1. Introduction
Steel beam-columns of double-tee section shape that are initially straight and untwisted, and subjected to compression and bending about the major principal axis y-y, exhibit failure-modes either in the plane of bending by deflecting along the oz axis, or out of that plane by deflecting laterally along the oy axis and by twisting about the member axis ox. The former failure-mode is related to the second-order bending that at the ultimate limit state corresponds to the member in-plane buckling resistance. In contrast, the latter failure-mode is associated with the bifurcation instability in the form of out-of-plane buckling that at the ultimate limit state corresponds to the flexural-torsional buckling (FTB) resistance.
The in-plane behavior and flexural-torsional buckling of beam-columns have been studied for many years, starting with the first solution obtained by Euler for flexural buckling of compressed elastic members, and then extended from the in-plane field of studies into the torsional and general flexural-torsional failure domains, including the elastic range of steel behavior, and then taking into account the elastic–plastic properties of steel. Problems related to different aspects of beam-column stability behavior are included in many textbooks, e.g., Timoshenko and Gere [
1], Vlasov [
2], Chen and Atsuta [
3,
4], Trahair [
5], Yang and Kuo [
6], and finally Trahair et al. [
7] and Kindmann and Kraus [
8], which are specifically related to the behavior of steel members and structures. For standardization purposes, the elastic-buckling solutions based on linear-buckling analysis (LBA) derived from the classical energy-equation are adopted.
In LBA of beam-columns under compression and major-axis bending, the small-rotation matrix is used for the formulation and evaluation of the first order stress resultants (the axial force and the major-axis bending moment). The displacement-field may be formulated in this case for the configuration being infinitesimally close to the initial one. Adopting an RHS Cartesian coordinate system, in which the axis
ox is the prismatic-member axis passing through the section shear-center, axes
oy and
oz are the section principal-axes (
oy is prescribed for the section’s greater moment of inertia and for the major bending-axis, and
oz is associated with the plane of the loading system), the displacement-field takes the form given by Pi et al. [
9]:
where
u,
v, and
w are the displacements of an arbitrary point within the member volume, the vector of
,
, and
contains the member-axis displacements of the bisymmetric open section, resulting from the elongation and rotations of the member’s elemental-length d
x,
is the section’s normalized warping-ordinate,
is the angle of twist rotation, and
is the derivative of the
x-dependent variables.
The first-order rotation matrix,
, in the undeflected configuration is given by:
where
and
are the angels of flexural rotations.
Figure 1 shows the kinematics of the member elementary-length, d
x, based on the first-order rotation matrix given by Equation (2).
For the purpose of stability formulation, the second-order rotation matrix may be adopted, in which the coupling between the twist rotation and the flexural rotations is taken into consideration for the evaluation of the displacement-field:
As a result, the classical energy-equation is obtained by using such a form of the rotation matrix as is explained in Pi et al. [
9]. A closed-form solution for FTB problems can only be obtained for the cases of uniform bending and uniform compression. For a symmetric-loading system, the solution is obtained by approximating globally the buckling displacements with the use of the functions satisfying the natural boundary conditions
in which the dimensionless coordinate
and
is the original length of the straight, untwisted member.
The linear-eigenvalue analysis (LEA) closed-form solution for beam-columns for the case of equal and opposite end-moments,
, and compression,
N, takes the well-known form
where
is the critical moment for uniform bending,
and
are the lowest bifurcation forces in relation to the independent out-of-plane buckling modes of flexural minor-axis form and torsional form, and
is the polar radius of gyration.
Cuk and Trahair [
10] developed an improved analytical FTB-solution for elastic beam-columns under unequal end-moments:
where
is the equivalent uniform moment-modification-factor and
is the end- moment of greater absolute value.
Replacing the
factor by
, which is valid for any shear-center loads, not only for unequal end-moments, the latter takes the following form:
where
,
are the maximum moments related to the symmetric- and antisymmetric-system load-components and
is the factor for any combination of end-moments and shear-center loads.
The factors
and
proposed by Cuk and Trahair [
10] for unequal end-moments correspond to those calculated from the equation provided by Serna et al. [
11].
Combined-load cases need a more complex analysis that nowadays is performed using LBA and the finite-element method leading to LEA, e.g., Papangelis et al. [
12]. In parallel investigations, a number of LEA formulations have been researched in order to obtain analytical elastic-buckling solutions for engineering applications. The approximate solutions for moment-gradient cases may be obtained by using the following global approximation of buckling modes:
with
given by Equation (5).
An LEA formulation and a general closed-form solution were presented by Giżejowski et al. [
13]. Such a general solution for the shear-center span loads takes the form of Equation (7), in which
:
where
,
are reciprocals of the classical energy-method conversion-factors;
,
are the field-moments for the symmetric and antisymmetric components, and
is the second-lowest flexural minor-axis bifurcation force.
The out-of-plane stability LBA formulation was further extended for the inelastic region. Bradford et al. [
14] presented a finite-element model for investigating the inelastic behavior of beam-columns. A large-displacement inelastic in-plane analysis was combined with the formulation of a non-prismatic elastic-line finite element with 7 degrees of freedom. An out-of-plane inelastic-buckling analysis was carried out for the evaluation of inelastic critical-loads for beams and beam-columns under selected in-plane loading systems. The effect of residual stresses was taken into account in the finite-element analysis, but the effect of geometric imperfections was not included.
Recently, a number of published papers have referred to more accurate nonlinear-buckling-analysis (NBA) formulations that include the effect of prebuckling displacements on the buckling state. Pi and Trahair [
15,
16] presented a general numerical model for nonlinear-buckling analysis (NBA,) using a beam finite-element approach. Pi and Bradford [
17,
18] continued these investigations, presenting an accurate rotation matrix that satisfies the orthogonality conditions
The terms of the rotation matrix are as follows:
where
is the Green measure of the member-axis extension, and
is the angle of twist rotation in the rotated axes,
and
.
Barszcz et al. [
19] adopted the rotation matrix developed by Pi and Bradford [
17,
18] for the lateral-torsional buckling (LTB) of beams, and showed that it allows for the LBA solution in the form of a quadratic-eigenproblem analysis (QEA). In conclusion, it was mentioned that the LEA buckling-moment,
, needs to be modified by the factor
, where
and
,
are the principal moments-of-inertia of the cross-section.
Mohri et al. [
20] used another rotation matrix for the virtual-work formulation that allowed for the effect of prebuckling displacements on the beam-column buckling state, by using the
factor. The trigonometric functions of the angle of twist rotation were retained in the formulation for the derivation process, starting from the rotation matrix and ending in the calculation of strain components and the equilibrium at the buckling state. The rotation matrix was presented in [
20], in the following form:
Finally, one has to mention that for uniform bending and compression, a more accurate expression of the critical moment was presented by Trahair et al. [
7]:
where
,
and
are the material’s elastic and shear moduli, and
and
are the section warping-constant for non-uniform torsion and the Saint-Venant’s constant for uniform torsion, respectively; for typical situations in steelwork, one may adopt
.
The purpose of this paper is to investigate the mechanics of flexural-torsional buckling in a different way than has been presented in the subject literature. The derivation is based on the rotation matrix formulated in the deflected configuration with the aid of transformation matrices, in order to represent the deformation components at buckling. A significant element of this research is in laying down the development of a refined stability criterion for FTB equilibrium as the second variation of the total potential energy. The derived equations are valid for any asymmetric in-plane loading system. The stability criterion includes the effect covered by the
factor, but neglects the effect covered by the
factor. The solution is obtained in an approximate way by using the superposition principle of amplified symmetric and antisymmetric moment-components. The analytical solution developed is verified for selected symmetric- and asymmetric-loading systems. Two hot-rolled steel sections, IPE 180 and HEB 180, exhibiting the different values of
for the former section and
for the latter section, are dealt with. For the purposes of verification, two finite-element flexural-torsional-buckling-analysis software programs are used, namely PRFELB, developed at the University of Sydney [
12] and LTBeamN, developed at CTICM [
21], the latter available for free in the public domain. It is concluded that the developed elastic-buckling equation for any asymmetric-loading system is an extension of the solution presented in Mohri et al. [
20]. The general solution developed in Mohri et al. [
20], based on the differential equilibrium equations and Galerkin method, was restricted to symmetric-loading systems. The general solution developed in this paper applies to any asymmetric-loading system and to the selected symmetric-loading patterns proven by Mohri et al. [
20].
4. Improved Solution Based on the Refined Energy-Equation
The improved solution of Equation (42) results from assuming the minor-axis curvature calculated from the differential equilibrium equation, as in the Timoshenko energy method; cf. Trahair [
5]:
Using the principle of superposition, the in-plane moments are taken as the sum of symmetric
and antisymmetric
moment components, corresponding to symmetric and antisymmetric load-components. In the proposed solution, the second-order in-plane moments
and
are approximated by amplifying the symmetric and antisymmetric first-order in-plane moment components,
and
, respectively. In a similar way, the second-order moments
and
are approximated by amplifying the symmetric and antisymmetric first-order out-of-plane moment components,
and
, respectively. Hence, the moment term in Equation (42) becomes
where
are two lowest bifurcation forces about the
y axis, corresponding to symmetric and antisymmetric buckling-modes.
Substituting Equation (44) into Equation (42), the buckling problem converts to a nonlinear-eigenproblem analysis (NEA). For the shear-center loading system, the solution becomes
The equivalent uniform moment-modification-factor takes the form
where
,
are reciprocals squared of the refined-energy-method conversion-factors.
One may notice that for laterally and torsionally unrestrained members, for which
, the following relationships hold:
For all the other solutions in which the applied loads are the off-shear center loads, the factor
is to be replaced by
, where the latter is related to the former as
The parameter
, representing the effect of off-shear span load
, distributed
and/or concentrated
, applied at a distance
from the shear centre,
or
, respectively, takes the form
In Equation (50), the off-shear load-dependent factor
is calculated as follows:
The coefficient , representing the effect of off-shear span loads on the FTB state is not dependent upon the formulation of the strain-energy equation.
Table 1 presents a summary of these coefficients and the equivalent uniform moment-modification-factors developed for the classical energy-method
and
as well as
and
, based on refined versions of the classical energy-method. This refinement is based on the inclusion of an additional term,
, in the strain-energy definition; cf. Equation (42).
One may notice that the moment conversion-factors
and
are generally greater than those of
and
, especially for the antisymmetric component of asymmetric-loading systems. Equation (8) with the factors
and
are used hereafter as a generalized Cuk–Trahair solution [
10] for asymmetric-loading systems other than those with unequal end-moments.
6. Discussion of Results
The results presented in the previous section, reinforced by their verification with results obtained from the analytical solution of Mohri et al. [
20] and from the computer programs LTBeamN and PRFELB show that for the considered symmetric-loading systems and for the steel section of IPE 180:
1. Results from the analytical models of the present study coincide with those of the analytical solution of Mohri et al. [
20]. Moreover, they are verified positively by the finite-element results from both computer programs, LTBeamN and PRFELB.
2. The section exhibits a low value of , which means the out-of-plane instability appears at relatively small prebuckling deflections. Therefore, the prebuckling deflections do not play an important role in the attainment of the LTB state. As a consequence, the problem of critical-moment assessment for the symmetric-loading cases may follow LEA, for which k1 is not included in the classical LEA-based formulations of the buckling criterion.
3. The computer programs considered herein are based on LEA; therefore, they give results slightly lower than those from the present study and from Mohri et al. [
20].
On the other hand, for the considered symmetric-loading systems and for the steel section of HEB 180, the following is observed:
1. Results of the present study coincide with those of Mohri et al. [
20], since both include the effect of prebuckling deflections on the critical-state assessment.
2. The section exhibits a distinctively higher value of than that of IPE 180, which means the out-of-plane instability appears for larger prebuckling deflections. Therefore, the prebuckling deflections play an important role in the attainment of the LTB state. As a consequence, the critical moment for the symmetric-loading cases evaluated by LEA is much lower than that calculated with the inclusion of both prebuckling stress-resultants and the prebuckling deflected profile in the analytical formulation of the buckling criterion.
3. The computer programs LTBeamN and PRFELB give results that do not coincide with those from the analytical solutions of the present study and from Mohri et al. [
20]. The results of the maximum-moment and the axial-compressive force at the buckling state, evaluated from the computer simulation, are well below those from the analytical models.
The other observations and discussions may be formulated on the basis of the results obtained for the asymmetric-loading systems. To the best of the authors’ knowledge, results for such loading cases have not yet been reported in the subject literature. Based on the results presented in this paper for the antisymmetric-loading cases, the following observations may be presented:
1. Results based on the LEA solution of Giżejowski et al. [
13] and of the present study constitute for antisymmetric-loading systems the bi-line curves. This indicates that the analytical models reflect a rather weaker interaction between the LTB mode in bending and the lowest buckling modes in compression than is indicated by the results of the computer simulations.
2. The LEA analytical solution presented in Giżejowski et al. [
13] and that of the present study give the results constituting the upper bound for the computer results; all of the results become rather comparable for such a loading asymmetry, in which the loads of both half-lengths of the beam-column are being described by positive values of
ψi.
3. The finite-element results from both the computer programs LTBeamN and PRFELB are identical or close to each other, but placed well below those from the present study analytical model that includes the effect of the prebuckling deflected-profile on the buckling load.
4. The general analytical solution derived in the present study and valid for any asymmetric in-plane loading system may be suggested for practical applications when the load asymmetry factor is positive. The solution presented needs an introduction of a semi-analytical buckling-modes interaction factor calibrated with the use of the results obtained from finite-element simulations and carried out with the use of the LTBeamN and PRFELB programs.