Engineering Structures 162 (2018) 121–134

Contents lists available at ScienceDirect

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

Inelastic large deflection analysis of space steel frames using an equivalent T

accumulated element
⁎
Ahmed H. Zubydan , Ashraf I. ElSabbagh, Tarek Sharaf, Abbad-Elrahman Farag
Faculty of Engineering, Port-Said University, Egypt

A R T I C L E I N F O A B S T R A C T

Keywords: This paper presents a new model to perform inelastic large deflection analysis of space steel frames depending on
Inelastic the spread of plasticity method. A stiffness matrix of a beam-column element with only two nodes and six
Large deflection degrees of freedom for each node is derived to represent the space frame member. The proposed matrix includes
Space frame the effect of section yielding along the member as well as the effect of large deflection. Stiffness degradation at
Nonlinear analysis
the cross-section due to yielded parts is calculated using a formula for the tangent modulus which is affected by
Spread of plasticity
Beam-column
the sectional internal forces. The proposed technique of accumulation of rigidity factors is submitted as a first
step to get exact first order stiffness factors for a member with variable cross sections. The effect of large de-
flection is included by considering the axial force while deriving the stiffness factors. Both cubic and higher order
shape functions are tried to produce an element that can represent the member without discretization. A finite
element program based on stiffness matrix method is developed to predict the inelastic large deflection behavior
of steel space frames using the derived stiffness matrix. The proposed finite element technique exhibits good
correlation when compared with the conventional spread of plasticity model results. Verification by solving
benchmarked steel structures is carried out. The analysis results indicate that the new model is accurate with
simple equations and it achieves a significant improvement to the run time.

1. Introduction exact function developed by Chan and Gu [9]. In addition to the well-
known stability functions, equations for the lateral torsion buckling effect
The use of computers in structural analysis helped carrying out a huge were included by Kim et al. [10]. Cubic shape function was used by many
amount of computations to explore more realistic response of structures researchers developing cubic Hermite element [11–14]. For asymmetric
by including effects that were ignored before. Nowadays, advanced thin walled sections, Chan and Kitipornchai [14], employed the cubic
analysis including effects of geometric and material nonlinearity is al- element for non-linear analysis. Meek and Tan [11], derived the cubic
lowed in some specifications [1,2]. Geometric nonlinearity was studied by element using the principle of minimum potential energy using arc-length
many researchers through the last five decades. Instead of using k factors method for non-linear solving. Unlike the expressions of stability func-
to check member stability, many beam-column elements were submitted tions, the expressions of cubic-Hermite element are simple and similar in
to merge the second order effect in element stiffness matrix. Large de- both tension and compression. The disadvantage of the cubic-Hermite
flection analysis of frames using a beam-column element can find the element is its low accuracy when using one element for the member. Chan
structure manner with little amount of computations comparing to other and Zhou [15] derived a new element based on a fifth order shape
finite element nonlinear analyses. The stiffness coefficients derived by function. As an initial imperfection, the member out-of-straightness was
Oran [3,4] formed a beam-column tangent matrix of stiffness containing included by Chan and Zhou [16] during the formulation of a fifth order
large deflection effect by adding stability functions. Despite of the accu- element. The previous shape function was modified by Zhou and Chan
racy of Oran’s stability functions, problems may occur because of the [17] to include member lateral loads. Seismic response of imperfect
different equations used according to the case of normal force. To im- member was studied by Liu and Chan [18] using an element based on the
prove Oran’s matrix, Kassimaly and Oran [5] added coupling terms and same higher order function. Using one element per member, acceptable
studied the response under dynamic loads considering large deformation. accuracy was achieved employing a fourth order element by So and Chan
Many researchers used the stability functions for large deflection frame [19]. Elastic and inelastic buckling analysis were studied by Iu and
analysis [6–8]. Member initial out-of-straightness was represented by an Bradford [20–22] producing a new higher order element. Third additional

⁎
Corresponding author.
E-mail address: [email protected] (A.H. Zubydan).

https://doi.org/10.1016/j.engstruct.2018.01.059
Received 6 August 2017; Received in revised form 20 January 2018; Accepted 22 January 2018
Available online 16 February 2018
0141-0296/ © 2018 Elsevier Ltd. All rights reserved.
A.H. Zubydan et al. Engineering Structures 162 (2018) 121–134

node was added by Chen and Chan [23] to generate a beam-column

element with springs at the ends and mid-span. Liu [24] included the span
loads and arbitrary location for the span point with rotational spring to
model the yield along the member span.
Ordinary finite element can represent the inelasticity at any location
of the member with the help of member dividing to shell or solid element.
To avoid huge amount of calculation, techniques based on beam-column
element were proposed by many researchers [25–27]. For the plastic a
hinge and plastic zone approaches, concentrated plasticity is considered
during plastic hinge analysis. Abbasina and Kassimaly [28], used plastic
hinges for studying space frames with large deformations. Chandra et al. Monitored sections
Internal segments
[29], employed plastic hinges to model inelasticity while developing a
technique for non-linear solving. Liew et al. [30], calculated forces at five
points along the member and permitted plastic hinge to form at member Node 1
Node 2
span. While Shungyo [31], used plastic hinge to model semi-rigid con- EI1 EI2 EIk EIn
nections of space frames. For seismic analysis, plastic hinges achieved
good results by Liu et al. [18]. Because of the saved analysis time by
plastic hinges, it was developed to give more accurate results by con-
sidering gradual yielding [32,33]. Chen and Chan [23], used refined b
plastic hinge (RPH) to model plasticity. Zhou and Chan [34], modified
Fig. 1. Proposed equivalent element. a. Member with yielded parts. b. Two nodes
fifth order shape function to include one RPH with arbitrary position
equivalent element with (n) number of internal segments.
along the member. The work in [34] was extended by Zhou and Chan
[35], to find the displacement function of a member with three hinges.
Movable RPH for member with span loads was studied by Kim and Choi only one proposed element, the frame member will be represented in-
[36]. Using flexibility-based inelastic analysis, Chiorean and Marchis cluding plasticity spread along the span. Fig. 1 shows the proposed
[37], used plastic hinges for tapered elements without dividing the equivalent element which will be used to model space frame member
member. Fiber hinge assumes the member to consist of one elastic part with yielded parts. The proposed element assumes the member consists
between two inelastic parts. At the mid length of assumed inelastic part, of n number of internal segments. Every internal segment starts with a
dividing section to many fibers helps to model inelasticity [38]. Accuracy monitored section and ends with another one. During analysis steps, the
of fiber hinge and its effective factors were investigated to get more internal forces at any cross-section along the member can be obtained
realistic modeling [39,40]. The main disadvantage of plastic hinge ap- from nodal forces and displacement function. At any span section, the
proach, assumed concentrated plasticity, can be avoided by plastic zone calculated internal forces provide accurate indication of section mod-
analysis. By discretizing the member, inelasticity at any location along the ulus (EI) which degrades due to the plasticity spread. The value of EIk
span can be captured to get reference solutions [41,42]. More calculations for every internal segment can be found using the average value of
are required for plastic zone than plastic hinge but more accuracy is section modulus at start and end monitored sections.
gained [43]. Jiang et al. [43], submitted a mixed element to obtain ad- Section 2.3 is concerning about deriving stiffness coefficients of first
vantages of two approaches. Zubydan [44–46], suggested formulae to order moments neglecting axial force effects. Material nonlinearity is
evaluate tangent modulus instead of section discretizing to save analysis included with first order stiffness coefficients by updating the degraded
time. Initial imperfections were included by Du et al. [47], using flex- values of (EI) for the monitored sections according to the plasticity
ibility based analysis. Using one element per member, the distributed spread. The stiffness coefficients derived in Section 2.3 can be used for
plasticity can be captured in the flexibility based analysis but it costs inelastic frame analysis neglecting effects of second order due to axial
much computations for the required integrations [47,50]. Residual force. As the proposed element has varying values of EI for its internal
stresses were documented by specifications like ECCE [48]. To get more segments, the displacement functions cannot exactly approach the ac-
realistic modeling, the residual stresses should be included in the analysis tual behavior of the element. So, the stiffness coefficients would not be
because the plasticity spread along the member span increases in the accurate if they have been derived using the displacement functions.
presence of residual stresses [52]. Because of the advantages of re- The nodal moments can overcome this problem and provide accurate
presenting the frame member with one element, it has been examined for relations between the displacements and the corresponding forces.
steel-concrete composite frames by Faella et al. [53]. Therefore, the nodal moments are used to derive the first order stiffness
This paper suggests a new element intended to capture effects of coefficients as shown in Section 2.3.
distributed plasticity along member with large deflections using only To include second order effects due to axial force, displacement
one proposed element per member. A higher order displacement functions are used to find only the second order terms in the stiffness
function is employed to represent the large deflections of the proposed coefficients (Sections 2.4 and 2.5). Because the first order terms in the
element. Accumulation technique is suggested to form stiffness coeffi- stiffness coefficients are based on the nodal moments, the effects of the
cients for inelastic member. The proposed stiffness coefficients are approximations due to using the displacement function are trapped in
calculated using closed form expressions without numerical integra- the second order terms only. Second order effect is included in the
tions. The proposed two-nodes element aims to represent a space frame derivation of stiffness coefficients by using Cubic displacement function
member with distributed plasticity during an accurate analysis with less (Section 2.4) and, alternatively, fourth order displacement function
computational costs. (Section 2.5).

2. Numerical models 2.2. Basic assumptions

2.1. The equivalent element 1. Plane sections remain plane after deformation.
2. All members are assumed to be restrained against lateral torsional
For a loaded frame member, the spread of plasticity is assumed to be buckling.
distributed along the volume of the member. Rigidity degradation at 3. All cross-sections are bi-symmetric and have no local buckling.
span cross-sections reduces the stiffness of the overall member. Using 4. Only nodal loads are permitted.

122
A.H. Zubydan et al. Engineering Structures 162 (2018) 121–134

5. Small strains but large displacements and rotations are assumed.

6. Shear deformations and warping are neglected.
7. Steel strain hardening is not considered.

2.3. Exact first order degraded stiffness

As shown in Fig. 1, between any two steps, each element is sub-

divided into n number of internal segments whose rigidities are eval-
uated according to the plasticity produced by recent calculated forces.
The degraded member can be treated as an elastic member with vari-
able cross sections during any step of analysis. In the present model, Fig. 2. Equilibrium condition for frame element bent about z axis.
Castigliano's second theorem is used to find the end-displacement from
the partial derivative of strain energy respect to the corresponding n
1 1 3 3
force. Assuming the member is flexed about z-axis, as shown in Fig. 2 Smz 2 = L ∑ ⎛ (ξ −ξ )⎞
k=1
EIzk ⎝ 3 k k − 1 ⎠ (12)
neglecting second order effects, the flexural strain energy U can be
expressed as a function of the corresponding moment as follows: n
1 1 1
Smz3 = L ∑ ⎛− (ξ 2−ξ 2 ) + (ξk3−ξk3− 1)⎞
1 Mz2 k=1
EIzk ⎝ 2 k k − 1 3 ⎠ (13)
U= ∫0 2EIz
Ldξ
(1) By solving Eq. (9) and Eq. (10) together, expressions for the end
where Mz is the bending moment about z-axis at distance x which is bending moments as functions of the end rotations and the corre-
given as sponding stiffness factors are produced as follows:

x Mz1 = Cmz1 θz1 + Cmz3 θz 2 (14)

Mz = Mz1 (1−ξ )−Mz 2 ξ where ξ =
L (2)
Mz 2 = Cmz3 θz1 + Cmz 2 θz 2 (15)
The expressions of θz1 and θz2 can be found by differentiating the
where Cmz1, Cmz2 and Cmz3 are expressed as
expression of strain energy in Eq. (1) respect to Mz1 and Mz2 as follows:
−Smz2
∂U 1 M ∂Mz Cmz1 =
θz1 =
∂Mz1
= ∫0 EIz ∂Mz1
Ldξ
(3)
2
Smz 3−Smz1 Smz 2 (16)

−Smz1
∂U 1 M ∂Mz Cmz 2 = 2
θz 2 =
∂Mz 2
= ∫0 EIz ∂Mz 2
Ldξ
(4)
Smz −
3 Smz1 Smz 2 (17)

Smz3
After simple mathematical transformations, the following relation- Cmz3 = 2
Smz −
3 Smz1 Smz 2 (18)
ships can be derived:
It can be easily checked that, if the member has the same value of EI
1 (Mz1 (1−ξ )−Mz 2 ξ )
θz1 = ∫0 EIz
(1−ξ ) Ldξ
(5)
for all internal segments, the stiffness coefficients in Eq. (16), Eq. (17)
and Eq. (18) will lead to the conventional stiffness coefficients 4EI / L ,
4EI / L and 2EI / L , respectively. For space frames, the derived stiffness
1 (Mz1 (1−ξ )−Mz 2 ξ )
θz 2 = ∫0 EIz
(−ξ ) Ldξ
(6)
coefficients can be extended for y-direction by the same methodology.
The subscript m at stiffness and stability coefficients indicates that, the
Because of the variation of EI, the integration is performed segment- shown coefficients do not include the effects of geometric nonlinearity.
wise throughout the element length. For n number of segments, each k Only material yielding effects can be included by the different values of
segment has EIk and integration limits from ξk − 1 to ξk where k ranges EI at the internal segments.
from 1 to the number of internal segments n. Multiplying moment
equation with the corresponding derivative helps to separate the coef- 2.4. Second order secant stiffness employing cubic displacement function
ficients multiplied with Mz1 and Mz2 as follows:
n n To include the second order effects, the co-rotational moment
ξ ((1−ξ )2) ξ (−ξ )(1−ξ )
θz1 = Mz1 ∑ ∫ξ k EIzk
Ldξ + Mz 2 ∑ ∫ξ k EIzk
Ldξ equation should contain P-v effect. Regarding the frame element shown
k=1
k−1
k=1
k−1 (7) in Fig. 2, the bending moment about z-axis including the axial force
n n
effect is given as follows:
ξ (−ξ )(1−ξ ) ξ ((−ξ )2)
θz 2 = Mz1 ∑ ∫ξ k EIzk
Ldξ + Mz 2 ∑ ∫ξ k EIzk
Ldξ Mz = Mz1 (1−ξ )−Mz 2 ξ −Pv where ξ =
x
k=1
k−1
k=1
k−1 (8) L (19)
The integration can be performed which leads to expressions for θz1 Assuming a cubic displacement function, the value of v at any point
and θz2 as follows: along the member can be obtained using the well-known cubic function
as follows:
θz1 = Smz1 Mz1 + Smz3 Mz 2 (9)
v = (ξ −2ξ 2 + ξ 3) Lθz1 + (−ξ 2 + ξ 3) Lθz 2 (20)
θz 2 = Smz3 Mz1 + Smz 2 Mz 2 (10)
For the member with variable values of EI, the strain energy can be
where Smz1, Smz2 and Smz3 are expressed as represented similarly as the expression in Eq. (1) but employing
n equation of moment in Eq. (19) to include second order effects.
1 1
Smz1 = L ∑ ((ξ −ξ )−(ξk2−ξk2− 1) + (ξk3−ξk3− 1)) To employ Castigliano's second theorem, the partial derivative of
EIzk k k − 1 3 (11)
k=1 moment equation respect to Mz1 and Mz2 are needed. In Eq. (19), the

123
A.H. Zubydan et al. Engineering Structures 162 (2018) 121–134

first and second terms can be partially differentiated as the included in this element has the same geometric limitations of one cubic-Hermite
Eq. (5) and Eq. (6) respectively. For simplicity, the partial derivatives of element per member during geometric nonlinear analysis [12].
the third term in moment equation (Pv) with respect to Mz1 and Mz2 are
neglected. 2.5. Second order secant stiffness using fourth order displacement function
By substitution from Eq. (19) and Eq. (20) into Eq. (1), Eq. (3) and
Eq. (4) the expressions of θz1 and θz2 can be presented as follows: Using a higher order displacement function produces an element
∂U
n
ξk (Mz1 (1−ξ )−Mz 2 ξ −Pv )(1−ξ ) which can accurately capture the effects of geometric nonlinearity by
θz1 =
∂Mz1
= ∑ ∫ξ EIzk
Ldξ using only one element per member [15,19,22]. Fourth and fifth order
k−1 (21)
k=1
displacement functions were used by many researchers. Both fifth and
n
(Mz1 (1−ξ )−Mz 2 ξ −Pv )(−ξ ) fourth order are accurate, but the fifth is more accurate at very high
∂U ξ
θz 2 =
∂Mz 2
= ∑ ∫ξ k EIzk
Ldξ values of axial force [15,20]. During inelastic analysis, section failure
k=1
k−1 (22) occurs before the axial force get very high, especially in the presence of
To guarantee simple final stiffness expressions, the value of EI in the initial residual stresses [21]. Fourth order displacement function can
second order parts is assumed to be constant with equivalent value accurately capture the second order effects due to the moderately large
(EIeqz ). The equivalent value (EIeqz ) is submitted for second order terms values of axial loads which are the common during the inelastic ana-
in stiffness coefficients, while the first order terms include material lysis. The displacement function in the following equation was pro-
nonlinearity using the variable values of EI. The calculations of the posed by Iu and Bradford [22] for elastic large deflection analysis with
equivalent value (EIeqz ) is discussed in Section 4.3. The end rotations one element per member. While many researchers included the higher
about z-axis can then be expressed as order displacement functions with the plastic hinge analysis, the pre-
n n sent work includes a higher order displacement function with the dis-
ξ (1−ξ )2 ξ (−ξ )(1−ξ ) tributed plasticity analysis.
θz1 = Mz1 ∑ ∫ξ k Ldξ + Mz 2 ∑ ∫ξ k Ldξ
k=1 k−1 EIzk k=1 k−1 EIzk
4(24 + q) 2 48 + 5q 3 2q
1 (−v (1−ξ ) v = ⎡ξ − ξ + ξ − ξ 4⎤ Lθz1
+P ∫0 EIeqz
Ldξ
(23)
⎢
⎣ 48 + q 48 + q 48 +q ⎥ ⎦
⎡ 48−q 2 3(16−q) 3 2q
n n + − ξ + ξ + ξ ⎤ Lθz 2
4
ξ (−ξ )(1−ξ ) ξ ((−ξ )2) ⎢
⎣ 48 + q 48 + q 48 + q ⎥ ⎦ (28)
θz 2 = Mz1 ∑ ∫ξ k Ldξ + Mz 2 ∑ ∫ξ k Ldξ
k=1 k−1 EIzk k=1 k−1 EIzk
Using the fourth-order displacement functions, two methods are
1 (−v (−ξ )
+P ∫0 EIeqz
Ldξ
(24)
employed to derive the end-moments, end-rotations as explained in the
following two subsections.
Expressions for Mz1 and Mz2 can be obtained by solving Eq. (23) and
Eq. (24). The integration is performed in parts for first order terms 2.5.1. First method for employing the fourth order displacement function
considering different values of EI to include material nonlinearity. In- The displacement function in Eq. (28) can be included to drive the
tegrating second order terms generates axial force parameter q which is element instead of the cubic in Eq. (20). The present fourth order dis-
expressed as follows: placement function, in Eq. (28), is submitted as the expression of v in
Eq. (19) to form the equation for the bending moment Mz . The same
PL2
q= methodology in Section 2.4 is repeated using Castigliano's second the-
EIeqz (25) orem. Also, the assumptions about excluding ∂Pv / ∂Mz1 and ∂Pv / ∂Mz2 are
After the integration including the displacement function v in Eq. submitted. The expressions of θz1 and θz2 in Eq. (21) and Eq. (22) can be
(20), the relationships between end-moments and end-rotations are obtained using the fourth order displacement function and considering
expressed as follows: the equivalent value of rigidity EIeq for the terms matched to the axial
force parameter as follows:
EIeqz 2q EI
Mz1 = ⎛Cmz1 + ⎞ θz1 + ⎛Cmz3− eqz q ⎞ θz 2 n n
⎜

L 15 ⎠
⎟ ⎜

L 30 ⎠
⎟
ξ (1−ξ )2 ξ (−ξ )(1−ξ )
⎝ ⎝ (26) θz1 = Mz1 ∑ ∫ξ k Ldξ + Mz 2 ∑ ∫ξ k Ldξ + Sgz1 θz1
k=1 k−1 EIzk k=1 k−1 EIzk
EIeqz q EI
Mz 2 = ⎛Cmz3−
⎜
⎞ θz1 + ⎛Cmz 2 + eqz 2q ⎞ θz 2
⎟ ⎜ ⎟ + Sgz3 θz 2 (29)
⎝ L 30 ⎠ ⎝ L 15 ⎠ (27)
n n
where the terms Cmz1, Cmz2 and Cmz3 are still similar to the previously ξ (−ξ )(1−ξ ) ξ ((−ξ )2)
θz 2 = Mz1 ∑ ∫ξ k Ldξ + Mz 2 ∑ ∫ξ k Ldξ + Sgz3 θz1
derived terms in Eq. (16), Eq. (17) and Eq. (18), respectively. The used k=1 k−1 EIzk k=1 k−1 EIzk
equivalent rigidity EIeq in the second order terms is discussed later in + Sgz 2 θz 2 (30)
Section 4.3.
It’s noticed that, the derived expressions in Eq. (26) and Eq. (27) can where the terms Sg1, Sg2 and Sg3 are declared as follows:
be directly obtained by including the well-known geometric factors of
cubic displacement function to the proposed first order stiffness coef-
ficients in Eq. (14) and Eq. (15). At any step of analysis, if the member is Sgz1 = P ∫0
1 (−(ξ − 4(24 + q) 2
48 + q
ξ +
48 + 5q 3
48 + q
2q
ξ − 48 + q ξ 4 ) (1−ξ ) ) L dξ
2
EIeqz (31)
still elastic and has the same value of EI for all internal segments, the
expressions in Eq. (26) and Eq. (27) produce the well-known cubic-
Hermite element. For space frame, the derived equations can be also
extended to y-direction getting the equations for My1, My2 . Although the Sgz 2 = P ∫0
1 (−(− 48 − q 2
48 + q
ξ +
3(16 − q) 3
48 + q
ξ +
2q
48 + q
ξ4 ) (−ξ ) ) L dξ
2
EIeqz (32)
employing of cubic displacement function generates simple equations,

124
A.H. Zubydan et al. Engineering Structures 162 (2018) 121–134

Sgz3 = P ∫0
1 (−(ξ − 4(24 + q) 2
48 + q
ξ +
48 + 5q 3
48 + q
2q
ξ − 48 + q ξ 4 ) (−ξ ) ) L dξ
2
EIeqz (33a)
Or

Sgz3 = P ∫0
1 (−(− 48 − q 2
48 + q
ξ +
3(16 − q) 3
48 + q
ξ +
2q
48 + q
ξ 4 )(1−ξ ) ) L dξ
2
EIeqz (33b)
Expressions for Mz1 and Mz2 can be obtained by solving Eq. (29) and
Eq. (30), that produces Eq. (34) and Eq. (35). The integration is per-
formed in parts for first order terms employing different values of EI to
include material nonlinearity, but second order terms related to the
axial force parameter q are integrated employing one equivalent value
of rigidity (EIeq ). The final relations between moments and rotations are
expressed as follows: Fig. 3. Coefficients of bending stiffness Cii and Cij.

Mz1 = (Cmz1 + Cgz1) θz1 + (Cmz3 + Cgz3) θz 2 (34)

they are being modified here to be valid for inelastic second order
Mz 2 = (Cmz3 + Cgz3) θz1 + (Cmz 2 + Cgz 2) θz 2 (35) analysis. These expressions are combined as they contain first and
second order stiffness values. Second order factors are obtained in-
where,
dependently by subtracting the coefficients of first order stiffness from
EIeqz 1 ⎛ (q3 + 144q2 + 4608q) ⎞ the total stiffness coefficients in Eq. (40) and Eq. (41). The subtracted
Cgz1 = Cgz 2 = ⎜ ⎟
L 15 ⎝ (q + 48)2 ⎠ (36) first order stiffness factors are replaced with the derived in expressions
from Eq. (16) to Eq. (18) these can include material inelasticity. So, the
EIeqz 1 ⎛ (q3−2304q) ⎞ produced expressions for the stiffness factors cαii , cαjj and cαij in Eqs. (40)
Cgz3 = ⎜ ⎟
L 30 ⎝ (q + 48)2 ⎠ (37) and (41) are modified and given as follows:

where the terms Cmz1, Cmz2 and Cmz3 are the same like those in Eq. (16), EIeqα 1 ⎛ 8q3 + 1008q2 + 32256q ⎞
cαii = cmα1 + ⎜ ⎟
Eq. (17) and Eq. (18), respectively. The derived equations can be ex- L 105 ⎝ (q + 48)2 (42)
⎠
tended to y-direction getting the equations for My1, My2 .
EIeqα 1 ⎛ 8q3 + 1008q2 + 32256q ⎞
2.5.2. Second method for employing the fourth order displacement function cαjj = cmα 2 + ⎜ ⎟
L 105 ⎝ (q + 48)2 ⎠ (43)
This method depends on the expressions that were derived by Iu and
Bradford [22] using the fourth order displacement function in Eq. (28).
To produce elastic stiffness coefficients, they used Castigliano's first EIeqα 1 ⎛ 5q3−16128q ⎞
cαij = cmα3 + ⎜ ⎟
theorem to get bending stiffness coefficients in the expressions from Eq. L 210 ⎝ (q + 48)2 ⎠ (44)
(38) to Eq. (41). While employing Castigliano's first theorem in Ref.
[22], the strain energy equation included the fourth order displacement where α indicates axis y or z.
function in Eq. (28).
M1 = (Cii ) θ1 + (Cij ) θ2 (38) 3. Tangent stiffness matrix

M2 = (Cji ) θ1 + (Cjj ) θ2 (39) Material inelasticity is included by using the different values of (EI)
where the stiffness factors are given as follows: through the exact first order bending stiffness of the proposed equiva-
lent accumulated element. Second order effect is included separately
EI ⎛ 9216 + 3456q/5 + 68q2 /5 + 8q3/105 ⎞ using its term related to the equivalent value of rigidity EIeq . The se-
Cii = Cjj = ⎜ ⎟
L ⎝ (q + 48)2 ⎠ (40) parated stiffness factors produce more accurate and rapid force re-
covery as explained in Section 4.6.
EI ⎛ 4608 + 576q/5 + 2q2 + q3/42 ⎞ During this work, the verifications contain inelastic analysis using
Cij = Cji = ⎜ ⎟
L ⎝ (q + 48)2 ⎠ (41) three proposed models to check the ability of the new proposed tech-
nique to represent material and geometric nonlinearity. Models No.1
For elastic member with internal segments have the same value of and No.2 are based on fourth order displacement function in Eq. (28),
EI, the derived expressions from Eq. (34) to Eq. (37) can be compared while model No.3 is based on cubic displacement function in Eq. (20).
with the equations from Ref. [22] shown from Eq. (38) to Eq. (41). As
shown in Fig. 3, the fourth order displacement function produces more
accurate coefficients than the cubic. Also, elastic stiffness coefficients 3.1. Fourth order displacement models
using derived expressions in Eq. (34) and Eq. (35) have an obvious
agreement with those shown in Eq. (40) and Eq. (41) for a wide range of The tangent stiffness matrix derived by Iu and Bradford [22] is used
axial load factor. During inelastic analysis, the axial load value usually for Models No.1 and No.2 as declared in Eq. (45). Model No. 1 uses the
doesn’t reach such high values at which the disagreement can be sig- modified expressions for the stiffness factor from Eqs. (42) to (44) for
nificant. both directions while Model No. 2 uses expressions from Eqs. (34) to
The equations based on fourth order function and the exact stability (37) to generate the stiffness factors cαii , cαij and cαjj . While every model
functions from Ref. [4] don’t have significant difference for moderately has different equations for the main coefficients cαii , cαij and cαjj , the
large axial load factor. Differences between equations increase for coupling terms G1α , G2α and Hα are the same for the two fourth order
P / Pe > 2, where Pe = π 2EI / L2 . The range of loading at which P / Pe > 2 models. As declared before, the assumption of equivalent stiffness for
is often controlled by the material yielding during inelastic analysis. second order terms helps to use the same axial force equation for the
As Eq. (38) to Eq. (41) are valid for elastic second order analysis, two models.

125
A.H. Zubydan et al. Engineering Structures 162 (2018) 121–134

⎛c + EIeqz G12z EIeqz G1z G2z EIeqz G1z G1y EIeqz G1z G2y EIeqz G1z ⎞ 4. Procedure of analysis
zii czij + 0
⎜ L Hz L Hz L Hz L Hz L LHz ⎟
⎜ EIeqz G12z EIeqz G2z G1y EIeqz G2z G2y EIeqz G2z ⎟
⎜ czjj + 0 4.1. Section tangent modulus
L Hz L Hz L Hz L LHz ⎟
⎜ EIeqy G12y EIeqy G1y G2y EIeqy G1y
⎟
⎜ cyii + cyij + 0 ⎟
KT4 = L Hy L Hy L LHy During analysis steps, it’s required to evaluate section degradation
⎜ ⎟
⎜ EIeqy G12y EIeqy G2y ⎟ due to plasticity spread at the section according to corresponding
⎜ cyjj + 0
L Hy L LHy ⎟ forces. Moment curvature relationship gives indication about section
⎜ ⎟
⎜
symmetrical T 0
⎟ tangent modulus. As shown in Fig. 4, fiber model depends on discrete
EI
⎜
⎝ L3H
⎟
⎠
fibers and monitoring every fiber individually and integrating the fibers
to evaluate the cross-sectional behavior. Zubydan [44–46] suggested
(45)
accurate formulae to calculate tangent modulus directly without using
where, fiber model to reduce calculation time. The formulae employ equations
G1α = 2b1 (θ1α + θ2α ) + 2b2 (θ1α−θ2α ) (46) for section plastic strength surface developed at the same work. For-
mulae are valid for H and I section axially loaded and bent about minor
G2α = 2b1 (θ1α + θ2α )−2b2 (θ1α−θ2α ) (47) or major axes [44,45]. For sections bent about both axis and axially
loaded, formulae are valid for H sections only as shown in Fig. 5 [46].
13824/5 + 864q/5 + 126q2 /35 + q3/40 The chart in Fig. 5 can be found for the cross-section using the formulae
b1 =
(q + 48)3 (48) in [46], where EItr = EItanget / EIelastic , Pr = Papplied/ Pyield and
Mr = Mapplied / MPlastic . Formulae can calculate tangent modulus with ⧹
4608 + 672q/5 + 66q2 /35 + 11q3/840 without residual stresses. The submitted residual stresses at the for-
b2 =
(q + 48)3 (49) mulae are the recommended by European Convention for Construction
N =z.y
Steelwork (ECCS) [48]. Similarly, Refs. [44,45] contain equations to get
Iα Iα charts coefficients for uniaxially bent frames with H- or I-sections. The
Hα =
AL2
− ∑ IN
b2′ (θ1N −θ2N )2
N (50) references show that, formulae accuracy was verified and achieved
∂b2 −3072q − 258048 good correlations comparing to fiber model. Calculations and time can
where b2′ = =
35(q + 48) 4
.
∂q be saved using these formulae. For numerical examples during the
The equations of bowing in Eq. (48) and Eq. (49) derived by Iu and present work, formulae in [44–46] are employed to calculate tangent
Bradford [22] to represent the effects of member bowing on the inter- modulus while monitoring sections. Fiber model is used only for
action between the axial load and rotational deformations. Those biaxially bent frames with I-sections, D/B > 1.2, and for rectangular
equations of bowing are submitted here to calculate the length cor- sections. At any step of analysis, employing the formulae in [44–46]
rection factor due to member bowing, so the axial force is calculated as with the proposed stiffness matrix makes the stiffness matrix to be to-
follows: tally based on closed form expressions without integrations.
Δu
P = EA ⎛ ⎞+ ∑ b1 (θα1 + θα 2)2 + b2 (θα1−θα 2)2
⎝ L ⎠ α = yz (51) 4.2. Plasticity spread

where, Δu is the difference between the deformed chord length and the For frame nonlinear analysis, incremental iterative strategy is used
original chord length. and it is based on employing the tangent and the incremental secant
Torsional stiffness is assumed to be constant along the member and moduli for displacements and force recovery calculations. Plasticity
is calculated as follows: spread along the member is captured by reducing the elastic modulus
GJ + Pri2 for a number of monitored sections along the member without using
T= additional nodes. According to matrices in Eqs. (45) and (53), the ele-
L (52)
ment stiffness matrix can be formed for the member with sections
where GJ is the torsional rigidity and ri is the polar radius of gyration. having different values of EI, Fig. 1b. For updating stiffness matrix
Structure tangent stiffness matrix can be assembled as during analysis, a number (n) of cross-sections are considered to have
KT = ∑elements L (T T KT4 T ) LT , where matrices [T] and [L] are to form the various tangent modulus EI due to plasticity spread.
[12 × 12] matrix and to transform from local to global coordinates, The member is represented as a number n of internal segments.
respectively, can be found in Ref. [15]. Every internal segment k starts with a monitored section at distance

3.2. Third order displacement models

The tangent stiffness matrix for Model No 3 which depends on the

cubic displacement function is given as follows:

⎛ czii czij 0 0 0 0 ⎞
⎜ c zjj 0 0 0 0 ⎟
⎜ c yii c yij 0 0 ⎟
KT3 = ⎜ c yjj 0 0 ⎟
⎜ ⎟
symmetrical T 0
⎜ ⎟
⎜ EA ⎟
⎝ L ⎠ (53)

where the stiffness coefficients cαii , cαij and cαjj are generated from Eqs.
(26) and (27).
Structure tangent stiffness matrix can also be assembled as
KT = ∑elements L (T T KT3 T ) LT . Superscripts 4 and 3 in KT4 and KT3 indicate
the degree of displacement function employed to derive tangent stiff-
Fig. 4. Cross-sectional fiber model.
ness matrix.

126
A.H. Zubydan et al. Engineering Structures 162 (2018) 121–134

Fig. 5. Inelastic tangent modulus ratios for cross section under biaxial bending moment and axial compression force [46].

ratio ξk − 1 and ends with a monitored section at distance ratio ξk . Forces

at any monitored section along the member are calculated using the
forces at member ends and the displacement function. For any mon-
itored section, it is required to calculate Mz and My using Eq. (19) which
includes the second order effects. Using the calculated forces at any
monitored section, its tangent modulus EI can be calculated as declared
in Section 4.1. The same methodology is repeated for the other mon-
itored sections along the member so the different values of EI at all the
monitored sections can be found.
For k internal segment, tangent modulus EIk is the average of tan- Fig. 6. Two nodes element with internal monitored sections.
gent modulus values at the two monitored sections at ξk − 1 and ξk . The
same concept is repeated along the member by changing k from 1 to n value of EI as an equivalent value to calculate second order terms
number of internal segments. Matrices in Eq. (45) and Eq. (53) employ during inelastic second order analysis of frames with initial prismatic
values of EIk to calculate first order terms of stiffness coefficients. members.
Monitored sections can be distributed along the span or located close to
member ends, as shown in Fig. 6. For members analyzed in presence of 4.4. Incremental secant modulus of section
residual stresses, plasticity spreads along the member span so, the
monitored sections shall be distributed along the span. In the absence of To calculate member internal forces during analysis steps, incre-
residual stresses, the monitored sections shall be condensed close to mental method is employed. Accurate calculations of internal forces
member ends. lead to accurate analysis. While using tangent stiffness to calculate in-
ternal forces may overestimate the structure strength, as shown in
4.3. Member equivalent stiffness Fig. 7. An incremental secant technique developed by Zubydan [44] is
used through this work to eliminate the error due to this over-
For second order terms, an equivalent value of EI is used. To find estimation. In Fig. 7, for load increment from step j to step j + 1, mo-
the equivalent value EIeq , many methods can be suggested to get an ment and curvature at the two steps are Mj , M′j + 1, φj and φj′+ 1 respec-
approximate value representing EIeq for member with degraded stiff-
tively with the relation M′j + 1 = Mj + (dφEtrj EI ) . Where, dφ is the change
ness. An approximation is considered by equating the summation of
in curvature between the two steps and EI is elastic flexural rigidity.
discrete flexibility factors ∑ (Lk / EIk ) to the flexibility factor of the Using internal forces at step j, value of tangent modulus ratio Etrj is
equivalent member (L/ EIeq) . Another alternative approximation may be calculated. Small increments help to consider Etr changes linearly with
considered by equating the change of elastic load of the member with
Mr from step j to j + 1 according to Eq. (54).

variable stiffness ∑ (MLk / EIk ) to that for the member with equivalent Etr = C1 Mr + C2 (54)
′ + 1 − Etrj
Etrj
stiffness (∑ MLk )/ EIeq . where C1 = Mrj′ + 1 − Mrj
′ + 1 and Mrj′ + 1 are the tan-
and C2 = Etrj−C1 Mrj . Etrj
For prismatic element, the member has constant value of EI during gent modulus ratio and the bending moment at point j + 1. Using
elastic stage of loading. Values of EI begin to change only when plas- equations in [44], incremental secant modulus ratio Esr can be calcu-
ticity spreads at the late stage of loading. For second order terms, the 1
lated using Eq. (55) considering M′j + 1 = C Mp (e D + C2) .
1
initial elastic value of EI can represent the approximate value of EIeq as
it is exact for a wide range of loading. Force recovery by a secant 1 M′j + 1−Mj
Esr =
strategy enhances accuracy of EIeq using the initial elastic value, which EI φj′+ 1−φj (55)
is assured through numerical examples.
For EIeq computed using the elastic value of EI or other modified where Mp is plastic moment for the cross section and
φ′j + 1 − φj
values, analysis results are very close. This work employees the elastic D = C1 φy
+ ln(C1 Mrj + C2) .

127
A.H. Zubydan et al. Engineering Structures 162 (2018) 121–134

2. The solution of the equilibrium equations for the displacement in-

crements.
3. The determination of the new updated stiffness and member forces.
4. The check of convergence.

The updated stiffness contains cross section modulus for the mon-
itored sections which can be assembled as discussed in Sections 2 and 3.
The tangent stiffness matrix is updated at every iteration, and Minimum
Residual Displacement method is used for updating the load vector to
trace nonlinear path. During incremental analysis, structure geometry
including all accumulated deformations is updated to form equilibrium
equations. The used incremental secant stiffness technique, explained in
[51], provided the analysis procedure with the required relations be-
tween the incremental displacements and the coordinates transforma-
tion matrix.
Unbalanced forces can be found by comparing the calculated in-
ternal forces, Section 4.6, and the applied loads. The calculated un-
balanced forces are considered as applied loads in the new iteration of
the current step of analysis.

4.6. Internal force recovery

The obtained incremental displacements vector helps to calculate

incremental flexural strains dφz and dφy these are required to the cal-
culated cross section incremental secant modulus ratios Esrz and Esry as
Fig. 7. Moment curvature curve for cross section. discussed in Section 4.4.
Fig. 9 shows the algorithm of internal force calculation in the pre-
4.5. Numerical solution sent work. Only first order moments are calculated using incremental
secant method with the incremental displacements, this provides more
As shown in Fig. 8, the sequence of analysis is based on four main accurate representing of the material yielding. The second order mo-
stages: ments are calculated using the total secant stiffness as presented in
Sections 2.4 and 2.5 depending on the total displacements. The use of
1. The formulation of the stiffness matrix. EIElastic for the equivalent value EIeq is sufficient for calculating second
order moments using secant method as EIElastic is the exact value during

Prepare the reference load vector, initialize the

displacement vector for the first increment. j =1 (all
monitored sections start with the value of EIelastic )

• Calculate the tangent stiffness matrix for each element • Calculate the tangent modulus EItangent for the
according to the derived matrices in Section 3. monitored sections and for the internal segments as
• Assemble the structure tangent stiffness matrix . declared in Sections 4.1 and 4.2.

• Solve the equilibrium equation to get the incremental

displacement vector.
• Calculate the load incremental factor.
• Evaluate the displacement increment.
• Calculate the load increment for
Next iteration i

the next step j = j +1.

• Update the coordinates of the structure nodes and the • Update the load vector.
total displacements and update the applied load
vector.
• Update the values of EIsr incremental secant) for the)
Next increment

monitored sections and for the internal segments as

j=j+1

declared in Section 4.4.

NO

• Calculate the internal force vector according to

Section 4.6 and Fig. 9. Is tolerance YES
• Calculate the unbalanced force vector from the achieved ?
internal and external forces.

Fig. 8. Analysis flow chart.

128
A.H. Zubydan et al. Engineering Structures 162 (2018) 121–134

Fig. 9. Internal force calculations.

the elastic range of the total displacements. The numerical examples coefficients while Model No. 2 uses the expressions from Eq. (34) to
assure the accuracy of this proposed technique. (37). On the other hand, Model No.3 used the stiffness coeffecints based
The axial force and the torsional moment are calculated depending on cubic displacement function in Eq. (53).
on the total displacements and the cross-sectional elastic properties. As
the expression of axial force in Eq. (51) contains the parameter q which
5.1.1. H-shaped cross-sectional columns considering residual stresses
is a function of P, an iterative procedure is used to compute P in which
The ability of proposed models to represent hot rolled column
the value of member axial force P which calculated from the last
considering residual stresses is evaluated for different cases of loading
iteration is used to improve the parameter q which is re-employed in a
and various slenderness ratios. Fig. 10 shows a cantilever column with
new iteration. This iterative procedure of calculating P achieves con-
hot rolled steel section and it is analyzed in presence of residual
vergence rapidly with satisfying tolerance.
stresses. The cross section of columns is W 310 × 143. Material Young’s
modulus and yielding strength are E = 200,000 MPa and
σy = 250 MPa, respectively. The inelastic large deflection analysis
5. Numerical examples
contains cases of buckling about major and minor axis. For the shown
figures, the Euler’s critical load is calculated as Pe = π 2EIα /4L2 where, α
5.1. Cantilever columns
indicates the axis about which the curvature occurs while for columns
under biaxial curvature, it indicates the weak axis.
Cantilever response is studied to evaluate validation of proposed
Analysis results of proposed one element employing Models No. 1,
models during different levels of axial load and slenderness. Two dif-
No. 2 and No. 3 are compared with results when the member is divided
ferent steel column sections are studied, hot rolled considering residual
into ten equals elements. For the proposed models, monitored sections
stresses and hot rolled neglecting residual stresses. The three models are
are also distributed at ten equal distances to capture the plasticity
examined in the analysis. Models No. 1 and No. 2 are based on the
spread without additional nodes. The analysis results are presented in
stiffness coeffecints based on fourth order displacement function. Model
Figs. 11–16. The measured displacement in case of biaxial loading is the
No. 1 uses the expressions from Eq. (42) to (44) for the main stiffness
resultant of x and y displacements. The value of factor r controls

Fig. 11. Load-displacement relationships for cantilever column. (L = 4 m, Fx = r P and

Fig. 10. Cantilever column with W 310 × 143 cross-section. Fy = 0).

129
A.H. Zubydan et al. Engineering Structures 162 (2018) 121–134

Fig. 12. Load-displacement relationships for cantilever column. (L = 4 m, Fx = r P and Fig. 15. Load-displacement relationships for cantilever column. (L = 8 m, Fx = 0 and
Fy = r P). Fy = r P).

Fig. 13. Load-displacement relationships for cantilever column. (L = 6 m, Fx = r P and

Fy = 0). Fig. 16. Load-displacement relationships for cantilever column. (L = 10 m, Fx = r P and
Fy = 0).

second order effect and consequently more contribution of first order

moment. At higher values of r and corresponding lower values of axial
forces, the three proposed models have obvious accuracy. On the other
hand, at lower values of r and corresponding higher values of axial
forces, Models No.1 and No. 2 still have obvious accuracy, while
weakness of Model No. 3 appears.

5.1.2. H-shaped cross-sectional column neglecting residual stresses

The cantilever column analyzed by Thai and Kim [49] is reanalyzed
in the present paper neglecting residual stresses as assumed in [49]. The
cross section of the columns is W8 × 31 and its height is 5 m. Material
Young’s modulus and yielding strength are E = 200,000 MPa and
σy = 350 MPa, respectively. Because of concentrated plasticity, mon-
itored sections are assumed close to the yielding node. Monitoring an
additional section at 0.05L from the column base is found to be suffi-
cient to trace plasticity spread. As shown in Fig. 17, the proposed
models based on fourth order displacement function (Model No. 1 and
Fig. 14. Load-displacement relationships for cantilever column. (L = 8 m, Fx = r P and Model No. 2) predicts well the full nonlinear behavior of the column
Fy = 0). while Model No. 3 overestimates the column capacity due to the use of
cubic displacement function.
horizontal forces and consequently controls level of axial load without
changing member slenderness. Controlling value of r can change the 5.2. Portal frame with rectangular section
contribution of both geometric and material nonlinearity in the in-
elastic buckling. Increasing value of r permits less contribution of Thai and Kim [49] studied one story portal frame with rectangular

130
A.H. Zubydan et al. Engineering Structures 162 (2018) 121–134

31.7 KN/m
10.23 KN
IPE 240

HEB 200

HEB 160
20.44 KN 49.1 KN/m

HEB 160
HEB 200
IPE 300

6 x 3.75m
49.1 KN/m
20.44 KN
IPE 300

HEB 240

HEB 220
49.1 KN/m
20.44 KN

IPE 330

HEB 240

HEB 220
49.1 KN/m
20.44 KN

HEB 220
HEB 260
IPE 330

20.44 KN 49.1 KN/m

IPE 400

HEB 260

HEB 220
1/450
Fig. 17. Load-displacement relationships for Thai and Kim cantilever column.

2 x 6m
Fig. 19. Vogel’s six story frame.

residual stresses. This portal frame is a type of structures which have

more contribution of second order effects at the inelastic buckling. To
evaluate the proposed technique in capturing plasticity spread, Model
No. 3 well not used through the next examples to avoid a misleading
evaluation of the proposed technique.

5.3. Vogel’s six story frame

Many researchers studied Vogel’s six story frame [42] as a bench-

mark frame to verify their geometric and material nonlinear analysis.
Modulus of elasticity and yield strength of material are 205 GPa and
235 MPa , respectively. The frame has an initial out of plumpness 1/450
and cross sections are defined as shown in Fig. 19. Plastic zone analysis
by Vogel is considered a reference solution in which residual stresses
were considered. Barsan and Chorian [52] also studied the frame in-
cluding and excluding residual stresses. To model beams distributed
load, beams are modeled with four elements since the proposed model
assumes nodal loads only while columns are modeled with only one
element. In addition to nodal sections, internal cross-sections are
Fig. 18. Analysis results of portal frame with rectangular section. monitored at distances 0.02L, 0.05L and 0.1L from every member end
to capture plasticity spread. As shown in Fig. 20, the results show ef-
ficiency of proposed technique to trace geometry and material non-
section with Young’s modulus and yielding strength are
linear behavior either including or excluding residual stresses.
E = 19,613 MPa and σy = 98 MPa, respectively. Using ABQUS, it’s re-
quired 20 beam elements to reach a sufficient accuracy for geometric
and material nonlinearities [49]. The frame and analysis results are 5.4. Two story space frame with rectangular section
shown in Fig. 18. This frame is resolved using the proposed technique
by Models No. 1, No. 2 and No. 3. Plasticity spread is captured by The two-story frame shown in Fig. 21 is a common benchmark
monitoring two sections at distances 0.05L and 0.1L measured from structure which solved by many researchers to assure their techniques
each member end. A fiber model is used to find section modulus during of analysis. De Souza [50] used force based method to solve it. Other
the analysis. Although the axial force is considered high, member researchers used this frame as a verification like Abbasina and Kassi-
geometric and material nonlinearities are traced with high accuracy mali [28], Liu [24] and Thai and Kim [49]. The rectangular section of
using Models No. 1 and No. 2 considering only a single element per elements has no residual stresses, so the plasticity is expected to be
member. One element of Model No. 3 shows a lake of accuracy because condensed closed to nodes. In the present research, the frame is re-
of the axial force which is permitted to increase due to absence of analyzed using monitored sections at nodes and at distances 0.02L,
0.05L and 0.1L from member ends which can represent plasticity spread

131
A.H. Zubydan et al. Engineering Structures 162 (2018) 121–134

2P
2P
2P
P Roof
2P Level

2.2 m
P 2P
2P
2P
P 2P 2nd Floor
Level

1.76 m
Fig. 20. Load-displacement curve at Roof of Vogel’s six story frame.
2.5
m
3m
P/2 P/2
3P u P Fig. 23. Two-story space frame with H-shaped cross section.
P/2 P
P/2 P/2
3P B-B accurately. Although bending moment is maximum at nodes, another
P/2 monitored section can be located at the mid span of the member to
40 cm
400 cm

guarantee tracing any yielding in this position due to axial force and
2P
moment produced by P-δ effect. Fig. 22 shows comparison between
2P
20 cm analysis results by proposed technique and other researchers results.
P The proposed technique seems to have a high ability to capture geo-
2P 2P P/4 metric and material nonlinearities.
B
P A-A
400 cm
20 cm

P/4 5.5. Two story space frame with H-shaped cross section
B
40 cm The two-story space frame with H-shaped cross section was studied
A A
in details by Coung and Kim [39]. The frame cross section is
ı y = 98 MPa
z H150 × 160×10 × 6.5 and its dimensions are declared in Fig. 23, and
E = 19613 MPa force P equals 80 KN. Young’s modulus and yield stress of material are
y
221 GPa and 320 MPa, respectively. The frame has initial out of
cm

x
0

plumpness as detailed in Ref. [39]. Coung and Kim included the re-
30

400 cm sidual stresses proposed by (ECCE) [48] in the finite element model.
They used a fine mesh of shell elements which produced a very accurate
Fig. 21. Two-story space frame with rectangular section. analysis. Researchers like Liu [24] and Li et al. [40] also tried their
techniques to get solutions for the frame.
In the present work, the frame is reanalyzed using one element per
member using Models No. 1 and No. 2. The monitored internal sections
are located at distances 0.05L, 0.2L and 0.5L form member ends. The
presence of residual stresses makes inelasticity spread along wide range
of member length, that explains monitored sections distribution. The
results of analysis presented in Fig. 24 show a high accuracy of the
proposed technique to capture geometric and material nonlinearities of
the frame. The comparison between present results with others shows
the advantage of the proposed model and its ability to capture yielding
at any point which helped to achieve accurate results closed to the
reference solution [39].

5.6. Six story space frame

The benchmark frame in Fig. 25 was solved by many researchers

like Jiang et al. [43], Liew et al. [30], Iu and Bradford [21], Coung and
Kim [39] and Liu [24]. The analysis in [43] by Jiang et al. depends on
spread of plasticity technique. Liew et al. [30] used improved plastic
Fig. 22. Load-displacement curve for two-story space frame with rectangular section.
hinge analysis and assured their results by reanalyzing the structure

132
A.H. Zubydan et al. Engineering Structures 162 (2018) 121–134

W 1 0 x 60
W 10 x 6 0
W 1 0 x 60

W 10 x 6 0
6 x 3.658 m = 21.948 m

W 12 x 8 7
W 12 x 120
W 12 x 8 7

W 12 x 120
W 12 x 87

W 12 x 87
y
a
x
a

b b
Fig. 25. Geometry of six story space frame a. Three-dimensional view. b. Frame plan.
Fig. 24. Load-displacement curve for two-story space frame with H-shaped cross section
a. Roof level. b. Second floor level.

using a computer program for steel offshore structures (USOFS). Liu

[24] used three-hinges element with higher order element to capture
geometric and material nonlinearities of the members. Refined plastic
hinge was used by Iu and Bradford [21], while fiber hinge model was
used by Coung and Kim [39]. Frame geometry, cross sections definition
and loads are detailed in Refs. [30,43]. The frame is reanalyzed during
the present work using only one element per member using the pro-
posed Models No. 1 and No. 2. To capture material yielding along the
span, cross sections at distances 0.02L, 0.05L and 0.1L from member
ends are monitored in addition to a cross section at mid-span. The re-
sults declared in Fig. 26 shows the accurate analysis performed using
the proposed technique. Jiang et al. [43] needed to divide member into
9 elements while the proposed technique used only one element per
member.

6. Conclusion

The present work introduced a new element for frame inelastic large
deflection analysis. The proposed element can capture second order
Fig. 26. Load-displacement relationships of six story space frame.
effects and spread of plasticity along member span using only one

133
A.H. Zubydan et al. Engineering Structures 162 (2018) 121–134

element per member. The proposed element saves the cost of calcula- [22] Iu CK, Bradford MA. Second-order elastic finite element analysis of steel structures using a
tions as it is based on closed form expressions without using integra- single element per member. Eng Struct 2010;32:2606–16.
[23] Chen WF, Chan SL. Second-order inelastic analysis of steel frames using element with
tions along the element length. According to displacement functions, midspan and end springs. J Struct Eng ASCE 1995;121:530–41.
three models were presented to form stiffness matrix. One model em- [24] Liu SW. Second-order design and advanced analysis of hybrid steel and concrete framed
structures. Ph.D. dissertation. Department of Civil and Environmental Engineering, The
ployed cubic displacement function and the two other models employed Hong Kong Polytechnic University; 2013.
fourth order displacement function. Accuracy and validation of the [25] Foley CM, Vinnakota S. Inelastic behavior of multistory partially restrained steel frames.
three models were examined by solving many numerical examples. The Part I. J Struct Eng ASCE 1999;125(8):854–61.
[26] Ngo-Huu C, Nguyen PC, Kim SE. Second-order plastic-hinge analysis of space semi-rigid
results of analysis showed satisfying accuracy of the proposed element steel frames. Thin-walled Struct 2012;60:98–104.
to capture the distributed plasticity along the member span. Using the [27] Ziemian RD, McGuire W. Modified tangent modulus approach, a contribution to plastic
hinge analysis. J Struct Eng ASCE 2002;128:1301–7.
higher order displacement function enhanced the accuracy of re-
[28] Abbasina R, Kassimali A. Large deformation elastic-plastic analysis of space frames. J
presenting second order effects and overcame the disadvantages of the Construct Steel Res 1995;35:275–90.
cubic element. The proposed element avoided the unrequired amount [29] Chandra R, Krishna P, Trikha DN. Elastic-plastic analysis of steel space structures. J Struct
Eng ASCE 1990;116(4):939–55.
of calculations in the force based analysis. Also, the new element [30] Liew JYR, Chen H, Shanmugam NE, Chen WF. Improved nonlinear plastic hinge analysis
overcame the unrealistic concentrated plasticity in the plastic hinge of space frame structures. Eng Struct 2000;22:1324–38.
analysis with one element per member. As a future work, the new [31] Shugyo M. Elastoplastic large deflection analysis of three-dimensional steel frames. J
Struct Eng ASCE 2003;129(9):1259–67.
element shall be developed to include span loads and other effects like [32] Chen WF, Liu EM. Effects of joint flexibility on the behavior of steel frames. Comput
lateral torsional buckling and initial out of straightness. Struct 1987;25(5):719–32.
[33] Liew JYR, Chen WF. Implications of using refined plastic hinge analysis for load re-
sistance factor design. Thin-walled Struct 1994;20:17–47.
References [34] Zhou ZH, Chan SL. Elastoplastic and large deflection analysis of steel frames by one
element per member. I: one hinge along member. J Struct Eng ASCE 2004;130:538–44.
[35] Zhou ZH, Chan SL. Elastoplastic and large deflection analysis of steel frames by one
[1] AS 4100. Australian standard. Standards Association of Australia Steel Structures. AS
element per member. I: three hinges along member. J Struct Eng ASCE 2004;130:545–53.
4100-1998.
[36] Kim SE, Choi SH. Practical second order inelastic analysis for three-dimensional steel
[2] AISC. Specifications for structural steel buildings. American Institute of Steel
frames subjected to distributed load. Thin-Walled Struct 2005;43:135–60.
Construction; 2010.
[37] Chorian CG, Marchis LV. A second-order flexibility-based model for steel frames of ta-
[3] Oran C. Tangent stiffness in plane frames. J Struct Div, ASCE 1973;99(ST6):973–85.
pered members. J Construct Steel Res 2017;132:43–71.
[4] Oran C. Tangent stiffness in space frames. J Struct Div, ASCE 1973;99(ST6):987–1001.
[38] Kim SE, Kim Y, Choi SH. Nonlinear analysis of 3-D steel frames. Thin Wall Struct
[5] Oran C, Kassimali A. Large deflections of framed structures under static and dynamic
2001;39:445–61.
loads. Comput Struct 1976;6:539–47.
[39] Coung NH, Kim SE. Practical advanced analysis of space steel frames using fiber hinge
[6] Kassimali A, Abbasina R. Large deformation analysis of elastic space frames. J Struct Eng
method. Thin-Walled Struct 2009;47(4):421–30.
ASCE 1991;117:2069–87.
[40] Li TJ, Liu SW, Chan SL. Direct analysis for high-strength steel frames with explicit-model
[7] Ekande SG, Selvappalam M, Madugula MK. Stability functions for three-dimensional
of residual stresses. Eng Struct 2015;100:342–55.
beam-column. J Struct Eng 1989;115:467–79. ASCE.
[41] El-Zanaty M, Murrary D, Bjorhovde R. Inelastic behavior of multistory steel frames. In:
[8] Kim SE, Lee J. Improved refined plastic hinge analysis accounting for local buckling. Eng
Structural engineering report no. 83. Alberta, Canada: University of Alberta; 1980.
Struct 2001;23:1031–42.
[42] Vogel U. Calibrating frames. Stahlbau 1985;10:1–7.
[9] Chan SL, Gu JX. Exact tangent stiffness matrix for imperfect beam-column member. J
[43] Jiang XM, Chen H, Liew JYR. Spread-of-plasticity analysis of three-dimensional steel
Struct Eng ASCE 2000;126(9):1094–102.
frames. J Construct Steel Res 2002;58:193–212.
[10] Kim SE, Uang CM, Choi SH, An KY. Practical advanced analysis of steel frames con-
[44] Zubydan AH. A simplified model for inelastic second order analysis of planar frames. Eng
sidering lateral-torsional buckling. Thin-walled Struct 2006;44(7):709–20.
Struct 2010;32:3258–68.
[11] Meek JL, Tan HS. Geometrically nonlinear analysis of space frames by an incremental
[45] Zubydan AH. Inelastic second order analysis of steel frame elements flexed about minor
iterative technique. Comput Methods Appl Mech Eng 1984;47:261–82.
axis. Eng Struct 2011;33:1240–50.
[12] Teh LH. Cubic beam elements in practical analysis and design of steel frames. Eng Struct
[46] Zubydan AH. Inelastic large deflection analysis of space steel frames including H-shaped
2000;23(10):1243–55.
cross-sectional members. Eng Struct 2013;48:155–65.
[13] Chajes A, Churchill JE. Nonlinear frame analysis by finite element methods. J Struct Eng
[47] Du ZL, Liu YP, Chan SL. A second-order flexibility-based beam-column element with
ASCE 1986;113(6):1221–35.
member imperfection. Eng Struct 2017;143:410–26.
[14] Chan SL, Kitipornchai S. Geometric nonlinear analysis of asymmetric thin-walled beam-
[48] ECCS. Essentials of Eurocode 3 design manual for steel structures in building. ECCS-ad-
columns. Eng Struct 1987;9:243–54.
visory committee, vol. 5(65); 1991. p. 60.
[15] Chan SL, Zhou ZH. Pointwise equilibrating polynomial element for nonlinear analysis of
[49] Thai HT, Kim SE. Nonlinear inelastic analysis of space frames. J Construct Steel Res
frames. J Struct Eng ASCE 1994;120(6):1703–17.
2011;67:585–92.
[16] Chan SL, Zhou ZH. Second-order elastic analysis of frames using single imperfect element
[50] De Souza R. Force-based finite element for large displacement inelastic analysis of frames.
per member. J Struct Eng ASCE 1995;121(6):939–45.
Ph.D. dissertation. Department of Civil and Environmental Engineering, University of
[17] Chan SL, Zhou ZH. Refined second-order analysis of frames with members under lateral
California at Berkeley; 2000.
and axial load. J Struct Eng ASCE 1996;122:548–54.
[51] Chan SL. Large deflection kinematic formulations for three dimensional framed struc-
[18] Liu SW, Liu YP, Chan SL. Push over analysis by one element per member for performance-
tures. Comp Methods Appl Mech Eng 1995;192:17–36.
based seismic design. Int J Struct Stab Dyn 2010;10(01):111–26.
[52] Barsan GM, Chorian CG. Influence of residual stress on the carrying-capacity of steel
[19] So AKW, Chan SL. Buckling and geometrically nonlinear analysis of frames using one
framed structures. Numerical investigation. In: Dubina D, Ivany M, editors. Stability and
element/member. J Construct Steel Res 1991;20:271–89.
ductility of steel structures. Elsevier; 1999. p. 317–24.
[20] Iu CK, Bradford MA. Higher-order non-linear analysis of steel structures. I: elastic second-
[53] Faella C, Martinelli E, Nigro E. Analysis of steel-concrete composite PR-frames in partial
order formulation. Adv Steel Constr 2012;8(2):168–82.
shear interaction: a numerical model and some applications. Eng Struct
[21] Iu CK, Bradford MA. Higher-order non-linear analysis of steel structures. II: refined plastic
2008;30:1178–86.
hinge formulation. Adv Steel Constr 2012;8(2):183–98.

134