U 336380
U 336380
A.P. Robertson
ABSTRACT
by
FEBRUARY
1983
ProQuest Number: U336380
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
uest.
ProQuest U336380
ProQuest LLC
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, Ml 48106-1346
's
y:v6 o>
ACKNOWLEDGEMENTS
The author wishes to express his sincere thanks to all those who
have contributed towards the work of this thesis in any way, large
or small.
ment, for his assistance with the experimental work, and to the
to the work might not have been possible. Thanks are extended parti
cularly to Mr. R. Widdowson and Dr. H. Jack for supporting the work
programme; to Mr. C. Grimley and Mr. A. Roper for technical advice; and
The author is very grateful to Helen Townsend for her excellent work in
Smith for his careful photography work; and to Doug Pratt, Noreen
SUMMARY
directly above one of the web plates. This web is subjected to in-plane
patch loading produced by the spread of a wheel load through the overlying
rail and flange. This study concerns the load carrying capacity of plate
panel.
The girder was loaded through various interfaces above the web by a wheel
the adjacent edges. Patch loads of various lengths are considered over
simply supported and clamped edges. Also considered are some non-
the support condition along the loaded edge which has greatest influence
Ill
modified form derived which reveals the transition region from collapse
initiated by direct web yielding for girders with stocky webs to failure
the analysis.
span model box-girders subjected to a wheel load above one of the webs.
The effect on the failure load of rail size, web thickness, panel aspect
was exhibited by several of the test web panels. From the results, a
The main findings of the work are used as abasis for a series of
CONTENTS
Page
Acknowledgement s i
Summary ii
Notation viii
1.1 Introduction 1
1.7 Terminology 13
2.1 Introduction 16
2.3.3 Results 33
2.3.4 Discussion 35
Page
2.7 Conclusions 68
3.1 Introduction 70
Page
Page
APPENDIX A 283
APPENDIX B 285
APPENDIX C 288
APPENDIX D 290
APPENDIX E 298
REFERENCES 306
Vlll
NOTATION
f value of u where e = t /2
w
f, coefficient of term in stress function $
k
f(Y) stress field function defined by Eqn.(3.21)
1 term in Eqn.(3.32)
t thickness of web
tg thickness of flange
t thickness of web
w
u distance from outer flange hinge towards patch load
rail respectively
line inclination
M bending moment
assembly
R angle
V dimensionless e : V = e /t
o o w
W work per unit length of web hinge
Z dimensionless z : Z = z/t
w
a, critical value of a,
bcr b
a critical stress under patch load
cr
yield stress of flange
a critical value of a
scr s
a maximum shear stress
sm
a ultimate shear stress
su
ultimate stress under uniform compression
LU cr
LU
cr cr Q
CD
Q
<
LU
X
é
m
cr
LU
> o
o (/)
o co
r
cr
Ü
y
LU
O)
Ll I
LU
CHAPTER 1
1.1. Introduction
This has been especially so with welded plate structures, for which the
bridge rail is positioned not centrally along the top flange but to one
side, directly over a web. This type of structure presents the designer
load which spreads through the rail and flange to produce a localised region
of higb compressive stress at the top edge of the panel. How this condition
might be allowed for in the design of such girders was the question which
is shown in Fig. 1.1. The two bridge girders, the length of which defines
the span of the crane, are mounted at their extremeties onto end-carriages.
The end-carriages contain the wheels on which the complete crane structure
runs along rails mounted on the crane gantry girders. This direction of
part of the building structure which supports it. The 'crab'* is the
name given to the trolley unit which accommodates the hoist mechanism.
It straddles the gap between the two bridge girders, through which passes
the hoist rope, and runs across the span of the crane on rails, one placed
(the least distances by which the hook approaches the walls of the building
housing the crane) which in turn describes the working area of the crane.
Control of the crane shown in Fig. 1.1 is achieved via the push-button
This enables the operator to control the crane from any position on the
floor beneath the crane span. On many larger cranes, the operator is
housed in a cabin suspended below girder level from either the crab or one
load.
Construction of overhead crane bridge girders can take various forms. This
span of crane, and duty of crane. The qualitative term 'duty' of crane is
a whole when in service and the extent to which the appliance is used to
carrying capacity, short span, and light duty, a girder constructed from
Such a girder may constitute a box-section but need not necessarily do so.
wide; in some special cases, however, a 'wide box’ is employed which has a
bottom horizontal flange plate and two vertical side web plates. Inter
spersed at regular intervals along the length of the box-girders are inter
full cross-sectional area bounded by the four peripheral plates of the box
except for a narrow horizontal strip across the bottom which arises from
the diaphragms being stopped just short of the tension flange. Unless the
girder is a 'walk-through' box with access holes cut in the diaphragms, one
outer plate of the box. Selecting the bottom edge as that to be left
Diaphragms maintain the integrity and shape of the box, thereby retaining
against shear buckling of the webs by serving to divide the webs up into
discrete panels. For a given web, the resistance to shear buckling of a
the diaphragm spacing. For cranes, diaphragm spacing is such that web
panel aspect ratios lie typically between 1.5 and 1.0, although ratios
The crab rails are mounted onto the top flange plates of the two bridge
girders at either of two locations; directly above the inner web or midway
between the two webs. It is now general practice to mount the rails above
the inner webs unless high wheel loads or other special conditions render
When the crane rails are mounted eccentrically, as shown in Figs. 1.1 and
crane structure than does its alternative. With centrally located rails,
diaphragms which are welded to the underside of the top flange and to the
two webs. The entity of the torsion-box, however, eliminates the requirement
for these additional stiffeners since each crab rail is supported along its
length by the inner web of the girder. With these girders, the full-depth
rail are subjected to vertical in-plane loadings from the crab wheels, as
shown in Fig. 1.3. Before entering the web, a wheel load is distributed
Crab
Torsion-box
In practice these panels have to withstand not only localised wheel loadings
but also global shear stresses and longitudinal in-plane bending stresses.
Bending of the bridge girders takes place both in the vertical plane and
horizontal plane, owing to inertial loadings. For high wheel loads, these
various loadings can combine to over-stress the seam weld between the web
and flange beneath the rail. It is for this reason that, currently, the
into the corner of the box where the rail is placed. This section replaces
the upper level of the web plate and the section of flange beneath the rail,
thereby removing the weld to a less highly stressed region further down
High concentrations of stress from wheel loads can arise also in crane
gantry girders when end-carriage wheel loads bear onto thin gantry webs.
When such loadings become excessive, bogies can be used instead of end-
can arise. Torsional loadings on the bridge girders are reacted by the
bogies which may cause undesirable inequalities in the bogie wheel loads
is employed. Although the webs are then free from wheel loads, the girder
diaphragms are then subjected to wheel loads, again distributed by the rail
and flange, centrally disposed on their upper edge. Any additional short-
design of EOT crane structures; their presence can determine some of the
provide reliable and accurate means for taking account of these loads.
This is especially so in the case of torsion-box web panels when not only
There now exist various British, European, and American Codes of Practice
and Standards which deal specifically with cranes (see Refs. [1] - [6])
and others on allied subjects such as bridges and the use of structural
steelwork in buildings (see Refs. [7] - [12]). Together they offer guidance
the codes, even when taken collectively, offer only very limited or
profile of the dispersed wheel load is shown schematically in Fig. 1.4 (a).
the panel are based on a rather simplified and somewhat arbitrary 'fan-
spread' postulation. This surmise involves only the height of the rail
and thickness of the flange. The wheel load is adjudged to spread linearly
and symmetrically from its region of contact with the rail, through the
rail and flange. The deemed angle of dispersal depends on the code concerned
but is generally between 30° and 45° to the horizontal. By assuming the
wheel load to act uniformly over the resulting panel length, the local
bearing stress can be evaluated and this used to check against yielding of
plate box-girders, a problem which also arises in the upper part of welded
crane gantries. For every cycle of load in a bottom, tension, flange there
are two or more cycles of local stress in the top flange produced by the
rail, and web centrelines, sideways thrusts onto the rail, uneveness of rail,
and poor fit between rail and flange. These are the factors attributed by
(a )
( b)
localised wheel loading effects are generally made with a good quality,
Both local yielding and fatigue cracking in the upper part of crane girders
rail and thicknesses of flange that are used; existing code recommendations
alone to carry the bending stresses with the webs being designed to meet
enables the second moment of area of the webs to be incorporated into the
deflection.
deep section and short-span, bending demands on the web panels around the
centre of the span generally exceed those of shear on the panels towards
demand attention and various codes, notably the German DIN 4114 [11], give
edge loading is now widely acknowledged and some research into the subject
loaded centrally by uniformly distributed patch loads (see Fig. 1.4 (b))
have been reviewed in a state of the art report by Rockey [14]. Although
wheel loads on EOT crane structures produce a very similar form of loading
(see Fig. 1.4 (a)) there is, to the author's knowledge, only one major code
web panels to buckle under wheel loadings. This is the Dutch code NEN
2019 [6]. Its coverage extends no further, however, than to make reference
in the codes on this subject and indeed the basic research information
girders. These panels are generally of a greater d/t than the webs of crane
common to all the panels. Although the overall performance of these panels
will depend upon the complex interaction of all the loadings to which they
girders, acted upon by a wheel load. Wheel loads applied at the midspan of
the panels are investigated since the load is then furthest from a vertical
a wheel loaded crane web panel requires a reliable means for assessing
the distribution of the load into a web for different rail, flange, and
of the web by a wheel load which acts through a rail and flange interface.
1.7. Terminology
the definitions listed below will be adhered to throughout the text unless
stated otherwise.
C3
c
TD
3
"
4"
c
o
00
00
a
Q1
oI
o
LJ
<
C£1
LU
LD
LO
cn
15
2) Patch load.
5) Codes.
6) Crippling.
crippling.
7) Dimensions.
CHAPTER 2
2.1. Introduction
Before the risk of web panel buckling under the action of a wheel load
load acting onto the top edge of the panel. Existing codes (see Refs.
[3] - [6] and [10]) estimate patch lengths by assuming a simple fan-
spread of the load through the rail and flange, indicating that lengths
depend only on the rail height and flange thickness. It is more likely,
however, that distribution of the load into the web will depend on the
ratio of stiffnesses of the rail and web plate, as in the elementary theory
Biot [17] has attempted a more exact solution to the problem of a beam
load and resting on an infinitely high and long foundation are in close
agreement with those produced by Parkes [IB] for a similarly loaded flange
of vertical stress between the flange and web of a beam. From the
distribution for a unit point load, the peak stress directly beneath the
17
load is given by
173Jt
where J is given by
(1+v) (3-v) 1 73
2 t
where I is the second moment of area of the .flange about its own
neutral axis and t is the web thickness. Assuming the vertical stress
a = 0.3191
m
A very similar expression has been derived from another long and
maximum load per unit length of web of a beam point loaded on the flange
P
^
= 2P
373 [ ]2E'I
Where E is Young's modulus for the web material and E ' that for the
flange. Converting this to a direct stress and assuming the web and
a = 0.3055
m
f [ 1 ]'
18
The analyses by both Parkes and Girkmann show that for a beam loaded
on the flange by a point load acting in the plane of the web, the
moment of area of the flange and the web thickness. This would appear
to offer a far more rational basis for estimating patch loading lengths
on crane web panels than that currently adopted by the design codes.
types and sizes of rail, are used in crane girders. Three basic rail
types are employed: the flat bottom rail, the bridge rail, and the
square bar rail. These are shown in Fig. 2.1. Two rails of different
section but similar height will therefore have different second moments
provided the capacity of the crane is not high or the duty severe. This
standardised crane products. For high wheel loads, local bearing stress
Bridge rail
from local buckling considerations may make it possible to fit only the
flat bottom type since these have a narrower base than comparable
bridge rails.
The dispersal of a wheel load onto a crane web is a rather more complicated
the load, there are in fact two: the rail and the flange. The extent to
Senior [20] and Senior and Gurney [13] have discussed the problems
both are intended to keep tra c k maintenance costs down. The 'fully hard'
underlay pad. In the fully hard method, the rail is attached in a manner
unit, thereby avoiding fretting wear. The fully soft approach, on the
way that the fastenings do not loosen and fatigue. Square bar rails are
clips are used unless the duty of the crane is arduous when soft spring
true patch loads on crane web panels, in order that buckling load
Some studies relating to patch loading on crane web panels have been
crane code (DIN 120), now superseded by DIN 15018 [5], are presented.
the other the spread is wider at 30°. Three theoretical curves are
elastically by the web plate, the other two are based on a Fourier
analysis but assume two different initial contact lengths, one 50 mm,
the other 230 mm. These distributions are compared with a curve
unclear what contact length was assumed between wheel and rail in the
analysis but it appears this may have been 230 mm which is surely very
unrealistic.
individual cranes and in both cases the type of rail is not specified,
has already been discussed and this throws into question the experimental
method. Concern has also already been expressed over the size of the
relatively inaccessible German work dating back to the 1920's and 193G's
present study.
Steinhardt and Schulz [22] have studied the effect on wheel load
severity of stress which can develop in the girder beneath the rail,
a flat bottom rail, the other to a square bar rail. These curves show
curves are presented, calculated using the same theory, which illustrate
CT = 0.318 P
t
obtained from the theory by Parkes [18], for determining the maximum
stress in the web when no rail underlay pad is present. Steinhardt and
Schulz point out that this expression can also be derived from the results
expression the second moment of area term is given as the algebraic sum
of that for the flange, I^, and that for the rail, I^. For cases where
rail cases are considered; a flat bottom rail (second moment of area =
24
888 cm^) and a square bar rail (second moment of area = 32 cm^). The
rails were loaded both with and without the rubber underlay present.
although the results for the flat bottom rail tended to be poorer than
those for the square bar rail. It is presumed that, since the rails
that only the central 300 mm of the 500 mm wide flange plate was effective
in transmitting the wheel load in the case of the small square bar rail,
but that the full width was effective in the case of the flat bottom
rail. They add, however, that incorporating this into the second
The main conclusion from the work by Steinhardt and Schulz is that by
using an elastic underlay pad, maximum local web stresses can be reduced
uniform patch loads suitable for buckling calculations on crane web panels.
Mendel [24] has developed the work of Girkmann [19] and Rieve [23] to
resemble that derived by Parkes [18] for the bending stress in the outermost
may be expressed as
= 0.4644 Py^ ‘/ 3
25
where is the distance from the neutral axis of the flange to the
rail is
T" 1.15 I. + I
f r
Ob = 0.485 Py^ V:
J.15 I^ + Ip
where y is the distance from the rail neutral axis to its outermost
r
fibre. These two equations apply to the case of a rail that is free
above except that, in each case, the inertia term (1.15 I^ + I^) is
for the factor 1.15 lies in the work of Rieve [23] as being an allowance
for the fact that the rail and flange are connected to the web. It
Senior and Gurney [13] have illustrated that the severity of local
stresses in the web, depend on wheel load alignment and the conditions
of fit-up between the rail and flange. In practice, crane rails are
▼
frequently rolled with a transverse camber on the underside. When
26
sometimes bearing onto the girder flange along lines towards the
edges of the rail base, when transverse bending of the flange about
the web takes place, and sometimes bearing onto the flange directly
above the web. Transverse flange bending has the effect of distributing
the wheel load over a greater web length than would otherwise result,
by Parkes [18] and Biot [17], for calculating local stresses in the
flange and web resulting from extremes of rail and flange fit. Their
Sedlmayer [26] have also presented studies of the effects of rail fit
needs to take due account of different crane rails and web thicknesses;
the classical theory by Parkes [18] may provide a sound basis for such
a method.
27
The model box section is shown in Fig. 2.2. It was fabricated from
mild steel plate using small fillet continuous welds. The central
(test) web panel, which is bounded on either side by the two inner
above the test web. The test panel experiences only symmetric shear
along its two vertical edges. Owing to the short span of the section,
O
O
Csl
ml
W
LU
ÜO
O û:
LU
Ln Q
Ûd
13
I
X
o
o CÛ
o
m
LU
Q
O
z:
u
<
LJ
m 00
I
00
CL
E
(N I
(N I
C7I
C—
JZ
CL
O
8
m
o
o
m
29
inside and outside faces of the web panel in order to monitor in-plane
rear web panel when the box was fabricated to enable gauges to be
mounted on the inner face of the test panel. The locations of the
face of the panel had a mirror image on the other. One gauge of each
direction. The upper line of rosettes was positioned close to the top
flange but clearing any weld irregularities. The two gauged positions
test their load distributing effects when mounted on the box. The three
rails are shown in Fig. 2.4 and their geometrical properties listed in
the accompanying Table 2.1. The two component parts of Rail 2 were
axes but different heights, while Rails 2 and 3 have the same height
be investigated.
Bolting was chosen as the method for attaching the rails to the box.
This models the hard clipping technique used in practice and readily
m Ln
g
CÛ
O
TD un
OJ LU
u
L.
u
o LO
o L.
(U
-Q O
E O
3 z
o C
C L3
O Z)
un <
LO O
O
CL
o- en <
c ÛC
l—
o o
00
un, m u
ou
O q: m
un. (Ni
en
31
CO
g
1 .
CD 'J-
ro CD en
! l l
C M- o ro CD
8
ui (b CO
m 10
z
I—
cr
o
ÜJ I—
û_ o
o LU
cc 00
CL cr>
°.-g
Z 2 E m en CM
CT-p E
(b t
X u
UJ LU
Q Q
O O
csi
en
If)
O (N
c
r O (N
<r MT
.0
00 ID LO
ro CM CM Lui
X
en
<JD CM ro
ô
5 cr S
S
32
along the entire length of the box section. This required small feet
given by
a = E (e + ve )
yy — yy %x
1-v*
The vertical and longitudinal strains, and are the average values
of the strains from corresponding pairs of gauges on the inner and outer
and then averaging, the facility to sum (or difference) pairs of strain
With the box positioned in the test frame beneath the ram as shown in
Fig. 2.2, a load was applied and the data logger activated. It recorded
the 24 outputs from respective gauge pairs together with the applied load
from an electric load cell positioned between the ram and box in 2.5 seconds,
Data were thus quickly compiled for a range of loads, both ascending and
descending.
33
Each of the three rails was tested in turn when bolted directly to
each case to avoid local yielding of the web. The procedure was then
repeated for each rail but with a 3 mm thick neoprene rubber underlay
incorporated between the rail and flange, the assembly again being
bolted together. For comparative purposes, the box when loaded directly
onto the flange, in the absence of any rail or underlay, was also
2.3.3. Results
stress at each of the upper gauged positions from the experimental data.
The program provided a graph plotting option which presented the results
in the form shown in Fig. 2.5 where the stress at each monitored position
is plotted against applied load. The stress for unit applied load is
Plotting these values with respect to position along the web panel gives
for unit applied load. Bending stresses in the panel were estimated
180
160
— 120
00
00 :
LÜ
cr
m 100 -
LOAD (kN)
Fig. 2.5. WEB IN-PLANE VERTICAL STRESS RECORDINGS
AGAINST APPLIED LOAD
35
to the box are shown in Fig. 2.6. The similarity between the plots
is apparent: they are symmetric about the line of action of the applied
which occur beneath the load, and, in each case, the distributions
The stress distributions for the three rails when mounted on the
broader and had a less pronounced peak than those of Fig. 2.6. In
plot with a pronounced peak. Fig. 2.7 compares typical stress distributions
in the box web arising from the three loading interfaces investigated:
loaded directly onto flange, loaded through rail (Rail 2), and loaded
along the length of the web panel at a depth of 15 mm below the underside
of the top flange. They correspond to an applied load of 1 kN; the units
2
of stress are N/mm .
2.3.4. Discussion
Generally there was good linearity between the vertical in-plane compressive
by the experimental points shown in Fig. 2.5 for recording position (J) .
In the tests where the rubber underlay was incorporated, however, there
increasing load. At some of the recording positions stress levels increased more
1kN
DISTANCE FROM MIDSPAN OF WEB PANEL (mm
100 50 50 100 150
RAIL 1
RAIL 2
RAIL 3
3 -0
VERTICAL STRESS
I N/ mmZ I
F i g . 2 .6 . STRESS D IS T R IB U T IO N S IN W EB PANEL
CO RRESPO NDING TO THE THREE RAIL
LOADING IN TERFACES
1kN
20
Loaded through
roil and underlay
Loaded through
rail
VERTICAL STRESS
( N/mmD
of stress along the panel as the load increased. Weitsman [27], and
Civelek and Erdogan [28], have reported on the tendency for a member
the rail bolts opposing this would account for the non-linear development
linearity were not excessive and developed at higher loads only. Linear
measures were thus taken as for the cases when the underlay was not
present and the resulting stress distributions for a rail and underlay
When the load was applied directly onto the flange, the linearised
a peak stress of 6.4 N/mm , as shown in Fig. 2.7. This is of the order
the plate is assumed uniform. For a load per unit thickness of P', the
a = 2P' cosR
; —
39
load and the line of length k between the point of load application and
the element. Now for the model test box, the vertical distance, including
and because the applied wheel load only approximates to a line force.
Fig. 2.7 shows that the presence of a rail (Rail 2) led to a reduction
in the recorded peak stress of 43% compared with the value for loading
underlay caused a further reduction of 33% compared with the rail only
and Schulz [22] whose web stress distribution profiles bear a close
web enables the vertical reaction force offered by the web to be determined,
For each rail case, this force is found to be 1.0 kN (- 4%), illustrating
that all the applied load is transmitted into the underlying web panel.
1.0 kN (- 8%).
40
The stress distributions of Figs. 2.6 and 2.7 illustrate the nature
of the patch loads that bear onto the upper edge of crane web panels.
the distance between the points where the stress drops to half its
maximum value. For the same total force to act, however, this does not
effective patch length is that length over which the peak stress would
have to act in order to balance the total applied force. This can be
expressed simply as
c =
a^t (2.1)
the top edge of the web. Thus, at a depth s below the rail surface, the
British crane codes BS 466 [ 3 ] and BS 3579 [4] specify a similar uniform
spread but with zero initial contact length and dispersal at an angle
50
c = 2s + 5 0
(DIN 15018)
c = 3s
( BS Z.66 and B S 3 57 9)
which gives a patch length of three times the depth, as shown in Fig. 2,
3s (2.3)
Patch lengths determined by Eqs. (2.2) and (2.3) depend, for a given
flange, only on the rail height. These expressions are typical of those
which appear in codes not only in Britain and Germany but also in
In 1971, a draft was released of the German code DIN 4132 [30] which
contains two expressions for the maximum direct stress in the web of
a crane gantry girder resulting from a wheel load. The expressions are
G.3_P Vs
t 1.151^ + I
f r
V :
1.151
fr
for a fixed rail. P is the greatest wheel load, including the appropriate
If these expressions for peak stress are substituted into Eg. (2.1), the
3.33 1.151^ 4. I^ V ,
(2.4 a)
43
and
c = 3.33 (2.4 b)
for determining the length over which a wheel load is distributed at the
web is given as
If + I, (2.5)
where A is a constant which takes the value of 3.25 when the railis
mounted directly onto the flange (there is, however, no mention of the
method by which the rail is assumed to be fixed) and 4.0 when a suitable
resilient pad is placed between them. This expression and those in the
draft DIN 4132 appear to originate from the works of Parkes [18],
Eg. (2.1) has been used to evaluate the effective lengths of uniform
patch loads on the test web panel equivalent to the experimental stress
distributions of Figs. 2.6 and 2.7. These are compared with the estimates
In applying the DIN 15018 method given in Eg. (2.2) to the model box and
rails it has been assumed that the 50 mm contact length between wheel and
c- O
CT3 00 ON ON
Pi
Q
•H
CO
oc
NO
CM •a CSJ O CS)
r\ 0
X
•H
U.
•H
u-
CD csj
co C ON UN
Pi •H
CO ON
o ■o
•H
I— I
CO
CO
CD Q.
On
0
Q
r* <3" NO NO
UN
0 MN
zc MN C7>
CJ D
CO
< CO 5
Q, Ü
sz
cn
c NO
C_) 0 NO ON
LÜ <f CM ON (ON
U. JO
CO 0
o CD
0
O Q_
z
o
Q-
I—
I—
!
a MN
< 0 MN o o
00 cn
3
O CO
CO UN CO
ÜJ
CM
r<N
CM O
JD
0 0
r— 1
JD
0
Q. CM UN '■O
X UN CD as ON MN
1— 1 10
0 0 0
4J w O
c 0
0 Pi CM O NO CM NO
E J-) z e CO NO UN
•H CO DA e
Pi NO KN CM
0 Di: z
a 0
X 0 Pi
w Q_ O
u_
TO
0
o_
+
KN CM
0
cc:
•H •H •H
0 0 0 0
cc QC QC QC
45
be applied directly to the model box situation. The draft codes from
which Eqs. (2.4) and (2.5) are obtained both consider crane gantries
A section symmetric about the loaded web has been assumed here (given by
twice the distance from the web midplane to the nearer flange edge),
always very much thinner than the rail is deep, the effect of the
assumed width on the total second moment of area term is only small
The experimental results in Table 2.2 show the relatively small variation
which is the length recorded for Rail 2. These patch lengths correspond
and 0.16 (the average value being 0.15) for the test web panel which has
a length of 600 mm. A much wider range exists between the cases for
The results from the DIN 15018 method of Eq. (2.2) reveal that it rather
lacks versatility to cater for anything other than standard rail sections
for the tall Rail 1 while underestimating for the other two rails, and
contact between curved and flat surfaces are given by Roark and
L = 2.15 P D'
c
rE
Thus the contact length is seen to depend on the wheel load P, the
wheel diamter D', the width of the rail surface r, and Young's modulus
for the materials E. For a crane with a 50 mm square bar rail, typical
2
values might be P = 150 kN, D ' = 700 mm, r = 50 mm, and E = 207 kN/mm ,
The results from the BS 466 and BS 3579 approach given by Eq. (2.3) follow
a similar trend to, but in all cases are greater than, those from the
thus achieved in all cases except for the predictions at the web weld
depth for Rails 2 and 3. Worst agreement occurs for Rail 1 with over
estimations for the patch length of 48% at the web edge and 100% at the
depth of the gauges. Schindler [21] has commented on the uniform fan-spread
Predictions from Eq. (2.4 a), derived from the draft DIN 4132 expression
for maximum web stress for a sliding rail case, and Eq. (2.5), the draft
good agreement with the experimental results for the three rails; the
predicted values are all within 15%. The draft BS specification for
the patch length but the code does specify "a suitable resilient pad not
less than 5 mm thick", which is a typical thickness for crane rail pads
(see Ref. [33j); the neoprene rubber underlay used in the tests, being
a fixed rail case, given by Eq. (2.4 b), overestimates for all the rails.
rail and flange assembly which is likely to be achieved only when the
directly mounted rails are best predicted by Eqs. (2.4 a) and (2.5).
It may well be, however, that for rails that are welded instead of bolted
to the flange then Eq. (2.4 b) would best predict the effective patch
note that the full DIN 4132 [10], published subsequent to the draft
can be taken of the various types and stiffnesses of crane rail pads
available (see, for example. Ref. [33]) by simply assigning a single value
on the basis of the present findings and those of Steinhardt and Schulz [22]
48
On a more general note, Parkes [18] has shown that although his analysis
Eqs. (2.4) and (2.5) can be considered to hold generally for crane webs
than the panel length - as the present test results indicate. This
length which, for a given crane box, suggests the patch length is not
changed then the patch length parameter changes also, the patch length
remaining constant. For example, had the aspect ratio of the model box
49
web panel been 1.0 instead of 1.5, the estimated patch lengths would
remain unaltered but the average value of the patch length parameter
for the three rail cases would then change from 0.15 to 0.23. Now
the British crane code BS 2573/1 [2] limits aspect ratios of web
panels with directly mounted rails lie typically between 0.15 and 0.23
times the panel length. Experience of rail and web combinations used in
practice, however, suggests that (using Eq. (2.5)) patch length parameters
as low as 0.07 may arise for a small rail supported by a thick web
panel of aspect ratio 1.5, but that they are unlikely to exceed 0.25.
given by
P = K n=D (2.6)
cr — T-
the value of which depends on the plate aspect ratio a = b/d, the patch
50
The relationship between the coefficients cited in Eqs. (2.7) and (2.8)
is given by
K = &K'
plotting over wide ranges of panel aspect ratios and patch length
buckling. Khan, Johns, and Hayman [34] have calculated values for this
Rockey and Bagchi [35], and Protte [36] have presented solutions for the
loaded plates with various edge conditions are presented which provide
the required information. In fig. 3.2, curves are given for the
is shown there that narrowing the patch load relative to the panel
cause buckling, but that the proportional reduction decreases the narrower
the patch load; changing the patch length parameter from 1 to 3/4 produces
revealed that the trend continues: coefficients for c/b = 1/16 were
lower than those for c/b = 1/8 by only 1% when b/d = 1, 2% when b/d
= 2, and about 5% when b/d = 4. The result for c/b = 1/33, and b/d = 1
This indicates that in the range 1.0 < a ^ 1.5 applicable to crane web
cause elastic buckling if B is reduced below about 1/4. Over this range,
Results from the preceding Section 2.4 indicated that crane girder
with the formulae of Eqs. (2.9) and (2.10); the second formula may
but the two equations give K values which differ by only 4% to 5%.
For the experimental web panel which measures 600 mm long, 400 mm deep,
and 3 mm thick, a typical effective patch length was 90 mm. This gives
a = 1.5, B = 0.15 and thus K = 2.6 approximately from Fig.3.2 or Eq. (2.10)
2
Therefore, taking E z 207 kN/mm and v z 0.3, Eq. (2.6) gives for the
2
From Eq. (2.7), the average stress when buckling occurs is z 122 N/mm .
2
A typical yield stress for the web is 250 N/mm and so, on this basis,
It is worth noting that had the panel been subjected to a narrower patch
load of say c/b z 0.04 (z1/25) instead of 0.15 then K would change
only to about 2.5 and the buckling load drop very slightly to 32 kN but
the corresponding average bearing stress on the web leaps to 444 N/mm -
well above the yield stress. Elastic buckling cannot then take place.
presence of the rail and flange which may have an influence on the
Rockey and Bagchi [35], in a paper mainly concerned with finite element
the panel would be expected to increase due to the flange causing some
rotation of the top edge of the web, which was previously considered
simply supported.
incorporate flange flexural and torsional properties for one case only -
that of a square web panel with a patch load of c/b = 0.2 on the overlying
The buckling coefficient is that defined by Eq. (2.8) but since the
from Fig. 3.2 or Eq. (2.9) to be approximately 3.4. Thus the Rockey
pass through this point also there must be an extremely abrupt change
usually not more than twice the web thickness and is often rather less
than that. In the absence of further data, particularly for t^/t^ < 2,
the test box web panel with rail and flange effects included, further
load, P, at a position 160 mm below the underside of the top flange and
two inner diaphragm positions. This is shown in Fig. 2.10 which gives
is shown in Fig. 2.12. When the results are arranged as the Southwell
plot in Fig. 2.13 the buckling load is readily estimated from the
direct loading of the box (no rail present) gave the growth of deflections
shown in Fig. 2.14 and the Southwell plot of Fig. 2.15. There is clearly
Southwell plots were also produced for the cases when Rails 1 and 3
were bolted to the box. Although the linearity of these plots was
not quite as good as that for Rail 2, inverse slope readings indicated
The Southwell method offers a very simple procedure for measuring buckling
with initial geometric imperfections often means that for the fundamental
of deflection well beyond the small deflection range. Roorda [37] has
ft
F ig . 2.1 0. G ENERAL V IE W OF E X P E R IM E N TA L E Q U IP M E N T
USED FOR ELASTIC IN V E S T IG A TIO N S
58
a
a
CNJ
CN
(N
O
CN
Z)
o
LO
CN E
00
CD
CD
LO
Ô
”o
CN
CN
CD
cn
LLI
CD o CD o o o
m cn (N
Go
CM
o
CN
CO
LU
LU
z:
CN
LL
o
00
CD
LU
CÛ
CD
O
LD
ro
O cn
(N
CD
O CD CD CD O O
LD PO CN
o
LU
LU
LU
O
LU
LU
LU
O
O
O
o
LU
_J
LL
CN
O
I/O
CO
•Q
o
-P
o
(N
CJ>
O 00 (N o
o
o
o
o
LU
LO
R
Ô
LO
LL
Ô
o
o
o
LO LO
O
Ô LO
LO o LO O
Ô
63
and Walker [38], in a paper which looks at various situations where the
Southwell plot data for plates and provide better estimates of buckling
loads. They classify cases where the Southwell plot fails to give a
effects can extend to data points for loads as high as 80% of the
critical load and that readings for still higher loads then have to
graphical and one numerical, have been taken from the work by Spencer
and Walker [38] and applied to the data for the model box panel. Each
imperfections. Only slight improvement to the plot of Fig. 2.15 for direct
04-
loading of the box was achieved by the graphical methods but this
pivot points. Moreover, when applied to the data of Fig. 2.13, the
of Fig. 2.15, may well be due to the very limited range of load (0-15 kN)
which was applied to the box in order to prevent yielding in the loaded
load is approached, which was not the case for the direct loading
test, and the modifications were designed more for data extending beyond
which were less linear than the original Southwell plots for the
with-rail cases, however, must throw into question their use in this
context.
For both the Southwell and the alternative 'Spencer' plot techniques,
stresses into the model box during fabrication by keeping the welds
initial imperfections.
b3
Nonetheless, good linear Southwell plots were obtained for the test
web panel when loaded through the test rails, particularly Rail 2.
The estimated buckling load for this case of 85 kN is more than double
the theoretical value calculated for the panel in the previous section.
results indicate buckling loads for the box panel when loaded through
a rail which are roughly 2 to 2.5 times greater than the value predicted
for the panel when assumed simply supported and subjected to a uniform
of the rail. However, it could be argued here that it is only once the
panel has buckled that any additional in-plane restraint can be provided
by the rail because it is not until then that rail flexure develops.
earlier illustrated that all the applied load was reacted by the web,
indicating the rail simply transmits (rather than carries) the load.
66
along the horizontal panel edges provided by the box flanges. As far
crane web panel situation; it was to rectify this situation that the
These were mounted on a special strut attached to the box flanges; the
the top and bottom flanges. Any rigid body movement of the strut relative
to the web plane caused by rotation of the top flange was monitored using
an eleventh transducer and the outputs from the remaining ten transducers
corrected accordingly.
Rail 2 mounted on the box are shown in Fig. 2.16; similar profiles were
obtained with Rails 1 and 3. Each of the profiles has a crest close to the
one-third depth position and the plots clearly indicate that the top and
under a partial edge loading calculated by Bagchi and Rockey [41] using
20 -
60-
90-
120-
P = 1 6 -7 0 kN
140
160
180-
E 200-
220 -
LU
2: 260-
280-
^ 300-
320
340 -
360
380
2.7. Conclusions
patch loads on their upper edge arising from the dispersal of a wheel
girder web panel have been presented which reveal the nature of these
loads and a simple expression has been used to equate them to equivalent
The results indicate that crane web panels are subjected to patch loads
distribution and, therefore, direct stress in the web are the second
moments of area of the rail and flange and the web thickness; however,
beneath a crane rail are the fit between rail and flange, and rail
For all the loading interfaces considered here, the vertical stress
distributions recorded in the web illustrate that all the applied load
method for the test panel when loaded through various rails by a wheel
theory for the panel when subjected to an equivalent uniform patch load
top and bottom edges where interaction with the flanges occurs.
these edges.
CHAPTER 5
3.1. Introduction
well constitute a safe design basis for plates with elastic moment edge
supported plates, mounted their test plates in slotted rollers that rota
edges of a web owing to interaction between the web and flange. The
the upper limit to coefficient values and when compared with coefficients
also presented here for simply supported plates gives the range of values
free edges). To assess along which edge or edges the provision of rota
have been computed for plates as successive plate edges are clamped, the
Coefficients for these loads are shown to compare very favourably with
those computed for the equivalent uniform loads, calculated by the method
The computing method used in the present analysis originates from the
but one in which the stress distribution throughout the plate need
Khan and Walker [43] developed this method and, with the additional
buckled stress distribution had, for the most part, discouraged any para
ments .
Balabukh led to this basic method later being further developed by Khan,
Johns, and Hayman [34]. They produced a refined version which, although
undoubtedly more complex than the earlier format by Khan and Walker,
obtained by other workers. The revised method, which this time assumed
The present author has developed the computer analysis method beyond the
parametric studies of patch loaded plates with some or all edges clamped.
that adopted by Khan, Johns and Hayman [34] has been used in the present
stresses on the vertical sides of the plate resist the applied load.
It should be noted that the x-direction is now along the central vertical
axis of the plate and the y-direction along the central horizontal axis.
From the work of Alfutov and Balabukh [44], the potential energy of a
written as
U = (3.1)
2C
2B
2L
here - then
.2
(B) dxdy (3.2)
9x^ 3y^/
The terms U''"'' and (2 ) are integrals which together represent the
sum of the membrane strain energy and the potential energy of the
external stresses.
^ . .2
9w * 9w
U (1) _
+ B dxdy (3.3)
2 t°x laxj
t * * 0
U (2) _ — (o + a )V^$dxdy (3.4)
E ^X y^ ^
Each of the three integrations is taken over the entire area of the plate
* * *
The values , and are components of a statically admissible
stress field, that is, they must satisfy the equilibrium equations
* *
da 9a
(3.5a)
* *
9q 9q
(3.5b)
This stress field need not, however, correspond to strains that satisfy
This stress function must also satisfy the homogeneous boundary conditions
finite series
n
w = I a^w^(x,y) (3.8)
i=l
Eqn.(3.7).
"ij = (3-10)
be singular is investigated.
U. . = U. + u. + U. (3.11)
1] 1] 1] ij
77
where the three summed terms are simply the second derivatives of
U. + U. = QU. (3.12)
13 13 ^ 13
(Q )
where U.. is a matrix not dependent upon the applied loads. The
1]
buckling load calculation thus reduces to the linear eigenvalue type
of problem
3w. 9 w. 9w. 9 w.
U. (
)1 ) -
1 1
13
f9w. 9w. 9w^ 3w,
_1
+ a dxdy (3.14)
xy 9x 9y 9y 9x,
(2 ']
The third term in Eqn. (3.11), , presents more difficulty to
m
(3.15)
"if = 1 'ijk "k
where
F = ^^^k (3.16)
ijk 9a.9a.
1 3
and
(3.17)
"k ■ ■
m
, I = 1 ,2 ,...m ; i,j=l,2 ,...n (3.18)
^ijk \ £ ^ij£
k=
where (3.19)
and not on its loading. This feature can be used to advantage when
loadings to be considered.
79
The method described above has been used to calculate buckling coeffic
Stress Function 0
The components of the stress function 0 have been split into two
where
*kX " ^ \ y =
coshA X cos X X
X (X) = -- ^ ------
r coshX cosX
r r
th
r+ 1
where X is the root of
r
tanhX + tan X = 0
r r
sinhX X sinX X
X (X) = - . , / ----- r - ^
r sinhX sinX
f ^th
r
where X is the root of
r
tanhX - tanX = 0
r r
80
therefore given by
coshA Y cosA Y
Y fY'i = E _ E_
s coshA cosX
tanhX + tanX = 0
s s
*
Stress Field a
+
The non-dimensional form of the stress field, a , is given by
+ * tL
. ° ° 2P
defined by
-^fCY) = ax C3.21]
where is the stress applied at the top edge of the plate and is
Y
h(Y) = f g(Y)dY (3.23)
81
where N = - —Lr k
16 C
(a ) = 2 + 3X - X'
^ X
a "*” = N (o (a
xy xy^ X xy Y
where N = JL k L
xy 16 C B
where
(o..) = X
that f(Y) is not defined for |Y | > C/L . For a uniform patch
f(Y) = 1
Displacements w
stress function components, been split into two parts. These are given
by
Wix(X) = sin p Y (X + 1 )
w^y(Y) = sin q y (Y + 1)
This corresponds to a clamped top edge with the bottom edge and two
tan g = g
This corresponds to clamped top and bottom edges with the sides remaining
simply supported.
- (P+1)
th
and for even-numbered terms of w^^ , g is the root of
tan g = g
The w^^ given in (3) are retained and the w^^ are similarly
assigned
simply supported edges could also have been incorporated but the four
In the previous work by Khan, Johns and Hayman [34] and by Khan and
simplicity of the functions which can be assumed for the simply supported
edge condition.
routine was incorporated into the program to compute all the required
work. The integration method adopted then, however, was less efficient
than the new approach and so appreciable savings were made in computer
tion procedures in the work of Ref.[34] were rewritten with only the
emphasis on versatility of the end product so that any of the four support
tions and their first and second derivatives, the stress functions and
their second and fourth derivatives, and the stress field expressions.
into their X-parts and Y-parts that arise from the differentiations in
Eqns.(3.13), (3.14], (3.17], (3.19] and (3.20] so that the numerical inte
The program was written in Fortran IV and run on the Leicester University
GDC Cyber 73 computer. The compilation time for the final program was
approximately 15 seconds. Since the main results were all run from a
Verification that the new refined calculation procedure had been success
and load geometry calculations performed by Khan, Johns and Hayman [34].
For various cases encompassing the whole range of L/B and C/L para
obtained by the two methods differed in only the fifth or sixth significant
86
figure when 32 strips were used in the new numerical integration program
and in only the sixth or seventh significant figure when 64 strips were
to within a very small fraction of one per cent of the values obtained
strips. The integrations performed along the length of the plate which
the central loaded region and the two unloaded regions. The new program
was therefore developed such that, whatever the nominal total number of
and unloaded regions of the plate pro rata, according to the C/L ratio.
This was an improvement on the previous method where the total number of
strips used in the numerical integrations depended on the C/L value, more
strips being required the smaller the ratio. For many geometries, more
tion terms reduces the value of the solution. However, when a non-exact
quently, as additional stress function terms are employed (which have the
the upper bound feature tends to be restored and generally the solution
87
1 X 3 3x3 11.902
2x3 3x3 6.862
3x3 3x3 6.592
4x3 3x3 6.558
5x3 3x3 6.537
6x3 3x3 6.521
7x3 3x3 6.510
3x5 3x3 6.588
3x7 3x3 6.585
5x5 3x3 6.523
5x5 1 X 1 5.859
5x5 2x2 6.049
5x5 4x4 6.514
5x5 5x5 6.518
5x5 6x6 6.520
!
increases. Convergence of the analysis thus depends on the number of
to achieve this accuracy depended on the geometry of the plate and on the
edge conditions. Generally it was found that the same number of terms
greater number of terms being required along the larger plate dimension.
Fewest terms were therefore required for the near square plates. For
plates simply supported all round it appeared that no more than seven
desired accuracy.
The effect of clamping the top edge of a plate was to demand additional
terms. For a square plate, series consisting of five terms in both the
For a tall, narrow plate with an L/B of V3 > nine terms were used in
45 X 45. For L/B values of 3.5 and above, five terms were employed
matrices which were the largest created in the study. Compared with the
edges demanded increased numbers of terms for L/B values in the range
0.25 , 0.5 , 0.75 and 1.0 are considered. The buckling coefficient
K is defined by
tions to the plate have been investigated and it was found that for each
C/L value the coefficients lay on a smooth curve. These curves are
shown in Figs.3.2 to 3.5 where for clarity the data points are not
where the numbers of displacement and stress function terms employed are
also shown.
Buckling coefficients for plates simply supported on all edges are shown
for each of eleven L/B values. These are listed in Table 3.2.
ween coefficient values and the plate aspect ratio. Consequently, the
number of L/B values for which coefficients were evaluated was increased,
from eleven to sixteen. Coefficients for plates clamped along the top
edge but simply supported on the other three edges are shown in Fig.3.3;
the computed values are given in Table 3.3. The case when the top and
bottom edges are clamped but the sides simply supported is shown in
.90
(NJ Csl
U
O
o
CO
ÜJ 9Z
§1
Li-
O
u in
IJL
o
o
CD <N| O 00 CSI o
iN31DIJd30D ONIIViOna
91
o
(NI (NI
O
O
&)
QZ
g
CN
§ ô o
o
CD
LU
gi
U
o
LD
LU
ÜJ
y
ü_
o
o
u J—
o
LU
Q
O
LD
(D
ro
cô
co o CD O
iN310IJd303 ONlIMOna
o
CM
CN
<N
O
ro
93
o g
LU
H- co
CO
8 if) p
CM <
I
CNI
U o o
LU O
O
LD
h-
O
LU
LO O
LO
Ô
ro
lll
CD O CD O
lN3IOIdd30D ONIIMOng
'j'5
CM
CN
8
ro
CN
LÜ
II
O If)
I'
Ô
If)
CN
CN
o
y
o
<
cr
u
LÜ
CL
CO
<
y o
s
Li.
If)
LÜ
CJ
O
O
O
CÛ
LO
00
lO OO LO O 00 LO CN
(N CN
M lN31Didd303 ONIIdOnS
Table 3.2 BUCKLING COEFFICIENTS FOR PLATE SIMPLY 24.
SUPPORTED ALL ROUND
Buckling Coefficient K
C/L
Ct r L / No. Terms
/b
x.y 1.000 0.750 0.5000 0.250 0.125
Table 3.3 BUCKLING COEFFICIENTS FOR PLATE CLAMPED ALONG TOP EDGE
Buckling Coefficient K
Buckling Coefficient K
C/L
No TArms
>C. y 1.000 0.750 0.500 0.250 0.125
Buckling Coefficient K
^ /l
a = Ly No. Terms
^B X. y 1.000 0.750 0,500 0.250 0.125
Fig.3.4 and Table 3.4. Coefficients for plates clamped along all four
plate simply supported all round when using four terms in the x-direction
and three in the y-direction to 520 seconds for a plate clamped all round
Several noteworthy points arise from the result of the present analysis.
It can be seen that for all support conditions, shortening the length
over which the patch load acts reduces the K value. Concentrating
the patch load therefore lowers the level of loading required to cause
Over the range of aspect ratios considered here, the K values for a
given C/L attain their maximum value for a tall, narrow, plate with an
L/B of V3 . For larger aspect ratios, the coefficients first fall steeply
and then, for all but the narrowest patch length, begin to rise again at
an L/B of between 1.0 and 2.0 . For cases where at least the top edge
of the plate is clamped, the curves then reach a local maximum at an L/B
the higher mode solutions obtained by the present method were, however,
required the use of greater numbers of terms to accommodate the more com
plex mode shapes; however, as several terms were already being used in this
region this was not possible because of the inherent increase in demand on
computer requirements.
the length of the plate was increased. For aspect ratios below that
at which the local maximum occurred, the solutions arose from entirely
symmetric mode buckling along the plate length and adding an antisymmetric
Because the determinant search method used to find the eigenvalues operated
necessary at these large aspect ratios to ensure that the search was run
used in this region, ensured that the lowest solution was always detected
eigenvalue solutions for plates with at least their top edge clamped can
a, = . (3.25)
^ SB^t
Now the elastic critical stress for a plate under pure bending is
obtained
4 ^b tt^D
'’c r " 3 ~ “ 2B • (3.28)
round, clamped along the bottom (tension) edge and simply supported
along the other three edges, and clamped along top and bottom edges and
simply supported along the sides. Minimum values for the three
of 39.6 for a plate clamped along the top and bottom edges is substituted
is given by
L r = 13.2 ^
This describes the full edge load on a plate that would cause buckling
for a plate of aspect ratio 4.0 , clamped along top and bottom edges,
and subjected to a uniform patch load of C/L = 1.0 can be seen from
for a patch loaded plate of L/B = 4.0 and C/L = 1.0 varies only between
13.2 and 13.3 depending on whether the top edge only is clamped or all
four edges are clamped. This illustrates that for patch loaded plates
where at least the top edge is clamped, the primary cause of buckling as
bending stresses.
Timoshenko and Gere also show that for a plate subjected to bending, the
wavelength of the buckle changes from an L/B = 0.67 when all edges are
simply supported to an L/B = 0.47 when the top and bottom edges are
clamped, although very little change occurs when only the bottom (tension)
edge is clamped. Thus, for a long plate with a clamped top (compression)
crests: for a plate of aspect ratio 4.0 with its upper edge clamped,
also explains the requirement for large numbers of terms in the longi
long plates with a clamped top edge. The observed convergence of solutions
solutions.
1üü
Although not presented here, a brief study of the coefficient curves for
simply supported patch loaded plates with aspect ratios greater than
4.0 revealed that they too begin to level off towards local maxima in a
similar manner to the curves for plates with a clamped top edge. This
It will be recalled from the previous chapter that crane web panels are
at larger aspect ratios, then the aspect ratio at which bending emerges
a plate acting as a beam simply supported at its two bottom corners but,
is now given by
M = ^
tt^D
cr ^b 3L
ir^D
Per = -ST -21- (3.30)
a = j . (3.31)
For plates with at least their top edge clamped, has a value very
close to 39.6 (see Ref.[46]); the K value, from Figs. 3.3 to 3.5, is
shown in Fig.3.6 for the case of a very narrow patch load (C/L = 0.125)
and in Fig.3.7 for a full patch load (C/L = 1.0). It can be seen that
appreciably compared with the values for a simply supported plate. Addit
ionally clamping the bottom edge has relatively little further effect on
plates of small aspect ratio, particularly in the case of long patch loads,
but has relatively little effect for plates of larger aspect ratio. For
aspect ratios above about 3.0 to 3.5 (depending on the patch loading
o
in
m
O
en
LiJ
üJ
co
u Q
in LU
CNI
ÏÏs Q
u_ o
h-
U3 O
LÜ O
ü_ LU
LÜ
Q Q
LO CO LÜ in
LO
lO
P
LO LO
LÜ
LO
(N O 00 tû o CD CNI O
CNI CNI
^ lN3101dd300 ONUXOng
103
o
ÜJ
H-
00
—J
O ÜJ Q >- II
O LU y
O o
O CNJ
<
cr
ÜJ
LO
LO
Q o
U LÜ ÜJ
CL
LO
LO <
ü_
ÜJ O
co
in
ô
o OO LD O 00 OJ
o
CM
^ lN3101dd300 ONUdOna
104
Edge conditions: K
S = Simply supported K K for simply
C = Clamped supported all round
2 93 1 00
6 ■ 52 2 23
6 - 87 2 34
7 - 48 2 55
1C5
the top and bottom edges become effectively indistinguishable from those
for plates clamped along the top edge only. These trends are further
L/B = 1.25 and C/L = 0.25 . For simply supported edges K is 2.93 .
and 2.55 .
results with other work. However, the following points help substan
obtained with the earlier version; the earlier version has itself
2. The curves for plates with clamped edges are of similar profiles
increase with increases in the C/L value, the C/L curves do not
aspect ratios has been observed to occur also for simply supported
the plate.
4. Khan [47] has calculated coefficients for a plate clamped along the
top and bottom edges and simply supported along the sides, but for
analysis.
The panels were subjected to a full edge load along the top edge which
was clamped, the other edges were simply supported. They considered
stress to bending stress. For the cases where patch loading only
on the vertical sides which may account for some of the discrepancy.
6 . The support condition along the loaded, compression, edge has been
Timoshenko and Gere [46] have shown, similarly, that the support
a far greater influence on the buckling load than that along the
opposite edge.
107
7. The ratios between coefficients for plates with clamped edges and
Numerical Example
Consider again the model box section used in the experimental work des
cribed in Chapter 2. The panel measured 600 mm long (2L), 400 mm deep
load on the web panel resulting from the dispersal of a wheel load
a =1= 1.5
B = ^ = 0.15
Et3
D =
12( 1- v 2)
Taking E = 207 kN/mm^ and v = 0.3 for the steel of the box web gives
D = 512 kN mm .
Therefore
tt^D
= 12.6 kN
2B
P = 12.6 K kN
cr
103
where K is selected for the particular L/B and C/L from the set
Thus, if the panel is assumed simply supported all round then, from
very closely with the theoretical predictions for the panel when at
least the loaded edge is considered clamped. Now it has been discussed
earlier that the boundary conditions to a crane torsion-box web panel are
are clamped, rather than simply supported, and the vertical edges simply
supported - the close agreement between the theoretical buckling load for
this hypothesis.
P
cr
cr 2Ct
For each of the four support conditions, the buckling loads for the web
panel equate to critical stresses of 122, 289, 317 and 319 N/mm^
respectively. A typical yield stress for the web would be in the region
109
of 250 - 300 N/mm^. It is clear from this example that the support
considered now is that arising at the top edge of a crane web panel,
Fig.2.6 ).
Adapting the method to cater for loads of this nature requires changes
value of Z is found from the condition that the total applied force
the stated values of m' and n' and upon integrating f(Y) to
becomes o 3
[y l '
f(Y) = f 1 •- (3.33)
110
2C
2B
2L
2C
2 0
This f (Y) and the associated g(Y) and h(Y) were incorporated into
35 P
a =
m 16 2Ct
or C = 0.4571 C [3.34]
uniform patch loading lengths for plates of different geometries and with
different sets of boundary conditions. For each load, the length of the
equivalent uniform load was calculated from Eqn.(3.34) and the buckling
Tables 3.7 and 3.8 show comparisons between and for non-
0.50 , 0.25 and 0.125 . Table 3.7 applies to a square plate, both
when all the edges are simply supported and when the top and bottom
edges are clamped instead. Table 3.8 concerns a plate of aspect ratio
1.5 both when the top and bottom edges are clamped and the sides simply
supported, and when clamped all round. In all cases the coefficients
for the equivalent load are lower than those for the non-uniform
load, but only marginally. For a non-uniform patch load occupying the
112
c
= 0.4571 £ K
L Knu e Knu
C ^e
= 0.4571 r
L Knu L L Ke Knu
for shorter patch loads the agreement is closer. That the equivalent
The results indicate that the method for estimating equivalent uniform
buckling loads for plates subjected to non-uniform patch loads of the type
3.6. Conclusions
the two adjacent edges. Coefficients are given for the cases when all
edges of the plate are simply supported and when successive edges are
clamped.
The results are presented graphically in a form suitable for design calcu
lations of plate buckling loads; wide ranges of plate and loading geo
metries have been considered such that the majority of practical situations
are covered.
When compared with the case when all edges of a plate are simply supported,
clamping the loaded edge has a marked effect on the buckling load of the
clamping the opposite edge has relatively little further effect, as does
clamping the adjacent edges except in cases where these are long compared
stresses.
Plate buckling under the action of non-uniform patch loads of the type
produced by wheel loads on crane web panels has also been investigated
The results show that a simple method presented in Chapter 2 for esti
mating equivalent uniform patch loads leads to very good and safe esti
CHAPTER 4
4.1. Introduction
web panels were assessed. This enabled the critical load of a test box
practice when at least the loaded edge was assumed clamped; for panel
aspect ratios applicable to crane web panels, clamping the other edges
also was shown to have relatively little effect on the predicted buckling
load.
Now German designers still insist - see Ref.[50] - that an elastic criti
cal load approach with a tangent modulus correction to allow for possible
stressing above the proportional limit (see Bleich [51]) is the only one
panels with any form of in-plane loading. However, in many other countries
panel.
For a very slender panel subjected to any in-plane loading other than a
and Ostapenko [48] have shown that for slender plate girders subjected to
116
typically three to four times the elastic buckling load (thus indicating
computed on the basis of clamped edges. For less slender panels, the
illustrated that, depending on the length of the patch load and the panel
the squash line and the Euler buckling hyperbola. The transition curve
arises from the tangent modulus correction and meets the Euler curve at
ultimate load, such as that offered by the Merrison Rules [12], shows the
DIN approach to be generous for all but very slender plates. This arises
cerned with ultimate strength and despite its being applicable only if the
bending and shear loadings on a plate. Over the slender plate region a
curve derived by Walker [52] is also shown which represents the ultimate
plate strength is also covered by Timoshenko and Gere [46] for plates
117
O
UO
00
ÜJ
cr
CL
z:
o
w LJ
u
cr
o
c
c:
L±J
CD
w z
Z)
00
CL LU
_J
Û.
z cr
CD o
u_
■O
00
-o z
o
L£2
cr
s
LJ
z
I—
LD
Z
LU
cr
I—
00
u_
o
z
L_J
00
o CD o
CD
31 >>
olo
118
under uniform compression.
tries, none gives entirely satisfactory predictions over the full range
of experimental results.
gives results that agree well with experimental data fromvarious sources
It is the purpose of this chapter to look further into tfie method presented
by Roberts and Rockey. Certain of the departures from rigour are dis
this chapter closely with the following Chapter 5 where details of a ser
under the action of a patch load, Roberts and Rockey [53] assumed the
in the girder flange and three yield lines in the web. The locations of
the web yield lines and flange hinges are defined by dimensions h and
is the plastic moment of the flange and the plastic moment per unit
load 6 v , the plastic hinges in the flange rotate by 6 v/j and those
in the web by 6 v/2 hcos^ (twice this value along the central yield line).
the load which is assumed to have yielded under direct compression and
to j gives
M-hcos^
. (4.2}
w
0.5 (b-c).
linearly from +M^ at an outer hinge to -M^ at the adjacent hinge then
Z
<
X
o
LU
ÜJ
o g
O
m Ck
o
Q
Z
<
co
■o I—
Ck
LU
CÛ
g
C
NJ
ch
lTI
121
is the second moment of area of the flange about its neutral axis.
MfjZ
= 2h(l-sin^] (4.3)
6ËÏT
cos*
(4.4a)
1 -sin^ 2
2 3
where M = a t /4 . For a standard plate girder, I» = b^t_/12 and
W W W ^ ^ f f f
2
= Ofb^t^/4 and Eqn.(4.4a) can then be expressed as
4Ea t ^
T?iTn* = ■ (4.4b)
The last three terms on the right hand side of Eqn.(4.1) represent the
determined
was based on experimental evidence which suggested that the overall depth
maximum value of 0.5 (b-c) and h is given by Eqn.(4.6b) but with the
value 0 .2 d .
123
becomes the more likely cause of stocky girder failure. This situation
there are still four hinges in the flange as in Fig.4.2 but the web is
Equating as before the work done by the applied load as it moves vertically
4M
" -j- + (j+c) (4.7)
4M-
j2 -
V w (4-8)
P = 4/M-u t + a t c (4.9)
u f w w w w
slender plate girders do not apply to the stocky girder solution. The
arising from bending failure and direct compression failure of the web.
4.2.3. Comments
Part of the appeal of the solution by Roberts and Rockey lies in its pot
cally the method gives results that correlate closely with experimental
data taken from various sources for 63 different girders. The ratios of
with a mean value of unity) of just over 1 1 % which is better than any
other method of prediction known to Roberts and Rockey. For the experi
mental data considered, the extremes of predictions from the solution are
girder flanges.
The present author has undertaken a study of the Roberts and Rockey solu
tion. Originally this was to assess whether the method could be applied
being conducted.
in both the flange and the web which determines the structural defor
assumed deformation just prior to collapse when the upper bound theo
posed .
sional ly correct and therefore does not give consistency with scaling
given by h = 25 t^ .
3) In the assumed mechanism, the distance h between the upper and cen
tral web yield lines is equal to that between the central and lower
that for these structures the vertical distance between the upper
and middle yield lines is usually somewhat less than that between the
web bending while elsewhere the web yield lines offer full plastic
beneath the applied load to full moment restraint towards the outer
flange hinges.
126
takes place.
0.2d or 2j. Now for cases where the plate is of short depth or
applied.
in cases where large deformations of the web develop, that the work
lines to either side of the patch load are of length j and that
they undergo the same rotation as the yield lines beneath the patch
and lower web yield lines are then 1.05 and 1.20 times the j
non-zero patch loading lengths and for panels with aspect ratios
possible that the work term for the web in the analysis may be some
what in error.
Fig.4.3. If the hinge axis 0 lies within the plate thickness (i.e.
tti" 6 6
tw/. tw/,
(q ) (b)
which becomes
ÔW = o + e^ 58 (4.10)
w
5W = a t e58 (4.11)
w w
Rockey mechanism except that the distance h^ between the upper and
z = h^8 ^ = h 2 0 2 (4.12)
dv = (z+2e^)(d8^+d8^) (4.14)
t£)
Z
<
X
o
LU
$
LU
IT>
en
LTI
131
spacing is lost.
u
e = e — (4.15)
o 1
where u is the distance from an outer flange hinge towards the adjacent
e takes the same value in each of the three web yield lines,
o
For the assumed mechanism, internal and external work can be equated for
(a) f t j 2
dui (2 d 8 ^+2 d0 2 )
I J J
This becomes
r 4 M1 ft ^ ft e 1
W . 0
P - — :— dv - 2a c + 2 j (d8 +d8 )
_u 3 J w I4 0 1,4 3j 1 2
4M, 2 a
P = _w _ 0
+ 2 j (4.16)
u 4 3;
ze rt^
0 0 w - w 0
2. ze (4.18)
J 3 ^ 3 4 4 2
general case.
f Pe u ^
P dv = 4M- ^ + a t e c(2d0 +2d0 ) +a t |2 - j- du|(2 de^+2 de^)
U f ] WWO l 2 WW
f
re U-)4-
0
+ o i2 duj(2d8i+2d02)
w I J J
where f =
2 e
zt‘
w
t^e + . - 3ze^ = 3cze^ (4.21)
wo 4 o 0
th
Solution of Eqns.(4.20) and (4.21) for e^ leads this time to a 5
thickness t thus
w
4M- 2a t r r
p = " [1 + V i] (4.22)
(Z+2V) T + V 2 |+ 2 j
4 3,
M (Z+2V)
where = (4.23)
21
o t
w w
and (4.24)
4M- 2a't V
p = y... c + f-1. + ----
1 ^
3 (4.25)
u 3 (Z+2V)
1 2 V2
2M V(Z+2V)
where j^ = (4.26)
1 2 '
a zero Z value. The equations for condition (b) V > 0.5 therefore
4M
These are the same expressions as those derived by Roberts and Rockey for
becomes
yZ + ZV - 3/4 = 0 V ^ 0.5 (4.28)
surface of the web plate when the middle hinge line of the mechanism has
When the values Z = 1.0 , V = 0.5 are substituted into Eqns. (4.23) and
(4.26) the same j value is obtained and the same collapse load is then
given by Eqns. (4.22) and (4.25). For Z values greater than 1.0 ,
condition (a) applies and collapse loads are found from Eqns.(4.22),(4.23)
and (4.24); for Z values less than 1.0 , condition (b) applies and
In the more general case when the patch length c is non-zero, computation
135
in Eqn.(4.24) are generated as are the 6 coefficients for the 5^^ order
For each Z value there exists only one applicable Vsolution for a
given structure and load. This V value is selected from the 11 solu
tions that result in the general case. If the value is less than 0.5
then condition (a) applies, jis evaluated from Eqn.(4.23) and the
0.5 , condition (b) applies and j is then evaluated from Eqn.(4.26) and
the collapse load from Eqn.(4.25). The results are presented in the form
of a collapse load curve drawn through the loads calculated for each Z
Using the new mechanism solution, collapse load curveshave been produced
for the 63 test cases that provided the collapse test data used by
Bergfelt and Hovik [60], [61]. All these girders have web slenderness
radios (d/t^) of between 150 and 400 . Table B.3 gives details of
collapse tests on stockier plate girders (the TTG girders) with d/t^
The general form of the test girders listed in Tables B.l and B.3 is
shown in Fig.4.6 . They have web panel aspect ratios of between 1.0
but have aspect ratios of between 4.8 and 8.0 . Tests Bl-BlO and
B21-B23, however, refer to two very long girders with continuous webs and
was subjected to a patch load with the girder supported on the bottom
different girder was supported at its ends and loaded at various positions
along a uniform top flange. These tests therefore refer to girders with
Collapse load curves have also been produced for the series of short-span
girder sections, each with a rail welded to the flange above either web, as
shown in Fig.5.3(a); test girders RHl, RH2A and RH2B incorporated a flat
steel strip between the square bar rail and flange as shown in Fig.5.3(b).
the flange) used in the calculation of has been taken as twice the
137
CL
LO
Z
ÜJ
•o
CL
uo
2
g
LD
<1 -, <
__I
CL
O
l/l
Q
\0
_o
cji
d
138
distance from the web midplane to the edge of the nearer flange outstand,
dinal web stiffener as shown in Fig.5.5; RHll had a small flat strip at
RH series are from 133 to 240 and aspect ratios of the panels are 1.5
except for girders RH9 and RHIO which have square panels.
dicted collapse load was evaluated for Z increments of 0.1 and the
plotted points then gave the smooth curves shown in the figures. Where
a curve intersects the horizontal broken line drawn through the 1.0
ordinate gives the Z value for which the new mechanism solution predicts
Figs.4.7 - 4.9 show the collapse load curves for girders TG5, TGll and
TG15 tested by Skaloud and Novak. The curve for girder TG5, see Fig.4.7,
little over 8 times the web thickness). Over this region, the slope
of the collapse curve is very shallow and the predicted collapse load is
for girder TGll, shown in Fig.4.8 , has a much steeper slope at the inter
section point with the broken line which occurs at a Z value of a little
in Z about this value. Although the curves for girders TG5 and TGll
both give a maximum squash (Z=0) load of approximately 2.7 P^^, their pro
files thereafter are quite different with the curve for TGll falling
o-
in
en CO
CL
LU
CD
CL
CD
<o
CL LA
O
LU
CJ
Q
O
LU
CL
CM
CD
CD
o
o in LP o U1 o LD o
CM O O
LD 00
0 \
cr
UJ
Q
cr
LD
Q
Z
LL
cr
o
Li_
ÜJ
CJ
Q
o ro
LU
in
o
CJ
o o in o o in o
ro ro r\j o
es
LO
LD 00
CJ
LU
en
ID
LJ
Q
<c
o
LU
en
eu
CM
o
CJ
o
o o LTI o in o tn o
rn CM O O
Fig.4.9 shows the collapse curve for girder TG15. Of all the 63 test
cases considered by Roberts and Rockey, this girder has, at just over 4 ,
the greatest ratio between squash load and experimental collapse load.
Otherwise, the curve is similar to that for girder TG5 in that it has a
shallow slope over the region where it intersects the broken line which
from small variations in their respective flange and web yield stresses,
girders TG15 and TGll differ only in the size of flange: girder TG15 has
a flange which is wider and nearly five times the thickness of that of
girder TGll. Girder TG5 has a similar sized flange to girder TG15,
tends to lead to a small Z value and to predicted loads that are more
Figs.4.10 and 4.11 show the collapse curves for girders STG7 ^ 8 and B12;
both girders have relatively small flanges. In both cases the Z value
is small and in a region where the slope of the curve is steep. Fig.4.12
shows the collapse curve for girder TTG2, one of a range of stockier
The collapse curves shown in Figs.4.14 - 4.17 are taken from the RH series
of girders. Figs.4.14 and 4.15 both apply to RHl. The first shows the
collapse curve when the patch length c is taken as zero, the second when
CO CO
Q S3
q :
o
LU
Q
O
LU
Q-
O
LJ
o
o m o in o Ln o in o
ro oo CM o o
C\!
ûO
CQ Ln
q :
Q
cr
Q o
en un
o
LU
LU
u
Q
O
LU
en
eu
OJ
O
eu
o
in o o LO o LD o
OvJ o o
CO
cr
LU
o
cr
eu
LU
Q
ÛL LD
O
LU
Ql
W
Q
O
LU
tn
Cl
OJ
o
eu
o
o in o in LH o in
o o
O
CM
cr oo
CQ
Q
UJ
ID
cr un
o
Li_
LU
C_j
CD
O
LU
en
D
OvJ
o
o un o m o
PO PO o o
CO
cr
LJ
G
LT
UJ
=)
cr
cr ui
o
Li_
LU
LU
LU
CD
CL
O
G
o
O LD O U1 o o LD O
r\) O o
o
o
oo
cr
UJ
CZl
cr
J]
LU
cr LO
o
u_
U -i
cr
LU
Ul
cr
C\J
o
LJ
LTî O in LTI O un o
OJ OJ o o
CO
cr
cr
LU
eu
en
CD S3
LU
cr
LP,
UJ
Q
O
LU
en
cr
OvJ
O
LJ
o
o in o Ln o in o in o
OvJ r\i O o
Ch
X 00
cr
cr
LU
Q
cr
eu
>o
LU
LD
o
LU
cr
LJ
CD
o m
LU
in
cr
OJ
o
CJ
o
o LOI o o in LD
ro C\J o o
over larger Z values. Figs.4.16 and 4.17 show curves for RH6 and RH9
which are typical of the curves for the RH series. In computing these
to the experimental collapse loads are given for all the girders listed
in Appendix B and for the RH girders detailed in Table 5.1. Also shown
dash denotes a case where the curve failed to reach the load specified
in the column heading as, for example, with girder TTG2 shown in Fig.4.12.
Now a full solution using the new mechanism approach requires some means
a comparison is now made between the new approach and that by Roberts and
Rockey.
z = h cos ip (4.30)
(4j+2c)M,,
n =
2M + a t z
w w w
2
a t
where (4.32)
2 j + c
Hence n = (4.33)
1 + (2z/tJ
4M- a t
f w w
(4.34)
'’u = — " TTlz
4M-Z
(4.35)
a t
w w
The solution for direct yielding of girders with stocky webs needs no
Using Eqns.(4.7) and (4.8), and Eqns.(4.34) and (4.35), collapse load
incrementing Z values in the same manner as were produced using the new
4.18 -4.21 for girders TGll, TG15, TTG2 and RH6 respectively. In each
figure two curves are shown : the continuous curve with a smooth transition
analysis and the curve with a squash load cut-off obtained with the method
o
oo
ex
û:
LU
Q
cr
LD
LU
LO
LU
cr
tu
CD
o
OvJ
o
LJ
o
o ui LO o LO LO
ro OvJ OvJ o o
es
CD co
q:
LU
Q
cr
CD
UJ
LO
LU
LJ
CD
O ro
ro
O
eu
o
O m o LO LO O LO o
r\j o o
CK
CD
CH
LU
Q
rv
-o
Q
en in
o
LU
cr
LJ
ro
OJ
O
eu
o
in o LO o LD O
OJ O O
CO
û :
LU
Q
cr
Cj 'O
Z
LU
cr
cr LD
o
LU
cr
(U
Q
o
LU
U1
cr
CM
O
(U
o
o o Ln LO in
ro CM CM O O
The horizontal squash load line shown in the figures arises from Eqn.[4.7)
and extends over the range of Z values for which the collapse load pre
curve obtained with the new approach can be seen to give lower estimates
load at Z=0 is obtained. Beyond the transition region, the curve from
the new analysis crosses the other and gives slightly greater collapse load
predictions but the two curves follow similar paths and converge towards
larger Z values.
Overall, the curves resulting from the two approaches are very similar.
such girders, however, that the two types of collapse curve are particu
larly similar, owing to the curves being very flat as shown for example
in Fig.4.20 for girder TTG2. For some girders, the restriction in the
original theory that c does not exceed 2 j has the effect of further
to apply; this restriction then lowers slightly the collapse curve from
the original solution over the region leading up to the squash line.
Thus it can be seen that the differences between the two analyses are un
4.5) and the same method can then be employed. From Eqn.(4.12) z is
now given by
z = h9 . (4.36)
of the flange as in the Roberts and Rockey method (see Eqn.(4.3)) and 9
cos 9 = 1 - 9^/2
V = 2h - h ( 2 - 9 2 ) + 4e 9 = M^j^/6 EI^
z% + 4e z = M hj2/6EI„ (4.38)
M h
Z' ' = 6 Ë ï V • (4.39)
f w
It will be recalled that the collapse load curves presented earlier were
(4.24); for V > 0.5 , V and j are found from Eqns.(4.26) and (4.27).
159
can be determined from the point where the two curves intersect. An
example is shown in Fig.4.22 for RH6 where the curve extending over the
the collapse load from Fig.4.21 then the predicted load can be seen to
tion of the new and original methods do not agree particularly well with
that obtained using the original analysis on its own. The increased
hybrid solution arises because the method used by Roberts and Rockey for
general application.
the squashing term 4e^0 . If this term is omitted then the equation
M h
Z' " 6 ËT:t-r jZ (4.40)
f w
160
NO
11
en
en
LU CD
Q
en
LD
CD
Z
\0
en
o
LU
en tn
LU
eu
eu
LU
LU
CNJ
<
O
o un o
o o
161
and this leads to selected Z values that give predicted collapse loads
The new mechanism solution has been shown to provide improvements to the
between stocky and slender collapse is achieved and the need to evaluate
the direct yielding term n is removed. Apart from over the region of
transition, however, the two solutions have been shown to lead to very
neglected.
web to one side of the loaded region. Now if, before loading, the web
were cut vertically at the positions aligning with the edges of the patch
load and there was no shearing between the web and flange, then the web
original flat plane of the web,the total area of the opening is as shown
in practice.
From the sectional view of the central zone of the mechanism shown in
gives
Hence v = h6 ^
V = -^ . (4.41)
A
163
C
(a I
lb )
( cl Id )
total stretched area when projected onto the original web plane is
A^D = DC = - vh
A^D = DC = h - v/2
Tn the same plane, let angle AOD be 4*1 • Let the angle between
a^ = j (1 - cos (p)
2
Therefore a^ = j[l - (l-(p^/2]] = ^
Now a^ = OA - OE
y = — —^ (4.44)
/j ^ + h^
165
_ ED _ h + (v/2)
sin (pi =
OD
/j ^ cos Y
cos Y = 1 - (4.45)
2 (jZ+h2 )
hv
- 2 - ~
(4.47)
j
greater than a^
h2
Extra Work = 2a t —r- v
w w ]
given by
2a t h^
G = ... . (4.48)
This term is similar to the term 4M^/j already present in the collapse
incorporated by modifying the first term on the right hand side of the
collapse equation to
—— ^ 2a t —— — —
] w w J ] f 2
/ViM 2M
^u ^ hcosl;'^ hcos^* h cos ip
8M 8jM
Pu " -j- • (4-49)
The significance of the web stretching term in the collapse equation can
! M h a t a t
w w w w
Test ^ Eqn.(4.6)
2 2 M^
Girder ; ^
X 10° Niran mra X 10^ Nmm
1
to the Mj. parameter. Table 4.1 shows the ratio a t h^/2M- for a
f w w f
number of girders.
trend in the values of the ratio for girders TGll - TG15 reveals that
web stretching is most significant for girders with light, thin flanges.
the same. Girder TGll with a thin flange has a web stretching term
which is over 5 times its value while TG15 with a very thick flange
has a value of only 0.17 for this ratio. In Table 4.1, h has been
strongly dependent upon this empirical estimation. For a given web and
which leads to a large ratio between the web stretching term and >
for girders with thicker flanges, is larger and the ratio decreases.
8 M h^
w
G = . . (4.51)
^wi
.2
j2 =h cos ip (4.53)
M t
w w
3 _ t^jCZj+c-n)
= (4.54)
8 cos ^
4M. 2M
P = + -— :—
4h_^_c£sJ^1 (4.55)
u J h cos ip 2 j + c - n +
J
4M, 2M
w (2 j+c-n)
P = 2 j + c - n +
u j h cos ip 2
4M, 3M
w
Hence P = - (2 j+c-n) (4.56)
u j h cos ip
vicinity of the patch load that yields in direct compression. The web
side of Eqn. (4.55) apart from the first. A new expression for n is
thus obtained by equating the web contribution to the force in the yielded
length n of web
2M
w 4h° cos ip
2 j + c - n + = a t n
h cos ip w w
before gives
MfjZ
= 2 h(l - sin ip)
6 EI^
12M t El-
1 - sini " M-(M-t +2M h^) " ^ (4.58a)
f^ f w w
cos* = (4.58b)
In the mechanism assumed for girders with slender webs, the web yield lines
are considered as plastic hinge lines. The second term on the right-
tion of these hinges along the lines to either side of the patch load
In evaluating this work it has been assumed that the lengths of the
inclined hinge lines are equal to j and that these hinges rotate through
The web mechanism to one side of the patch load is shown in Fig.4.24(a)
where h is not small compared with j . The inclined web yield lines
OC OC ZTTTr— T-9
= + j2 -— (4.60)
sin ({>3 sin(90+#2) cos <t>2
2 ^ 4h^ + Sh^jZ +
Hence cosec 4)3 = ----- ^2 j 2 ---- —
Then 6 = z/h
a
2 -1
Rotation x Length = j 8 ^ =
172
0 A
9b
Id ) Rotation 6d along OC
AK = j sin (p2
where =' = §§ = = ^
i. 4 .j2
Rotation x Length = z
h hj
BM = OB sin 4)3
*d■ôrârîT ■TT
Hence Rotation x Length = OC 8 ^ = (OC)^ = (4h^+j^)
Summation of the work terms for the inclined web yield lines gives
Rotation x Length =
174
multiply the work term associated with rotation along the inclined web
H = (4.62)
‘ • I F
1.0 can arise, as shown in Table 4.2. It will be recalled from Section
4.8.1. that the first two terms on the right-hand-side of the collapse
terms, its effect on the collapse load prediction can be seen to be signi
ficant. From Table 4.2 it appears that, as with the web stretching term
that develop for girders with light flanges. The trend in H values for
becomes
M - h cos ^
2 "f 3^2 2h3 cos *
(4.64)
M 2 t
w w
175
Test h j
Girder nun nun
8 h^ cos 3h^
+ - (2j + c - n) = 0 (4.65)
n is now given by
24M t EI_
cos Ip ^ W W f ________________ .y-x
1 -sin ip M -[2M-t + 4M h^ + (3M t h/cos ip)]
f f w w w w
further.
Development of the Roberts and Rockey solution has been taken no further
design purposes.
Roberts and Rockey, it has been shown in Refs.[53] and [54] to give
generally very good collapse load predictions for the test girders detailed
P^^ , is again achieved. With one exception, the predictions lie within
than the others which are approximately half scale models of typical full-
RHl are between about 0.6 and 0.8 times the corresponding dimensions of
RH2A and RH2B, typically the scaling factor is 0.7. Material yield
stresses of the components of RHl are similar to the yield stresses of the
corresponding components of RH2A and RH2B. The Roberts and Rockey solu
tion predicted the collapse loads of RH2A and RH2B very closely, but failed
approximately 0.7 times the size of that predicted for RH2A and RH2B needs
Table 4.2 that this is not the case. Dimension h for RHl is approxi
mately 0.4 times that for the RH2 tests, j is approximately 0.54 times
(approximately 0.7 times the h value for RH2A and RH2B) instead of
now also approximately 0.7 times the j value for the RH2 cases). The
erroneous h value (and hence erroneous collapse load) in the case of the
small RHl girder compared with the RH2 girders. The above example
girder size, and this then leads to collapse load predictions that vary
determined E value (see Chapter 5) of 210 kN/mm^ has been used. The
cos ip term was evaluated from Eqn.(4.4a) using the integral rail and flange
term I^^ (given in Table 5.2) for the I^ term in the equation. It
substituted into Eqn.(4.6) were those values given in Table 5.1; this
then leads to the good general agreement between predicted and experimental
loads shown in Table 4.3. In fact, the major component of the ’flanges'
with these girders are the rails which have much higher yield stresses
(a^ in Table 5.1) than do the flange plates. If the values are used
P
Test P P u
u ex
Girder P
kN kN ex
since Eqn.(4.4a) rather than Eqn.(4.4b) was used in the solutions for the
of the flange material yield stress used to find h as this is then the
One further point relating to the evaluation of h and the strong influence
girders RH3, RHll and RH12, which are all similar except that RHll has a
longitudinal web stiffener at the mid-depth position while RH12 has one at
from Table 4.3 that there is a slightly increasing trend in collapse load
from RH3 to RHll and from RHll to RH12. If the full web depth is used
obtained for girders RHll and RHl2, as well as for RH3. ' If, however, the
4.11 Conclusions
A detailed study has been made of the plastic mechanism solution presented
by Roberts and Rockey [53] for predicting collapse loads of plate girders
A new mechanism solution, based on that by Roberts and Rockey, has been
developed and this has been used to produce collapse load curves. The new
direct web yielding for girders with stocky webs and failure by a mechanism
also removes the requirement (in the solution for girders with slender webs)
analysis and producing collapse load curves in the same manner as were pro
duced using the new solution. It is shown that the curves differ notice
ably only over the relatively small region of transition between stocky and
the new solution do not justify the increased complexity entailed in opera
ting it. With both methods, prediction of collapse loads requires an esti
general application and is not particularly well suited to the new method.
effects of web stretching and of the inclination of the web yield lines.
the collapse load expression but further work is required to ensure true
girders tested to collapse under the action of a wheel load produced very
aroused further concern over the validity of the empirical expression for
accuracy depending on the size of the girder. Predictions for two half
solution over-estimated by nearly 50%. This casts some doubt over the
CHAPTER 5
5.1. Introduction
the validity of the method may be checked. Walker [42] has drawn
Although, over recent years, some collapse test programmes have been
torsion-box-girder.
and 2.5 , whereas for conventional girders this ratio generally lies
the crane girder web arises from a wheel load : at the rail surface the
load is of a length close to zero but the load then spreads through the
top edge of the web. Most tests on conventional girders have considered
184
patch loads uniformly distributed over some length of the flange, the
load then spreading through the flange only before entering the web.
the two types of girders are similar. Crane girders generally have
web slenderness ratios in the range 100 - 250 and web panel aspect
ratios between 1.0 and 1.5, both ranges being within those occupied by
tical reasons, scale model specimens were used (these generally being
of half scale size). The purpose of the tests was to study the physical
ratio of the loaded web panel, the size of the rail, and the presence
A further reason for the tests was to determine whether the mode of
forms the basis of the Roberts and Rockey [53] mechanism solution for
to design and construct a test frame to house the available ram. Facil
connections.
The frame was designed to BS 449 [9] for a load capacity of 400 kN (to
Design details are given in Appendix D: Figs.5.1 and 5.2 show the
girder sections, each with a rail welded to the top flange above either
web. Two collapse tests were obtained from each box by loading either
Fig.5.3 illustrates the construction of the boxes and Fig.5.4 shows the
employed on each box; the flange thickness and yield stress were there
fore common to the two tests on any particular box. The web plate and
rail on one side of the box could be either similar to-, or different
positions remote from the test panel, thereby avoiding the introduction
are attached to each flange; these act as backing-bars to the webs and
can be seen from Fig.5.4, this feature was retained in the design of the
model test boxes in order to provide locations for the webs and, thereby,
facilitate the construction of a true box with the webs having straight
The first box to be constructed and tested was the smallest of the series
crane girder and bore a close resemblance to the box section used in
rail was attached to the flange, as can be seen in Figs.5.1 and 5.2
189
(_
JD c c
nr
:-‘E E
00
ÛL
00 <
cr cr
LU
CD
QC
<] ID
I I
X
< z
cr o
o I
CO cr cr
â â
cn I
LU
u_ cr cc
o
00 I
00
â
o
00
LU
Q
nn
LT\
di
<3
yI
5
1
191
where the box is shown positioned in the loading frame. The rail
the other side of the box could be used for a second test. Another
rail, in this case identical to the first, was therefore welded to the
box above the unloaded web and this side tested also. Since the two
webs were cut from the same steel plate, two tests involving the same
(RHlb). Variation about the mean value was therefore less than 2% and
these two tests are consequently referred to jointly as RHl when the
mean value of the collapse loads is quoted. Roberts and Rockey [53]
half as large again as the first and thereby constituted a half scale
model girder section. The two rails were of equal size and again of a
first, was nominally of the same construction on either side, the two
webs came from different steel plates and were found to have slightly
consequently referred to as RH2A and RH2B. Rather heavy main seam welds
were used in the construction of this box and this resulted in particu
test facilities since the collapse loads were only slightly below the
selected with greater confidence that the collapse loads would not exceed
The remaining box sections were all nominally half scale models and were
with RH2A and RH2B. Each side of each box provided a different combina
welded directly to the flange (as shown in Fig.5.3(a) and 5.4) was em
ployed in each case. Panel aspect ratios were 1.5 in all cases except
for RH9 and RHIO where the aspect ratio was 1.0. Slenderness ratios of
shown in Fig.5.5. RHll had a small steel flat web stiffener attached
fections arising from the box fabrication process. RH12 had a heavier
on the outside, rather than the inside, of the box; this was to facili
small plates were welded to the outside of each web at the inner diaph
ragm locations.
193
OJ
cr»
c
<
LT»
OO
o_: m
c
O CL
S:E <
3 CT»
c CL
Cn4
m %:
cr
■o LH
Csl Q
z
<
cc
CL
00
00
CL
00
CD
LU
3
uni
uri
CT»
194
For all the half scale models, the outer (support) diaphragms were so
positioned that the span of each girder was 1380 mm (a little over
5.4. Preliminaries
Every steel plate from which a web or flange plate was cut had an
Similarly, samples were taken from each of the square bars used for the
test rails. The tensile specimens were made and tested according to
tion in yield stress with the direction of rolling, two samples were
tested for certain of the plate thicknesses, one having been cut
stresses o^, of, and oj- for, respectively, the web, flange, and rail
All plates were grade 43 steel except the thinnest - the 2.45mm thick
plate used for the webs of RH5 and RH6 , however, was of grade HR4 since
The square bar rails were of grade En3B. The actual yield stresses
the minimum quoted in the relevant code. From the stress versus strain
plots obtained from the tensile tests, an E value of 210 kN/mm^ was
estimated and this value has been assumed throughout this chapter.
Prior to loading, every care was taken to position the box and moveable
box supports within the frame so that the load was applied at the mid
span and in the midplane of the test web panel. The girders were
0.008 mm/second.
As can be seen from Figs.5.1 and 5.2, load was applied to the boxes via
foot and the actuator ram. Vertical displacement of the top flange of
From the outputs of the load cell and transducers, curves of load against
face of the test web panel along the vertical centre line (it was not
possible to mount gauges on the inner face owing to the closed con
struction of the box). The purpose of the strain gauging was to aid the
Since the principal aim of all the tests was to determine the collapse
the transducer was positioned at the vertical centre line of the web
panel and as close to the web as was possible. This procedure for
TD c
CJ
TD OJ
E
o
II E O
L_
W
00
H—
00
O
g k
-5.-8 “ 00
I (/) TD
CL
ë § <
ÛC
O
C C L CL 00
CL
LU
LJ
ZD
Q
LTl 00
Z
<
cr
LJ
CL
CL
Q
TD OJ
W LL
o
13
—— ( q !: r —
ZD
O
cn VO
cn uri
CT
LJ CL
LU CD
198
the depth to A from the underside of the top flange was varied
between 35 ram and 120 ram, the depth to B was varied between 125 ram
The beam housed a long linear potentiometer which monitored the position
of the trolley. By levelling the beam over a box rail (or, with the
box lying on its side, over the midspan of a web panel) and traversing
box-girders are given in Table 5.1 where the relevant girder dimensions,
width bf of twice the distance from the web centre-line to the nearer
00 00 O lO uo CO CO •d" CN UO uo CN CN
2? ^ CN CN •d- O O NO VO 1-4 O o 1-M r4
M- C C O4 M r- CO CO P-.
rN g P^
M p- 1
—4
CNJ
L g m m lO lO 1
—4 m CO U1 CO uo CO UO uo
to S o> ON •d- 00 •d- •d- p- •d" p^ •d- •d"
vO UO lO NO UO NO uo VO uo NO uo VO NO
S’
co CN
00 0—04 0 OO OO 00
4-1 g —04 0
—0 00 00 00 00
ON ON
w to B vO o o m 1 14 1
—4 1
—4 1-M ^4 r-4 1—4 r-4
CO CO CO m en CO CO CO CO CO CO CO CO
S
I
&3
ÿ M O- CN •d- •d- CN CN 1-M UO •d" •d" UO uo
1-M O CN CN CN •d- •d- UO CN CN CN CN CN
CO CO CO CO CO CN CN CN CO CO CO CO CO
S
I
M W
■p
g
B
vO 00 OO
i
X X X
X X
o o 1 1 1 1 1 1 1 1 1 1
^ 1 lO g 00
è in lO
U 1 o\ ON 00 ON UO ON UO ON uo ON ON
co r—4 00 r-H CO 1-4 CN 1—4 CN f-4 CN r—1 r4
hJ
CN
I X
4-1
4->
4M
g
E
X
g
lO
X
NO
X
VO
vO
X
VO
vO
X
r-
vO
X
p-
NO
m
X
vO
X
uo
vO
X
vO
NO
X
p-
NO
X
p-
vO
X
p-
vO
M
X
vO
X
p-
rO E r- r—4 r-4 lO UO r» UO uo uo uo UO uo
lO r-M f—1 uo uo
0>
1—4 CO CN 00 m m m uo O NO U1 uo VO vO
.Û o O o ON ON •d- •d- o ON ON ON ON ON
H ^ 1 CO •Cl •d- CN CN CN CN •d" CN CN CN CN CN
lO io lO lO if) uo U1 uo uo O O UO UO
O
r—H r—^ r-H f“4 1
—4 1-4 1
—4 1
—4 1—4 r-4 r4 1—4
o O o o o o
T3 ^ o g O g g o o o g g g o
<- I VO NO vO vO NO VO VO NO NO VO vO
O O O O O O o
.O 1 g o g o O g g o o g o o
vO i ON ON ON ON ON ON ON NO NO ON ON
M
(U PQ O 1—4 CN
4J T3 1
—4 1 CN CO œ ON 1
—4 1—4 r4
CO M
<U -H § § § § §
H O
200
tests on girders RHl and RH2A and RH2B were preliminary tests
(see Fig.5.6 ).
flange.
With RH3, snap buckling of the web panel first occurred at a load of
at a slightly lower load than that at which the snap buckle on loading
occurred. Loading beyond the elastic limit of the web had the effect
and 5.9).
Fig.5.7 illustrates that after collapse the load remained very nearly
LO
z; o
CL
CL
X
O
(f)
Û3 CCI
CO
LU
in oc
LU
1—
CL
O
LL
t— o
Z
LU
Z
LU
(_)
<
CL LD
en
CM
CD
<
O
O
1—
CL
LU
>
<L in
CL
t—
Z
LU
CJ
en o
z
<
CO
c
Cü
< in
o
o
CO
LU
z
CL o
CL
C o
CO o
S o o o o o o
_i 00 o C\J
If)
CD
iZ
202
o
LO
in CL.
X cn
o
CÛ
t— CO
cn
LU
O
cr
o
u_ in
o
ro
UJ
z: in
LU
CJ
en
LU
CO
cr
LU in
CO
cr
o
cn
<c LTl
CD
O
Q
O O
a o
O
LU o
O o o o O O
CO OJ Ln
Q.
CL
O
204
displacements towards the centre of the box have been taken as positive),
ment growth, the new (outward) direction then being followed throughout
which shows the displacement at the upper of the two recording positions.
The load was therefore removed (no reverse snap buckle occurred), the
This snap buckle was therefore not recorded on the experimental plots.
It can be seen from the plots in Figs.5.10 - 5.12 that no further snap
buckling took place during the test. The load against vertical dis
peak was exhibited and the load then increased slightly with increasing
displacement; the load peak was recorded as the collapse load. The
Fig.5.13 shows the vertical displacement plot for RH5. It can be seen
that, as with RH3, the load remained approximately constant after collapse
Transverse displacements for RH5 are shown in Figs. 5.14 and 5.15;
approximately 35 kN, a snap buckle took place from which there was no .
o
<r
3:
ü:
CL
X 01
o
QÛ U1 Q
en CL
LU LU
q:
o o
LU
21
LU
U
-c LT,
_J
Q- CM
01
c O
o
CM
cr
LU
in
c
û:
LU
o
o
O o
Q
LU
Û_ o
CL
o
Q o o O O
<c 'O <r CM o o o o O
o <— <r tM
T— '— >— cc o
lO
6
li.
206
O
o o O
CM O O o
CO CM
o_
X
CM en
o
ÛO o
en
CD
LU c
oc
Ctl
O
U_
'O
LU
21
LU
O
■<
—I
o_
en
o 00
LU
CO
CC
LU
>
en
< o
cc
(n
<c
LO CM
<
Q
<C
O
Q
LU
û_
CL
■<
10
d>
11
207
V: o
Q o O O
o o o o
O CO
ÜC
CL
X en
o
ÛQ r\j
D
CO
en
LU C
cc
a:
o
LU
<r
I
CO
LU 'O
5
U
I
C
_J
Q_
en
LU 00
en
g
>
en
z
<
ce:
o
en 1
z
<c
(_3
<
O
< C\i
O
s
CL
CL
<=c
CM
lO
à
Ü_
208
CM
CM
O
in
X
q:
X
O-
o 00
QQ
CD (X.
LU LU
cc
o
Li_
LU
X
LU
(U
< CM
—i
Û-
cn
o
<
o
q: 00
LU
o
<
q:
I—
5 o
O
cn
o
■<
<
CM
O
<
o
o v: o
LU
o
Q. LTi O O L'n O LTl O LH o LI O
Q. o 'O XD LTl LTl U- U" M iaj LTl
c
CO
lO
d
IL
in
X
q:
CL
X CO
o
CD o
CO
cn
LU
cn o in o in o LTl O in O LD o c
c O LTi LTl <r f'O C\| C\) T '— in cc
QC o
O
Ll_
CM
LU
O
■c
d-
cn
Q
UJ
CO
QC
lU
>
CO
z
QC
CO
-c
CD
Q
LÜ
D_
CL
C
lA
g
iZ
210
Ln
q:
in CL
X CO
o LTl
CD Q
I
— CO
CD O
c
in CL
O'
o in
LL
CD
LA
O
-<
__I
CL
O
to
LU U1
cn
CL
LU
>
CO
z o
CL OvJ
LA
CO
<
CO O
<
Q
<
o
LA
Q O
LU
Q_ O
CL
<
O LA O LA O LA O U1 o LA o LA o
lO o -O LA LA 'cr -T ro NO r\] A nJ '— LA I
lO
d
u.
211
collapse point and a near constant load level thereafter. During post
collapse loading of RH5, a web hinge was observed to form just above
Figs.5.16 and 5.17 show the vertical displacement plots for RH6 and RH7
collapse; the load then remained nearly constant for further vertical
The transverse displacement plots for RH7 show a corresponding but more
seen from the plots that permanent set in the web is associated with the
the RH7 plots arose as a result of the girder having a fairly thick web
but small rail (it will be recalled from Chapter 2 that this combina
yield stress of the web was relatively low, produced significant web
yielding before the collapse load was attained (see Section 5.6.5).
212
'■o
X
cr in
CL
X on
o
CD
o
CD q;
LU LTl LU
cr LTI
o
LL
LU O
X
LU
CJ
C
_J
Q. LT!
en
Q
LU LO
C\J
C
CD O
LU r\j
O
LD
en
<
CD
c O
o
in
o
LU o
û- o
Q.
< o
C O O o o o O
(0 O fv '-o in 'j- ro CM
lO
d
LL
213
o
r-o
i Q.
X en
o
OÜ
Ln
U1 CM
et
LU LU
>
(X
o
ü_
o
LU
C
_J
n_
C/l
LTi
<
o
et
LU
■C
et O
LU
O
en
z
<
r_D
<C LTi
Q O
<
O
Q- O
CL
O o
o rsi o o
co C'
lO
6
LL
215
JZ
ce
Q_
X cn
o
03 C3
CNJ CO
CD
LU c
ce
ce
o
u_
03 O
<
1
—
z
UJ
z
LU CD
CJ
—I
Ci_
CO
o
LU
cn
ce
LU
>
CO
z
<
ce
1—
01
z
c
CO
<
C3
<
o
_J
C3
LU
—I z
CL
CL
<C
Q o o
■C
0) o r\j o o o o O
CC' <r Oj
in
O)
il
216
three plots for this test shown in Figs.5.20 - 5.22, On first loading,
loading, the panel reverse snap buckled, but not until the load had
fallen nearly to zero. For the second cycle of loading, the panel
and B in the RH8 test were as for RH7 and were thus 10 mm higher than
in the RH5 test. The two transverse displacement plots for RH6 (meas
ured at the same positions as in the RH5 test) were broadly similar to
Fig.5.22.
Figs.5.23 and 5.24 show the vertical displacement plots for RH9 and
RHIO, the two girders with square web panels. Both plots display an
manent set and then a well defined turnover at collapse, further dis
with RHIO). The transverse displacement plots for these girders (meas
ured at depths of 60mm and 140mm for RH9, and 45mm and 125mm for RHIO)
were broadly similar to those of the RH8 test but without evidence of
2 1 7
00
3 =
oc
Q_
X CD
o
CÛ Q
t
— O
CD CC
LU
oc
o
LL
LU
2 :
LU
C_)
<
Q_
en
C
•<
O
QC
LU r-n
GC
LU
CJ
(AJ
CD
CO
c
■<
o
Q
LU
O.
CL
Q
<C 0 0
O O r\) 0 0 0 0 o
<N co 0 r\j
10
à
11
21 8
LD
O
00
X
Œ o Cl
X 00
o o
QO o o CO
CO
<c OJ o o o o
00
00
o 00 ■o Csj
LU «=c
o CL
OU
O
o
U1
LU
X
o
5
c
_l
CL
00
U1
co
LU
fM
en
CL
LU o
:>
f/l
Z
c
CL
LT,
00
c
co O
-<
Q
<
O
LD
CL
CL
-<
C\i
U5
Ô
LL
219
'30
oo
31
cn
Q-
X
o
co
CO a
f—
cn
(/I
UJ <c
t— cn
cn
o
u_
CO
•
t—
c
[—
z
L_
LÜ
LJ
<
_i
CL
cn
c
LU 00
cn
cn
UJ
cn
-c
cn
t—
cn
z
-<
CO
<
Q
<
_1 CM
Q
UJ
_l
CL
o_
<
CO O
c
OJ o O o o o
CS|
CO -sC (M
in
O)
il
221
un
o o
E LP
CC
CL
X <J1
o
co LD
(O oc
UJ UJ
O
g
UJ
o
<c
—I
CL
cn O
5
CC
UJ
O
QC r\j
Z
UJ
CJ in
cn
c o
CD
O
<
o
in
c o
UJ
CL V: o
CL
«C
Q o
-C o o
O r\j o o o o o
CM '— >— co o -T rsj
id
c»
222
S
G
X en
o
ÛD G
LTI
CO G
LU G
et
O
LL
LU O
LU C\J
O
CL
CD
G
U1
<
CJ
QC
LU
QC o
LU
CJ
en
<
CD
< LTI
G O
C
O
G
LU
G O
G
O
G O
10 O O O o o O
cg co o <r Os)
in
g
il
223
o
o
LD
ÛC
un
X
§ Q
O
ex
LU
>
CC un
LU
LU
O
LU
LTI
en
o
CJ
û: LA
LU
r \i
c
CC o
LU
CJ
L— un
en
O
CJ
o
LA
LU O
eu o
o
o
o o o o o o
00 o r\j
224
X to
Q
03
4 rc
!.n
LU
CD o
CC
LU
LD
cr
03
Cl
o
o
o
no CM
225
o
o
o o o o
C\I m
o
X
o o
ÜC Q
tn
LU ui
CL
o
o
CÛ
C-J
LU
ù)
Wj Ul
lu
cn
CL
en O
U~l
Q
Ul
O
O
Q
LU
CL
CL
<
226
RHIO test grew more rapidly at low loads than at high loads, but not so
Test girder RHll was very similar to RH3 except for the addition of a
The vertical displacement plot for RHll in Fig.5.25 shows a rather more
RH7) and also shows the load to fall-off gradually after collapse.
RH3 but for the addition in this case of a structural stiffener at the
Transverse displacement was monitored both above and below the stiffener.
was at a depth of 280 mm. It can be seen from these two plots that dis
Snap Buckling
The snap buckling phenomenon noted in several of the RH tests has been
pattern of the buckled surface of some of their test girder web panels.
for the buckled surface of a web to follow the shape of the imper
at the top edge of a web by loads with an effective line of action that
turning moment increases and may oppose development of the preferred mode
causing a sudden change in the buckled shape of the web. Loading eccen
wheel, rail, and web centre-lines, or due to a wheel load acting non-
menon of web snap buckling. Skaloud [63], and Evans [63] have both
occurring in various web buckling tests at loads well below the collapse
girder failure. Both Skaloud and Evans link the snap-through pheno
menon to the critical load of the web: Skaloud envisaged multiplying the
limit for webs of steel bridges; Evans suggested the serviceability limit
228
should ensure that web stresses are not allowed to exceed the critical
The loads at which snap buckling occurred with the RH test specimens
are given in Table 5.2, together with elastic buckling loads calculated
for the panels. Pg^ denotes the snap load on loading, that on
Pcrs denotes the elastic buckling load calculated on the basis of all
panel edges being simply supported, and Pcrc ^^e buckling load calcu
can be seen that for RH3, Pgg was very close to ?crs » whereas for t h e
girders. Against the introduction of such a limit is the fact that there
that snap buckling did not occur in either of the tests where longitudinal
web stiffening was incorporated, did not occur in cases where the web
slenderness ratio was 150 or below, and did not occur in either of the
tests where the panel aspect ratio was 1.0 (as opposed to 1.5). These
U 00 o\ vO -Cf CO 00 00 CN vO CN If) co
X u CN vO VO 0 0 CN if) 1—4 «d- CJv CN 1—4 CN
(U u
PU PU rH f™4 p—4 CN co <N CN 1-4 CN f-4 CN CN CN
3 Z 00 co OV 00 CN 00 «St «d- ov CO 0 0
PU CT> 00 a\ X if)
1—4
If) VO 0 0 00 00
I—H p-4 r4
I
o PU
u
00
00 co
r—4
co
CN
10
'd’
If)
0
CO
1—4
co
0
co
1-4
co
in
P~
If)
00
1
0
co
UO
CO
UO
s
CO
4 z st CO VO 0 CN CN CN 0 1—4 P- p~ r—4 f—4
j 2 CO If) UO CN CN ^4 ^4 If) CN CN CN CN CN
PU
I
ë VO 0
.0 2 r4
CO 1 1 1 1 1 1
I
CO 9 1 1 1 1 1
PU UO if)
0
CJ)
M 0
(d '7 f—4 1—4 if) 00
R co 9 1 1 1 vO co 1 1 «et 1 1 ■ 1 1
PU r—4
I CN
CO 0 0 0 0 0 0 0 0 0 0 0 0
CO If) If) 0 0 *cf if) 0 0 0 0 0
o 1—4 CN CN CN CN r4 CN CN CN CN CN
CN
m X OV 00 0 CO 00 CN •d- 00 OV OV
1—4 1—H i“H 0 I—4 f—4 f-4 0 1-4 p—4 1— 4 0 0
0)
r— I
0 0 0 0 0 0 0 0 0 0 0 0 0
g
CN 0 0 CO If) 00 If) •et 00 co 00 CO CO
u 0 1 0 LO 00 vO 00 0 00 0 00 00
r—4 p-H r-H 1—4 1—4 ^4 1—4
'd’
co 1
—* f—4 CO 1—4 VO Ov VO 00 co 00 co co
0 0 CO vO «cl- Ov 0 VO co vO co co
M <" If) o\ Ov OV 00 OV vO Ov vO ov vO ov (OV
cw 0
M i-H X r-H CO 0 CO 00 CO 00 co 00 co CO
X co co co
M
<ü < PQ
UO r-> 0
0 ov
0 r—4
1-4
CN
f-4
4_» t3 CN CN co - vO
CO M
0) -H
H CJJ
230
girder produces twisting which tends to move the line of load applica
tion towards the edge of the rail nearer the centre of the box; this
produces a turning moment on the top edge of the web which acts so as
outward direction.
Buckling Loads
was not possible because factors such as the presence of snap buckling
Theoretical elastic buckling loads for the test panels have been calcu
the buckling loads was firstly to evaluate the Ifj;- terms given in
Table 5.2 - the same bf dimensions (see Fig.5.3(a) and Table 5.1)
were used for this as were used to evaluate the Mf parameters for the
then evaluated using Eqn.(2.4b) - the expression for girders with welded
rails - and this enabled the patch length parameters B = c/b , also
lie in a range from 0.08 to 0.18 which is within the range of values
box-girder web panels. Having the 6 values and knowing the panel
Figs.3.2 and 3.4 (note that in these figures the patch length parameters
The buckling loads presented for RHll and RH12 ignore the presence of
^crc exceeds Pcrs tiy a factor of between 2.5 and 2.6 for panels with
a = 1.5 , and b y a factor of between 2.1 and 2.2 for panels with
a = 1.0 (see Table 5.1 for aspect ratios of panels). As can be seen
from Table 5.2, the experimental collapse loads Pg^ exceeded the Pcrs
loads by a factor of between approximately 3.1 and 7.5, and exceeded the
Pcrc loads by a factor of between approximately 1.2 and 3.0; the largest
factors occurred with RH4, the girder with a particularly heavy rail.
Thus, in all cases, the collapse load exceeded both calculated elastic
buckling loads.
Owing to the number of variables and limited number of tests in the pres
indeed any exist. The difference between these two loads may be con
the ratios Pex/^Crs ^ex/^crc RH7, RH3, and RH5 then indicates
or - since the girders have a common web depth - that the reserve of
load Pq at wh ich the onset of web yielding directly beneath the load
takes place. This load, which ignores any effects of web buckling, is
given simply by
Pg ~ ^W^^W^ (5.1)
where c represents the distributed length of the patch load at the web
edge (see Section 5.6.4 for procedure for calculating c). The values
Table 5.2.
The ratios Pgx/^o lie in the range 1.0 - 2.1. The lowest value
occurred with RH4, the girder with a heavy rail, and the highest value
with RH7, the girder with a relatively thick web of fairly low yield
stress. Comparing the ratios for RH3, RH8 and RH4, and for RH9 and RHIO,
size - see Table 5.1). Thus Pq falls below Pg^ for girders with
light flanges, but approaches (and may possibly exceed) Pg^ for girders
loads.
It can be seen from Table 5.1 that RH3, RH4, and RH8 differed signifi
and the Mf values for these two sets of girders. It is evident from
Comparison of the collapse loads for RH3 and RH9 and for RH8 and RHIO
in Fig.5.29 or Table 5.1 shows that a reduction in the web panel aspect
112 kN; RH8 (a = 1.5) collapsed at 128 kN and RHIO (similar except
160
150
140
130 RHIO,
kN RH8
120
RH9
110
RH3
100
0 1 2 3 4 5 6 7 8 9 10 11
160
150 RH7
140
130
Rex
kN 110
RH3
100
RH5
Mw N
Fig. 5. 30. EFFECT OF WEB PLASTIC MOMENT Mw ON COLLAPSE LQ4D
235
It will be recalled that RHll and RH12 resembled RH3 except for the
divides a panel u p into two distinct sub-panels, such that the upper sub
panel only is acted upon by the patch load, then these results may be
It is pertinent here to quickly review some of the previous work that has
loaded plate girders. Rockey, Samuelsson and Wennerstrom [64] have used
the compression zone of the plate. For the particular case considered
the panel edge, the buckling load was maximised when the stiffener was
Following this work, Rockey, Bergfelt and Larsson [65] carried out
increased the collapse load by 7.5% in one case and 18% in the other.
by Rockey, Evans and Porter [66], some more-recent findings are summ
tively weak. For girders with stiffer flanges, however, the flange
aspect ratio of 1.0 the ultimate load was increased by 42.5%, for
another with an aspect ratio of 2.0 the ultimate load was increased by
located within the buckled region that forms in a restricted zone close
On a more general note, Rockey, Evans and Porter add that by employing
The present findings support this comment in so much as the snap buckling
exhibited by RH3 was not exhibited also by RHll and RH12 (girders
237
different size, made from materials of the same yield stress and sub
Reference to Table 5.1 shows that RH2A and RH2B were approximately
half as large again as RHl; of the two, RH2B was slightly nearer a 3/2
241 kN but it should be noted that both the web and flange thicknesses
of RH2B were less than 3/2 times the corresponding RHl values, and
lower.
5mm
O
O
vO
RH4 RH5
5 mm
o
VD
241
Plate 1 shows RHlb after collapse; the residual profile of the vertical
w e b profiles in F i g . 5.31 (for RH3, RH4 and RH5), and the web profiles
in Fig.5.33 (for RH6, RH8, RH9 and RHIO), were recorded after the girders
residual rail profile for RH3 in Fig.5.34, and the rail profiles in
Fig.5.35 (for RH6, RH8, RH9 and RHIO), were also recorded after the gir
After the residual web profiles given in Fig.5.31 for RH4 and RH5 had
again; these profiles are shown in Fig.5.32. The residual rail profiles
given in Fig.5.34 for RH4 and RH5 were recorded after the additional d e
formations had been applied. Plate 4 shows RH5 after having undergone
C3 to
•
— o
Q CL
E
E
o
CD
C7S
CC
O
00
Q
00
cr
m
LTI
CT»
243
en
CL •—
un
__i
<
cr
u_
o
00
o
6 q:
CL
E
o
o
cr\
Q
00
LU
cr
m
m
LO
en
CTn
JI □c
cr cr
vO co
z: %:
ÛC c
e
4--;:\. - ) ^ ^
•''''.2’ ^ V" ■•• •'.'^:-_
RHlb
Plate 1
Plate 2
RH5
Plate 3
t
RH5
Plate U
I
«
RH6
Plate 5
Plate 6
f
Plate 7
Plate 8
m
rl
Plate 9
Plate 10
254
collapsed girders and is the mechanism which formed the basis of their
upper part of the web, and hinges develop in the rail and flange. Not
opposed to the two central hinges shown in Fig.4.2 for a girder subjected
Plates 1 - 1 0 shows, however, that the three hinge lines in the RH webs
formed such that the spacing between the upper and middle lines (h^)
was generally less than that between the middle and lower lines (h2 ) -
several of the tests reported in Refs{56], [58], [62] and [65] on con
(as opposed to equal) spacing became lost as the analysis was developed.
It should also be noted that the web profiles in Figs.5.31 - 5.33 show
the webs which tends to maintain the web edge perpendicular to the flange.
The residual rail profiles given in Figs.5.34 and 5.35 illustrate the
Comparing the profile for RH5 with that for RH3, and the profile for
RH6 with that for RH8 , shows the distance between the two outer hinges
dual profiles for RH3, RH8 and RH4, and for RH9 and RHIO, shows the
255
distance between the two outer hinges to increase as the rail size
increases. It can also be seen that for the girders with the larger
rails, rail hinge formation is often rather poorly defined with gradual,
case of RH4 (Fig.5.34) with its particularly heavy rail, formation of the
outer hinges is very poorly defined and smooth rail flexure extending
trates that, associated with this, the RH4 web mechanism extended over
the full length of the panel and further spread was constrained by the
vertical diaphragms. Comparing Plates 7 and 8 (for RH9 and RHIO res
pectively) shows that the relatively larger rail of RHIO caused the
outer flange hinges of RHIO to form nearer to the diaphragms and led the
Although the mode of failure of all the RH girders was broadly similar,
beyond the value at which the maximum load occurred or beyond the value
not until the girders had been subjected to fairly extensive post
became discernible.
The point is illustrated by comparing the residual web profiles for RH4
deformation) with the profiles given for the same two girders in Fig.5.32
RH4, it can be seen that evidence of hinge line formation (web rotation)
is the delayed formation of the lower hinge line in the case of RH5.
RH5 exhibited no lower web hinge line but instead displayed a smooth,
position (Fig.5.31 shows this to be the case for RH3 also). Comparing
than those shown in Figs.4.2 and 4.5 since these mechanisms appear to be
collapse loading.
The collapse load results presented in Table 5.1 enable empirical solu
assumed that the collapse load depends on the plastic moment Mf (of
the flange for a conventional girder and ofthe rail and flange
assembly for a crane girder), the web yieldstress ,the web thick
X U
Pua ^f “
^w ^w ^
2
Mf a af bf tf
257
Ç X C 2C ,P
Pua ® cjf a-v °f tf tv d
for structural steels is small. Also neglected is the ratio b/d, but b/d
for crane girders varies typically only between 1.0 and 1.5 and in
; + X = 1
Ç -2x + u + p = 0
If c = 1/3
then X = 2/3
and n + p = 1
and
Pua ^ ^w
. 2'4
258
2
The choice of powers results in a solution giving a (length)
2V
?ua = 5.5 t^(Mfa^) ' (5.2)
The predicted collapse loads for the RH test boxes obtained using
Eqn.(5.2) are given in Table 5.3. As can be seen, the worst prediction
is a 25% underestimation in the case of RH7. The predictions for the two
RH12 and 12% for RHll. For the remainder of the girders, the pre
few are on girders broadly similar to the RH specimens. Girders Bl, B3,
B5, B7 and B9 tested by Bergfelt [59] (Table B.2) bear a close resemblance
in that they were subjected to a patch load of zero length but, unlike
web panels of excessive aspect ratio. The predictions for these girders
using Eqn.(5.2) are also listed in Table 5.3. Generally good (and
with the experimental collapse load is achieved with girders B7 and B9,
the girders which have the heaviest flanges of the five and which have
parable test where the patch loading length is very small compared with
the panel length is that on girder BR3, tested by Bagchi and Rockey [41],
listed in Table B.3. It can be seen from Table 5.3 that Eqn.(5.2) also
259
Table 5.3. PREDICTED COLLAPSE LOADS Pya FROM EMPIRICAL FORMULA: EQN.(5.2)
RH5 67 68 1.01
RH6 80 81 1.01
Bl 95 67 0.70
B3 105 80 0.77
B5 121 98 0.81
B7 126 108 0 . 8 6
BR3 89 82 0.92
260
5.8. Conclusions
panel buckling but snap buckling of the panels was frequently in evi
Snap buckling first occurred at loads generally fairly close to the panel
was provided in all cases by the panel buckling load Pcrs calculated
on the basis of all panel edges being simply supported. There is some
The collapse loads of the test girders exceeded the Pcrs loads by a
factor in the range 3.1 - 7.5, and exceeded the Pcrc loads by a factor
the load calculated for the onset of local web yielding - onset of local
either the web or flange is increased. The collapse load also increases
slightly as the panel aspect ratio is reduced from 1.5 to 1.0. Longitu
dinal web stiffening increases the collapse load but the effect of the
10% increase in the collapse load. The test results also suggest that
patch loading.
initiate collapse and it was not until the girders had undergone exten
strains exist within the mechanism zone of the girder and suggests that
collapse analysis than that indicated by the fully plastic residual pro
files. However, since the load against vertical displacement plots devel
mechanism solution of collapse loads should not prove too sensitive to the
A simple expression has been developed using the collapse load data and
with the aid of dimensional analysis which gives generally good collapse
load predictions for the RH test boxes and for conventional plate girders
CHAPTER 6
6.1. Introduction
are now presented to help fill the current void in design information
girders.
1) Local yielding
2) Elastic buckling
3) Ultimate load
and the maximum wheel load be restricted to the lowest limit implied
considered in turn.
suggested that the effective length of the patch load at the top edge
i/3
= 3.25 p -£ (6.1)
expression be used
. _rif + Ir ^
=
c = (6 .2 )
263
be estimated from
J _ rl-15 I f r i
c =
- 0.3 [ t J
I
c = 3.49 (6.3)
For an I-section girder, the full section of the flange may be consid
onable assumption for the effective flange width is therefore the full
flange projection on the outer side of the rail supporting web (since
line (equal to twice the distance from the centre-line to the nearer
the flange outstand is less than 12 tf. In many cases where the rail
the If term from Eqns.(6.1) and (6.2) which leads to still simpler
264
expressions for c .
The expressions for the patch length c enable the patch length para
following section) and also enable the maximum direct stress in the
On, = — (6.4)
ct
gives
° lir t
for the direct stress in a web where the rail is clipped and no underlay
In all cases, the term P for the wheel load should incorporate the
As regards the shear stress between the flange and web due to the wheel
load, draft DIN 4132 [30] gives the maximum shear stress ag^ as
This can be used to check the strength of weld required between the
consider the horizontal edges clamped and the sides simply supported,
= (6.9)
where
E t5
D =
1 2 (1 -u2 )
For a particular set of edge conditions and a given aspect ratio, the
from Eqns.(6.1) - (6.3) and the panel length b will be known, hence
given in Table 6.1. The formulae give K in terms of the panel aspect
of 0.125. The present work has indicated that for crane web panels.
266
patch length parameters are typically around 0.125, with a range 0.07 -
0.25, and that over this range the buckling coefficient changes only
slightly. The formulae in Table 6.1 thus give K values which are
Panel
Aspect
Ratio a
Elastic buckling interaction curves for patch loading with bending and
patch loading with shear are given by Khan and Johns [45] and by Rockey,
patch loading and bending than between patch loading and shear; the
Both Khan and Johns [45] and Rockey, El-gaaly and Bagchi [6 8 ] show the
with the patch loading length. Khan and Johns illustrate the effect
of aspect ratio on the interaction for aspect ratios between 1/3 and
Pb
^ = 1.0 (6 .10)
cr bcr
Eqn.(6.10) is given by
^bcr = % ^ (6.11)
d^t
268
and bending exists when the longitudinal edges are clamped than when
all edges are simply supported. Although then rather more conservative,
Khan and Johns [45], and Rockey, El-gaaly and Bagchi [6 8 ] show that,
shear is almost independent of the length of the patch load. Khan and
Johns show the dependence on aspect ratio for aspect ratios between 1/3
vative, fit for all the curves shown is given by the parabola
a = 1.0 ( 6 . 12)
scr
5.34 + 4^ (6.13)
°scr a2 d^t
For interaction between patch loading and shear, no results are given
by Bagchi and Rockey [69] for the case of a panel clamped along its
action than for a panel simply supported all round. This indicates
shear. Assuming this to be so, Eqn.(6.12) may be used for panels with
Three-way interaction curves are given by Khan and Johns [45] for
partial edge loading acting over 1/4 of thepanel length, and a panel
The curves again show there to be little dependence on the aspect ratio
further information, the curves for the square panel be used if a three-
collapse loads of plate girders under patch loading was studied and
cast doubt on the confidence with which the solution might be applied
From the results of the collapse tests reported in Chapter 5, and with
2 Vs
Pu ~ 5.5 (Mf (6.14)
(which represents the plastic moment of the rail and flange combination),
the same assumptions on the effective flange section may be made as were
an additional check could be made using the Roberts and Rockey solution
[53] (see Chapter 4) and the more conservative estimate then be adopted
Roberts and Rockey [53] have considered this loading combination and
proposed use of the following interaction formula which has been incor
= 1.0 (6.15a)
u w
which gives
Fb = Fu 1 -
(6.15b)
.^ 1 ]
[14] is
f°bi
= 1.0
■u w
but since this expression was derived from experimental results on thin-
L.8 1
rp, '1 /O \
[Fb = 1 . 0 (6.16)
IP^J .^su
required.
It may be helpful to the crane designer if the present findings are now
the model box sections reported in Chapter 5 back to the full-size cases
Of the model boxes, RH5 and RH8 serve as two appropriate examples since
known to the author. These girders are employed in cranes with working
full-size cranes be known as FS25 and FS50. The crabs for these two
This gives a wheel load of 72 kN for FS25 and 147 kN for FS50 when a
272
class 1 crane impact factor of 1.1 is used (as given by BS 2573/1 [2]).
Ultimate Load
The wheel load that would cause collapse of the underlying region of
load given in Table 5.1 for the corresponding half scale model girder
(see Section 5.6.10). This gives a collapse load of 268 kN for FS25
yield stresses of the steels in the full-size cranes were comparable with
It has been shown in Chapter 5, however, that such stiffeners have rela
ratio between maximum wheel load and estimated collapse load is 0.27
that this level of patch loading has relatively little effect on the
Local Yielding
Table 5.2 shows the patch length c at the web to be 8 8 mm for RH5 and
108mm for RH8 . The patch lengths for the full-size girders would be
expected to be twice the size, that is 176mm for FS25 and 216mm for
The maximum direct stress can be estimated from Eqn.(6.4); this gives
82 N/mm^ for FS25 (5 mm thick web) and 113 N/mm^ for FS50 (6 mm thick web).
from Eqn.(6.7). Both bearing stresses are well below the permissible
Elastic Buckling
check of elastic web panel buckling to be made for the two chosen examples,
The stiffener divides the panel into an upper and a lower sub-panel;
the patch load acts on to the upper sub-panel but this panel is no longer
and FS50, the buckling load could readily be estimated from Eqn.(6.9).
176mm for FS25 and 216mm for FS50. FS25 has an overall web depth d
of 1150 mm, d for FS50 is 1450 mm. If the web panel aspect ratios for
the two girders were 1.0, the patch length parameter c/b would be 0.15
for both girders; if the web panel aspect ratios were 1.5, c/b would
Per = 20.1 K kN
Per = 27.6 K kN
for FS50.
If clamped horizontal edges are assumed, Table 6.1 gives K = 7.0 for
a = 1.0 , and the buckling load is given as 140.7 kN for FS25 and
193.2 kN for FS50; for a = 1.5 , K = 6.5 and ?er becomes 130.7 kN
If the same applied wheel loads were to act on to the unstiffened girders
as were estimated earlier (72 kN for FS25 and 147 kN for FS50) then the
51-55% for FS25, and 76-82% for FS50, depending on aspect ratio.
If simply supported panel edges are assumed instead, then buckling loads
lower than the applied loads are obtained: for a = 1.0 , Table 6.1 gives
K = 3.3, thus P^r = 66.3 kN for FS25 and P^,^ = 91.1 kN for FS50; for
a = 1.5 , K = 2.3 and P^.^. becomes 46.2 kN for FS25 and 63.5 kN for
FS50.
the web, approximate estimations of the upper sub-panel buckling load may
Per = 80.6 K kN
Per = 110.4 K kN
for FS50.
are simply supported or the top edge only is clamped and the remaining
edges simply supported. If the full panel were of aspect ratio 1.0
then the upper sub-panel will be of aspect ratio 4.0. From Table 6.1,
K is then 2.3 for simply supported edges and the buckling load is
given as 185.4 kN for FS25 and 253.9 kN for FS50; for a clamped top
edge, K = 5.9 and P^r becomes 459.4 kN for FS25 and 629.3 kN for FS50.
Buckling coefficients for panels with aspect ratios greater than 4.0
have not been investigated in the present work. However, Figs.3.2 - 3.5
show that for a narrow patch load, the buckling coefficient is nearly
constant for aspect ratios above about 1.5 and an analysis presented in
least the top edge clamped, and up to an aspect ratio of about 7.2 for
a panel with all edges simply supported. Thus, in the absence of more
relevant data, the coefficient values for a = 4.0 could be used within
the above limits to estimate buckling loads for panels with aspect ratios
above 4.0.
for this form of longitudinal loading are given in DIN 4114 [11] for
276
6 .6 . Concluding Remarks
The indications are that wheel loadings on present day cranes are
CHAPTER 7
wheel load which spreads through a rail mounted on the girder flange
directly above one of the webs. This form of loading also arises at
where purlins transmit loads to principal frame members and during the
Owing to its origins, this study has focused attention on the patch load
sions are therefore specific to this situation while others apply to patch
The main conclusions, of both a specific and a general nature, will now be
The girder was loaded through various rails and other loading interfaces
above the web by a wheel load. The profiles were found to be consistent
27 8
in that the total force on the web plate deduced from the measured
stresses was, in each case, very close indeed to the applied wheel load.
tensity of the true distribution and whose overall magnitude is the same
as that of the true loading (the length over which the equivalent uniform
patch load acts being taken as the effective length of the true patch
load).
tions by this method were not well predicted by the linear fan-spread
the second moment of area of the rail and flange and the web thickness.
featured in any major design code, although two draft codes have contained
give expressions for estimating the maximum direct stress in the web
It has been shown that the patch loads to which crane web panels are sub
patch load centrally disposed on the upper horizontal edge and supported
0.125 to 1.0 times the plate length have been investigated for plates
with aspect ratios between 1/3 and 4.0. Four sets of plate boundary
supported and clamped edges. The coefficient results have been presented
condition along the loaded edge which has greatest influence on the
buckling load. For long plates with at least their top edge clamped,
a local mode of buckling under the patch load to a mode produced essen
The computer program has also been used to determine buckling coefficients
edges attached to the flanges as clamped, with the vertical edges simply
Although the short patch loading lengths to which crane web panels are
subjected mean the buckling load is not very sensitive to the precise
patch length, the buckling load does vary significantly depending on the
A study has been presented for predicing the collapse load of patch
loaded plate girders. The study was based on an upper bound plastic
has previously been shown to give good collapse load predictions for a
analysis has been developed here which reveals the transition region
from collapse initiated by direct web yielding for girders with stocky
The ultimate load study also showed that certain of the simplifying
the method to include the omitted terms has shown that the need for an
(as is done for the other length parameter of the mechanism). However,
complex solution, and to ensure that the derived expressions for the two
281
In several of the tests, snap buckling of the loaded web panel occurred.
This was usually at loads fairly close to the panel buckling load calcu
lated on the basis of clamped horizontal panel edges, but on one occasion
was at a load well below this buckling load. Snap buckling occurrences
were in all cases at loads greater than the panel buckling load calculated
factors which influence snap buckling of web panels under wheel loadings
Collapse loads of the test girders increased as the rail size or web
The measured collapse loads exceeded the calculated elastic buckling loads
panel edges were exceeded by a factor between 1.2 and 3.0; the buckling
relationship between the measured collapse loads and the loads estimated
The mechanism of collapse of the test box-girders was similar in all cases
and resembled that which has been observed in conventional plate girders
and hinge lines in the web. However, in many cases, the full plastic
mechanism did not develop until the girders had undergone considerable
the type developed by Puthli, Supple and Crisfield [71] (using the finite-
loading.
From the present collapse results, a simple expression has been developed
girders .
283
APPENDIX A
ASSIGN CONSTANTS
Ac =
Bq = 2 A(,
Cc = 9c2/Bc
Dq = 3c2 /4A(.
Er ~ ________
INCREMENT Z
YES
1 r
SOLVE POLYNOMIAL
F.yPRF.SSTON TN V
1r
WORKING
NO
V VALUE
FOUND ?
/2MfV(Z + 2V)
J= ^ t ^(V +(1/12)]
c + j(l +
12 V2
284
GENERATE POLYNOMIAL
COEFFICIENTS FOR V < 0.5
F(l) = Dc
F(2) = 2(ZDc-l) IF c = 0
F(3) = Dc(Z2+0.25) - 52
F(4) = Z(Dc-4Z) + 3
F(5) = Z(4.5-Z:)+0.25Dc(3Z=-1.25 SET F(l) F(2)
F(6 ) = 0.5(3Z=-0.75ZDc-2.25) F(2) F(3)
F(7) = 0.1875(0.25Dc ~3Z) ETC.
F(7) = 0
SOLVE POLYNOMIAL
EXPRESSION IN V
WORKING
V VALUE
FOUND?
MfCZ + 2V)
a^t^CCV V3) + (I/12);
4Mf 2a,.,t
Pu = T " z
NO
CHECK UNIQUE P^
FOUND FOR EACH Z
1 r YES
WARN
PLOT Pu/Pex AGAINST z|
APPENDIX B 285
1
b d t c a a Mf
w V w r ^ex
]irdei
mm mm mm mm mm mm N/mmZ N/mmZ x10&Nmm kN
TGI 1000 1000 2.50 160 5.50 100 298 342 0.414 51.50
TG2 M II II II
200 10.09 299 253 1.288 63.76
TG6 2000 1000 3.00 160 6.29 100 290 294 0.465 81.90
TG11 2000 1000 3.00 160 6.29 200 290 294 0.465 93.19
SIG7&8 1000 500 2.00 50 4.97 100 ■ 192 294 0.091 35.56
Table B.2 TEST DATA OF BERGFELT [39] AND BERGFELT AND HOVIK [60], [61]
Girder b d t c
w “« 2
mm mm mm mm mm mm 2
N/mm N/mm xIO Nmm kN
B1 2400 700 3.26 150 6.10 0 326 347 0.484 95.16
B11 2400 300 2.00 100 6.00 0 294 294 0.265 57.88
821 9800 700 3.40 250 10.00 0 280 280 1.750 89.68
- --
b d t c a
w P
Girder mm mm mm mm mm mm ^2 "^2 ex
N/mm N/mm
XIO Nmm kN
TTG1 300 300 3.97 49.4 10.00 30 285 269 0.332 130.00
TTG4 450 450 3.97 49.3 10.00 45 257 267 0.329 120.00
TTG7 600 600 3.57 50.5 10.00 60 257 274 0.346 140.00
TTG1 ' 300 300 3.97 49.4 10.00 45 285 269 0.332 150.00
TTG4' 450 450 3.97 49.3 10.00 60 257 267 0.329 136.00
TTG7' 600 600 3.57 50.5 10.00 30 257 274. 0.346.... 119.00 .
BR1 660 635 3.25 152 12.70 75 • 250 250 1.532 141.20
864 II II II II 50 II II II 124.00
BR2
II II II II 50 II II II 89.40
BR3 1270
288
APPENDIX C
Table C.3 Z VALUES FOR TTG AND Table C.4 Z VALUES FOR RH
BR GIRDERS GIRDERS
Z Value Z Value
Girder Girder
Pex Pex Pex 1'0 Pex ^^ Pex 0.9 Pex
TTG4' 0.23
TTG6' - - 0.32
TTG7'
TTG8' - —
APPENDIX D
the base unit is shown in detail in Fig. D.2, and details of the remaining
The test boxes were seated in the frame on a pair of beams which spanned
the width of the base unit; onto the top of each support beam was mounted
a half-round bar. The beams were independently movable along the base
ram was mounted centrally on the underside of the top cross-members and
the test boxes were positioned eccentrically in the frame such that the
The main bending stresses in the frame were thus carried by the two
longitudinal base members, with the two vertical members being tension
members. The frame was designed to BS449 specification [9] for aload
The two longitudinal members were fixed relative to each other by two
cross-members, one at either end of the base unit. To ensure the full
2) Vertical Member
Tensile stress p. = P^
2A
3) Top Cross-Member
M = P X L where L = 920 mm
2 4
4 X 203.2 X 8.9
4 X 303.8 X 7.2
t = web thickness
W = 200 kN
V3d = /3 X 239.2 = 41
t 10.2
p^ = 138 N/mm^
W = 200 kN
B = 304 (Minimum)
W = 200 kN
/3d = /3 X264.6 = 64
t 7.2
p = 1 2 2 N/mm^
B = 304 (minimum)
6) Bolted Connection
Load/bolt = 200 kN = 25 kN
8
Minimum bearing plane thickness = 7.1 mm
2
Bearing stress = 25000 = 147 N/mm
24x7.1 2
Shear stress in bolt = 2500 x 4 = 55 N/mm
7t X 24^
= 0.45 X 1 X 144 kN
1.4
Load/bolt = 25 kN.
Q
ch
»
0
oo«
i
6
n
li
i
;
ii
iiii;
3® » « S ?S® f
I
' 1 ^ : w
Q
iiÏs ch
III IT
■fi
II
tr 2i
“5n 5 â , x ? 5n. W -
ill
jlil
in: SI#
rr l.i,
r«
- +
4/
#
«
-
;t '
-1
t
+ pii
I '
- ' I
4
!*
II
i9
I
:±5
ifr
m
In a
li!
CJt
I
Jilfi
\
3 j
c
Iliwmujvi-iui j
)L, I late-Jruoi^
!i!
| 5in yi
iiiii H
---
HI M
M
III
Tl
-j'j ill
~7W-0*TI 1*f
- oȍ -
IlH
298
APPENDIX E
gd
0 1 2 3 4 5 6 7 8 9 10 11 12 20
A
jg d
B
C ■■■ ■
0 1 2 3 4 5 6 7 8 9 10 11 12
A 0 . 0 0 -0.3C -0.64 -0.81 -0.83 -0.78 -0.97 -0.75 -0.46 -0.38 -0.59 -0.53 -0.49
B -0.30-0.60 -1.76 -2.60 -2.87 -2.46 -1.62 -0.13 0.75 0.84 0.29 -0.32 -0.61
C -0.62 -1.27 -2.89 -4.28 -4.70 -3.86 -2.09 0.48 2.18 2.44 1.55 0.32 -0.54
D -0.94 -1.96 -3.60 -5.05 -5.50 -4.36 -2 .Of 0.96 2.98 3.42 2.44 0.82 -0.47
E -1.27 -1.94 -3.52 -4.94 -5.40 -4.18 -1.84 1.14 3.09 3.56 2.60 0.92 -0.47
F -1.37 -1 . 6 8 -2.89 -4.14 -4.58 -3.51 -1.51 1.03 2.59 2.91 2 . 0 2 0.60 -0.51
G -1.64 -1.44 -2.13 -2.95 -3.31 -2.5S -1.2S 0.45 1.39 1.53 0.90 0.09 -0.59
POINTS ; gd = 100
0 2 4 6 8 10 12
1 2 3 4 5 6 7 8 9 10 11
A 0 .0 -0 .5 -1 .0 -0 ;5 -0 .5 -0 .5 0 .0 0 .5 1 .0 1 .0 0.5
B -0 .5 -2 .5 -4 .0 -4 .5 -4 .0 -3 .0 - 0 .5 2 .0 3 .0 3 .0 1.5
C -1 .5 - 4 .5 - 7 .5 —8 .0 -6 .5 -4 .5 0 4 .0 6 .0 5 .5 3 .0
D - 3 .0 -6 .5 - 9 .5 -1 0 .0 -8 .5 -5 .0 0 .5 5 .0 7.5 6 .5 3.5
E -3 .0 - 7 .0 -1 0 .5 -1 0 .0 —8 .0 -4 .5 1 .0 6 .0 8 .0 7 .0 3.5
F -2 .5 —6 .0 - 8 .0 -8 .5 - 7 .0 -4 .0 1 .5 6 .0 7 .5 6.5 3.5
G -1 .5 -4 .0 -5 .5 -5 .5 -4 .5 -2 .5 1.5 5.5 6 .0 5 .0 2 .5
1 2 3 4 5 6 7 8 9 10 11
A 0 .0 0 .0 0 .0 0 .0 0 .5 1 .0 0 .5 0 .0 0 .0 - 0 .5 - 0 .5
B 0 .0 0 .5 1 .0 1.5 1.5 1 .0 0 -1 .0 - 1 .0 -1 .0 - 0 .5
D 2 .0 3 .5 4 .0 4 .0 3 .0 1.0 - 1 .5 - 2 .0 -3 .5 -2 .5 - 1 .0
G 1.5 3 .0 3 .0 2 .0 1 .0 - 1 .0 - 2 .5 -3 .5 -3 .0 -2 .0 - 1 .0
301
0 1 2 3 4 5 6 7 8 9 10 11 12
B 0.68 0.70 1.34 1.94 2.23 2.20 1.01 - 0.23 - 1.15 -1 .0 2 - 0.63 - 2.13 - 5.43
C 0.54 1.77 3.24 4.16 4.33 3.61 1.54 -0 .8 7 - 2.51 - 2.52 -1 .8 6 - 2.35 -4 .0 0
D 0.48 2.43 4.32 5.37 5.41 4.30 1.65 - 1.38 - 3.58 -4 .1 2 - 3.70 - 3.05 -2 .6 2
E 0.57 2.33 4.15 5.30 5.37 4.22 1.53 -1 .6 4 -4 .1 2 - 4.90 -4 .1 9 - 2.38 - 0.98
F 0.60 1.58 3.09 4.25 4.51 3.61 1.26 - 1.67 -4 .0 0 - 4.56 -3 .3 5 -1 .4 0 - 0.56
Table E . 6 RH4: gd = 75
0 1 2 3 4 5 6 7 8 9 10 11 12
F 0.05 - 1.08 - 3.05 - 5.05 -6 .2 0 -6 .4 4 -5 .4 2 - 2.65 1.04 3.59 3.63 2.00 1.00
G 0.42 0.28 - 1.16 - 2.85 -3 .7 7 - 4.25 -3 .7 7 - 1.90 0 .85 2.80 3.02 2.15 1.89
302
0 1 2 3 4 5 6 7 8 9 10 11 12
B 0.09 0.23 -0.1 7 -0 .8 4 -l.lt -0 .6 9 0.51 1.48 1.67 1.65 1.14 0.35. . -0 .1 6
C 0.05 0.16 —0.68 -1 .9 0 -2.41 -1 .4 6 0.57 2.39 2.98 3.02 2.07 0.70 -0 .2 6
G 0.13 1.05 0.45 -0 .9 4 -1.55 -0 .6 4 0.94 2.28 3.04 3.17 2.69 1.69 -0 .3 4
Table E . 8 RH6 ; gd = 75
0 1 2 3 4 5 6 7 8 9 10 11 12
A 0.00 0.08 0.42 0.64 0.78 0.62 0.55 0.55 0.62 0.50 0.10 -0 .1 0 -0 .1 5
B 0.38 0.27 0.75 1.58 2.21 2.20 1.52 0 .8f 0.29 -0 .4 8 -1 .1 9 - 0.80 —0 .l8
0 1 2 3 4 5 6 7 8 9 10 11 12
A 0.00 0.08 0.30 0.70 0.88 0.82 0.68 0.65 0.50 0.45 0.32 0.13 -0 .1 6
B -0 .0 1 0.17 0.93 1.69 2.01 1.56 0.54 -0 .4 7 -1 .1 9 - 1.43 - 1.29 - 0.89 - 0.74
C -0.0^ 0.81 2.05 3.19 3.47 2.63 0.79 -1.3C -2 .7 7 -3 .2 1 - 2.74 - 1.72 -0.9 2
D -0.2C 0.84 2.52 3.93 4.30 3.23 0.88 -1.77 -3 .7 3 - 4.31 - 3.72 - 2.38 - 1.31
E -0 .1 4 0.98 2.7C 4.14 4.50 3.42 1.09 - 1.71 -3 .7 7 -4 .4 1 - 3.76 - 2.23 - 0.85
F -0 .1 4 0.97 2 .4 i 3.67 4.02 3.18 1.20 -1.26 - 3.08 - 3.58 - 2.85 -1 .3 9 - 0.23
0 1 2 3 4 5 6 7 8 9 10 11 12
A 0.00 - 0.85 - 0.90 —0.60 -0 .4 4 -0 .4 6 - 0.23 0.20 0.66 0.78 0.77 0.96 0.81
C - 0.03 - 1.80 -3 .6 4 - 3.49 -1 .1 0 2.42 5.12 6.62 7.18 7.14 6.20 4.31 2.01
D 0.07 -2 .0 2 -4 .0 2 - 3.71 -0 .7 7 3.43 6.84 8.70 9.22 8.87 7.45 5.00 2.30
E 0.15 - 1.85 - 3.67 -3 .1 2 0.09 4.50 7.93 9.73 9.98 9.25 7.48 4.94 2.45
F 0.12 -1 .3 3 -2 .7 7 -2 .0 1 1.27 5.53 8.52 '9.79 9.62 8.51 6.56 4.33 2.51
G 0.30 -0 .2 2 -1 .0 2 -0 .1 0 2.80 6.22 8.21 8.73 8.18 6.91 5.10 3.52 2.76
304
2 3 4 5 6 7 8 9 10
2 3 4 5 6 7 8 9 10
0 1 2 3 4 5 6 7 8 9 10 11 12
A 0.00 -0 .0 1 0.33 0.78 0.56 0.31 0.39 0.61 0.58 0.16 - 0.07 0.16 0.16
B 0.16 0.30 0.66 1.08 1.13 1.06 1.11 1.33 1.14 0.73 0.48 0.38 0.22
C 0.05 0.27 0.62 1.08 1.23 1.37 1.43 1.55 1.32 0.99 0.66 0.37 0.03
D 0.11 0.17 0.46 0.86 1.10 1.26 1.32 1.34 1.15 0.92 0.64 0.29 -0.12
E 0.15 0.17 0.50 0.92 1.17 1.13 1.07 1.06 0.90 0.85 0.58 0.16 - 0.36
F 0.10 0.22 0.77 1.42 1.59 1.47 1.21 1.14 0.99 1.01 0.87 0.39 -0.42
G 0.15 0.15 0.91 1.78 1.87 1.62 1.21 1.15 1.04 0.96 0.88 0.40 -0.57
0 1 2 3 4 5 6 7 8 9 10 11 12
A 0.00 -0 .1 7 0.23 0.60 0.14 0.16 0.32 0.17 0.20 0.44 0.37 0.35 0.22
B - 0.05 0.08 0.25 0.36 0.17 0.09 - 0.03 -0 .0 4 -0 .0 4 0.17 0.28 0.32 0.12
C 0.06 0.14 0.21 0.26 0.12 -0.10 -0 .4 0 -0 .4 0 -0 .4 6 - 0.31 -0.08 0.03 - 0.27
D —0.18 0.39 0.96 1.25 1.06 0.46 -0 .4 1 - 1.29 -1 .8 6 -1 .7 0 - 1.06 - 0.34 - 0.26
E 0.03 0.93 1.92 2.46 2.25 1.36 -0 .1 2 - 1.58 - 2.57 - 2.52 - 1.62 - 0.58 - 0.05
F 0.02 1.06 2.19 2.90 ' 2.68 1.73 0.08 - 1.52 - 2.59 - 2.63 -1.77 -0 .7 5 - 0.17
G 0.06 0.88 1.70 2.35 2.32 1.58 0.20 - 1.05 - 1.92 -2 .0 2 - 1.45 - 0.85 - 0.59
305a
APPENDIX F
gives good collapse load predictions for the RH girders and for
very narrow patch loads, equally good predictions can be obtained from
0.5 r 1.5 n
w
P = 0.5 t ‘
u w tf/tw
306
REFERENCES
works." 1963.
[5] German Standard, DIN 15018, "Cranes, principles for steel structures,
[10] German Standard, DIN 4132, "Crane runways; steel structures; design
1952/3.
[12] Merrison, A.W., "Inquiry into the basis of design and method of erec
London, 1973.
307
[13] Senior, A.G., and Gurney, T.R., "The design and service life of
London, 1977.
pp. 417-467
[20] Senior, A.G., "New methods of mounting crane rails on overhead run
pp. 132-134.
girder type". Translated from Stahl und Eisen 79 (1959), No. 26,
Foundrymen.
[25] Mendel, G., "Uber die Querbiegung des Obergurts von Kranbahntragern"
[29] Timoshenko, S., and Goodier, J.N., "Theory of elasticity", 2nd Ed.,
[30] Draft German Standard, DIN 4132, "Crane-runway girders, design and
[32] Roark, R.J., and Young, W.C., "Formulas for stress and strain",
[34] Khan, M.Z., Johns, K.C., and Hayman, B., "Buckling of plates with
ASCE, Vol. 103, No. ST3, Proc. Paper 12798, March, 1977,
pp. 547-558.
[35] Rockey, K.C., and Bagchi, D.K., "Buckling of plate girder webs under
ally stiffened web plates under local loading and under loading
from the main bearing member), Techn. Mitt. Krupp. Forsch. - Ber.,
[37] Roorda, J., "Some thoughts on the Southwell Plot", Journal of the
[38] Spencer, H.H., and Walker, A.C., "Critique of Southwell Plots with
[45] Khan, M.Z., and Johns, K.C., "Buckling of web plates under combined
[48] Bossert, T.W., and Ostapenko, A., "Buckling and ultimate loads for
plate girder web plates under edge loading". Report No. 319.1,
Division, ASCE, Vol. 8 8 , No. EM5, Proc. Paper 3297, Oct. 1962,
pp. 67-86.
[50] Scheer, J., and Nolke, H., "The background to the future German
[53] Roberts, T.M., and Rockey, K.C., "A mechanism solution for predicting
plane patch loading", Proc. Instn. Civ. Engrs, Part 2, Vol. 67,
[54] Roberts, T.M., and Rockey, K.C., "An assessment of the accuracy of
Press, 1979.
pp. 805-819.
[58] Skaloud, M., and Novak, P., "Post-buckled behaviour of webs under
[59] Bergfelt, A., "Studies and tests on slender plate girders without
[60] Bergfelt, A., and Hovik, J., "Thin-walled deep plate girders under
[61] Bergfelt, A., and Hovik, J., "Shear failure and local web
[62] Drdracky, M., and Novotny, R., "Partial edge load carrying capacity
and 6.43-6.44.
[64] Rockey, K.C., Samuelsson, A., and Wennerstrom, H., "The buckling
[65] Rockey, K.C., Bergfelt, A., and Larsson, L., "Behaviour of longi
Division, ASCE, Vol. 98, No. ST12, Proc. Paper 9409, Dec. 1972,
pp. 2739-2752.
313
[69] Bagchi, D.K., and Rockey, K.C., "Web plate buckling under combined
Section 3.