International Journal of Non-Linear Mechanics 46 (2011) 7995

Contents lists available at ScienceDirect

International Journal of Non-Linear Mechanics

journal homepage: www.elsevier.com/locate/nlm

Large-deection and post-buckling behavior of slender beam-columns with

non-linear end-restraints
Carlos Vega-Posada a, Mauricio Areiza-Hurtado b, J. Dario Aristizabal-Ochoa c,n
a
Department of Civil and Environmental Engineering, Northwestern University, Evanston, IL, USA
b
ACCIONA Infraestructuras, R&D Department, Composite Group, Madrid, Spain
c
School of Mines, National University of Colombia at Medelln, Colombia

a r t i c l e in fo abstract

Article history: The large-deection analysis and post-buckling behavior of laterally braced or unbraced slender beam-
Received 27 February 2008 columns of symmetrical cross section subjected to end loads (forces and moments) with both ends
Received in revised form partially restrained against rotation, including the effects of out-of-plumbness, are developed in a
16 July 2010
classical manner. The classical theory of the Elastica and the corresponding elliptical functions
Accepted 23 July 2010
utilized herein are those presented previously by Aristizabal-Ochoa [1]. The proposed method can be
used in the large-deection analysis and post-buckling behavior of elastic slender beam-columns with
Keywords: rigid, semi-rigid, and simple exural connections at both ends including linear and non-linear inelastic
Bars connections like those that suffer from exural degradation (such as exural cracking and elasto-plastic
Beams
connections) or exural stiffening. Only bending strains are considered in the proposed analysis. Results
Beam-columns
from the proposed method are theoretically exact from small to very large curvatures and transverse
Buckling
Columns and longitudinal displacements for laterally braced or unbraced slender beam-columns under bending
Composites caused by end loads. The large-deection analysis and post-buckling behavior of slender beam-columns
Cracking with both supports partially restrained against rotation and with sway inhibited or uninhibited are
Large deections complex problems requiring the simultaneous solution of two coupled non-linear equations with
PD analysis elliptical integrals whose unknowns are the limits of the integrals. The validity of the proposed method
Post-buckling and equations are veried against solutions available in the technical literature. Three comprehensive
Non-linear analysis
examples are included that show the effects of linear and non-linear connections at both ends on the
Second-order analysis
large-deection analysis and post-buckling behavior of slender beam-columns.
Semi-rigid connections
Stability
& 2010 Elsevier Ltd. All rights reserved.

1. Introduction

The non-linear large-deection analysis and post-buckling behavior of bars and beam-columns are of great importance in physics,
structural engineering and engineering mechanics particularly when they are slender and made of high-strength elastic materials like
FRP or composites. This problem has been investigated using the exact expression for the curvature in the differential equation of the
deection curve (i.e., the Elastica approach, Timoshenko and Gere [2], pp. 7682) or using the approximate second-order analysis such
as the Finite Element Method (FEM) with large deections and with or without large strains ([36], among others).
The exact post-buckling elastic behavior of a perfect beam-column under a gravity load is known; however, its response under other
load combinations and the relationships between the applied axial load and the transverse deections, particularly when the member
suffers from imperfections (such as lack of straightness, out-of-plumbness or eccentricity in the applied axial loads) become even more
complicated [7]. Khamlichi et al. [8] presented the different formulations to the solution of the large-deection problem of a perfectly
hingedhinged elastic bar without sidesway under end axial loads, and discussed the effects of the axial strains and shear deformations
using the asymptotic expansion technique on the post-buckling behavior.
Aristizabal-Ochoa [9] developed an approximate algorithm based on the classical Timoshenko stability functions for the large-
deection small-strain analysis of beam-columns with semi-rigid connections including the effects of out-of-plumbness, axial strains and
lateral bracing. Recently, Aristizabal-Ochoa [1] presented an analytical method for the non-linear large-deection stability of a slender
beam-column of symmetrical cross section with semi-rigid connections (linear and non-linear) and sidesway uninhibited under end

n
Corresponding author.
E-mail address: jdaristi@unal.edu.co (J. Dario Aristizabal-Ochoa).

0020-7462/$ - see front matter & 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijnonlinmec.2010.07.006
80 C. Vega-Posada et al. / International Journal of Non-Linear Mechanics 46 (2011) 7995

Nomenclature x0 a and y0 a deections of end A along the x0 - and y0 -axis,

respectively (Fig. 1a)
The following symbols are used in this publication:
ka and kb stiffness of the exural connection at ends A and B,
respectively
A cross-sectional area of beam-column member l + yo angle between the x- and x0 -axis
E Youngs modulus of the material x current angle between the tangent to the elastic curve
h initial length of the beam-column AB at s and the x0 -axis
I principal moment of inertia of the beam-column xa and xb end slopes of member AB measured with respect to
cross-section about its axis of bending the x0 -axis (Fig. 1b)
M bending moment at s along the beam-co- jcon relative rotation at the connection
lumn  Pa(y0 a  y0 )  Ma jcona and jconb relative rotation at ends A and B, respectively
Ma total bending moment applied at A joa and job plastic rotation of reference at ends A and B,
Mcon moment applied at the connection respectively
Mua and Mub ultimate moment applied at A and B, respectively ra and rb xity factors of the linear exural connections at
na and nb shape parameter of non-linear connections at A and ends A and B, respectively
B, respectively s bending stress at s along the beam Mz/I
Pa applied load at A yo initial out-of-plumb angle with respect to the beam-
Pe Pcr Euler load column axis (Fig. 1a)
Ra and Rb stiffness indices of the exural connections at ends A
and B, respectively

loads including the effects of axial load eccentricities and out-of-plumbness using the classical theory of the Elastica. Phungpaingan
and Chucheepsakul [10] investigated the post-buckling behavior of a simply supported elastic column with various end elastic rotational
restraints subjected to compressive end axial load. More recently, Aristizabal-Ochoa [11] developed a complete algorithm based on the
modied Timoshenko stability functions for the large-deection analysis and post-buckling behavior of beam-columns with semi-rigid
connections subjected to conservative and non-conservative end loads including the effects of out-of-plumbness, shear deformations, the
induced transverse component of the applied axial force, axial strains and lateral bracing.
The main objective of this publication is to present the large-deection analysis and post-buckling behavior of slender beam-columns
of symmetrical cross sections with both supports partially restrained against rotation and sidesway totally inhibited or uninhibited
subjected to end loads (forces and moments) including the effects of out-of-plumbness. The classical theory of the Elastica and the
corresponding elliptical functions are utilized as proposed by Aristizabal-Ochoa [1]. The main limitation of the Elastica is that only
exural strains are considered. The proposed method can be used in the large-deection elastic analysis of slender beam-columns with
rigid, semi-rigid (including non-linear inelastic connections that suffer from exural degradation or exural stiffening including elasto-
plastic connections), and simple connections at both ends. The large-deection stability analysis of slender beam-columns with both
supports partially restrained by rotational springs of different stiffness and with sidesway totally inhibited or uninhibited becomes a
complex problem requiring the simultaneous solution of two coupled non-linear equations with elliptical integrals whose unknowns are
the limits of the integrals. The validity and effectiveness of the proposed method and equations are veried against two different
methods proposed by Aristizabal-Ochoa [9,11] and with analytical results reported recently by Gurel and Kisa [12] and Phungpaingan
and Chucheepsakul [10]. Three examples are included that show the effects of linear and non-linear rotational connections at both ends
on the large stability analyses and post-buckling behavior of slender beam-columns.

2. Structural model

Assumptions: Consider the 2-D prismatic beam-column that connects points A and B shown in Fig. 1a. The element is made up of the
beam-column itself AB, and the exural connections AA and BB at the top and bottom ends, respectively. The exural connections AA

Fig. 1. Beam-column with semi-rigid connections at both ends and with initial out-of-plumbness yo: (a) structural model and applied force Pa and moment Ma and
(b) coordinate systems and notation.
 C. Vega-Posada et al. / International Journal of Non-Linear Mechanics 46 (2011) 7995 81

and BB have stiffness (given in force  distance/radian) in the plane of bending of the beam-column. It is assumed that the member AB:
(1) bends about one of the principal axis of its cross section with a moment of inertia I and span h; (2) is made of a homogeneous linear
elastic material with a modulus of elasticity E; (3) its centroidal axis is a straight line with an initial out-of-plumbness yo with respect to
the vertical axis; (4) is loaded in the plane of bending at A with Pa and Ma both of which can vary independently.
For convenience the following two parameters are introduced:
1 1
ra and rb 1a2b
1 3EI=h=ka 1 3EI=h=kb
where ra and rb are called the xity factors. Since for linear elastic connections the xity factor can vary from 0 (for perfectly hinged
connections) to 1 (for perfectly clamped connections), it is more convenient to use in the elastic analysis of structures with semi-rigid
connections [1,9,11].

3. Proposed equations

3.1. Beam-column with non-linear semi-rigid connections at both ends

The three-parameter model, as described by Chen et al. [13], is used in the non-linear momentrotation relationships of the end
connections AA0 and BB0 as shown in Fig. 2. This model is dened by Eq. (2) as follows:
For jcon =jo and Mcon =Mu 4 0:
Mcon jcon =jo
2
Mu 1 jcon =jo n 1=n
Applying Eq. (2) at end A (i.e., sh):
Mcona jcona =joa
3
Mua 1 jcona =joa na 1=na
Using Eq. (40) derived in the Appendix:
 
dx 2bk Ma
cos ca  4
ds a h EI
Substituting Eq. (3) into Eq. (4):
Mua jcona =joa tan ca
b 5
2 EIh 1 jcona =joa na 1=na sin xa =2

1.2
a

c
1.0

0.8
b
Mcon
Mu

0.6

0.4

0.2

0
1 2 3 4 5
con
u

Fig. 2. Linear, non-linear and elasto-plastic bending connections.

82 C. Vega-Posada et al. / International Journal of Non-Linear Mechanics 46 (2011) 7995

Applying Eq. (2) at end B (i.e., at s0):

Mconb jconb =job
6
Mub 1 jconb =job nb 1=nb
Knowing that Mb Ma Paya0 and Pa b2EI/h2 and using Eq. (48) derived in the Appendix:
   
jconb =job Mub jcona =joa Mua 2b EI=h sin xa =2 cos ca cos cb
7
1 jconb =job nb 1=nb 1 jcona =joa na 1=na sin ca

C
Calling : jconb 8
1C=job nb 1=nb

where
0 1
B jcona =joa Mua 2bEI=hsinxa =2cos ca cos cb C job
C @h  na i1=na A
sin ca Mub
1 jcona =joa

Using Eq. (37) derived in the Appendix

xa xb
sin cb sin sin ca sin 9
2 2
Since jconb l + yo  xb and replacing it into Eq. (8)
(    1=nb )
sin cb x 1 C nb
sin a sin l yo C= 1 10
sin ca 2 2 job

Notice that jcona xa  yo  l at end A (i.e., s h); and jconb l + yo  xb at end B (i.e., s0).
Eqs. (5) and (10) represent the proposed solution to the problem of stability at large deections of a beam-column with non-linear end
connections and with sidesway uninhibited between its ends.
Linear and elasto-plastic end connections can also be modeled using expression (2) as follows:

1) For linear elastic end connections: By making n N, jcon o jo and Mu kjo (where k is the stiffness of the exural connection), Eq.
(2) becomes Mcon kjcon which represents a linear elastic semi-rigid connection. Now using expressions (1a) and (1b), Eqs. (5)
and (10) are reduced to

3ra tan ca
b x yo l 11
21ra sinxa =2 a


sin cb x 1 r 1rb 2b1rb cos ca cos cb sinxa =2
sin a sin yo l a yo xa  12
sin ca 2 2 rb 1ra 3rb sin ca

Eqs. (11) and (12) represent the solution to the problem of stability at large deections of a beam-column with linear elastic
connections at both ends and with sidesway uninhibited between the ends. This solution is identical to that proposed by
Aristizabal-Ochoa [1].
2) For elasto-plastic end connections: By making nN the values of Mu and jo represent the moment and the corresponding rotation
at yielding as shown in Fig. 2.
Once the values of ca and cb are determined using the pair of Eqs. (5) and (10) or Eqs. (11) and (12) for members with non-linear
and linear connections, respectively, the values of b, Pa, xua and yua are calculated using Eqs. (13)(16) as follows:
Z ca
dc
bk, ca , cb q Fk, ca Fk, cb 13
cb 1k2 sin2 c

2 EI
Pa b 14
h2
Z ca q!
h
xua 2 1k2 sin2 c h 15
b cb

hk
yua 2 cos ca cos cb 16
b
where k sinxa =2=sin ca sinxb =2=sin cb . Notice that Eqs. (13), (15) and (16) are identical to Eqs. (42), (47) and (48) derived in
the Appendix, respectively.
 C. Vega-Posada et al. / International Journal of Non-Linear Mechanics 46 (2011) 7995 83

3.2. Proposed numerical algorithm

The following steps are suggested in the iterative procedure in the solution of Eqs. (5) and (10) or Eqs. (11) and (12) for beam-columns
with non-linear or linear elastic connections, respectively:

1. Choose the value of xa (i.e., the angle made by the tangent at end A with respect to the x0 -axis).
2. Using a graphical method plot both functions:
Z ca
dc Mua jcona =joa tan ca
F1ca, cb q EI na 1=na sinx =2
cb 1k2 sin c 2 2 h
1 j cona =j oa  a

and
(   nb 1=nb )
sin cb x 1 C
F2ca, cb sin a sin l yo C= 1
sin ca 2 2 job
using the same axis for F1 and F2 and then determine the traces of the surfaces F1 and F2 on the cacb-plane (i.e., F1F20).
Then the intersections of these two traces become the solutions to the system of non-linear equations (5) and (10) that
govern the problem.
For the case of linear elastic connections:
Z ca
dc 3ra tan ca
F1ca , cb q yo xa
cb 2 2 21ra sinxa =2
1k sin c

sin cb x 1 r 1rb 2b1rb cos ca cos cb sinxa =2
F2ca , cb sin a sin l yo a yo xa 
sin ca 2 2 rb 1ra 3rb sin ca

Then using the NewtonRaphson method the point on the cacb-plane closest to the origin (i.e., ca cb 0) that satises
both equations F10 and F20 is determined. The Gaussian quadrature method with 8 interpolation points is used to
Rc q
evaluate the integral c a dc= 1k2 sin2 c in both cases.
b

3. Once ca and cb are determined, the corresponding values of b, P, xua and yua can be calculated using Eqs. (13)(16),
respectively as well as the bending moment and shear force along the member.

Note: The analyst must be aware that the traces of both functions F1 and F2 on the cacb-plane are families of numerous curves
presenting multiple solutions, as shown by Fig. 3. This is because F1 is a function of tan ca and F2 of sin cb/sin ca and cos ca  cos cb,
respectively. Although each one of the solutions (i.e., point 1, 2, 3 and 4 in Fig. 3) yields to different values of applied load Pa, the solution
point (ca, cb) for a particular value of xa is that corresponding to the lowest load Pa and must be generally closest to the origin of the
cacb-plane (i.e., ca cb 0). The lowest applied load Pa corresponds to the rst mode of buckling and the higher loads correspond to
higher buckling modes. To determine this solution point (ca, cb) for a particular value of xa, a bi-dimensional iterative NewtonRaphson
method with a suggested initial point (cao, cbo) is used. The convergence of this method is improved substantially if the initial point
(cao, cbo) is located close to the solution point and near to the origin of the cacb-plane.

o
Fig. 3. Example 1A: Traces of functions F1 and F2 on the cacb-plane (initial solution for ra 0.50, rb 0.90 and xa 20 % closest to the origin is shown within a circle).
84 C. Vega-Posada et al. / International Journal of Non-Linear Mechanics 46 (2011) 7995

The bi-dimensional iterative NewtonRaphson method uses the following expression in the (i + 1) iteration:

ci 1 ci J1

i ci
2 3
@F1 @F1
" #
ca 6 @c @cb 7
6 a 7
where ci ; Ji 6 for i1, 2, y, n.
cb 4 @F2 @F2 7
5
i
@ca @cb
i
The partial derivatives for F1 or F2 with respect to ca and cb are as follows:
@F Fca Dca , cb

@ca Dca
@F Fca , cb Dcb
and with Dca Dcb 0.001.
@cb Dcb
The Gaussian quadrature method with 8 interpolation points is also used to evaluate the integral equation (15).
Note: The authors recently learned that there is an exact numerical expression to solve the elliptic integrals of rst and second kind
[16]. Although the results obtained using the Gaussian quadrature method with 8 points can be considered accurate, considering the
number of signicant gures used for the value of xua andyua (maximum 4), it is up to the analyst to select the most desirable method.

3.3. Approximate method by Aristizabal-Ochoa

The results of the lateral deection of the top end of the member and the elastic curve obtained with the proposed model and
equations are compared with those using the approximate method and equations proposed by Aristizabal-Ochoa [9] which is much less
cumbersome since it requires the solution for y only using the following equation:
 
Ma Mb Hz
P 1 2 2 sin y 0 17
h 4Pt h
q
where j Pt =EI=h2 ; Pt P cos y; and
If Pt Z0 (compression)
8  h i 9
<3ra 1rb 3rb tanj=2=f yyo EIh f2 1ra 1rb f2 3rb 1f=tan f M =
Ma  18
: 1r 1r f2 3r r 2r r 1f=tan f 9r r tanf=2=f=21 ;
a b a b a b a b

( )
2
3rb 1rb 3ra tanf=2=fyyo EI=hf 3rb 1ra 1f=sinfM
Mb  2
19
1ra 1rb f 3ra rb 2ra rb 1f=tan f 9ra rb tanf=2=f=21

Hz fMa2 Mb2 1=tan f f=sin2 f2Ma Ma 2 2fMa Mb 1 f=tan f=sin f 20

If Pt o0 (tension)
( )
2 2
3ra 1rb 3rb tanhf=2=fyyo EI=hf 1ra 1rb f 3rb 1f=tanhfM
Ma 2
21
1ra 1rb f 3ra rb 2ra rb 1f=tanhf 9ra rb tan hf=2=f=21

( )
2
3rb 1rb 3ra tanhf=2=fyyo EI=hf 3rb 1ra 1f=sinh jM
Mb 2
22
1ra 1rb f 3ra rb 2ra rb 1f=tanh f 9ra rb tanhf=2=f=21

2
Hz fMa2 Mb2 1=tanh f f=sinh f2Ma Ma 2 2fMa Mb 1 f=tanh f=sinh f 23
The deected shape of the member with respect to its chord (i.e., the elastic curve) can be obtained using one of the following
expressions [9]:
If Pt Z0 (compression)
yu Mb =Pt sinf1xu=h=sin f1xu=hMa =Pt sinfxu=h=sin fxu=h 24
If Pt o0 (tension)
yu Mb =Pt sinhf1xu=h=sinh f1xu=hMa =Pt sinhfxu=h=sinh fxu=h 25
The deections x, y along the member (i.e., the deected shape) with respect to the axis parallel to the axis of the applied force P are
calculated using the following expressions:
x xu cosR p=2 yu sinR p=2 26

y xu cosR p=2yu sinR p=2 27

1
where R tan yua =xua
 C. Vega-Posada et al. / International Journal of Non-Linear Mechanics 46 (2011) 7995 85

Fig. 4. Example 1A: Large-deection analysis of a vertical column with linear semi-rigid connections at both ends. The Elastica for different angles xa at the top and
deections xa/h and ya/h in terms of P/(p2EI/h2) for: (a) ra 0.25, rb 0.50; (b) ra rb 0.50; and (c) ra 0.50, rb 0.90.
86 C. Vega-Posada et al. / International Journal of Non-Linear Mechanics 46 (2011) 7995

4. Comprehensive examples and verication

4.1. Example 1. Large-deection stability of an unbraced column with both ends unequally restrained with linear, non-linear, and elasto-plastic
exural connections subjected to gravity load at the top (step-by-step calculations)

Consider an unbraced column subjected to a vertical gravity load at the top end with both ends unequally restrained. Assume that the
column is initially: A) vertical (yo l 0) and B) out-of-plumb with yo 0.01 radians. Determine the large-deection stability behavior
(i.e., the variations of xa/h and ya/h of the top end A against the applied axial load P) and the Elastica assuming that the member is
partially restrained against rotation with linear, no-linear and elasto-plastic springs of initial stiffness ka and kb at ends A and B,
respectively. Compare the calculated results for the particular case of linear elastic springs with those obtained using the approximate
method proposed by Aristizabal-Ochoa [9] where ka 3 EI/h and kb 27 EI/h.
Solution:

0
A) Initially vertical column (ho 0). To gure out the values of ca, cb, b, P, xa ,yua and the Elastica (i.e., x and y values along the member)
for a given beam-column (h, E, I, yo, ka and kb) and for a chosen value of xa the following steps are suggested:
Step 1: Select the value of xa .
Step 2: Solve for the surfaces F1 and F2 as indicated above for different pair of values (ca,cb). Then using the same axis for F1 and F2
determine the traces of these surfaces on the cacb-plane (i.e., F1F20). These traces are shown in Fig. 3 for the selected value of
xa 20 and for ra 0.5 and rb 0.9 (corresponding to elastic springs ka 3EI/h and kb 27EI/h, respectively). Notice that families of
traces are obtained as shown in Fig. 3.
Step 3: The closest initial solution (intersection of the two traces F1 and F2 is shown within a circle in Fig. 3point 1) to the origin is
cao 0.602 and cbo  2.975. This point is taken as the initial point in the iterative process using the NewtonRaphson method
(previously discussed) which generally converges to the lowest load P.
Step 4: To determine the solution point (ca, cb) for the particular value of xa 201, an iterative bi-dimensional NewtonRaphson
method with the suggested initial point (cao, cbo) was used. The solution obtained is ca 0.678 and cb  3.052.
Step 5: The corresponding values of b, P, xua and yua calculated using Eqs. (13)(16) are as follows:
Z ca
dc
b q Fk, ca Fk, cb 2:432
cb 1k2 sin2 c
2 EI EI
P b 2 5:915 2
h h
hk
yua 2 cos ca cos cb 0:904h
b
Z ca q!
h
xua 2 1k2 sin2 c h 0:403h
b cb

Fig. 4ac shows the variations of xa/h and ya/h with P/(EI/h2) as well as the Elastica of the beam-column for three different
pairs ra and rb (Fig. 4c corresponds to the column with ka 3 EI/h and kb 27 EI/h) and for different values of the slope xa at
end A. The obtained results compare well with those calculated using the method proposed by Aristizabal-Ochoa [9]. Notice
that the large-deection response and stability of beam-columns are highly sensitive to the type and stiffness of the exural
end-restraints.

Fig. 5. Example 1A: Large-deection analysis of a vertical column with non-linear connections at ends A and B according to those shown in Fig. 2b.
 C. Vega-Posada et al. / International Journal of Non-Linear Mechanics 46 (2011) 7995 87

Fig. 6. Example 1B: Large-deection stability of an out-of-plumb column (yo 0.01 radian) subjected to gravity load with both ends elastically restrained: (a) ra 0.25,
rb 0.50; (b) ra 0.50, rb 0.50; (c) ra 0.30, rb 0.90; and (d) ra 0.50, rb 0.90
88 C. Vega-Posada et al. / International Journal of Non-Linear Mechanics 46 (2011) 7995

The column of Fig. 2 was also analyzed with both ends restrained against rotation by two non-linear rotational springs as described by
Fig. 2 curve b. The momentrotation relationships of the two end rotational springs are dened by the three-parameter model [13] as
follows:
At end A: Mua EI/h, joa 0.333 and na 1.
At end B: Mub EI/h; job 0.037 and nb 1.
Fig. 5 shows the Elastica for different values of the slope xa at end A and the variations of xa/h and ya/h with P/(EI/h2) for the beam-
column restrained at both ends with non-linear semi-rigid connections just described and yo 0. Again, the obtained results are very
sensitive to the type and stiffness of the exural end-restraints. Notice that initially the column reaches a critical load identical to that
obtained for the case of linear elastic connections. This behavior was expected since the initial slopes of the connections (shown in
Fig. 2 curve b) were chosen to be exactly ka 3 EI/h and kb 27 EI/h at ends A and B, respectively. However, once the column starts to
buckle, its load capacity is reduced substantially as the column deects laterally from zero to 0.6 h (see the right side of Fig. 5). This
example shows that columns subjected to gravity load at the top with bending restraints at the supports that become less stiff as the
member deects (like those shown in Fig. 2 curve b) are inherently unstable once the buckling takes place. In addition, the
phenomenon of snap-through occurs at the bifurcation point, indicating that if the load is applied using load control, the column
would jump to the right as soon as the applied load reaches buckling at the bifurcation point.
B) Initially out-of-plumb column (with ho 0.01 radians). Fig. 6 ad shows the variations of xa/h and ya/h with P/(EI/h2) for different
values of the slope at the free end xa for a beam-column restrained at both ends with different combinations of linear semi-rigid end
connections.

Notice that: 1) as in Example 1A the obtained results compared well with those using the approximate method proposed by
Aristizabal-Ochoa [9] up to their maximum lateral deection ya; (2) their maximum lateral deection generally occurs at loads near to P2Pcr
r
40 8r2a r2b ra rb ra rb 3ra rb 34
where Pcr p2 EI=ke h2 with ke 34ra rb ra rb ra rb . At higher axial load levels (i.e. P42Pcr), the lateral deection ya is
reduced as the load increases with part of the column under tension (i.e., the overhanging part near to top end). This overhanging is also
captured by the approximate method proposed by Aristizabal-Ochoa [9]; and (3) the large-deection response and the stability of beam-
columns are not only sensitive to the stiffness of the exural end-restraints but also to the magnitude of the initial out-of-plumb angle yo.
Fig. 7a and b shows the variations of xua =h and yua =h against P/(p2EI/h2) for a column with linear, non-linear and elasto-plastic end
connections and with yo 0, and Fig. 7c and d shows them for yo 0.01 radians, respectively. Curve 1 corresponds to an initially vertical
column (yo 0) with linear end connections (ra 0.5 and rb 0.9). As expected, buckling occurs when the axial load reaches its lowest

1.0 1.0
2 2* 3*
0.8 0.8
1 2*
3
0.6 0.6
2EI
2EI

6
Ph2

2
Ph2

4 3* 5 3*
0.4 0.4
3

0.2 0.2
3
0.0 0.0
0.2 0.4 0.6 0.8 1.0 1.2 0.2 0.4 0.6 0.8 1.0 1.2
_X'a _X'a
L L

1.2 1.2
2* 3* 3 2
1.0 1.0

0.8 0.8
O 1 A D
1 D
2EI

2EI
Ph2

Ph2

2* 6
0.6 0.6
2 5 E
4* 4
3 E
0.4 0.4 3*
C
B

0.2 0.2
Snap through Snap through

0.0 0.0
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
_Y'a _ a'
Y
L L

Fig. 7. Example 1: (a) P/(p2EI/h2)-versus-(xa/h) and (b) P/(p2EI/h2)-vs-(ya0 /h). Legend: Linear connections with yo 0 (path 1); linear connections with yo 0.01 radians (path
22*); non-linear connections with yo 0.01 (path 33*); non-linear connections with yo 0 (path 44*); elasto-plastic connections assuming Mua EI/L, fua 0.33;
Mub 1.5 EI/L, fub 0.0556 (path 5); elasto-plastic connections assuming Mua 1.5 EI/L, fua 0.5; Mub EI/L, fub 0.037 (path 6).
 C. Vega-Posada et al. / International Journal of Non-Linear Mechanics 46 (2011) 7995 89

critical value indicated by the bifurcation point O of the loaddeection curve with branches or paths shown in Fig. 7b (notice that the
column remains stable after buckling). Curves 22* correspond to paths of a column with an initial out-of-plumb of yo 0.01 radians and
linear elastic connections (ra 0.5 and rb 0.9) for which there is not bifurcation of the loaddeection curve as shown by Fig. 6ac.
Notice that the column remains stable along the right path or curve 2. However, the phenomenon of snap-through occurs at a minimum
load denoted by point A along the left curve of path 2* (left path) with the column jumping to the right.
Curves 33* and 44*correspond to loaddeection paths of a column with non-linear connections (Mua EI/h, jua 0.33 radians,
Mub 1.5 EI/h, jub 0.0556 radians and na nb 1) with an initial out-of-plumb angle yo 0.01 and 0 radians, respectively. Notice that the
phenomenon of snap-through occurs at a maximum load denoted by point C, much lower than that of the bifurcation point O (along
the right path of curve 3), indicating that if the load is applied using load control, the column would jump to the right as soon as the
phenomenon of snap-through occurs at a minimum load denoted by point B along curve 3* (left path), indicating applied load exceeds
slightly that indicated by point C. However, this occurs when load control is used. Notice that in the case of an initially vertical column
(already discussed using Fig. 5 with yo 0) indicated by curves 44*: 1) column buckling occurs by bifurcation point O as shown by the
loaddeection curve with both branches descending (Fig. 7b); 2) the column becomes unstable just after the rst buckling; and 3)
the bifurcation axial load is identical to that obtained assuming linear elastic connections at both ends with stiffness corresponding to the
initial tangents of the momentrotations curve (Fig. 2 curve b).
Curves 5 and 6 describe the behavior of a column with elasto-plastic connections at both ends (Mua EI/h, jua 0.33 radian, Mub 1.5 EI/h,
jub 0.0556 radians and na nb N) and with an initial out-of-plumb angle yo 0.01 and 0 radians, respectively. Notice that: 1) the initial
behavior indicated by curves 5 and 6 are identical to those of curves 33* and 44*, respectively; 2) once one the end connections yields the
column becomes unstable (indicated by the peak points D in Fig. 7d) with the curves descending after that; and 3) once the other end connection
yields the column becomes a mechanism (indicated by points E in Fig. 7d) with plastic hinges at both ends and with sway uninhibited.

4.2. Example 2. Effects of mid-span section reduction on the stability and post-buckling behavior of a rectangular beam-column having at ends
and compressed by two horizontal rigid Planes

Determine the effects of the magnitude of a mid-span section reduction on the large-deection stability of an elastic rectangular
beam-column perfectly vertical having at ends and compressed by two horizontal rigid planes, as shown by Fig. 8a. Assume that: 1) the
member is initially straight and perfectly vertical with properties E, A, I, 2b and L2h; 2) bending takes place about one of its principal
axes with sway inhibited between the bar ends; and 3) the applied axial load P is always vertical. Compare the results with those
presented by Gurel and Kisa [12] for the buckling load of a column with reduced section at mid-span and Aristizabal-Ochoa [1] and
Feodosiev [15] for the large displacement and post-buckling behavior of a member without a reduced section.
Solution: Using symmetry the structural model is that shown in Fig. 8d with the reduced mid-span section modeled as two exural
springs kb connected in series (i.e., 1/k 1/kb +1/kb or kb 2k) with a exibility index given by the following expression developed by
Shifrin and Ruotolo [14]:
Cf 5:346hf w 28

where w a/(2b); asection reduction depth; Cf 2EI/kb; and

f w 1:8624w2 3:95w3 16:375w4 37:226w5 76:81w6 126:9w7 172w8 143:97w9 66:56w10 29

P P P P

A A P
2b
A

x' x'

B B
L = 2h B y'
y'

P P P

Fig. 8. Example 2: Rectangular beam-column with a mid-span exural crack having at ends and compressed by two horizontal rigid planes.
90 C. Vega-Posada et al. / International Journal of Non-Linear Mechanics 46 (2011) 7995

Note that the model shown in Fig. 8d is identical to that of Fig. 1a, with the y-axis being the axis of symmetry and l yo 0 with
exural connection at end B of stiffness kb and exural connection at end A varying from ka N (perfectly clamped end) when the whole
section of the column of width 2b is in full contact with the top rigid plane (Fig. 8c) to ka 0 (perfectly hinged end) when the column has
already bucked with a top edge in contact with the top rigid plane where the vertical load is applied eccentrically, as shown by Fig. 8c.

6 6
PL2 PL2

5 2EI 5 2EI
0.3375

C* C** C*** 0.04

4 4 =b/L
0.3375 0.02

0.04 0.0
3 b/L= 3
0.02
0.0 2
2 E*
D*
1.173
1 B* 1 B*
A*
b = 1 b = 1

-0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

6 6
PL2 PL2
5 EI
2
5 2EI

C* C** C*** 0.3375

4 4
0.3375 0.04
= b/L
3 0.04 0.02
b/L = 3
0.02 0.0

2 0.0 2 E*
D*

1 B* 1 B*
A*
b = 0.9515 Grel (2004) b = 0.9515

-0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

6 6
PL2 PL2

5 2EI 5 2EI

0.3375
4 4 C* C** C*** 0.04
= b/L
0.02
0.3375
3 3 0.0
0.04
b/L =
0.02 E*
2 2
0.0 D*

1 B* 1 B*
A*
b = 0.9 b = 0.9

-0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Xa' Ya'
L L

Fig. 9. Example 2: Effects of a mid-span exural crack on the large-deection stability of a rectangular beam-column having at ends and compressed by two horizontal
rigid planes: (a) ra 1; (b) ra 0.9515; and (c) ra 0.9.
 C. Vega-Posada et al. / International Journal of Non-Linear Mechanics 46 (2011) 7995 91

Results calculated using the proposed model and corresponding algorithm were compared to those reported by Gurel and Kisa [12] for
a column with following properties: 2b 0.15 m; L2h4.5 m; E2  1010 N/m2, a 0.0375 m. Therefore w 0.25, f(w) 0.095438,
Cf 0.076532 m, kb 2EI/Cf 58.79EI/h and rb 1 3C1 =2h 0:9515 using Eq. (1b).
f

A total of three columns cases were studied with linear elastic exural connection at mid-span: kb N, 58.79EI/h and 27EI/h
corresponding to rb 1, 0.9515 and 0.90, respectively. Note that the case of rb 0.9515 was analyzed by Gurel and Kisa [12] and the case
of rb 1 was reported by Aristizabal-Ochoa [1]. After the column bends two types of buckling are possible as shown in Fig. 8ac. The rst
type occurs at axial load Pe 4p2EI/L2, 3.86 p2EI/L2 and 3.72 p2EI/L2 for the three cases under consideration (i.e., rb 1, 0.9515 and 0.9),
respectively. This type of buckling occurs when the at ends of the column are very wide (Fig. 8b) and do not rotate remaining horizontal
and with sidesway inhibited. The second type of buckling occurs when the at ends are very narrow rotating simultaneously with one
edge of the cross section at each end bearing against the rigid horizontal walls, as shown by Fig. 8c. If the ends are of zero width (i.e.
b/L 0), the second type of buckling occurs at a force smaller than Pe of the rst type reaching buckling loads of magnitudes p2EI/L2,
0.97p2EI/L2 and 0.93 p2EI/L2 for rb 1, 0.9515 and 0.9, respectively. Notice that Pcr 0.97p2EI/L2 for the case of rb 0.951 is identical to
that reported by Gurel and Kisa [12].
The large-deection behavior corresponding to second type of buckling was also calculated following the steps listed in Example 1
assuming hL/2, Ma  Pb cos xa for the three column cases under consideration with rb 1, 0.951 and 0.9, respectively. Fig. 9ac shows
the results of this analyses for three different values of b/L0.02, 0.04 and 0.3375, respectively. The results indicate that the critical loads
Pe corresponding to the rst type of buckling and the large-deection stability corresponding to both types of buckling are highly
sensitive to the reduction of the cross section at mid-span and the width 2b of the cross section.
Notice that the results shown in Fig. 9a coincide with the results reported by Feodosiev [15, p. 288] and are also identical to those
presented Aristizabal-Ochoa [1] for the member with rb 1 and that the results shown in Fig. 9b coincide with the results reported by
Gurel and Kisa [12]. Important features of the loaddeection curves like those shown in Fig. 9ac are fully discussed by Aristizabal-
Ochoa [1]. Additional features of the loaddeection curves in Fig. 9ac are the following:

 The load P decreases rapidly at small deections yua and then, at point A*pstarts
to increase but at low rate up to point B*.
 At point B , all curves cross the curve corresponding to b/L0 with k1/ 2 and ca 901. This means that at B*, the end moment is
*

zero and the end rotation is 901 (i.e., the bar reaches the deformed shape shown in Fig. 8e, regardless of the value b/L with
xa ca 901).
 None of the curves in Fig. 9ac crosses the horizontal axis (i.e., P/Pe 0). Notice, for instance, that the curve for b/L0.02 on the left
side of Fig. 9a becomes asymptotic vertically to the line (yua =L) 0.04. Similar behavior is shown in Fig. 9bc.
 For values of (b/L) a0 the bar is always stable in the small, provided that the instability occurs in the second type of buckling
(Fig. 8c). However, when the bar is loaded with PPe 4 p2EI/L2, 3.86p2EI/L2 and 3.72 p2EI/L2 for the three elastic exural connections
under consideration (i.e., rb 1, 0.951 and 0.9, respectively) the bar buckles in the rst type of buckling (Fig. 8b). That is, when the bar
is loaded with a force lower than Pe, the bar retains its vertical straight-line shape (Fig. 9a). However, when the bar is deected from
the vertical by assigning some curvature to its axis, the bar, when left to itself will return to the original vertical position if the
imposed deection is small. Otherwise, if the imposed deection is sufciently large [say (yua =L) 0.053 corresponding to PPe/
22 p2EI/L2 for the case of rb 1 and b/L 0.02 shown in Fig. 9a], the bar will assume a new shape corresponding to point E*. This
behavior is typical in all three cases under consideration.
 Also, for values of (b/L) a0 the second type of buckling (Fig. 8c) starts to takes place when the loaddeection curves crosses the
corresponding curve of the rst type of buckling, as indicated by points C*, C** and C*** in Fig. 9ac. Then, the axial load decreases
rapidly down to a minimum value indicated by point A* from which the load starts to increase again.
Depending on the magnitude of the forced deection, the bar may or may not return to its original vertical position. The snap-
through phenomenon occurs at a force P larger than a specic value for each case. For instance for the column with rb 1 and
b/L 0.02 according to Fig. 9a, this happens for both P 41.173 p2EI/L2 (point A*) and Po4 p2EI/L2 (point C*, where buckling takes place
irrespective of the magnitude of the imposed deection). Therefore, a range of possible buckling loads for the column with rb 1 and
b/L 0.02 from which a transition to a new position of equilibrium can be established: 1.173Pe rPo4 Pe. When the applied load
PZ4Pe, the transition to a new form of equilibrium is inevitable. Similar behavior, with different ranges of loads, is shown for the
columns with a reduced section at mid-span shown in Fig. 9bc. The ranges of possible buckling are also affected by any initial
imperfection of the bar. Therefore, research on the effects of initial curvature imperfections is needed.

4.3. Example 3. Large-deection stability of an out-of-plumb beam-column with linear elastic connections and braced laterally

Determine the stability and post-buckling behavior of an out-of-plumb beam-column subjected to compressive vertical load P of
constant exural rigidity EI, and span length h with both ends A and B restrained against rotation and braced laterally, as shown in Fig. 10.
Compare the calculated results using the proposed method with those reported by Phungpaingam and Chucheepsakul [10] for the
following particular cases: 1) ka kb 5(EI/h) and 2) ka 1000(EI/h) and kb 10(EI/h) both with yo 0.
Solution: Since the top end A must move along the x-axis as the beam-column is loaded and deects:
ya yua cosl yo xua sinl yo 0 [using Eq. (50) of the Appendix]. Then the angle l can be determined directly from this boundary
 
yua
condition as follows: l sin1 p
2 2
yo
xua yua
In addition, as the beam-column is loaded a reaction force V perpendicular to the x-axis is induced as shown by Fig. 10. Therefore, the
resultant force Pa, and the induced reaction force V shown in Fig. 10 at A can be expressed in terms of the applied force P as follows:

P
Pa 30
cos lsin l tan yo
92 C. Vega-Posada et al. / International Journal of Non-Linear Mechanics 46 (2011) 7995

Fig. 10. Example 3: Large-deection stability of an out-of-plumb beam-column with linear elastic connections at both ends and braced laterally: (a) initial conditions and
(b) deected shape.

180
ka = 1000EI/h kb = 10EI/h
ka = kb = 5EI/h
1 = 60
160
2 = 45
C 1
2nd Mode 1st Mode 3 = 30
140 4 = 15
5 = 0
1
1
120
1
Snap-through 2
2
1 3
100
P/(EI/h2)

4
2 1st Mode
2 3 3 5
4 2
80 5 4
3 5 3
4 B D
60 4
5
5 Snap-through
1 = 60
40 B 2 = 45
3 = 30 A 2nd Mode 1st Mode
20 4 = 15
A
Phungpaingam and 5 = 0 Phungpaingam and
Chucheepsakul [10] Chucheepsakul [10]
D
0
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
u/h u/h

Fig. 11. Example 3: Large-deection stability behavior of a beam column with (a) ka kb 5(EI/h); and (b) ka 1000(EI/h) and kb 10(EI/h).

sin l P
V 31
cos yo cos lsin l tan yo
Two beam-columns are analyzed: 1) ka kb 5(EI/h) and 2) ka 1000(EI/h) and kb 10(EI/h). The calculated results are presented in
Fig. 11a and b, respectively. Each gure shows the variations of P/(EI/h2)-versus-(u/h) [where uh hx
h
a
1 yuha sinlyo  xuha coslyo ] for
ve different out-of-plumb angles (yo 01, 151, 301, 451 and 601, respectively).
The calculated results presented in Fig. 11a and b for the perfectly vertical column (case 5 or yo 0) are identical to those reported by
Phungpaingam and Chucheepsakul [10] who discussed in detail the effects of the rotational elastic restraints on the large-deection
stability of the beam-column of Fig. 10. They suggested that the loaddeection curves shown by Fig. 11a and b indicate that the large-
deection stability behavior can be divided into two categories, beam-columns with (1) ka kb and (2) ka a kb.
For beam-columns with ka kb (see Fig. 11a), the equilibrium path starts from point O at which the load P is zero, and then P increases
along path OA until the load reaches the critical load at point A. Beyond point A the equilibrium path diverges. The path where both the
load P and the displacement u both increase is stable along which a secondary bifurcation point B is found. There are two possible
equilibrium paths that emanate from point B. The one that follows path BC is unstable; the other that follows path BD is also unstable and
only becomes stable along DE.
For beam-columns with ka a kb (see Fig. 11b) the equilibrium paths are different from the case of ka kb. The load P increases from
point O to the critical load at point A beyond which the load P slightly increases with the displacement u until it reaches the limit load
point B (depending on the values of ka and kb). Beyond point B, the path becomes unstable with P decreasing as the displacement u
 C. Vega-Posada et al. / International Journal of Non-Linear Mechanics 46 (2011) 7995 93

increases. The load P decreases to zero and continually decreases to the limit load at point C. After point C, the load P increases as
displacement u increases (stable path CDE). In addition, the phenomenon of snap-through occurs at u/h E0.90 jumping to u/h E1.65
when load P is set as the control variable. Further details on the types of buckling and their characteristics are given by Phungpaingam
and Chucheepsakul [10].
Based on the obtained results presented in Fig. 11a and b the following additional conclusions can be made:

1) The bifurcation point A is very sensitive to the out-of-plumb angle yo with the corresponding buckling load increasing with yo.
2) Fig. 11a indicates that the secondary bifurcation point B occurs at u/h 1.11 regardless the value of yo with the corresponding load P
increasing with yo.
3) The path of the second mode of buckling is not stable (descending curve) beyond the bifurcation point and it crosses the horizontal
axis of Fig. 11a at u/h 0.92 and u/h 1.18 regardless the value of yo. Notice that the transition from compression to tension force P
occurs at u/h E1.07 and u/hE1.50 for the rst mode, and at u/hE0.85 and u/LE0.85 for the second mode.

5. Summary and conclusions

The non-linear large-deection and post-buckling analysis of laterally unbraced and braced slender beam-columns of symmetrical cross
section subjected to end loads (forces and moments) with both ends partially restrained against rotation, including the effects of out-of-
plumbness, is developed in a classical manner. The classical theory of the Elastica and the corresponding elliptical functions utilized herein
are those presented recently by Aristizabal-Ochoa [1]. The main limitation of the Elastica is that only exural strains are considered. The
proposed method can be used in the large-deection and post-buckling analysis of slender beam-columns with rigid, semi-rigid, and simple
connections at both ends including linear and non-linear inelastic connections like those that suffer from exural degradation (such as exural
cracking) or exural stiffening. The validity and effectiveness of the proposed method and equations are veried against two different methods
proposed by Aristizabal-Ochoa [1,9] and with analytical results reported recently by Gurel and Kisa [12] and Phungpaingan and
Chucheepsakul [10]. Three comprehensive examples are included that show the effects of linear and non-linear connections at both ends on
the large stability analyses and post-buckling behavior of slender beam-columns with and without lateral sway.
The advantages of the proposed method are: 1) the non-linear large-deection elastic analysis of slender beam-columns subjected to
any combination of end forces and/or bending moments with rigid, semi-rigid and simple connections at both ends including linear and
non-linear inelastic connections like those that suffer from exural degradation (such as exural cracking) or exural stiffening can be
carried out without major difculties; and 2) the effects of semi-rigid connections (linear or non-linear) at both ends are built in the non-
linear governing equations.
Analytical results indicate that: 1) large-deection and post-buckling response of a slender beam-column is affected not only by its elastic
properties, the magnitude of the applied loads, boundary conditions and the characteristics of the end connections, but also by the bowing or
curvature effect caused by the applied forces and moments. Signicant increases in the lateral deections are caused by the geometric non-
linear effects; 2) the large-deection analysis and post-buckling of a beam-column with exible connections at both ends becomes a complex
problem requiring the simultaneous solution of at least two highly non-linear equations with elliptical integrals whose unknowns are the
limits of the integrals. The solution of this problem becomes even more complex when the end connections are non-linear or the direction of
the applied load changes. These effects are all included in the present analysis; and 3) the approximate method proposed by Aristizabal-Ochoa
[9] gives excellent results when compared with those from the proposed method and it is highly recommended in the large-deection analysis
of a beam-column with linear elastic connections. Further research is needed to investigate the effects of bracing (lateral and longitudinal
restraints), axial and shear strains, initial curvature imperfections and inelastic material behavior.

Acknowledgments

The research presented in this paper was carried out at the National University of Colombia, School of Mines at Medelln. The authors
are thankful to the anonymous reviewers for their valuable comments. Their opinions helped to improve the quality of the paper. The
author wants to express his appreciation to DIME of the National University of Colombia for the nancial support and encouragement.

Appendix. Derivations of proposed equations

The proposed method is that presented by Aristizabal-Ochoa [1] including the effects of non-linear semi-rigid connections at both
ends of the member. The derivations of the proposed equations are reproduced below for easy reference.
Two coordinate systems both located at the bottom end B (Fig. 1b) are used: the xy-system is orientated along the tangent and normal
to the elastic curve at B; and the x0 y0 -system orientated along the line of action of force Pa. The large-deection equations are developed in
the plane of bending about one of the principal axes. Denoting by l + yo the angle made by both x and x axes, and x the current angle
between the tangent to the elastic curve at s and the x0 -axis, the governing equation of the member under bending is simply:
dx M
 32
ds EI
where M  Pa(y0 a y0 ) Ma. Notice that Ma includes all external moments applied at end A and addition to the moment caused by the
connection at end A (see Examples 13):or
d2 x Pa dyu
 33
ds2 EI ds
94 C. Vega-Posada et al. / International Journal of Non-Linear Mechanics 46 (2011) 7995

2 Pa
Calling b EI=h2
and since dy0 /dssin x, then Eq. (33) can be written as follows:
2
d2 x b dyu
2 0 34
ds2 h ds
Eq. (34) which is identical to the non-linear free-vibration of a simple pendulum can be rewritten asor
 
dx 2 x x
h2 d 2b sin cos ds 35
ds 2 2
 
Multiplying both sides of Eq. (35) by dx=ds and integrating once:
 2  
dx 2 x
h 4b C1 sin2 36
ds 2
Calling C1 k2 and making the following change of variables:
x
sin k sin c 37
2
The derivative with respect to s of both sides of Eq. (37) is
   
1 x dx dc
cos k cos c 38
2 2 ds ds
or
dx cos c dc
2k q 39
ds ds
1k2 sin2 c
q
But from Eqs. (36) and (37): h ddsx 2b C1 sin2 2x 2bk cos cor
dx 2bk
cos c 40
ds h
Equating Eqs. (39) and (40):
b dc
ds q 41
h
1k2 sin2 c

and integrating Eq. (41)

Z c
s dc
b q Fk, cFk, cb Fk, cb , c 42
h cb 1k2 sin2 c
Rc dc
where Fk, c Fk,0, c 0
p
2
elliptic integral of the rst kind.
2
1k sin c
Let us now determine the equations for the elastic curve in both coordinate systems. It is known that
dx0 cos x ds and dy sin x ds or in terms of x/2:
 
x
dxu 12sin2 ds 43
2

x x
dyu 2sin cos ds 44
2 2
Using Eqs. (37), (38) and (41), Eqs. (43) and (44) can be expressed in terms of c as follows:
q
h
dxu 2 1k2 sin2 c dcds 45
bk, c

hk
dyu 2 sin c dc 46
bk, c
Integrating these two last expressions:
h 
x0 2 Ek, cEk, cb s 47
bk, c

h
yu 2 kcos ccos cb 48
bk, c
R c q
where Ek, c Ek,0, c 0 1k2 sin2 c dc elliptic integral of the second kind.
In the xy system, the coordinates of any point of the elastic curve can be determined using Eqs. (49) and (50):
x xu cosl yo yu sinl yo 49

y yu cosl yo xu sinl yo 50

The deected shape (i.e., the values of x0 , y0 or x, y along the member) can be obtained using Eqs. (47)(50), and the moment and
bending stress using the expressions M Pa(y0 a  y0 )  Ma and s Mz/EI.
 C. Vega-Posada et al. / International Journal of Non-Linear Mechanics 46 (2011) 7995 95

References

[1] J. Dario Aristizabal-Ochoa., Large-deection stability of slender beam-columns with semi-rigid connections: the elastica approach, J. Eng. Mech. ASCE 130 (3) (2004)
274282.
[2] S. Timoshenko, J. Gere, in: Theory of Elastic Stability, 2nd ed., McGraw-Hill, 1961 (Chapter II).
[3] M.A.M. Torkamani, M. M. Sonmez, J. Cao., Second-order elastic plane-frame analysis using nite-element method, J. Struct. Eng. ASCE 123 (9) (1997) 12251235.
[4] A. Hori A., A. Sasagawa, Large deformation of inelastic large space frame I: analysis, J. Struct Eng. ASCE 126 (5) (2000) 580588.
[5] A. Hori, A., A. Sasagawa, A, Large deformation of inelastic large space frame. II: application, J. Struct. Eng. ASCE 126 (5) (2000) 589595.
[6] M.S. Wendel, E.M. Richard, Non-linear FE analysis of steelconcrete composites structures, J. Struct. Eng. ASCE 126 (6) (2000) 662674.
[7] J.P. Smith-Pardo, J. Daro Aristizabal-Ochoa, Buckling reservals of axially restrained imperfect beam-column, J. Eng. Mech. ASCE 125 (4) (1999) 401409.
[8] A. Khamlichi, L. Elbakkali, A. Limam, Postbuckling of elastic beams considering higher order strain term, J. Eng. Mech. ASCE 127 (4) (2001) 372378.
[9] J. Dario Aristizabal-Ochoa., Non-linear large-deection-small strain elastic analysis of a beam-column with semi-rigid connections, J. Struct. Eng. ASCE 127 (1) (2001)
9296.
[10] B. Phungpaingan, S. Chucheepsakul, Post-buckling of an elastic column with various rotational end restraints, Int. J. Struct. Stability Dyn., 5, (World Scientic
Publishing Company), 2005, pp. 113123.
[11] J. Dario Aristizabal-Ochoa, Large deection and postbuckling behavior of Timoshenko beam-columns with semi-rigid connections including shear and axial effects,
Eng. Struct. Elsevier J. 29 (6) (2007) 9911003.
[12] M.A. Gurel, M. Kisa, Buckling of slender prismatic columns with a single edge crack under concentric vertical loads, Turkish J. Eng. Environ. Sci. 29 (2004) 185193.
[13] W.F. Chen, Y. Goto, J.Y. Richard Liew, in: Stability Design of Semi-Rigid Frames, John Wiley & Sons, Inc., 1996 (Chapter II).
[14] E.I. Shifrin, R. Ruotolo, Natural frequencies of a beam with an arbitrary number of cracks, J. Sound Vib. 222 (3) (1999) 409423.
[15] V.I. Feodosiev, in: Selected Problems and Questions in Strength of Materials, MIR Publishers, Moscow, 1977.
[16] H.T. Davis, in: Introduction to Non-linear Differential and Integral Equations, Dover Publication, New York, 1962, pp. 129155.