Leadership in Sustainable Infrastructure

Leadership en Infrastructures Durables

Vancouver, Canada
May 31 – June 3, 2017/ Mai 31 – Juin 3, 2017

LATERAL TORSIONAL BUCKLING ANALYSIS AND DESIGN OF STEEL

BEAMS WITH CONTINUOUS SPANS

Sahraei, Arash1, Pezeshky, Payam2 and Mohareb, Magdi3,4

1 Graduate research assistant, University of Ottawa, Canada
2 Graduate research assistant, University of Ottawa, Canada
3 Professor, University of Ottawa, Canada
4 [email protected]

Abstract:

Design standards do not provide provisions to account for the interaction between adjacent spans of
continuous beams. In the absence of such provisions, the designer may opt for calculating the lateral
torsional buckling capacity for each span separately by applying the moment gradient factors provided in
standards and adopting the smallest critical moment as the one governing the design. The Salvadori
hypothesis of isolating a member from the rest of the structure is assessed in the present study. The elastic
lateral torsional buckling resistance for continuous beams is investigated based on finite element analysis.
Comparisons are made between two types of solutions: (1) those neglecting interaction effects between
adjacent spans, and (2) those considering span interaction. Also examined is the effect of presence of
lateral/torsional restraints at intermediate supports of continuous beams. The results illustrate the merits of
adopting the FEA solution in accounting for span interaction when determining the elastic lateral torsional
buckling capacity of continuous beams.

Keywords: lateral torsional buckling, finite element analysis, interaction effects, moment gradient factors,
continuous beams

1. INTRODUCTION AND LITERATURE REVIEW

Present provisions in design standards (e.g., CAN/CSA-S16-14 (2014), EN 1993 Designer’s guide
(Gardner and Nethercot 2011), AS-4100 (1998) and ANSI/AISC-360-16 (2016)) do not account for the
interaction between various segments of a continuous beam when determining their lateral torsional
buckling (LTB) resistance. In the absence of such provisions, designers may resort to treating each
segment as a separate span and calculate the individual LTB capacity for each segment separately by
adopting moment gradient factors and critical moment equations provided in standards, and then,
conservatively, adopt the smallest critical moment as the one governing the LTB capacity of the continuous
beam. The above procedure omits the interaction effect between adjacent members. Procedures that
account for such interaction effects were proposed in the work of (Nethercot and Trahair 1976), (Trahair
1977) and (Trahair and Bradford 1988) as reported in SSRC guide (Ziemian 2010). Such procedures are
iterative and based on an analogy between the LTB buckling of continuous beams and flexural buckling of
continuous columns. According to (Ziemian 2010), in most cases, these procedures lead to conservative
estimates for the critical moments while in cases of high moment gradients, they can overestimate the LTB
buckling strength. The present study documents a LTB Finite Element Analysis (FEA) that captures

EMM635-1
interaction effects in continuous beams. To establish the validity of the FEA model, the FEA is first adopted
to determine the critical moments for a number of single span problems under a variety of loads and
comparisons are made against the critical moment predicted by design standard provisions. The finite
element is then adopted to investigate the span interaction effect on the lateral torsional buckling of
continuous beams and identify conditions where interaction effects are significant. The model is then used
to quantify the effect of lateral and torsional restraints that may be present at intermediate supports.

2. OVERVIEW OF FINITE ELEMENT FORMULATION

The formulation is based on the following assumptions:

1. Material is linearly elastic, isotropic and homogenous.

2. The beam follows the Vlasov beam kinematics (Vlasov 1961), i.e., the section is assumed to move
in its own plane as a rigid disk and the shear strain in the middle surface of the beam is negligible.
3. Pre-buckling deformations and local warping effects are neglected.

The beam finite element in (Barsoum and Gallagher 1970) is chosen to conduct the study. The element
has two nodes, each having three pre-buckling and four buckling degrees of freedom (DOF) as shown in
Figure 1a. The pre-buckling DOFs are the longitudinal displacement u , the transverse displacement w
and the strong-axis rotation w  and the buckling displacements are the lateral displacement v , the weak-
axis rotation v  , the angle of twist  x and warping deformation  x . In Figure 1a, displacements are shown
with red single arrows, rotations are depicted with blue double-headed arrows while the warping
displacements are illustrated by green triple-headed arrows.

(a) (b)

Figure 1: (a) Buckling DOFs and adopted right-handed coordinate system, and (b) Internal force
interpolation bending moment

The strong-axis bending moments M y 1 and M y 2 at the ends of each member are obtained from the pre-
buckling analysis. Assuming the beam element length L is small, the bending moments within the element
are linearly interpolated (Figure 1b), i.e., My  x   My 1 1 x L   My 2  x L  . The total potential energy in
going from the point of onset of buckling to the buckled configuration is expressed in terms of buckling
displacements as

[1] eb  Ub  Vb  Uv  Usv  Uw   Vm  VPL  VqL 

where the internal strain energy U b is the sum of three components; Uv due to the lateral displacement v
, Usv due to the conventional Saint-Venant torsion, and Uw due to warping torsion. Also, the destabilizing
load potential energy term Vb consists of three components; Vm due to the bending moments, VPL due to
the load height above the shear centre for point loads, and VqL is load height effect above the shear centre
for line loads. By expressing the above six terms in terms of the lateral displacement v and the angle of
twist θ x , one obtains

635-2
 1L 1
L
1
L
 L np
1 1
L

[2] eb    EIz v   dx   GJ x2dx   ECw x2dx      My  x  xv dx   Pi zi xi2   q  x  zq x2dx 
2

20 20 20  0 i 1 2 20 

where G is the shear modulus, J is the Saint-Venant torsional constant, I z is weak-axis moment of inertia,
Cw is the warping constant, zi is the vertical distance between the section shear centre and the point of
application of transverse load Pi  i  1,..., np  taken as positive in the positive direction of coordinate z ,
q  x  is the member transverse line load, and zq is the vertical distance between the shear centre and the
line of application for q  x  , taken positive in the positive direction of coordinate z . The lateral displacement
v  x  and angle of twist  x  x  are related to the nodal displacements using Hermitian polynomials. By
substituting into Eq. [2] and evoking the stationarity conditions, one obtains the linearized eigenvalue
problem:

[3] K    K U  0

G

where  K  is the elastic stiffness matrix,  K G  is the geometric stiffness matrix, the eigenvalue λ is a load
multiplier that determines the buckling load(s) at the which the system buckles and the eigen vector U 
is the buckling configuration.

3. VALIDATION STUDY

Three beam geometries were considered to assess the validity of the FEA. These are: 1) a W410x39 with
a 4.5m span, 2) a W410x39 with an 8m span, and 3) a W310x52 with a 5.7m span. All beams were simply
supported with respect to the lateral displacement and twist (i.e., v  0   v  L    x  0    x  L   0 , but free

to undergo weak-axis rotation and warping, i.e., v   0   0,v   L   0, x  0   0, x  L   0 as may be the
case in beams connected to columns through simple shear connections. A variety of loading conditions
were investigated; end moments  M,  M  , where   1,0.75,0.5,0.25,0, 0.25, 0.5, 0.75, 1 (runs #1-9),
uniformly distributed loads with and without fixed end moments (runs #10, 11), mid-span point load with
and without fixed end moments (runs #12, 13), and two point loads acting at third spans (run #14). A mesh
study indicated that eight finite elements are needed to achieve convergence. Thus, eight elements were
used to mesh all beams. In each case, the critical moments Mcr  Mcr  FEA  based on FEA were computed
and the corresponding moment gradient C  C  FEA  was determined from the equation


EI y GJ   E L  Cw I y
2
[4] Mcr  CMu  C
L

where Mu   L  EI y GJ   E L  Cw I y is the critical moment for the idealized case of a simply supported
2

beam v  0     0   v  L     L   0 
x x subjected to uniform moments. The corresponding moment
gradient values are provided in Column 2 of Table 1. The critical moment expression in Equation [4] is
universal in most standards. For example, the Eurocode guide adopts equation [4] in conjunction with
moment gradient factors C  C  EUR  prescribed for various loading conditions as provided in Column 3
of Table 1. Also, the Canadian standards CAN/CSA-S16-14 adopt equation [4] with a moment gradient
factor C  C CAN   2 given by

635-3
[5] 
C CAN   4Mmax 2
Mmax 
 4Ma2  7Mb2  4Mc2  2.5

where Ma , Mb , Mc are moments are the quarter, mid-span, and three-quarter span points, and Mmax is the
peak moment within the span. The corresponding values are provided in Column (5) of Table 1. The ratios
C  EUR  C  FEA  of the moment gradient factors of the Eurocode guide to those predicted by the FEA are
provided in Column (4) and show close agreement except for run #11 and to a lesser degree for run #13,
both involving fixed end moments. A comparison for C CAN  C  FEA  is provided in Column (6). The
mean value of C CAN  C  FEA  is 0.936 with a standard deviation 0.0703. In all cases, except run #14,
the Canadian moment gradient equation tends to under-predict the critical moments compared to the FEA
results.

Table 1: Comparison of MFG of FEA, CAN/CSA-S16-14 and EN 1993

C  FEA  C  EUR  C CAN 
Run Bending Average C  EUR  / C CAN  /
# Moment
(1) Pattern
value of
(3) C  FEA  (5) C  FEA 
FEA
(2) (4) (6)
(1) 1.000 1.000 1.000 1.000 1.000
(2) 1.140 1.141 1.001 1.131 0.992
(3) 1.320 1.323 1.002 1.293 0.980
(4) 1.548 1.563 1.010 1.497 0.967
(5) 1.854 1.879 1.013 1.746 0.942

(6) 2.186 2.281 1.043 2.028 0.928

(7) 2.556 2.704 1.058 2.286 0.894

(8) 2.821 2.927 1.038 2.405 0.853

(9) 2.720 2.752 1.012 2.309 0.849

(10) 1.144 1.132 0.990 1.130 0.988
(11) 2.694 1.285 0.477 2.359 0.876
(12) 1.362 1.365 1.002 1.265 0.929
(13) 1.723 1.565 0.908 1.414 0.821
(14) 1.040 1.046 1.006 1.131 1.088
--- Mean N/A N/A 0.969 N/A 0.936
--- Standard Deviation N/A N/A 0.140 N/A 0.0703

4. CASE STUDY 1: DEVELOPING DESIGN AIDS FOR CONTINUOUS BEAMS

Consider four continuous beams with equal spans where the spans L vary in the range L   4,5,6,7,8m  .
Beam cross-section in all cases are W250x58 with the properties
 
Iyy  1.88  105 m4 , J  4.09  107 m4 ,Cw  2.68  107 m6 . All supports provide lateral and torsional

635-4
restraints (as marked by the red X) in addition to a transverse restraint. Case (a) involves two spans and is
subjected to mid-span point loads P (Figure 2a). Case (b) involves two spans and is subjected to a
uniformly distributed load (UDL) (Figure 2b). Case (c) involves three spans and is subjected to mid-span
point loads P (Figure 2c) and Case (d) involves three spans and is subjected to UDL (Figure 2d). The
pertinent bending moment diagrams (BMD) are shown below each Case. All loads are applied at the shear
centre. It is required to determine the critical moment in each case, and use the results to develop design
aids by developing moment gradient factors or effective length factors. In a manner similar to the validation
study, the moment gradient factor C  FEA  is obtained from the FEA predicted critical moment Mcr  FEA 

through the relation C  FEA   Mcr  FEA   L  EI y GJ   E L  Cw I y  . Alternatively, one can obtain the
2

 
effective length factor by solving the equation Mcr  FEA    kL  EI y GJ   E kL  Cw I y for k . Figure 3a
2

depicts the relationship between the moment gradient factor C  FEA  and the torsion parameter
  L GJ ECw and Figure 3b depicts the relationship of the effective length factor k versus the torsion
parameter  . The results show that the moment gradient factors are nearly constant, suggesting the
independence of moment gradient of the torsion parameter. In contrast, the effective length factors show a
mild dependence on the section torsional parameter. This observation suggests that the moment gradient
factor provides a simpler approach to estimate the critical moments than effective length approach. The
values provided in Table 2 show that the moment gradient factors is constant for Case (a). Thus a value of
C=1.82 is recommended for Case (a). In a similar manner, the recommended moment gradient factor is
2.29 for Case (b). Cases (c) and (d) show a rather weak dependence on the torsion parameter, and the
smallest values of 1.65 are recommended for Case (c) and 1.75 for Case (d). For run #11 involving three
spans with L=4m, the critical moment based on the Canadian Standard equation is 515.3 kNm for the
exterior span and 736.3 kNm for the intermediate span. Since there is no direct means of accounting for
span interaction in the standards, the critical moment is conservatively taken as the smaller value of 515.3
kNm. This value compares to 891.4 kNm based on FEA. As expected, the FEA solution predicts a higher
critical moment since (a) it accounts for interaction effects whereby the stronger span delays the buckling
of the weaker span, and (b) the moment gradient equation in the standard is only approximate. For Run #1
involving two- spans with L=4m, the critical moment for both spans predicted by the Canadian Standard
equation is 515.2 kNm. An identical value is obtained by modeling a single span, suggesting no interaction
effects between identical spans under identical loads. This value compares to 704.9 kNm as predicted by
the present FEA solution, suggesting that for the present loads, the CAN-CSA-S16-14 moment gradient
equation provides an overly conservative critical moment prediction.

(a) (c)

635-5
 (b) (d)

Figure 2: Continuous beams of Example 1: (a) Two-span beams under mid-span point loads, (b) Two-
span beams under UDL, (c) three-span beam under mid-span point loads, and (d) three-span beam
under UDL

2.5 0.8
0.7
Moment Gradient

Effective Length
2
Factor C(FEA)

0.6
Factor k

1.5 0.5
0.4
1 0.3
2-span under UDL
2-span under point load 0.2
0.5 3-span under UDL 0.1
3-span under point load
0 0
2 4 6 8 2 4 6 8
Torsional Parameter χ Torsional Parameter χ
(a) (b)

Figure 3: (a) Moment gradient factor vs  , and (b) Effective length factor vs 

Run #5 is chosen to illustrate the remaining steps of the design. Given that the W250x58 cross-section
meets class 2 requirements for a yield strength Fy  350MPa , the plastic moment is calculated as
Mp  Zx Fy  270kNm . For run #5, the critical moment is Mu  275.6kNm (Table 2). Given that
Mu  0.67M p , the design is governed by inelastic lateral torsional buckling and the resistance is given by
Mr  Mp 1  0.28 Mp Mu   174.8kNm  Mp . If one assumes the service dead load PD is equal to the
service live load PL , one has P  1.25PD  1.5PL  2.75PL or PL  0.364P . The peak factored moment
corresponding to P as provided in Figure 2a is Mf  0.188PL . Equating the resisting moment
Mr  174.8kNm to the factored moment Mf  0.188PL yields a factored load P  116kN which
corresponds to a service live load PL  0.364P  42.2kN . The corresponding peak displacement within the
span is found as   11.3mm which corresponds to L /   708 which is lower than the threshold value of
L /   360 given in appendix D of CAN-CSA-S16-14, i.e., the live load deflection meets the requirement
of Appendix D.

635-6
 Table 2: Summarized results of Example 1

Critical
Case Run # Span (m)  moment C  FEA  k
(kNm)
(a) 1 4 3.07 704.9 1.82 0.688
2 5 3.83 508.9 1.82 0.670
3 6 4.60 396.8 1.82 0.655
4 7 5.37 325.1 1.82 0.643
5 8 6.13 275.6 1.82 0.631
(c) 6 4 3.07 655.4 1.69 0.718
7 5 3.83 469.8 1.68 0.706
8 6 4.60 364.0 1.67 0.695
9 7 5.37 296.4 1.66 0.686
10 8 6.13 249.9 1.65 0.678
(b) 11 4 3.07 891.4 2.30 0.600
12 5 3.83 643.2 2.30 0.582
13 6 4.60 501.3 2.30 0.565
14 7 5.37 410.6 2.29 0.550
15 8 6.13 347.9 2.29 0.538
(d) 16 4 3.07 699.9 1.81 0.690
17 5 3.83 500.4 1.79 0.678
18 6 4.60 386.8 1.77 0.667
19 7 5.37 314.5 1.76 0.657
20 8 6.13 264.9 1.75 0.650

5. CASE STUDY 2: INTERACTION EFFECTS IN CONTINUOUS BEAMS WITH UNEQUAL SPANS

Consider a continuous beam with a W250x58 cross-section subjected to loads  P1, P2 , P3  as shown in
Figure 4a. The beam is laterally and torsionally restrained at the three support locations. It is required to
determine the elastic LTB capacity of the beam. Two types of solutions are sought: (1) Neglecting
interaction effects between both spans and (2) Considering interaction effects.

(a)

(b)
Figure 4: (a) Continuous beam restrained laterally and torsionally at all three supports, and (b) BMD
under applied loads

Solution (1) - Omitting interaction: Since CAN/CSA-S16-14 does not account for the interaction between
both spans, the designer may opt to calculate the LTB capacity of each span separately by applying the

635-7
moment gradient factors based on the quarter-point formula (Eq. 5). The procedure involves the following
steps: Step 1: Determine the bending moment distribution (Figure 4Figure 4b), Step 2: Determine the critical
moment for each span using the critical moment M cr expressions in Equation 4, where L is the distance
between lateral torsional supports (i.e., the left span has L  4m and the right span has L  8m ). Step 3:
Determine the critical moments for both spans by applying the quarter-point moment gradient equation (Eq.
5) and conservatively taking the smaller value (i.e., neglecting the interaction between both spans). Table
3 provides a summary of the results. For comparison, FEA predictions are provided for the case where the
interaction between both spans is omitted, by modelling each span alone in a separate buckling FEA model
while using 8 elements for each span.

Solution (2) - Incorporating interaction effects: The FEA solution provides a natural means of incorporating
the interaction between the two spans. This is done by modelling both spans in a single model. For the
present problem, this interaction is beneficial for the weaker right span (which governs the design in Solution
1). The stronger left span provides some restraint to left span and delays its buckling. By incorporating
interaction effects and using 8 elements for each span, the critical moment of the system is found to
increase by 28.6% from the Canadian standards predictions (Table 3). As expected, the artificial isolation
of each member leads to overly conservative predictions of the critical moments.

Table 3: Comparison of critical moment predictions for a continuous beam

Estimated
critical
Left Span Right Span moment for
the
structure

Bending Moment
Diagrams

Span L  m  4m 8m
Mu  Eq.4  386.9 kNm 151.8 kNm
M A  kNm  +5 +5
MB  kNm  +10 +10
MC  kNm  -5 +15
Mmax  kNm  -20 -20
C CAN  Eq.5  2.219 1.746
Mcr CAN   C CAN  Mu
858.5 265.0 265.0
 kNm 
Mcr FEA1 kNm -
neglecting interaction
(treating each span in a 1134 295.1 295.1
separate model)

M FEA2 kNm -
accounting for interaction 340.7
340.7 kNm
(treating both spans in a
single model)

635-8
6. CASE STUDY 3: EFFECT OF LATERAL AND TORSIONAL RESTRAINTS AT INTERMEDIATE
SUPPORTS

In the previous case study, the middle support was assumed to be laterally and torsionally restrained. In
the present example, it is required to investigate the case where intermediate support provides vertical
displacement restraint but no lateral nor torsional restraints (Figure 5) as may be the case for a beam
supported by a column that is pinned at the base with no lateral members framing into the beam at the
beam to column junction. All other loading and end boundary conditions remain unchanged. Since the
boundary conditions regarding the transverse displacements are identical to the past case study, the BMD
remains identical (Figure 4b). However, in the absence of lateral and torsional restraints at the middle
support, the unsupported span of the beam becomes 12m and quarter point moments become
MA  5kNm, MB  5kNm, MC  12.5kNm and Mmax  20kNm . The critical moment calculations based
on the Canadian standard equation is found to be 211. 2kNm and the steps of the calculation are provided
in Table 4. Also, provided for comparison are the critical moments predicted by the FEA which is 191.0
kNm. The Canadian moment gradient equation is found to overestimate the critical moment by 10.5% in
this case. A comparison with the previous case study indicates that, although the bending moment
distribution remains identical in the two problems, the critical moment is found to drop from Mcr  340.7kNm
for the case where lateral and torsional supports are provided at mid-span to Mcr  191.0kNm for the case
where such restraints are removed.

Figure 5: Lateral torsional restraints removed at middle support (Example 3)

Table 4: Critical moment comparisons for Example 3

Results
Span L  m  12m

  E 
2

Mu  EIy GJ    Cw Iy 95.20
L  L 
M A  kNm  -5.0
MB  kNm  +5.0
MC  kNm  +12.5
Mmax  kNm  -20
C CAN  Eq.5  2.219
Mcr CAN   C CAN  Mu  kNm  211.2

Finite Element solution M  FEA  191.0

635-9
7. CONCLUSIONS

The following conclusions have been drawn from the current study:

1. The present study implemented and established the validity of the beam finite element of Barsoum
and Gallagher (1970) as part of a larger project to integrate the element within the commercial S-
FRAME analysis software and the S-STEEL steel design software.
2. The formulation was successfully used to derive moment gradient factors in agreement with the
Eurocode Guide recommendations and CAN-CSA-S16-14 moment gradient factors. The element
was then used to capture interaction effects between adjacent spans when calculating the LTB
capacity of continuous beams.
3. Moment gradient factors were proposed for the design of two-span and three-span beams under
either mid-span point loads or UDL. The moment gradient factors account for span interaction.
4. The study has quantified the effect of span interaction when determining the elastic LTB resistance.
Span interaction was shown to play an important role for three-span beams and two-span beams
with unequal spans. In the majority of the cases considered, neglect of interaction effect by adopting
the moment gradient factor in the CAN-CSA-S16-14 was shown to lead to conservative predictions
of the critical moments.
5. The study quantified the effect of lateral and torsional restraints at intermediate supports. For Case
studies 2 and 3, removal of such restraints were shown to result in a significant drop in the critical
moments.

8. ACKNOLEDGEMENTS

The authors wish to express their gratitude to Mr. George Casoli, Dr. Feng Rong, Dr. Siriwut Sasibut, and
Dr. Marinos Stylianou, from S-FRAME Software Inc. for their instructive feedback and effort. Financial
support from the S-FRAME Software Inc. and matching funds from the Natural Sciences and Engineering
Research Council (NSERC) of Canada are also gratefully acknowledged.

9. REFERENCES

Standards Association of Australia (SAA). 1998, Steel Structures, AS4100-1998, SAA, Australian Institute
of Steel Construction, Sydney, Australia.
CSA (2014). "Limit states design of steel structures." Standard CAN/CSA-S16-14, Canadian Standards
Association, Mississauga, Ontario.
ANSI/AISC-360-16 (2016). "ANSI/AISC 360-16." Specification for structural steel buildings, American
Institute of Steel Construction (AISC), Chicago, IL.
Barsoum, R. S., and Gallagher, R. H. (1970). "Finite element analysis of torsional and torsional–flexural
stability problems." International Journal for Numerical Methods in Engineering, 2(3), 335-352.
Gardner, L., and Nethercot, D. A. (2011). DESIGNERS’ GUIDE TO EUROCODE 3: (EN 1993-1-1, -1-3 and
-1-8) DESIGN OF STEEL BUILDINGS, Imperial College London, UK.
Nethercot, D. A., and Trahair, N. S. (1976). "Lateral Buckling Approximations for Elastic Beams." Structural
Engineering, 54(6), 197-204.
Trahair, N. S. (1977). "“Lateral Buckling of Beams and Beam-Columns,’’ in Theory of Beam-Columns."
McGraw-Hill, New York.
Trahair, N. S., and Bradford, M. A. (1988). The Behaviour and Design of Steel Structures, Chapman & Hall,
London.
Vlasov, V. Z. (1961). Thin-walled elastic beams, 2nd Edition, Israel Program for Scientific Translation,
Jerusalem.
Ziemian, R. D. (2010). Guide to stability design criteria for metal structures, John Wiley & Sons, New York.

635-10