B.

PLATED STRUCTURES
Limit states design in structural steel, CAN/CSA-S6-00 (Canadian Highway Bridge Design Code), Crane Code (CMAA
Specifications #70 & #74).

B.1 Introduction
The high bending moments and shear forces associated with carrying large loads over long spans will frequently
exceed the capacity of universal beam sections. In this situation, plate girders may be fabricated, their
proportions being designed to provide a high strength to weight ratio. In a fabricated plate girder, the primary
function of the flanges is to resist axial compressive and tensile forces arising from the bending moments. The
primary function of the web is to resist the shear force. For an efficient plate girder design, the web depth d
should be increased as far as possible to give the lowest flange force for a given bending moment. To reduce
self-weight, the web thickness t should be reduced to a minimum. The consequence of these requirements is that
the web has a high d/t ratio and tends to buckle in shear if stiffeners are not provided. For an economic design,
advantage should be taken of the post buckling reserve of strength commonly known as “tension field action”.
BS 5950-1 does allow this reserve of strength to be taken into account. It is inevitable that the increased
efficiency of designs to BS 5950-1 leads to some additional complexity of design calculations. There are special
requirements for the ends of the plate girders in order to anchor the “tension field action”.
Steel girders are typically
1. rolled beams 2. steel plate or 3. box girders.
A rolled beam is a steel girder which has been formed by hot rolling. The most common type of rolled beam
used is the I-type. Rolled beams have the advantage of quick erection, straight forward fabrication, and an
overall simplicity of design.
A plate girder, like a rolled beam, has an I-type cross section. Rather than being hot-rolled, however, the girder
is constructed from steel plate elements which are connected together with welds, bolts, or rivets. A greater
economy of material can be obtained as the designer has the ability to specify the section properties of the
stringer to accommodate the local forces (e.g. flange thickness).
Plate girders gain an advantage over rolled beams as span lengths become great.
The box girder is a form of plate girder which utilizes four or five rather than three plates. Since box girders
possess excellent torsional stiffness, they do not usually require secondary members to provide bracing. Bridge
superstructures are a common application of steel plate and box girders. Box girders while more expensive to
fabricate than plate girders because of their complexity, have a number of significant advantages, particularly
for longer spans. Firstly because of the shape of the box, the top flange itself can act as the decking without the
need for a concrete deck. They can also be designed with an aerodynamic shape, again making them ideal for
long spans.
Advantages of plate girders over box girders:
1. Easier to fabricate
2. Easier to handle in workshop
3. Jigs not required
4. Fewer field splice bolts since bottom flange is narrower than for a box girder
5. Lower unit price
6. Lighter piece weight erection where crane capacity is a concern
Advantages of box girders versus plate girders:
 1. More efficient load distribution due to high torsional stiffness
2. Efficient where girder depth must be minimized
3. Efficient for curved alignments
4. Less area exposed to airborne road salts
5. Less horizontal surface onto which corrosion products can deposit
6. Fewer bearings possible with multiple box girders
7. Fewer pieces to erect
8. Improved aesthetics

Typical sections

b
T

weld
t d

Rolled beam/section Plate girder Box girder

5.0 PLATE GIRDERS

5.1 Plate Thickness and Proportions
The choice of plate thickness is related to buckling. If the plates are too thin they may require stiffning to
restore adequate stiffness and strength, which costs more.
The following are the recommended plate thicknesses and proportions for plate girders:
 Maximum depth -to-thickness ratio (d/t) of the webs of plate girders;
/
120 120 where, pyw is the design strength of web plate

 Width-to-thickness ratio of the compression flange (b/T);

/
13 13 where, pyt is the design strength of compressive flange

 Vertical or transverse stiffness may be provided but hardly horizontal stiffness

 To avoid serviceability problems, minimum web thickness values;
- For unstiffened webs,

- For a > d, where, a is the web width

 /
- For a ≤ d,

 To avoid buckling of compression flange into the web;

- For unstiffened web,

- For transversely stiffened webs, for a >1.5d;

/
for a ≤ 1.5d;

5.2 Buckling of Plates

The webs and flanges of both steel plate and box girders are comprised of flat steel plates. Failure could be due
to local buckling, yielding or an interaction of the two. A designer must ensure that both local and global do not
happen.
It should be noted however that it is desirable for global buckling to be the first mode of failure reached as the
loads on a girder are increased past the anticipated service loading. Global buckling is associated with large
deformations which give warning of failure and is generally less brittle and sudden than local buckling.
In order to understand the design of steel plate and box girders, it is important to understand the principles
behind plate buckling.
5.2.1 Buckling of Unstiffened plates with Uniaxial Uniform Compression:

For a plate to be considered slender, the in-plane dimensions, a, and, b, need to be significantly greater than the
plate thickness, t. The dimension, b, is usually taken as the direction transverse to the main direction of in-plane
loading and the expression, m = a/b is called the aspect ratio.
(a) Long rectangular plates

The elastic critical stress of a long plate segment, σc, is determined by;
 the plate width-to-thickness ratio, b/t,
 the restraint conditions along the longitudinal boundaries,
 the elastic material properties.
It is expressed as
#$ %
! "
( '($ )( / )$

k, is the buckling coefficient which is determined by a theoretical critical load analysis, and is a function
of plate geometry and boundary conditions.
 σ

Segment of long plate having thickness t, width b, and

various edge conditions as tabulated below

σ
Case Description of edge support k b
1 Both edges simply supported 4.00

2 One edge simply supported, the other 5.42

fixed
3 Both edges fixed 6.97

4 One edge simply supported, the other 0.425

free
5 One edge fixed, the other free 1.277

(b) Short plates:

When a plate is relatively short in the direction of the compressive stress (i.e. a/b<<1), the critical stress,
σc may be conservatively estimated by assuming that a unit width of plate behaves like a column.
5.2.2 Buckling of Unstiffened plates due to Pure bending:
#$%
The critical buckling load equation above, i.e ( ! " ( '( $ )( / )$
) is used with substitution of the appropriate
buckling coefficient, k.
Pure bending plate buckling coefficients:
k = 23.9 edges simply supported
k = 39.6 unloaded edges fixed
k = 0.85 top edge free, bottom edge simply supported
k = 2.15 top edge free, bottom edge fixed
The buckling coefficient for a plate in bending is very significantly influenced by the fact that half (in linear
response) of the load is applied in tension.
 σ σ

-σ a -σ

Plate subject to pure bending

5.2.3 Buckling of Unstiffened plates due to Pure Shear:

In a plate subject to pure shear, there exists tension and compression stresses equal in magnitude to the shear
stress and inclined at 45º. The destabilizing influence of compressive stresses is resisted by tensile stresses in
the perpendicular direction.

, :<
;
*
a

Plate subject to pure shear

#$%
The critical stress can be obtained by substituting, τc and ks for σc and k in, ! "
( '( $ )( / )$

#$ %
i.e; *! "+
( '($ )( / )$

Where shear buckling coefficients, ks are;

(i) For plates simply supported on four edges;
.
, 1: "+ 4.00 0
1$
.
, 1: "+ 5.34 0
1$

(ii) For plates clamped on four edges

4.54
α ≤ 1: "+ 5.60 0 1$
.8
, 1: "+ 8.98 0
1$

(iii) For plates clamped on two opposite edges and simply supported on the other two edges
4.54
α ≤ 1: "+ 5.61 0 9 1.99,
1$
 .8 .55
, 1: "+ 8.98 0 9
1$ 1$

5.2.4 Buckling of Unstiffened plates due to Combined Stresses:

The Crane Code (CMAA Specifications #70 & #74, revised 2000) gives useful equations for the buckling
coefficient for simply supported plates subject to combined in-plane bending and compression according to the
cases shown in the Table below
Case Loading
1 Compressive stresses, varying as a
σ1 σ1
straight line. 0 < ψ < 1

ψσ1 a=αb ψσ1

2 Compressive and tensile stresses;

σ1 σ1
varying as a straight line and with
the compression predominating. -
1< ψ < 0 b

ψσ1 a=αb ψσ1

3 Compressive and tensile stresses;

σ1 σ1
varying as a straight line, with
equal edge values, ψ = -1 or with
predominantly tensile stresses. ψ < b
-1
- σ1 a=αb - σ1

σ1 σ1

ψσ1 a=αb ψσ1

Buckling coefficients for combined bending and compression

The critical stress is then calculated using the buckling coefficient, k

 4.
Case 1: α>1 "
=> .

.
α<1 " ,0
1 => .

Case 2: " ?(1 0 @)"′B 9 (@"") 0 ?10@(1 0 @)B

k’ is the buckling coefficient for ψ = 0 (case 1)
k’’ is the buckling coefficient for ψ = -1 (case 3)
Case 3: α ≥ 2/3 k = 23.9
.4
α < 2/3 " 15.87 0 0 8.6,
1$

With predominant tension, replace the width of the plate, b, by 2 times the width of the compression zone for
calculation of α and σc.
The crane code also gives an interaction failure criteria for plates subject to in-plane bending, compression and
shear. First the comparison stress, σ1k, is calculated;
√G $ > H$
E JKL O RSM O $ T $
I M NIO N>QI M O N >IT N
P P P

If the resulting critical stress is below the proportional limit, σp, buckling is said to be elastic. If the resulting
value is above the proportional limit, buckling is said to be inelastic. For inelastic buckling the comparison
stress is reduced to σ1kR;
G GJV $
EU . 4 8G $ >GJV $
G
.
The comparison stress is then used to calculate a safety factor, ϑB, which is then compared with allowable
design factor values, DFB, for each load combination.
Elastic buckling
GJV
WX YZ[
√G $ > H$
Inelastic buckling
GJV\
WX YZ[
√G $ > H$
Design factor DFB requirements:
Case 1 DFB = 1.7 + 0.175(Ψ - 1)
Case 2 DFB = 1.5 + 0.125(Ψ - 1)
Case 3 DFB = 1.35 + 0.05(Ψ - 1)
5.3 Buckling of stiffened Plates:
 σ4

σ3
d1
d2 σ2

d3

σ1 τ
Stress states in a longitudinally stiffened plate
The Figure above shows the stress state in each sub panel for a stiffened plate subject to bending and shear
(compressive stresses can easily included by modifying the ratio of longitudinal stresses at the edge of each
panel). Each sub panel can be checked for local buckling subject to these stresses using the buckling
coefficients given in combined stresses.
A second mode of failure also needs to be checked, which is local buckling of the plate as a whole including the
stiffeners. The third and final mode of failure involves the local buckling of elements of the stiffeners
themselves. This is usually avoided through adhering to slenderness limits.
5.4 Beam Classification
5.4.1 Moment-Curvature (M-κ) relationships for beams:
From elementary strength of materials, an elastic beam subject to a bending moment will have a moment
_ a _b
curvature relationship given by: κ →
%` b %`

From this it can be seen that the strain distribution will be linear, increasing from zero at the neutral axis to a
maximum at the outer most fibre. If the strain does not exceed the yield strain of the material, the stress
distribution will also be linear, given as the product of the stiffness multiplied by the strain at any given point,
this is shown as stress state 1 below

σ
ε

M
M
y

State 1 Stress

State 2 Stress

b State 3 Stress
Beam section Strain Stress
Strain and Stress in a beam subject to Pure bending
As
the moment in the section is increased, the strain in the outermost fibre increases and eventually reaches the
yield strain εy at which point the moment is given by:
 db be

where, S = I/y → section modulus

As the moment increases, the strain exceeds the yields strain εy, the stress cannot increase above the yield
strength (if work hardening is ignored), and hence yielding begins to penetrate from the extreme fibres towards
the interior of the cross section (stress state 2)
The strains will continue to increase until the cross-section is fully plastified, at this stage the full plastic
moment, Mp, has been developed (stress state 3). The plastic moment is computed by summing up the moments
caused by the forces on each fibre at a stress σy to obtain:
Mp σyZ
where the plastic modulus, Z, represents the first moment of area of the tension and compression zones about
the neutral axis. Theoretically an infinite strain is required for a section to reach Mp, but experiments have
shown that for most wide flange shapes, 98% of the plastic moment Mp is obtained at strains of twice εy.
The diagram shown below represents the moment - curvature relationship of the beam element, as the moment
is increased.

Residual strains neglected

Strain hardening
M considered

Mp
My

0.7My Residual strains considered

EI
k

Moment – Curvature relationship for beam section

For moment values below My, the relationship is elastic and the slope of the M - k curve is equal to EI. Above
My, yielding begins to penetrate through the flanges and the section ‘softens’. As the section continues to
deform, the moment closely approaches Mp.
5.4.2 Load- deflection relationships for beams:
(a) Local buckling:
The moment-curvature relationship reflects only the behaviour of a short element of beam length. To predict the
behaviour of a complete member, the curvatures corresponding to a given bending moment distribution must be
integrated along the member to determine the slopes and deflections. The idealized load deflection relationship
for a beam subject to a central concentrated load, P is shown
 op
Plastic moment
dn Ideal behaviour
4

Mp
My
Compact (class 2)
Plastic design section
Non-Compact (class 3) (class 1)
0.7My

Slender (class 4)
EI
Δ

Load deflection relationships for beams

However, the idealized load-deflection relationship is not attained for all beams, as local buckling of the
component members making up the element may occur prior to the attainment of the full plastic moment, Mp.
This leads to the categorization of beams into four primary classes according to what load-deflection
relationship their component elements will allow as shown above.
(b) Class 1: Plastic design section
The component elements should be such that the beam has the capacity to both attain its full plastic moment,
Mp, and undergo large inelastic rotations, enabling the complete structure to redistribute bending moments and
reach the load carrying capacity as anticipated on the basis of a plastic analysis.
dk ∅d ∅mZb
(c) Class 2: Compact section
The component elements should be such that the beam has the capacity to attain its full plastic moment, Mp, but
is not required to undergo large inelastic deformations.
dk ∅d ∅mZb
(d) Class 3: Non compact section
The component elements should be such that the beam is capable of developing a moment resistance just equal
to the yield moment My.
dk ∅db ∅eZb
(e) Class 4: Slender section
The component elements are such that they buckle locally at a moment less than My and the moment resistance
is a function of the width-to-thickness ratios of the component elements. This class is subdivided into three
categories.
(i) contains those sections having both flange and web plates falling within Class 4.
(ii) contains those sections having flanges meeting the requirements of Class 3 but having Class 4 webs.
(iii) contains sections having web plates meeting the Class 3 requirements, but with compression flanges
exceeding Class 3 limits.
for Class 4(iii);
dk ∅db ∅eq Zb
where, Se is the effective elastic section modulus determined using an effective flange width. The
effective width is 670t/√ Fy for flanges supported along two edges parallel to the direction of stress and 200t/√
Fy for flanges supported along one edge parallel to the direction of stress. For flanges supported along one edge,
in no case shall b/t exceed 60.
For Class 4(ii) sections, equation dk ∅db ∅eZb is used to calculate the basic moment resistance, which is
then reduced to account for the effects of local buckling (softening) of the slender web. Most plate girders fall
into this class of section.
The Table below show Width – to -Thickness Limits
Class Plate girder flange Web in compression Box girder flange
in compression under compression
;n / 145/rZb s ;n / 525/rZb
Class 1 ∗w
t ru
;n / 170/rZb s ;n / 525/rZb
Class 2 ∗w
t ru
;n / 200/rZb s 5 ;n / 670/rZb
Class 3 ∗w
t ru
With
w x1 9 0.39 ∗ yz /({ ∗ yb )|
w x1 9 0.61 ∗ yz /({ ∗ yb )|
w x1 9 0.65 ∗ yz /({ ∗ yb )|

5.4.3 Lateral Torsional Buckling

Beams subject to flexure have much greater strength and stiffness in the plane in which the loads are applied
(major principal axis) than in the plane of the minor axis. It has been assumed thus far that the strength of the
beam is determined by the capacity of its cross-section, and this, in turn, is dependent on the local buckling
capacity of its plate elements.
However, if the beam is laterally unsupported, the strength may be governed instead by lateral-torsional
buckling of the complete member, as shown

Position before buckling Position after buckling

Lateral-Torsional buckling motion

Beams are especially prone to this type of buckling during the construction phase, where lateral bracing are
either absent or different in type from their permanent ones.
the elastic lateral buckling strength of doubly symmetric beams;
~$ #
d}
•
r€•b ‚ƒ 0 („€/p) •b yt

where, … 1.75 0 1.05" 0 0.3" 2.5

k, is ratio of smaller bending moment to large bending moment at opposite ends of the unbraced length.
This equation is accepted as the basis for the design of slender members and is thus valid until Mu reaches two-
thirds of Mp for Class 1 or 2 sections, and My for Class 3 or 4 sections.
The following is valid for doubly symmetrical members:
For d} 0.67d (slender members);
dk ∅d} for class 1 or 2
For d} > 0.67db (stocky members)
. 4_‡
dk 1.15∅d 19 ∅d for class 1 or 2
_ˆ

For d} 0.67d ;
dk ∅d} for class 3 or 4
For d} > 0.67d
. 4_
dk 1.15∅db 1 9 _ˆ
∅db for class 3 or 4

Lateral torsional buckling can be avoided by properly spaced and designed lateral bracing, or by using cross
sections which are torsionally stiff, such as box-shaped sections or open-section beam groups connected
intermittently by triangulated lacing or by diaphragms or by ensuring that the required design moment does not
exceed the lateral-torsional buckling capacity.

5.5 Design of plate girders

There are a variety of factors which the designer must play off of each other in order to realize the greatest
economy of materials. Questions such as; should a thicker web be used or a thinner one with fewer stiffeners?
Should the flange thickness be varied or kept constant? Should the girder be haunched?
The web of the plate girder can fail due to buckling, yielding or a combination of the two in either shear or
bending. The web dimensions and stiffener spacing are chosen in order to ensure that there is an adequate safety
margin with respect to these failure modes. The flange of the plate girder is sized to prevent local buckling or
yielding.
5.5.1 Preliminary sizing
A preliminary design of a plate girder takes the weight of steel and the amount of fabrication into main
consideration. The optimum depth for moment resistance according to allowable stress design is used to obtain
an expression for the web depth, h.
/
dz
ℎ ≈ 540 ” •
Zb
Once an approximate web depth is chosen, an approximate flange area, Af (area of one flange) can be
calculated, assuming that
 - lateral torsional buckling will not govern the design,
- the flange material will be able to reach yield,
- the flange areas are concentrated at the top and bottom of the web and
- the contribution of the web to bending resistance can be neglected.
dz
–z ≈
Zb ℎ
Once an approximate web depth is chosen, an assumption that the web carries the entire shear will enable the
web thickness, w, to be determined;
—z
–t ˜ℎ
∅Z+
The shear resistance, Fs is a function of the web slenderness, h/w, and whether there web is stiffened however
the maximum strength is given by
Z+ 0.66Zb
By considering the maximum allowable web slenderness, the minimum web thickness, wmin, can be obtained.
The minimum web thickness for corrosion protection is 4.5mm. By considering the minimum web thickness for
no reduction in flange stress and hence no reduction in moment resistance due to web slenderness, w2, can be
obtained. A value of web thickness can then be chosen between these two values as a starting point for the
detailed design.
u s
˜™š›
4

ru s
˜
5

5.5.2 Design of cross section for flexure

A plate girder subject primarily to bending moment usually fails by lateral-torsional buckling, local buckling of
the compression flange, or yielding of one or both flanges. Buckling of the compression flange into the web
(vertical buckling) has been observed in many tests.
(a) Lateral-torsional buckling
The Standard specifies for laterally supported members:
Class 1, 2: dk ∅d ∅mZb
Class 3, 4(ii)*: dk ∅db ∅eZb
Class 4(iii): dk ∅eq Zb
where, Se is the effective elastic section modulus determined using an effective flange width of 670t/√ Fy for
flanges supported along two edges parallel to the direction and an effective width of 200t/√ Fy for flanges
supported along one edge parallel to the direction of stress. For flanges supported along one edge b/t should not
exceed 60.
* For Class 4(ii) sections the moment resistance may need to be reduced.
The Standard specifies for laterally unsupported doubly symmetric members:
 ~ #
d} r€•b ‚ƒ 0 („€/p) •b yt with all the earlier stated equations for type and class of
$
•
members following.
(b) Web buckling under pure flexure
Due to the initial out-of flatness of the web, buckling of slender webs (Class 4) due to bending occurs at low
stress levels. This does not exhaust the panel capacity, as stresses are redistributed in the post buckling range,
allowing the girder to continue to take increasing bending moment. However the web does not resist the stresses
given by conventional bending theory, but ‘throws off’ some of its load to the stiffer flange. Hence the web is
less effective in resisting bending and the flange experiences a higher stress than that given by the code
resulting in a reduction in ultimate moment capacity of the girder.
A linear reduction to this maximum attainable value is then applied, which is a function of web slenderness, the
relative proportions of the flange and web, and the buckling load of the web.
ž s 5
dk œ dk •1 9 0.0005 ž t
9 ¡
Ÿ r_Ÿ /∅

If a longitudinal stiffener is provided which will prevent the lateral deflection of the web, or if the web
slenderness h/w 1900/ √(Mf/∅S) then this reduction need not be applied. However if a reduction is
required, then Mr’ should be used in place of Mr in all subsequent calculations.
(c) Vertical buckling of the web
As the curvature which accompanies bending occurs, a vertical force is transmitted from the flanges into the
web. This force may cause vertical buckling of the web on the compressive side.
By equating the buckling resistance of the web to the applied compressive force and by making some
simplifying assumptions and substituting suitable values for E and υ leads to a requirement on the web
slenderness (h/w) in order to prevent vertical buckling of the web.
s 4
t u

5.5.3 Design of cross section for shear

The code identifies 3 limiting states for determining the shear capacity of the web;
- shear buckling,
- shear yielding or
- a combination of the two.
The mode of failure that will occur first is dependent on the web slenderness and stiffener spacing.
(a) Shear yielding:
When steel is subject to a combined stress condition, the yield stress in shear, Fy, is normally approximated by
the Von-Misses value, which is increased to allow for the strengthening effects of strain hardening:
Zb
Z+ ¥ → 0.66Zb
√3
(b) Shear buckling:
The standard equation for buckling of a plate subject to pure shear is used to calculate the resistance to shear
buckling after substituting the correct notation for dimensions of plate girder webs.
E# $ %
*!k ( '($ )(s/t)$
 .
For a/b ≥ 1.0, for simply supported edges, it is found that: " 5.34 0 (
/s)$

(c) Combined shear yielding and buckling:

The equation for this curve is given and was chosen mainly on the basis of experimental evidence.
s
1. 439r"¦ /Zb
t

Z+ 0.66Zb

With a/h < 1 then

.
"¦ 40( (shear buckling coefficient)
/s)$

or a/h ≥ 1 then

"¦ 5.43 0 (
/s)$
(shear buckling coefficient)

2. 439r"¦ /Zb (ℎ/˜) 621r"¦ /Zb :

5 r u E§
Z+ Z!kš (s/t)

3. (ℎ/˜) > 621r"¦ /Zb :

Z+ Z!kq 0 " ¨0.5 ∗ Zb 9 8.66 ∗ Z!kq ©

with
4 E§
Z!kq (s/t)$
and
" (aspect coefficient)
r >( /s)$

The code also imposes a limit on slenderness:

s 4
t u

The capacity of the section is calculated by multiplying the ultimate shear stress for the web multiplied by a
resistance factor φ and the web area Aw.
Vr φ Aw Fs
(d) Transversely stiffened Girder webs:
The upper limit of the strength of a girder web stiffened by vertical supports will be the same as that of the
unstiffened girder, that is, the strength corresponding to shear yielding. As for unstiffened webs, stiffened webs
may fail due to shear buckling before shear yielding occurs, however in stiffened webs it is found that
significant amounts of shear past the theoretical buckling load can be carried, due to ‘tension field action’.
Subsequent to buckling the stress distribution in the web changes and considerable postbuckling strength may
be realized because of the diagonal tension that develops. This is called the ‘tension field action’. Even without
transverse stiffeners a plate girder can develop a shear stress significantly greater than the shear-buckling stress,
although this is not recognized by the code.
5.6 Design of stiffeners

Longitudinal stiffener

Transverse stiffener

5.6.1 Bearing (Transverse) stiffeners:

The code specifies that bearing stiffeners are required where factored concentrated loads or reactions exceed the
factored compressive (vertical) resistances of the plate girder web. Failure due to concentrated loads can be due
to local buckling of the web in the region where it joins the flange, or overall buckling of the web throughout its
depth.
5.6.2 Longitudinal stiffeners:
Longitudinal stiffeners can greatly increase the bending strength of plate girders. This additional strength can
be attributed to control of the lateral deflection of the web which increases the flexural stress that the web can
carry and also improves the bending resistance of the flange due to greater web restraint.
5.6.3 Intermediate transverse stiffeners
Transverse intermediate stiffeners increase the critical buckling stress of the web by providing anchorage for
tension field action, and thus allow a reduced web thickness. As the number of transverse stiffeners is increased,
the web thickness can be reduced further, however there is an optimum number of stiffeners before the
fabrication cost of adding an extra stiffener outweighs the cost saving through a reduced web thickness.
5.7 Design of box girders
Box girders are generally rectangular or trapezoidal in cross section, and can be either open or closed.
Trapezoidal cross sections offer several advantages over rectangular cross sections. The narrower bottom flange
allows steel savings in low bending areas and offers plate stockiness in compression zones of multi span
bridges. A trapezoidal cross section is also more aesthetically pleasing.
Rather than a limit states approach, where the resistance of the section is calculated for several limit states and
the lowest value used for comparison with the applied loading, a form of working stress design will be used.
This involves the calculation of stresses due to the applied loading which are then used to check for buckling or
yielding of any individual plate element or the cross section as a whole, accounting for the effects of stress
interaction.
5.7.1 Design of cross section for flexure
A box girder subject primarily to bending will normally fail due to buckling of the compression flange. Unlike
plate girders, lateral torsional buckling rarely governs for practical box girders. Internal bracing and diaphragms
are typically used to increase resistance to lateral torsional buckling and maintain the cross-sectional shape.
Failure is deemed to have occurred when the extreme fiber flange stress reaches the calculated ultimate stress of
the compression flange, or the yield stress of the tension flange.
Traditionally box girder flanges were stiffened in both the longitudinal and transverse directions, however the
current trend is towards the use of longitudinal stiffeners only. The unstiffened flanges of narrow box girders
can be treated as plates, by using effective widths to account for the effects of buckling.
5.7.2 Design of cross section for shear
The key difference between plate and box girders which may influence the shear strength of the webs is the use
of relatively thin flanges in box girders at the boundaries of the webs. Caution is needed in applying available
tension field models, derived and verified in the context of plate girders, to the design of the webs of box
girders.
Of major concern is the relatively small amount of support against in-plane movement which may be afforded
to the web by the thin flange of a box girder, compared with the restraint offered by the thicker and narrower
flange of a corresponding plate girder. In the latter case the out-of-plane bending rigidity and in-plane
extensional rigidity of the flange to resist movement perpendicular to and parallel to the flange/web junction,
respectively, is more effectively mobilized than in the case of thin flanged box girders.
5.7.3 Design of cross section for Torsion
In the design of webs, additional shear stresses caused by torsion may be added to those associated with
bending when calculating the total stresses applied to the web.
Allowance can also be made for the effects of the shears induced by torsion in the flanges, by considering the
effect on stability of the flanges.
5.7.4 Design of stiffeners
The flanges of webs of box girders have traditionally been stiffened both longitudinally and transversely.
However, the design using three or more longitudinal stiffeners and designs using both longitudinal and
transverse stiffeners are seldom economical. Hence the current trend for box girders is for the webs and flanges
to be stiffened longitudinally only.
5.7.5 Combined shear, moment and torsion
Each plate element of a box girder is typically longitudinally stiffened and may also be stiffened transversely.
For detailed and accurate analysis finite element methods are utilized. However for an approximate and quicker
solution, there are several structural idealizations that can be made to simplify the design. In increasing order of
complexity, a stiffened plate element can be treated as:
1. A plate of uniform effective thickness
2. A series of disconnected struts
3. An orthotropic plate
4. A discretely stiffened plate
5.8 Design Approaches for Box girders
5.8.1 Plate effective thickness approach:
The basis of this approach is to smear the longitudinal stiffeners to create a flat plate of length equal to the
transverse stiffener spacing and with an effective thickness, allowing the use of plate buckling formulas. The
thickened plates are assumed to be simply supported between transverse stiffeners or internal diaphragms. The
effective thickness is calculated to give the same moment of inertia as the stiffened plate.
The applied loads on the webs or flanges are used to check for buckling of the effective plates and sub panels
(between longitudinal stiffeners). Yielding of the section is also checked considering the effect of combined
stresses. This is the simplest method and is adopted in many formatted design spreadsheets for the design of box
girders.
5.8.2 The strut approach:
The basis of this approach is to treat a plate stiffened by several equally spaced longitudinal stiffeners as a series
of unconnected compression members or struts, each of which consists of a stiffener acting together with an
associated width of plate that represents the plate between stiffeners. Knowing the cross-sectional properties of
the strut, the calculation of the buckling strength of the strut can be obtained from a column buckling formula.
5.8.3 Orthotropic plate approach:
An orthotropic plate is one whose material properties are orthogonally anisotropic. A uniformly stiffened plate
is reduced to this case by effectively ‘smearing’ the stiffness characteristics of its stiffeners over the domain of
the plate. Clearly, the theory is best applicable when the spacing of the stiffeners is small.
5.8.4 Discretely stiffened plate approach:
Analytical studied have been made of stiffened panels in which account has been taken of the discrete nature of
the stiffening and which incorporate non-linear geometric and material effects. Both finite difference and finite
element numerical formulations have been used of this purpose.
5.9 Fabrication considerations
Evaluation of the economics of design often includes a perception that least weight and least cost are
synonymous. Although cost of a structure is related to the weight of steel material, there are numerous other
considerations in purchasing, fabricating, shipping, and erection and effective use of material locally which may
override the decision to aim for a least weight structure. Some of these considerations require familiarity with
purchasing, fabricating and erecting processes. Various fabricators have their own processes, and it is difficult
for a designer to produce a design to satisfy everyone. Fabricators should be allowed flexibility in detailing,
with designer approval, to make adjustments to the number and location of splices. Material content is only one
element in the cost equation and will represent about 20% to 30% of the total ‘in place’ cost in fairly standard
bridges. The total rate per tonne of steel depends on several factors, including:
1. Complexity of details
2. Quality control requirements
3. Amount of welding, including grinding, type and amount of inspection etc…
4. The amount of repetition and reuse of assembly jigs
5. Size and number of individual pieces to be fabricated
6. Other demands on shop space, particularly when large box girders are involved
7. The access for erection
8. Number of girder field splices
5.10 Erection considerations
A well conceived economical steel bridge requires consideration of its erection at two stages in the design
process. Firstly, erection must be considered at the concept stage because it typically represents about 30% of
the superstructure cost and therefore the most economical arrangement cannot evolve without its consideration.
Truss versus girder, curved versus parallel chords and flanges, continuity, main member dimensions, drop in
spans, pier arrangements, etc… all have significance at this stage.
Secondly, erection must be considered at the detail stage. Details of splices, diaphragms, bracing and pier
members are very significant contributors to erection cost.
Those elements which are in the control of the designer should be designed to facilitate construction wherever
possible. Field labor is very expensive, therefore keep things simple. Realistic tolerances must be built into the
system wherever shop fabricated elements meet field construction. Access to splices, anchor bolts and bearings
and adequate space to install jacks is very necessary for proper installation, inspection and future maintenance.
5.10.1 Constant depth or Curved chords:
Strictly from an erection point of view, constant depth girders have the advantage. Pier sections of haunched
girders frequently require extraordinary effort in shipping, handling and turning because of their increased bulk.
Constant depth girders are much easier to ship, to turn and to lift and block.
5.10.2 Plate girder or box girder:
From an erection point of view, the box girders are usually preferable to the plate girders because there are
fewer pieces of girder and less bracing. Box girders are reasonably stable in shipping, handling and free
cantilever, whereas plate girders, particularly if slender in flange width, can pose stability problems in shipping,
and handling and frequently require top chord stiffening trusses in cantilever erection.
Plate girders can often be nested during shipping whereas the internal diaphragms present in common box
girders prevent nesting. The particular configuration of box selected has a very significant effect on the erection
cost of the bridge. Unless circumstances dictate boxes larger than about 3.5 m in width should be avoided
because they will cause excessive shipping and handling problems and, in the limit, will require a longitudinal
splice.
Box girders having more than two webs should be avoided except for special situations such as an axial girder
cable stayed bridge.
Flange width has an impact on the stability of the girder during handling and erection. According to an industry
rule of thumb, Igirders will be stable if their length is less than or equal to 60 times the flange width. If this is
exceeded the erector and fabricator may need to use temporary bracing to handle and erect the girder.
5.10.3 Splicing:
The maximum economy will result if the fabricator/erector is permitted freedom to choose the splice locations
that best suit his equipment. If the strength requirements of the splices are spelled out in general terms in the
drawings and specifications, then the Contractor can detail the bridge with his preferred splice locations for the
Engineer’s approval.
It is common to have all holes drilled or punched sub-size and then reamed to full size in full girder assembly of
not fewer than three girder sections, laid on blocking corresponding to the cambered shape. If this method is
performed accurately, all components should fit precisely in the field and the required bridge geometry will be
attained. This method has the advantage of minimum time spent on field fitting and rework, as well as optimum
quality in the connection. However, it should be noted that the large assemblies tie up a lot of shop space and
reaming is very time consuming.
5.10.4 Diaphragms and bracing:
While angle bracing and diaphragms are very cheap to fabricate, they are generally very expensive to erect, due
to the cost of the crane and labor for erection. In order to minimize these costs, the designer should not use
bracing and diaphragms indiscriminately, but only were strictly necessary.