 108.63, 108.63, 167.08, 318.41]
Iteration F-count f (x) Maximum constraint Directional 200
violation derivative
0 7 0 6.16E+08
1 14 8,297.79 5.13E+08 8,300 100
2 21 9,382.46 1.45E+08 1,080
3 28 9,443.04 2.22E+07 60.6
4 35 9,467.05 6.11E+05 24 0
5 42 9,462.55 1.29E+03 −4.51 0 1 2 3 4 5 6 7 8 9
6 49 9,282.62 9.47E+05 −180 Iteration number
7 56 9,296.96 6.70E+02 14.3
8 63 9,296.97 5.07E−04 0.00131 (b) Design parameters related to neutral axis location
9 70 9,296.97 5.96E−08 3.94E−10
Fig. 12 Convergence plots of design parameters against iteration
F-count = cumulative number of function evaluations, f (x) = numbers for the case with starting point x0 = [0, 0, 108.63, 108.63,
function value at current point 167.08, 318.41]
168 H.J. Lee et al.

Reinforcement Amount
7000
objective and constraint functions are convex functions,
then the Kuhn–Tucker equations are both necessary
Design parameter value (mm )
2

6000 and sufficient for a global solution point (MATLAB


5000 2006). As shown in Fig. 8, the constraint functions are
not convex and thus the function fmincon might only
4000
give local solutions in specific cases.
3000 The solution of design parameters from the MAT-
LAB optimization for the case with the ACI strength
2000
reduction factor is x = {As , A′s , c1 , c2 , c3 , c4 } = {3,244.83,
x(1)
1000 6,052.13, 89.42, 139.09, 166.92, 253.58}. This solution
x(2)
0
was found in the combination with Load Combination
0 1 2 3 4 5 6 7 8 1 in Domain I, Load Combination 2 in Domain I, Load
Iteration number Combination 3 in Domain II, and Load Combination 4
(a) Design parameters related to reinforcement amount in Domain IV. This solution is a global solution point
Neutral Axis Location
of this particular problem as shown from the graphical
400 solution of Fig. 8. The same solution was obtained
x(3)
using two different starting points; each starting point
Design parameter value (mm)

x(4)
x(5) was an infeasible point. One is simply the case with
300 x(6)
As = A′s = 0 and the other is the case with As = A′s =
2,000. The initial neutral axis value used with these
200 starting points was the median of the lower and up-
per bound values used in the domain partition. Both
of the starting points converged to the same optimal
100
solution as shown in Fig. 11. The optimization was
terminated with the condition that first order optimality
0
measures were satisfied to the specified tolerance and
0 1 2 3 4 5 6 7 8 maximum constraint violation was less than the speci-
Iteration number fied tolerance. Output history information is presented
(b) Design parameters related to neutral axis location in Tables 1 and 2. Since the case with the start point
As = A′s = 2,000 is closer to the optimal solution, it
Fig. 13 Convergence plots of design parameters against iteration requires fewer numerical iterations and function eval-
numbers for the case with starting point x0 = [2,000, 2,000, 108.63,
uations. Convergence plots of design parameter values
108.63, 167.08, 318.41]
against iteration numbers are presented in Figs. 12 and
13. Three iterations provide a good approximation of
the optimal solution. Search directions corresponding
subproblem is solved and an estimate of the Hessian to the sequence of points computed by fmincon are
of the Lagrangian is updated at each iteration, using presented in Tables 3 and 4. As the number of iterations
the BFGS formula (MATLAB 2006). The standard increases, search directions show less change.
algorithm uses a solution to the Kuhn–Tucker (KT) For the case with the invariant strength reduction
equations to ensure optimality. If the problem is a factor of 1.0, three divisions of domains in terms of the
so-called convex programming problem, that is, the neutral axis depth are necessary because the strength

Table 3 Search directions Iteration x (1) x (2) x (3) x (4) x (5) x (6)
corresponding to the
sequence of points 0
computed by fmincon for 1 2.84E+03 5.46E+03 −6.58E+01 7.48E+01 −3.72E+01 −1.15E+02
the case with a starting point 2 3.75E+02 7.09E+02 2.66E+01 −5.61E+01 2.74E+01 4.50E+01
x0 = [0, 0, 108.63, 108.63, 3 2.12E+01 3.93E+01 1.57E+01 8.70E+00 −2.31E−02 2.14E+00
167.08, 318.41] 4 9.93E+00 1.41E+01 3.26E+00 2.97E−01 −1.36E−01 −8.78E−02
5 4.31E−01 −4.94E+00 1.22E−01 8.65E−02 2.81E−01 9.69E−02
6 1.11E+00 −1.81E+02 9.55E−01 2.88E+00 9.46E+00 3.30E+00
7 −3.65E−02 1.44E+01 −5.05E−02 −1.59E−01 −2.17E−02 −2.41E−01
8 1.98E−04 1.12E−03 1.06E−04 2.86E−04 1.04E−05 1.11E−04
9 3.52E−10 4.18E−11 8.24E−11 2.93E−10 1.84E−11 2.78E−11
RC column reinforcement considering multiple load combinations 169

Table 4 Search directions Iteration x (1) x (2) x (3) x (4) x (5) x (6)
corresponding to the
sequence of points computed 0
by fmincon for the case with a 1 1.12E+03 3.93E+03 −4.04E+01 4.54E+01 −4.78E+00 −8.57E+01
starting point x0 = [2,000, 2 1.09E+02 3.48E+02 1.69E+01 −2.07E+01 −9.40E+00 1.62E+01
2,000, 108.63, 108.63, 167.08, 3 1.06E+01 3.16E+01 2.86E+00 1.60E+00 8.56E−02 −5.52E−02
318.41] 4 3.72E−01 −1.25E+01 1.30E−01 2.05E−01 6.14E−01 2.31E−01
5 1.61E+00 −2.75E+02 1.41E+00 4.27E+00 1.34E+01 4.98E+00
6 −7.76E−02 3.13E+01 −1.09E−01 −3.40E−01 −4.06E−02 −5.23E−01
7 9.25E−04 4.56E−03 4.98E−04 1.34E−03 4.88E−05 5.36E−04
8 9.06E−09 1.59E−09 1.87E−09 6.55E−09 4.80E−10 6.72E−10

reduction factor is not involved in the domain divisions. factor specification and the loading combination.
This modification is easy and thus is not presented in Thus, it is desirable to consider strength reduction
this paper. The solution of design parameters from the factor variation, loading combination, and control
MATLAB optimization for the case with the invariant of failure mode to improve design solutions. This
strength reduction factor of 1.0 is x = {As , A′s , c1 , can be done effectively using Reinforcement Sizing
c2 , c3 , c4 } = {3,238.77, 3,077.41, 115.72, 205.09, 236.73, Diagrams.
249.70}. The results are almost identical to the graphical • Reinforcement Sizing Diagrams can help a designer
solutions of the preceding section. Minor differences to explicitly control the failure mode of the section;
results from the numerical precision of the assumed • The Load Combination Reinforcement Diagram
neutral axis depth values in the graphical solutions. was developed for determining optimal design so-
Therefore, it can be concluded that the general nonlin- lutions for the examples with multiple load com-
ear optimization formulation works for column design. binations. The feasible design solution zone under
In the formulation presented, the domains are es- multiple load combinations was explicitly repre-
tablished presuming± that the neutral axis depth, c, is sented on the Load Combination Reinforcement
greater than d′ β1 and less than d. It is trivial to include Diagram. The Load Combination Reinforcement
additional domains to extend the approach to cover Diagrams were effective for determining optimal
all possible neutral axis depths. Since the column de- reinforcement solutions. Hence, an optimized de-
sign problem is trivial for domains beyond this range sign solution under multiple loading combinations
(because the column is mainly subjected to axial load may be determined using Load Combination Rein-
(compression or tension), a design check for these situ- forcement Diagrams.
ations can be easily done separately. • A mathematical formulation for the optimization
problem was presented and was shown to pro-
vide results that match those determined using the
6 Conclusions graphical approach of the Load Combination Rein-
forcement Diagram.
In this study, the optimal design problem of a rein- • Since the feasible solution region for multiple load
forced concrete column section under multiple load combinations can be determined by more than
combinations was investigated. Due to nonlinearity in one load combination, the solution generally does
the strength reduction factor, the stress–strain rela- not coincide with the optimal design solution for
tionship of reinforcement, and design formulations for any one load combination. Generally, the optimal
the compression force carried by concrete, the opti- design solution under multiple load combinations
mal design problem was formulated as a constrained requires more reinforcement than that under any
nonlinear optimization problem. A general approach to individual load combination.
obtain the optimal design solution of the column under
multiple combinations of axial load and moment was
proposed in this study. In addition, a graphical method
to obtain the optimal design solution under multiple References
load combinations was also illustrated. The following
is a summary of results from this study: ACI 318 (2002) Building Code Requirements for Structural
Concrete and Commentary. ACI Committee 318, American
• Concrete Institute, Detroit, MI
Optimal design solutions generally do not coin- ACI 318 (2005) Building Code Requirements for Structural
cide with the symmetric design solution; the op- Concrete and Commentary. ACI Committee 318, American
timal solution depends on the strength reduction Concrete Institute, Detroit, MI
170 H.J. Lee et al.

Adamu A, Karihaloo BL (1994a) Minimum cost design of RC Hernández-Montes E, Gil-Martín LM, Aschheim M (2005) The
beams using DCOC Part I: beams with freely-varying cross- design of concrete members subjected to uniaxial bending
sections. Struct Optim 7:237–251 and compression using reinforcement sizing diagrams. ACI
Adamu A, Karihaloo BL (1994b) Minimum cost design of RC Struct J 102(1):150–158
beams using DCOC Part II: beams with uniform cross- Kanagasundaram S, Karihaloo BL (1990) Minimum cost
sections. Struct Optim 7:252–259 design of reinforced concrete structures. Struct Optim 2:
Adamu A, Karihaloo BL (1995a) Minimum cost design of RC 173–184
frames using the DCOC method Part I: columns under uni- Kanagasundaram S, Karihaloo BL (1991a) Minimum-cost rein-
axial bending actions. Struct Optim 10:16–32 forced concrete beams and columns. Comput Struct 41(3):
Adamu A, Karihaloo BL (1995b) Minimum cost design of RC 509–518
frames using the DCOC method Part II: columns under Kanagasundaram S, Karihaloo BL (1991b) Minimum-cost de-
biaxial bending actions. Struct Optim 10:33–39 sign of reinforced concrete structures. Comput Struct 41(6):
Adamu A, Karihaloo BL (1995c) Minimum cost design of RC 1357–1364
beams with segmentation using continuum-type optimality MATLAB R2006a (2006) Optimization toolbox 3 user’s guide.
criteria. Struct Optim 9:220–235 The MathWorks, Inc., Natick, MA
Adamu A, Karihaloo BL, Rozvany GIN (1994) Minimum cost Whitney CS, Cohen E (1956) Guide for ultimate strength
design of reinforced concrete beams using continuum-type design of reinforced concrete. ACI J 28(5):445–490; Nov.
optimality criteria. Struct Optim 7:91–102 (Proceedings V. 53)

You might also like