15
Steel & Composite Structures Proceedings of the 4th International Conference
generating the first and full yield surfaces is illustrated in section 5. Finally, two examples are
presented for demonstration of the validity, accuracy and advantages of the proposed method.
2. BASIC ASSUMPTIONS AND DEFINITIONS
In this paper, some basic assumptions are adopted in the crosssection analysis as: (1) Plane
sections before deformation remain plane after deformation, that means a linear strain
distribution exists across the depth of the section; (2) the bondslip between the concrete and
steel is not considered, it is assumed that full strain compatibility exists between the steel and
the surrounding concrete; (3) the tensile strength of concrete is neglected in computations; (4)
steel reinforcement embedded in concrete does not buckle under compression; (5) compressive
stresses and strains are taken to be positive and (6) the effect of concrete cracking or crushing
on the torsional stiffness is ignored.
In the formulation of the beamcolumn element, the following assumptions are taken: (1) The
EulerBernoulli hypothesis is valid and warping is neglected; (2) strains are small but the
deflection can be large; (3) plane section normal to the centroid axis before deformation remains
plane after deformation and normal to the axis; (4) the concept of lumped plasticity is employed,
i.e., the yielding of material is assumed to be concentrated at the both ends of beamcolumn
element and (5) loads are conservative and shear distortions are negligible.
3. GEOMETRIC NONLINEARITY
The P and P effects are two principal parameters needed to be considered in the second
order or advanced analysis. In the topic of formulating a design element capturing the P effect
of a member, Chan and Zhou [48] developed several elements with different features to simplify
the analysis procedure and make the advanced analysis practical. In this paper, the pointwise
equilibrating polynomial (PEP) element proposed by Chan and Zhou [4] is adopted. The PEP
element is capable of modelling the member initial curvature which is mandatory for buckling
design by secondorder analysis in various codes. The basic forcedisplacement relations in an
element are illustrated below and more details about its formulation can be referred to the
original papers. The equilibrium condition can be stipulated as follows.
Figure 2 : The basic forces vs. displacements relations in an element
16
Brian Uy, Zhong Tao, Fidelis Mashiri, Xinqun Zhu, Olivia Mirza & Ee Loon Tan
1
2 1
0
..
)
2
( ) ( M x
L
L
M M
v v P v EI +
+
+ + = (1)
in which E is the Youngs modulus of elasticity, I the second moment of area, L the member
length, v the lateral displacement due to applied loads, v
0
the initial member deflection, P the
axial force, and M
1
and M
2
the nodal end moments. A superdot represents a differentiation with
respect to the distance x along an element.
The secant stiffness matrix, which relates the equilibrium equations between forces, moments,
displacements and rotations, can be obtained by the energy principle. For incrementaliterative
nonlinear procedure, the tangent stiffness matrix which relates the incremental forces, moments
to rotations and displacements is needed and can be formed by the second variation of the total
potential energy functional.
The remarkable advantage of proposed advanced analysis by PEP element is its automatic
computation of primary linear and secondary nonlinear stresses such that the assumption of K
factor or effective length factor is avoided. The influence on member stiffness in the presence of
axial load is also allowed for in the stress computation and analysis. The first eigenvalue
buckling mode shape is used to determine the direction of local member imperfection and the
global frame imperfection due to outofplumbness which is normally taken in the range from
1/1000 to 1/200 of the building height H. The member initial imperfection can be obtained from
codes such as Table 5.1 in the Eurocode3 [2] and Table 6.1 in the HKSC [9] and the later
imperfections are adopted herein.
4. MATERIAL NONLINEARITY
4.1. Stressstrain constitutive relationship
To consider the inelastic behaviour of hybrid section, the constitutive relationships of steel and
concrete shown in Figure 3 are adopted. The steel material is assumed to be ideally elastic
plastic, while the stressstrain curve of concrete by Saenz [10] is adopted to consider the
material nonlinearity. The tensile strength of concrete is neglected as specified in many codes.
(a) Steel and reinforcement (b) Regular and confined concrete
Figure 3 : Stress vs. strain curve for steel and concrete
17
Steel & Composite Structures Proceedings of the 4th International Conference
The equation of the constitutive relationship for concrete is taken as,
in which
0
E ,
yield
and
yield
are the initial youngs modulus, initial yield strain and yield stress.
4.2. Crosssection properties
To evaluate the stress resultants in any shape section, the section can be divided into 3
components, i.e., the unconfined and confined concrete, steel and void areas. More details of
the stress resultants of each component is discussed in section 5.
(a) Overview
=
(b) Concrete
+
(c) Steel

(d) Void
Figure 4 : Components of cross section
To reflect the gradual strength reduction and stiffness deterioration, the first and full yield
surfaces should be identified for secondorder inelastic analysis. Here, the generalized stresses
such as axial force (P) and moments (My and Mz) are monitored in the nonlinear incremental
iterative process to check the state of section capacity.
For reinforced concrete and steelconcrete composite sections, the first yield surface is defined
as the strain of external fibre reaching the initial yield strain
yield
of concrete, while the full yield
surface is determined when the external fibre exceeds the strain
cu
of concrete.
The two yield surfaces divide the loading space into three zones, i.e., elastic, elastoplastic and
plastic hardening zone as shown in Figure 5. If the loading combination is within the first yield
surface, the section can be ideally treated as elastic without reduction in either strength or
stiffness. If the loaded coordinate falls into the elastoplastic zone, the section is under gradually
yielding stage and its strength and stiffness are discussed in the followings. Once the loading
combination exceeds the full yield surface, the sectional hardening appears.
0
2
0
1 ( 2)
s yield yield
E
E
E
=
+ +
( 2)
h
18
Brian Uy, Zhong Tao, Fidelis Mashiri, Xinqun Zhu, Olivia Mirza & Ee Loon Tan
Figure 5: First and full yield surfaces under given axial force
4.4. Plastic hinge method
For inelastic analysis, it is necessary to assume a yield function to monitor the gradual
plastification of a section. This refined plastic hinge approach introduced by Chan and Chui [11]
is revised and adopted in this study. The nodal rotations of a deformed PEP element with
pseudosprings at the end nodes are shown in
Figure 6.
Figure 6: Internal forces of a PEP element with endsection springs
The zerolength spring elements belonging to the internal degrees of freedom of beamcolumn
element can be eliminated by a standard static condense procedure such that the size of the
element stiffness matrix will not be increased. The final incremental stiffness relationships of the
hybrid element can be formulated below and more details can be referred to references 12 to14.
^
DD
Plastic moment
Elastic moment
Elastic zone
Plastic hardening zone
Elastoplastic zone
19
Steel & Composite Structures Proceedings of the 4th International Conference
2
1 1 1 22 2 1 2 12 1
2
2 1 2 21 1 2 11 1 2
/ 0 0
0 ( ) / /
0 / ( ) /
e e
e e
P EA L L
M S S K S S S K
M S S K S S K S
= +
+
(3)
in which S
1
and S
2
are the stiffness of the end springs, K
ij
are the flexural stiffness of the PEP
element considering the presence of axial force, P is the axial force increment,
e
M
1
and
e
M
2
are the incremental nodal moments at the junctions between the spring and the global node and
between the beam and the spring, L is the axial deformation increment,
e
and
e
are the
incremental nodal rotations corresponding to these moments, and
11 1 12
11 1 22 2 12 21
21 22 2
( )( ) 0
K S K
K S K S K K
K K S
+
= = + + >
+
(4)
The section spring stiffness, S, can be calculated by the following equation,
6
( )
pr
er
M M
EI
S
L M M
= +
( 5)
where EI is the flexural constant, L is the member length and M
er
and M
pr
are respectively the
first and full yield moments reduced due to the presence of axial force and the represents the
strainhardening parameter. From this equation, the section stiffness varies from infinity to a
small strainhardening value which represents three sectional stages, i.e., elastic, partially plastic
and fully plastic with strainhardening stages.
5. CROSSSECTION ANALYSIS
In this section, the numerical procedure of sectional analysis for arbitrary reinforced concrete
and encased composite sections are discussed.
The arbitrary section studied here is shown in Figure 7. Two variables, which are the orientation
n
and the depth d
n
of neutral axis, should be determined to calculate the stress resultants in the
cross section. By changing the
n
from 0
o
to 360
o
, a quasiNewtion iterative procedure is
introduced to determine d
n
under a given angle
n
.
20
Brian Uy, Zhong Tao, Fidelis Mashiri, Xinqun Zhu, Olivia Mirza & Ee Loon Tan
Figure 7 Arbitrary shape crosssection
n
21
Steel & Composite Structures Proceedings of the 4th International Conference
Figure 8 : Stress block of concrete in first yield stages
     
1
1
) (
1 1 1
+
+
= = =
= = =
i
i
j
j
v L L
v
v
u
u
c
i n
j
n
i
zci
n
i
zc zc dudv N P N (6)
+
+
= =
= = =
1
1
) (
1 1 1
i
i
j
j
v L L
v
v
u
u
c
i n
j
n
i
uci
n
i
uc vdudv M M
(7)
+
+
= = =
= =
1
1
) (
1 1 1
i
i
j
j
v L L
v
v
u
u
c
i n
j
n
i
vci
n
i
vc ududv M M
(8)
in which n
L
is the number of the compression layers, n
v
(i) is the number of iteration points in the
layers. is equal to +1 when
0 > zc P
and equal to 1 when
0 < zc P
.
It should be pointed out that the strain of the external layer is taken as the crushing strain
cu
in
the full yielding stage. Here, the equivalent stress block (see Figure 9) is introduced to calculate
the stress resultants.
Figure 9 : Stress block of concrete in full yield stages
     
1
) (
0
1 1
+
= =
= = =
i
i
c c
u
u
u v
c
n
i
zci
n
i
zc zc
v dud N P N (9)
22
Brian Uy, Zhong Tao, Fidelis Mashiri, Xinqun Zhu, Olivia Mirza & Ee Loon Tan
+
+ = =
= =
1
) (
0
1 1
)] ( [
i
i
c c
u
u
u v
n c
n
i
uci
n
i
uc
v dud v v M M
(10)
+
= =
= =
1
) (
0
1 1
i
i
c c
u
u
u v
c
n
i
vci
n
i
vc
v udud M M
(11)
in which n
c
is the number of vertices of the compression zone,
( ) v u
is the linear equation of
boundary line equal to v(u) v
n
with v
n
being the coordinate of the neutral axis in vaxis.
(2) Stress resultants in steel
Each rebar is treated as an individual fibre, while the structural steel will be meshed into several
fibres with rectangular areas. The stress resultants of steel section and reinforcement can be
computed as,
sk sk
ns
k
rj rj
nr
j
zs
A A N
= =
+ =
1 1
(12)
sk sk sk
ns
k
rj rj rj
nr
j
us
v A v A M
= =
=
1 1
(13)
sk sk sk
ns
k
rj rj rj
nr
j
vs
u A u A M
= =
+ =
1 1
(14)
In which the subscript r and s represent respectively reinforcement and steel fibres.
(3) Void area in the crosssection
The negative area approach is used to remove the void due to steel, rebar components or
opening in the section.
(4) Total force and moments sum up and transformation
The bending moments obtained from the above equations are summed and transformed to the
xoy system by the following transformations.
uo us uc u
M M M M + = ,
vo vs vc v
M M M M + = (15)
n v n u x
M M M sin cos = ,
n v n u y
M M M cos sin + =
(16)
23
Steel & Composite Structures Proceedings of the 4th International Conference
5.3. Iteration scheme
By rotating the orientation
n
of neutral axis from 0
o
to 360
o
with the change of depth d
n
, the
sectional capacity can be precisely determined as seen in Figure 7. The RegulaFalsi numerical
method is utilized to satisfy the equilibrium, compatibility, and constitutive relationships.
The axial force capacity N
z
is iterated with respect to d
n
by the following equation with
n
kept
unchanged.
) (
'
'
, z zd
z z
n n
n k n
N N
N N
d d
d d
+ =
(17)
where d
n,k
is the updated neutral axisi depth, d
n
and d
n
are the neutral axis depth where the
axial capacity is smaller and greater than the design value respectively, N
z
and N
z
are the axial
force capacity calculated at d
n
and d
n
and N
zd
is the current design axial loading.
The whole procedure of sectional analysis is illustrated in Figure 10.
Figure 10 : Flowchart of analysis procedure
yz Input data
Get the geometry centroid
Calculate N
z
, M
y
and M
z
n
d
z
N
zd
N
adjust
n
Initialize
n
Initialize d
n
yes
0
360
yes
Output stress resultants
End
24
Brian Uy, Zhong Tao, Fidelis Mashiri, Xinqun Zhu, Olivia Mirza & Ee Loon Tan
6. VERIFICATION EXAMPLES
6.1. Initial and full yield surfaces verified by design codes
In this example, a reinforced concrete section and an encased composite section shown in
Figures 11 (a) and 12 (a) respectively will be studied to verify the accuracy of the proposed
sectional analysis method. The sections are subjected to axial force (P) and uniaxial bending
moment (M). The PM interaction curves obtained from the proposed method, HKCC [15] and
HKSC [9] are shown in Figures 11 (b) and 12 (b).
(a) General dimensions (b) PM interation curve
Figure 11 : Sectional analysis of reinforcedconcrete section
(a) General dimensions (b) PM interation curve
Figure 12 : Sectional analysis of encased composite section
25
Steel & Composite Structures Proceedings of the 4th International Conference
For reinforced concrete section, the proposed method produces almost the same results as the
tedious hand calculation by code as seen Figure 11. It should be noted that the concrete area
occupied by reinforcement is commonly neglected in the hand calculation and therefore a
slightly unconservative result is obtained, especially for cases of high steel ratio. The proposed
approach uses the negative area method to remove the concrete areas occupied by rebar.
For encased composite section, PM interaction curve from the proposed method is also very
close to the result calculated by code as can be seen in Figure 12. It should be pointed out that
only four key points (i.e. A, B, C and D in Figure 12) are given in HKSC [9] to form the design P
M interaction curve. The small difference of the maximum bending moment (see point D)
between the proposed method and the design code is due to the use of different failure criteria.
In the proposed method, the failure of the section is controlled by the concrete strain at extreme
fibre, which is the basic assumption in many concrete design codes. However, the design
formula for the encased composite section in the steel code representing the failure of the
section is controlled by fully plastic state of all components. As a result, the proposed method
provides a more conservative result for design.
6.2. Numerical analysis for three simple portal frame
In this example, a simple portal frame with three types of sections, i.e., bare steel, reinforced
concrete and composite, is studied. The geometrical layout and the section properties are
shown in Figure 13 below.
Material Properties:
Concrete : C45 , fcu = 45 N/mm
2
, m = 1.5
Steel section: S355, fy = 355 N/mm
2
, m = 1.0
Reinforcement: R460, fy = 460 N/mm
2
, m = 1.15
Case Beam Column
1
2
3
Figure 13 : Geometry of portal frame and section propertities
26
Brian Uy, Zhong Tao, Fidelis Mashiri, Xinqun Zhu, Olivia Mirza & Ee Loon Tan
As discussed in Section 3, initial imperfections such as local member initial curvature and global
frame imperfection should be considered in a practical secondorder analysis. Here, the initial
member imperfections of 1/300 and 1/400 of member length are assumed for columns and
beams respectively, while the global imperfection is taken as 1/500 of building height. The
equivalent axial stiffness (EA)
e
and (EI)
e
are calculated by the following equations.
( )
e c c s s r r
EA E A E A E A = + + (18)
r r s s c c e
I E I E I E EI + + = 5 . 0 ) ( (19)
The loaddeflection curves of the top of right column for three different types of frame obtained
from the proposed secondorder inelastic analysis are plotted in Figure 14.
Figure 14 : Loaddeflection curves of portal frames
From Figure 14, it can be seen that ultimate load resistance of composite frame is much higher
than the other two frames and the bare steel frame shows lowest load resistance. This figure
also indicates that both steel and composite frames experience excellent ductility while
reinforced concrete frame shows relative poor ductility with maximum lateral deflection at about
175mm.
It should be emphasized that as both the P and P effects as well as initial imperfections have
been taken into account in the proposed secondorder analysis. Only the section capacity check
for cross section of a structural member is needed without assumption of effective length.. The
procedure leads not only to timesaving, but also safety as the error for assuming an effective
length is eliminated. Also, the tedious member design by different codes can be avoided since
the proposed sectional analysis method has the capability to exactly determine the yield and
failure surfaces automatically.
E
E
E
27
Steel & Composite Structures Proceedings of the 4th International Conference
7. CONCLUSIONS
In this paper, a practical secondorder analysis is proposed for design of hybrid frame structures.
By using an initially curved beamcolumn element in the robust nonlinear incrementaliterative
procedure, the secondorder effects of individual members and the structural system can be
modelled. Further, a divergenceproof iterative procedure is used to exactly calculate the yield
surfaces of arbitrary shape reinforced concrete and composite sections subjected to axial force
and biaxial bending. Thus, the proposed method checks the stability and section strength in the
process of structural analysis and as a result no additional individual member design by code is
needed. The remarkable advantage is that assumption of effective length for member buckling
checks is no longer required. Also, the member design by different codes is avoided and only
the material stressstrain curve is needed to check the section capacity.
The proposed method is further extended to inelastic analysis so that this method can be
applied to advanced analysis, plastic design, seismic design and progressive collapse. To
capture gradual yielding behaviour, the first and full yield surfaces of reinforced concrete and
steelconcrete composite section are computed and defined in the proposed method for a
secondorder inelastic analysis.
8. ACKNOWLEDGMENT
The authors acknowledge the financial support by the Research Grant Council of the Hong Kong
SAR Government on the projects Secondorder and Advanced Analysis and Design of Steel
Towers Made of Members with Angle Crosssection (PolyU 5115/08E ) and the project
Advanced analysis for progressive collapse and robustness design of steel structures (PolyU
5115/07E).
9. REFERENCES
u uu Lnv 1 L C
S
u uu Lnv 1 L C S
u uu Lnv 1 L
C S
C S L Z P Z l l NA l !
S LA
C S L Z P Z O
NluA ! C S 8
C S L Z P Z N
L S
Z Z P S L C 5 NO A 5 I l !
L MA
Z Z P S L C l
O ! S LA
c l l 1 5 u 5 8 u A 8 u
P k SA8 C
S L u u k ! ACl
C S L 1 C A
L S
28
Brian Uy, Zhong Tao, Fidelis Mashiri, Xinqun Zhu, Olivia Mirza & Ee Loon Tan
C W l S L C NO l A 5 l u M 5
! S LA
P W M C S L C 5 8A l c 5 l !
S LA
? C ? S L C l 5 A l c 5 l 55 M
! S LA
c l 5 u c 5 8 u P k SA8
C
29