SUPPLEMENT TO THE WELDING JOURNAL, AUGUST 1973

Sponsored by the American Welding Society and the Welding Research Council

^ _ ^
l'R^J

Experiments on Castellated Steel Beams

Results of the study permit an evaluation of existing
design theory and provide specific limits for several
design factors
BY M. U. HOSAIN AND W. G. SPEIRS

ABSTRACT. This report briefly s u m marizes the results of experiments

performed on 12 simple castellated
steel beams. The objective of the investigation was to study the effect of
hole geometry on the mode of failure
and ultimate strength of such beams.
The effect of changes in the number
of panels on the performance of
beams having the same span and expansion ratio was investigated. An attempt was also made to study the
phenomenon of web buckling due to
compression and due to shear in the
framework of the existing approximate method of design.
The specimens were all fabricated
from 10B15 beams and were exp a n d e d to 1.5 t i m e s the o r i g i n a l
depth. With the exception of four
specimens, which were fabricated
from CSA G40.12 steel, all other
specimens were of ASTM A-36 steel.
The test results indicated that the
optimum hole geometry requires a
minimum length of the throat which
makes the beam less susceptible to
failure due to V i e r e n d e e l m e c h anism, i.e., formation of hinges at the
four re-entrant corners. Failure in
such a beam may be caused by a
"flexure mechanism," which is formed
due to yielding of the flanges in the
region of high bending moment, or by
the rupture of a welded joint due to
shear.
M. U. HOSAIN and W. G. SPEIRS are
Associate Professors in the Department of
Civil Engineering of, respectively, the University of Saskatchewan, Saskatoon, and
Nova Scotia Technical College, Halifax,
Canada.

The elastic behavior of the specimen was consistent with the results
p r e d i c t e d by the f i n i t e e l e m e n t
method.
Introduction
Scope
Castellated steel beams have been
the subject of considerable research
during the past decade. Most of the
investigations, however, were basical-

ly of analytical nature and involved the

study of the elastic behavior of such
beams. (Refs. 1-7). An approximate
statical elastic analysis of castellated
beams was reported, in 1957, by Altfillisch, Cooke and Toprac (Ref. 8).
This analysis, which is based on the
assumption that points of inflection
are located at the midpoints of the
members, is used extensively for design purposes. Toprac and Cooke
(Ref. 9) carried out an investigation of

Fig. 1 Failure by web shear

WELDING

RESEARCH SUPPLEMENT!

329-s

nine open-web beams expanded to

depths varying from 1.33 to 1.78 of
the original 8B10. They found that the
elastic stress patterns predicted by
the simple statical method are consistent with the observed values.
Since then, different researchers
have either supported (Ref. 10) or
contradicted (Ref. 11) the validity of
simple statical analysis. The most recent study (Ref. 12) concerning the
elastic behavior of castellated beams
was reported in July, 1971. The
authors presented an analysis based
on the theory of elasticity for two
separate loading cases: a moment
loading and a shear loading. Results
obtained by this method were compared with those calculated by the
simple statical method and those
determined experimentally. In most
cases the theory of elasticity solutions indicated close agreement with
the observed values which were considerably different from those calculated by simple elastic theory.
The elastic behavior of castellated
beams has been studied extensively.
This report discusses the results of 12
experiments performed to study the
post-elastic behavior of castellated

beams with emphasis on their failure

mechanism and ultimate strength.
Previous Investigation

Plastic or limit analysis had not

been applied to castellated beams
until recently. An experimental investigation carried out by Halleux
(Ref. 13) indicated the possibility of
failure caused by two different collapse mechanisms. The first failure
mechanism, normally referred to as
shear failure or Vierendeel mechanism, is caused by the formation of
plastic hinges at the tee sections
touching the four re-entrant corners
of a panel. This type of collapse generally occurs near a concentrated
load or a support where shear in the
beam makes a significant contribution to the formation of collapse
mechanism.
The second mechanism is the flexure mechanism in which the upper
throat section of the critical panel becomes completely plastic in compression at the hinge, while the lower
throat section behaves similarly in
tension. Halleux observed such
failure in castellated beams subjected to equal concentrated loads at

Table 1 Specimen Dimensions

d =,25
1 1

I
<

'

\-AA-

SPECIMEN

n,

in.

6.5

B-l
B-2

5.0

B-3

A-t.

j
m, in.

0.269"

w = C .23"

? de = 15" * *

A-l
A-2

M>

< m

SERIES

-r-

0
degrees
(approx)

SPAN
L, in.

23.0

45

138.0

23.0

45

69.0
63.0

15.75

60

15.75

60

63.0

15.75

60

94.5

G-l

1.75

15.0

41 O'

120.0

G-2

1.375

10.0

54 3'

120.0

G-3

1.125

7.5

62 18'

120.0

4.0

18.0

45

54.00

3.5

12.75

60

51 . 0 0

2.687

15.375

45

307.50

2 00

9.75

60

273.00

330-s I A U G U S T 1 9 7 3

the third points. The critical panel, in

all cases, was located at the middle
third of the span which was under
pure bending.
Halleux used the statical energy approach to predict the limit loads for
beams likely to fail by a Vierendeel
mechanism. However, he neglected
the effect of axial force and shear on
the plastic moment capacity of the tee
section. This limited the application of
his method to cases where the influence of these forces is small.
In 1968, R. G. Redwood (Ref. 14)
presented an ultimate strength design method developed for beams
with rectangular openings in the web
but applicable to castellated beams
after slight modifications. He considered three types of failure mechanisms: single hole failure, interaction failure including two or more
holes and failure through shear in the
web post. Having developed equations for each type of failure, he calculated the ultimate load necessary
for each failure mechanism to occur.
He assumed that the beam being
analyzed would fail through the mechanism which required the smallest ultimate load to form. Redwood compared the results obtained by his
method with those determined experimentally by Toprac and Cooke
(Ref. 9), Halleux (Ref. 13) and Sherbourne (Ref. 15). In most cases, discrepancies between experimental
and theoretical values were quite
large.
In 1969, Hope and Sheikh (Ref. 16)
presented an interaction method for
calculating the ultimate load of castellated beams which are expected to
fail by Vierendeel mechanism. These
authors were the first to point out that
when a four-hinge collapse mechanism is completely formed in a panel, the point of contraflexure is no
longer located at midspan of the
throat section as assumed in approximate elastic theory. Their
method, however, does not require a
knowledge of the location of the point
of contraflexure and essentially consists of plotting the interaction curve
for the tee section of the castellated
beam being analyzed. The interaction curve for the tee section is obtained by varying the location of the
neutral axis and plotting the resulting
values of moment and normal force.
The effect of shear stress on the plastic moment of the tee section was
neglected on the assumption that its
influence would be cancelled by the
influence of strain hardening on the
ultimate load capacity. Hope and
Sheikh compared the theoretical
failure loads calculated by this
method with the experimental values
reported by Altfillisch, Cooke and
Toprac (Ref. 8) and Toprac and
Cooke (Ref. 9). The agreement be-

HYDRAULIC

RAM

HYDRAULIC PRESSURE LINE

50-TON LOAD CELL

2 NOS STANDARD
CHANNELS
TEFLON BLOCK

STAINLESS STEEL
PLATE

STEEL
HOLDER

- TEST SPECIMEN

SUPPORT

ROLLER
TIE BAR

isS&S^S^^Sv

Fig. 2 -

Photograph of experimental setup and diagram showing lateral bracing system

tween theoretical and experimental

values was excellent in most cases.

Table 2 Material Properties

Length

Beam
No.

1
1
1
1
2
3
3
4

A-2
B-1
B-2
B-3
A-1
G-1
G-2
G-3

Objectives

Although considerable research

work has been done on castellated
steel beams, there is a lack of information concerning the following
topics.
1. Ultimate strength of castellated
beam on the basis of web weld failure.
2. The effect of hole geometry on
the mode of failure and ultimate
strength.
3. Optimization in elastic design. It
is generally kno n that a castellated
beam with shorter throat length,
which is normally equivalent to more
castellations in a given span, is likely
to result in a stiffer design against
failure due to Vierendeel mechanism. Subject to the practical limits
of fabrication, optimization in elastic
design for a given span and expansion ratio is expected to be achieved
by increasing the number of panels.
4. The problem of web buckling
due to compression and due to shear.
Experimental work concerning the
strength of castellated beams on the
basis of web weld failure (see Fig. 1)
has been completed and reported
(Ref. 17) by the authors.
The present investigation was undertaken to study the effect of hole
geometry on the mode of failure and
ultimate strength. The elastic optimization procedure has been studied experimentally by testing three
beams of same length and expanded

Yield stress,
F , ksi
Flange
Web

Ultimate stress,
Fu ksi
Web
Flange

Avg stress,
ksi
F
F

48.59

64.79

48.59

64.79

62.25

64.80

80.82

80.95

63.52

80.88

44.26
60.89

46.36
57.25

68.36
79.75

69.37
77.28

45.56
59.07

68.86
78.51

depth but with the number of panels

equal to 8, 12 and 16 respectively. An
attempt was also made to explore the
effect of changes in size of stiffeners
and loading conditions on the possibility of web buckling due to compression and due to shear.
Description of the Experimental
Program
Test Program

A summary of the geometric

properties of the 12 test specimens is
given in Table 1. The specimens have
been divided into four series based
on their expected mode of failure. The
five specimens listed in Series 1 were
expected to fail by the formation of
plastic hinges at the four re-entrant
corners of the critical panel. Three
test beams with shorter throat lengths
constitute Series 2. These specimens
were designed by balancing the
allowable flexural stress at the critical
section of the throat with the allowable shearing stress along the critical

welded section. The calculations were

based on an allowable stress of 0.6 Fy
for flexure, and 0.4 Fy for shear,
where Fy is the specified minimum
yield stress. These beams were tested
to investigate optimization in elastic
design by varying the number of
castellations in a given span.
As shown in Table 1, the span
length was held constant at 10 ft but
specimens G-1, G-2 and G-3 had
panel lengths, m, equal to 15.0, 10.0
and 7.5 in. respectively. The number
of panels was equal to 8,12 and 16 for
G-1, G-2 and G-3 respectively. Because of the method used to design
these specimens,
specific mode
of failure was not predictable.
The design of the two specimens C
and D, listed in Series 3, was similar to
that of specimens G-1, G-2 and G-3 of
Series 2. The only difference was that
partial stiffeners, 4 in. long, were
placed just below the top flange at
the load-points in specimens C and
D instead of full depth stiffeners as

W E L D I N G R E S E A R C H S U P P L E M E N T ! 331-s

another research p r o j e c t not reported in this paper.

The specimens were all fabricated
from 10B15 beams and were exp a n d e d to 1.5 times the original
depth. Specimen A-1 of Series 1 and
the three test beams of Series 2 were
fabricated from CSA-G40.12 (Ref. 18)
steel, which has a specified minimum
yield stress of 44 ksi for 10B15 rolled
section. All other specimens were of
ASTM-A-36 steel.
All specimens were simply supported. Specimens A-2 and B-1 of
Series 1 and C and D of Series 3 were
subjected to a two-point loading. For
these beams, lateral bracing was provided a t , t h e load points and at the
reaction points. The other eight specimens were tested under a c o n centrated load at midspan. For these
beams lateral bracing was provided at
midspan, at quarterspan and at the
reaction points.
Each s p e c i m e n had f u l l - d e p t h
stiffeners at the reaction points. Five
of the specimens, viz. A - 1 , B-3, G - 1 ,
G-2 and G-3, had full-depth stiffeners
at load points as well. Specimens C
and D had, as mentioned earlier, 4 in,
long stiffeners at the load points.
Three specimens (A-2, B-1 and B-2)
had no stiffeners at the load points.

Fig. 3 Failure by Vierendeel mechanism

Test Set-up and Instrumentation

Fig. 4 Sequence of yielding in Specimen A-1

used in Series 2. Specimens C and D

were designed to test the efficiency of
short stiffeners, and web buckling due
to c o m p r e s s i o n was e x p e c t e d to
cause failure.
332-s I A U G U S T

1 973

Series 4 consisted of two long

beams which were tested as simply
s u p p o r t e d beams well within the
elastic limit, before testing them as
continuous beams in connection with

Figure 2 shows a typical experimental setup which was used to test

the specimens of Series 2. A 32 ft long
30WF 108 beam was used as the test
b e d . The test s p e c i m e n s were
mounted on steel rollers and were
loaded by a hydraulic ram which was
operated against a loading head c o m posed of standard angles and cover
plates. The loading head was secured
to the test bed through two 1V4 in.
diam steel rods. Hydraulic pressure to
the ram was provided by an air-driven
pump, and regulated by adjustment
of the air pressure and hydraulic line
valves at a control panel. A Philip RR1 type load cell of 50 long-ton capacity was used to record the applied
loads.
Vertical centerline deflections were
recorded by Baty 0.001 in. dial gages,
and Baldwin SR-4 type strain gages
and AR-1 type rosettes were used to
measure strains. A B&F 161 minimultichannel digital strain indicator
was used to record the strains.
The experimental setup used for
the other beams was exactly similar to
the one described above, except that
a Gilmore 48 channel recorder was
used to plot the strains and a 50 kip
capacity load cell was used to record
the loads which were applied by a 100
kip capacity hand operated hydraulic
jack.
Figure 2 also shows the details of a
bracing unit which was used for all
tests. The teflon blocks against the
stainless steel plates allowed the

beam to deflect freely in the vertical

direction but prevented lateral movement effectively.
Testing Procedure
In the elastic range, loading was in
predetermined increments of the applied load. Increments of 4 kips were
normally used until the first sign of
yielding was noticed. After that, the
applied load was increased by 1 to 2
kips until the m a x i m u m load was
recorded. In the plastic range, the
load was applied to achieve predetermined rotation increments. Each
specimen was deformed well into the
unloading region.

LOCAL

BUCKLING

Material Properties
Table 2 summarizes the material
properties obtained from laboratory
tension tests on flat tensile specimens conforming to ASTM specification A370-69 (Ref. 19). Two coupons
each were cut from the top flange, the
bottom flange and the web for each
length. For length 1, however, all six
coupons were cut, inadvertently, from
the web.

END

ROTATION,

6,

DEGREES

Fig. 5 Load-end rotation curve for Specimen A-1

Results a n d Discussions
Structural Behavior
The structural behavior of the specimens is discussed on the basis of the
following observed modes of failure.
1. Failure by Vierendeel mechanism
2. Flexural failure
3. Failure due to instability
Failure by Vierendeel
Mechanism.
Three specimens ( A - 1 , A-2 and B-3)
exhibited this type of failure. These
specimens failed by the formation of
four plastic hinges at the re-entrant
corners of the panel adjacent to the
load point and in the part of the beam
where both shear and moment are
present. A typical mode of failure is
shown in Fig. 3.
Figure 4 shows the sequence of
yielding in specimen A - 1 . The loadend rotation relationship for the same
beam is presented in Fig. 5. When the
total load was about 32 kips first sign
of yielding was observed along the
line marked (1) in Fig. 4. As the load
was being raised to 34 kips, yielding
started first at the re-entrant corners
marked (2) and then along lines (3).
Yielding at these locations became
prominent as the load was gradually
increased and a well-defined Vierendeel m e c h a n i s m had o c c u r r e d
when the maximum load of 40.2 kips
was recorded.
The beam sustained the m a x i m u m
load only briefly. Web buckling occurred in the first panel on the right
hand side of the load and the beam
started to unload shortly thereafter.

Fig. 6 Mode of rupture in Specimen B-3

L a r g e p l a s t i c d e f o r m a t i o n s , as
recorded by the end rotations, were
observed and at an end rotation of 2
deg 57 min the beam could support
only 26 kips. At that point the applied
load was released in three installments. A total elastic recovery of 31
min was noted, with a final permanent end rotation of 2 deg 26 min.
The behavior of specimens A-2 and
B-3 was similar to that of A-1 up to the
attainment of the m a x i m u m load.
However, in these specimens unloading was not accompanied by web

buckling. Failure was caused entirely

by Vierendeel mechanism and culminated with the rupture of the two
critical re-entrant corners in tension
as shown in Fig. 6.
In another specimen (B-1), web
b u c k l i n g under the c o n c e n t r a t e d
loads caused unloading before the
Vierendeel m e c h a n i s m was fully
formed.
The load deflection curves for
specimens A-2, B-1 and B-3 are presented in Fig. 7.
The observed failure loads were

WELDING RESEARCH SUPPLEMENT!

333-s

Pu=

compared with those calculated by

Halleux's method (Ref. 13) and the int e r a c t i o n m e t h o d (Ref. 16). T h e
results are presented in Tables 3 and
5 respectively.
Halleux's method is based on the
kinematic theorem and assumes a
failure mechanism illustrated in Fig. 8.
The limit load, which is derived by
equating the internal work due to
rotation and the work done by the external forces, can be expressed by
Eq. (1).

52

"(limit) =

where Z t
Fy
n

0 4

0.2

DEFLECTION, $ INCHES
Fig. 7 Load-deflection curves: Specimens A-2, B-1 and B-3

SYMETRICAL
ABOUT

Plastic
Centroid

fy

(b)

Fig. 8 Halleux's shear failure mechanism

Table 3 Comparison of Experimental Ultimate Loads with Those Calculated by

Halleux's Method

Specimen
A-1
A-2"
B-1
B-3

Span, L,
in.
138.0
69.0
63.0
94.5

Loading

la)

P at L/2
P/2 at L/3
P/2 at L/4
P at L/2

Lin- it load, P , kips

ExperiHalleux
mental
54.52
41.70
54.27
54.21

37.5
41.4
48.5
41.8

Ult. load,
Pu,kips,
experimental
40.2
45.0
52.0
44.0

(a) All beams were simply supported. Concentrated loads (P= single load, P/2 =two equal loads) were located at points
measured from each end of the span.

334-s | A U G U S T

1 973

(8Z

Fy)/n

(1)

= plastic modulus of
the tee section
= yield stress of steel
= width of the
welded joint

In Table 3, P refers to the maximum load recorded

experimentally and P, refers to the limit load.
The theoretical value of P, was
calculated by Eq. (1) with F y =
63.52 ksi for specimen A-1 and Fy =
48.59 ksi for specimens A-2, B-1
and B-3. The experimental value of
P, was obtained from the load deflection curve by applying Halleux's tangent intersection method (Ref. 13).
This load represents "the load beyond
which deflections become large in
relation to purely elastic deflections".
It appears from Table 3 that the experimental and theoretical results for
Pi disagree considerably with the predicted values always on the unsafe
side. In discussing the results it must
be taken into consideration that Eq.
(1), adapted from Halleux, represents the upper bound of the actual
failure load and ignores the effect of
axial force and shear on the plastic
moment M p of the throat section.
Table 4 was prepared, on the basis of
the experimental value of P, , to find
out roughly if the effects of axial force
and shear can be ignored or not. It
appears that the vertical shearing
stress is far less than its yield value
and, therefore, the effect of shear on
M p may be i g n o r e d . However, it
seems that N 7 N P is more than 0.15 in
all four specimens and therefore the
axial force must be taken into account.
For specimens B-3 and A - 1 , the
values of N 7 N P are 0.766 and 0.767
respectively, and this accounts for the
wide disagreement between experimental and theoretical results for
these two specimens. In specimen
A-2, on the other hand, the effect of
axial force is much less. In specimen
B-1, the wide disagreement, in spite
of low value of N 7 N P , may be due to
the fact that failure was accompanied
by local buckling of the web below the
load points.

Table 4 Effect of Normal Force and Shear on the Tee-section

N p = Fy (area
Specimen
A-1
A-2
B-1
B-3

Limit load
P , kips
37.5
41.4
48.5
41.9

Avg. shear stress

at Tee-section, T
ksi
5.90
6.51
7.63
6.59

In Table 5, the ultimate loads predicted by the interaction method show

a much closer agreement with the observed values. This is expected since
the interaction method of Hope and
Sheikh (Ref. 16), takes into account
the effect of axial force on the plastic
moment of the tee section. With the
exception of specimen B-1, which did
not quite attain the possible ultimate
load capacity for Vierendeel mechanism because of web buckling, the
predicted ultimate loads are always
lower than the observed values. This
is expected since the interaction
method is based on the assumption
that the effect of shear force on the
plastic moment of the tee section and
the influence of strain hardening on
the ultimate load capacity are of equal
magnitude and opposite in sense and
cancel each other. This assumption is
not quite true for the specimens
tested since the effect of shear stress
was comparatively low. The effect of
strain hardening is not cancelled and
reflects on the observed ultimate load
values.
Flexural Failure
(a) Load-Rotation Characteristics.
The three specimens in series 2, G1, G-2 and G-3, which were designed
by balancing the allowable flexural
stress at the critical throat section with
the allowable shearing stress at the
critical welded joint, exhibited flexural failure. The characteristic feature
of this failure mechanism, as illustrated in Fig. 9, is the yielding of the
flanges in the region of high bending
moment. The yielding pattern is,
therefore, similar to that of a beam of
solid I or WF section. Halleux (Ref. 13)
reported such failure in the middle
third portion of castellated beams
subjected to t w o - p o i n t loading
system. Under pure bending, a castellated beam is not likely to fail by Vierendeel action since no secondary
bending effect is present.
Specimens G-1, G-2 and G-3,
which were tested to investigate optimum design, were subjected to a
moment gradient. However, these
specimens were less susceptible to
failure due to Vierendeel action be-

Normal force, N'

on Tee-section.
kips

Tee-section)
kips

y = 0.55F y
ksi

34.94
26.72
26.72
26.72

77.42
17.09
13.71
59.23

cause of their narrow throat widths. It

appears, therefore, that by guarding
against the basic weakness of a castellated beam, its performance may
be greatly improved; i.e., it may be
made to behave more like a solid I or
WF beam.
Figure 10 shows the complete loadend rotation curve of specimen G-1.
The first sign of yielding was observed, on the inside face of the compression flange near the load point,
when the load was 24 kips. Yielding of
the flanges became pronounced as
the load was gradually increased and
at 30 kips yielding was noticed along
the welded joints. Unloading started
after the load reached a maximum
value of 37.9 kips. Rate of unloading
was gradual at the beginning but at
36.5 kips web buckling occurred at
the panel adjacent to the load point
(see Fig. 15) and unloading became
abrupt. At an end rotation of 2 deg 5
min the total applied load came down
to 30 kips. At that point the applied
load was released in two installments. An elastic recovery of 26 min
was observed, with a final permanent
end rotation of about 2 deg. The behavior of specimens G-2 and G-3 was
similar to that of G-1, except that in
these beams the unloading was
triggered by local buckling of the
compression flange.
The load-rotation curves for specimens G-1, G-2 and G-3 are presented in Fig. 11. The ultimate loads,
Pu, for the unexpanded 10B15 section corresponding to specimens G-1
and G-2, which were fabricated from
the same length, and specimen G-3

100.94
77.20
77.20
77.20

T /ry

N7N,

0.169
0.244
0.286
0.247

0.767
0.221
0.178
0.766

Table 5 Comparison of Theoretical and

Experimental Ultimate Loads

Specimens
A-1
A-2
B-1
B-3

Ultimate loads , kips

Cheng's
ExperiInteraction
method
mental
34.98
41.24
53.74
37.42

40.2
45.0
52.0
44.0

are shown by two solid lines. Pu was

calculated using measured flange and
web yield stresses. The increase in
ultimate load is about 62%, 64% and
66% respectively for specimens G-1,
G-2 and G-3. It may be pointed out
here that for specimen A-1 (Fig. 5),
which was designed to fail by
Vierendeel mechanism, the corresponding increase was only 41%.
(b) Effect of Elastic Optimization.
As mentioned earlier in section 2.1,
the expanded depth and the span
length of specimens G-1, G-2 and G-3
were held constant but the number of
panels, N, was varied to study optimization in elastic design. N was
equal to 8, 12 and 16 for specimens
G-1, G-2 and G-3 respectively.
The moment-rotation curves for
specimens G-1, G-2 and G-3 are
shown in Fig. 12. The midspan moment, M, has been non-dimensionalized as M/M p to bring the results for
specimens having different yield
stress levels to a common basis. M p
represents the moment capacity of
the castellated beam at which the top
and bottom tee sections are fully

Table 6 Web Buckling Due to Compression

Specimen

Web weld
length,
n, in.

KL/rfor
assumed
column

Allowable
compressive
stress, Fa ,
ksi

Allowable
transverse
load, P ,a >,
kips

Observed
maximum
transverse
load, Pu ,
kips

B-2
C
D

5.0
4.0
3.5

150.61
150.61
150.61

6.56
6.56
6.56

15.09
12.07
10.56

42.0
20.0
20.0

FOS (b >
with
respect
to
observed
Pu
2.78
1.66
1.89

(a) P = 2nwF a
(b) FOS = factor of safety

W E L D I N G R E S E A R C H S U P P L E M E N T ! 335-s

plastified. M P was calculated using

measured cross-section dimensions
and the flange and web yield stresses
obtained from tension tests.
It appears from Fig. 12 that there is
no significant increase in ultimate
strength due to an increase of n u m ber of panels from 8 to 12. The relative ultimate strength of specimen G3, with 16 castellations, shows a
decline. However, there is substantial
increase in the rotation capacity as
the number of panels is increased.
This factor may be of importance for
plastically designed members. An
examination of Fig. 11 reveals that the
elastic stiffness remains unaffected.
It may therefore be concluded that
elastic optimization does not seem to
have any drastic effect on the ultimate strength or the elastic stiffness
of beams designed on the basis of
'balanced stresses' but considerably
increases their ductility.

Fig. 9 Failure by flexure mechanism

WEB

END

Failure due to Instability

(a) Premature Failure due to Web
Buckling Caused by Compression.
Three beams (B-2, C and D) failed
p r e m a t u r e l y b e c a u s e of w e b buckling. Specimen B-2 did not have
stiffeners below the load point and
specimen C and D had 4 in. deep
stiffeners below the load points.
Specimen B-2, whose dimensions
were exactly similar to that of specimen B - 1 , was tested under a midpoint concentrated load and failed at
42 kips, c o m p a r e d to 52 kips in the
case of specimen B-1 which was
tested under a two-point loading system and failed due to excessive flexural stress coupled with web buckling
below the load points.
Specimens C and D, which were
tested under a t w o - p o i n t loading
system, failed in exactly the same
manner due to web buckling below
the load points. Fig. 13 shows specimens C and D after failure.

BUCKLING

R O T A T I O N , 6 , DEGREES

Fig 10 Load-end rotation curve for Specimen G-1

Table 7 Web Buckling Due to Shear (a)

Specimens

Properties

A-1

$
w
n
h
m

=
=
=
=
=

45deg \
0.23 in.I
6.5 in.
5 in.
j
23 in. )

G-1

$
w
n
h
m

=
=
=
=
=

3 8 d e g , 3 min\
0.23 in.
1.75 in.
5 in.
15 in.

Allowable
compressive
bending stress
Fa , ksi

Allowable
shear force
along webweld,
F(all.), kips

Allowable
mid point
load,
P(all.), kips

Observed
maximum
load
Pu . kips

16.40

20.14

24.39

40.2

5.59

10.38

16.40

(a) For the equations used in these calculations and for an evaluation of the results, see text discussion of Table '
(bl FOS = factor of safety.

336-s i A U G U S T

1 973

38.5

FOS (b|

1.65

3.71

Blodgett (Ref. 20) has suggested an

approximate elastic analysis which
treats the nonprismatic solid web as a
column having a length equal to the
clear height of the hole, h; a width
equal to the web weld length, n; and a
thickness equal to the web thickness,
w. Referring to Fig. 14, the compressive stress, fa, in the assumed column
may be expressed by Eq. (2).

f.

=2 x n x w

(2)

where P is the total transverse load

acting on the panel.
Table 6 was prepared using Eq. (2)
to calculate the factor of safety
against web buckling due to c o m pression with respect to the o b served load at which web buckling occurred. The allowable compressive
stress, F a , was calculated according
to CSA S p e c i f i c a t i o n S 1 6 - 1 9 6 9 ,
Clause 16.2.2 (Ref. 21). The value of
Fa depends upon the member's effective slenderness ratio, K x L / r where
K x L is the effective column length
and r is the radius of gyration of the
column cross section. In the present
case, the value of K was assumed to
be 1.0 as recommended by the CSA
specifications for columns which are
essentially pinned at both ends. The
value of r for the assumed rectangular cross-section is given by w/(12 1/2 )
where w is the web thickness. Since
specimens B-2, C and D were all
fabricated from A-36 steel, a nominal
yield stress of 36 ksi was used for the
calculations.

END

Fig. 11

ROTATION,

Load-end rotation curves: Series 2

The results indicate that the actual

factor of safety (FOS) in elastic design
is not consistent. The lowest value of
FOS (specimen C) is just equal to the
basic factor of safety of 1.67. For all
three specimens, the allowable c o m pressive stress, F a , was calculated by
the f o l l o w i n g f o r m u l a for slender
columns:
149,000
(KL/r) 2

(3)

END

ROTATION,

DEGREES

Fig. 12 Moment-end rotation curves: Series 2

Eq. (3) is a d a p t e d f r o m
column buckling equation:

Euler's

(KL/r) 2
Table 8 Factor of Safety for Elastic Design: Series 1
and is based on a FOS of 1.92.
Moreover, in the present case the
nonprismatic solid web was assumed
to be pin-ended and the value of K in
Eq. (3) was taken as 1. In his presentation Blodgett (Ref. 20) did not specify any value for K. If one assumes that
the web is semi-hinged or fixed the
calculated FOS will be greatly reduced.
(b) Web Buckling due to Shear
Force.

Elastic
design
P(all.) for
Specimen
A-1
A-2
B-1
B-3

o ;-6F

Experimental

kips
8.90
9.44
12.43
9.80

FOS with
respect to
experimental

31.0
36.0
42.0
36.0

3.48
3.81
3.38
3.67

WELDING RESEARCH SUPPLEMENT!

337-s

bending stress, ar (max), may be expressed by Eq. (4).

(max)

<4>

3 Ftan 6
4 w X n (f?2)

Blodgett recommended that the maximum bending stress must not exceed the allowable stress, Fa , given
by Eq. (5).
(5)
1.0 -

where C r ={

4 3 4

2rr2E

( - \

0.6 F

and

w = web thickness,
Fy= nominal yield stress

Fig. 13 Specimens C and D after failure

In specimen A - 1 , which failed by a

Vierendeel mechanism, and also in
specimen G - 1 , which failed by a flexural mechanism, rapid unloading was
triggered by web buckling caused by
the shear force along the web weld.
The characteristic features of such
web buckling are illustrated in Figs.
15 and 16. The shear force, F, acting
along the web weld, stresses the web
in bending.
In Fig. 16, the fibers along AB are in
tension whereas those along CD are
in compression. In both beams m e n tioned above, edge buckling (along
CD and EF) occurred only after the
failure mechanism was completely
formed and certain amount of un-

loading had already taken place because of local buckling of the plastified compression flange. Therefore,
web buckling due to web shear may
not be regarded as causing premature failure as far as ultimate load is
concerned. However, it does prematurely terminate the rotation capacity of the beam and, therefore, is important for plastically designed m e m bers.
An exact solution of the web buckling problem is not available at pressent. However, Blodgett (Ref. 20) has
presented an approximate elastic
m e t h o d of a n a l y s i s b a s e d o n
Olander's Wedge method (Ref. 22).
Referring to Fig. 16, the maximum

SYMMETRICAL

Fig. 14 Load causing web buckling

338-s I A U G U S T 1 9 7 3

Table 7 was prepared by using Eqs.

(4) and (5). Since both specimens A-1
and G-1 were fabricated from CSA
40.12 steel, a nominal yield stress of
44 ksi was used for the calculations.
Note that the allowable compressive
binding stresses, F a , shown in Table
7, were obtained directly from Eq. 5
above, and that the allowable shear
forces, F (all.), were obtained by substitution from Eq. (4):
4 w X n (92) , c ,
F(all.) =
<Fa)
3 tan d
The equation used to determine the
allowable m i d p o i n t
concentrated
loads, P(all.). w a s :
P(all.) =

4 F(all.) 6.93
m

In discussing the results it must be

recognized that the factors of safety of

ABOUT

Fig. 75 Web buckling due to shear in Specimen G-1

Table 7 are not in the usual sense of

the word since the actual factor of
safety for these beams should be based on flexural yielding in the flange.
Web buckling was only a secondary
design criterion.
Nonetheless, the limited results
shown in Table 7 tend to indicate that
Blodgett's method is not valid here.
We are dealing with a trapezoidal plate
rather than a tapered beam. The plate
proportions are more consistent with
the shear buckling of a plate. In which
case, the m a x i m u m shear force along
the web weld will depend upon the
critical shear stress and the average
width of plate available. The maximum shear force obtained from experiments are 31.6 and 19.1 kips respectively for specimens A-1 and G - 1 .
The ratio is 1.65 which is approximately equal to the ratio of average
plate widths and the ratio of hole
spacings, m, for the two specimens.

COMPRESSION

TENSION
EDGE

EDGE

COMPRESSION

TENSION

EDGE

EDGE

Strain Distribution
The observed elastic flexural strain
distributions across the tee section at
the midlength of the throat confirm
the findings of Cheng et al (Ref. 7).
The strain distribution is uniform (Ref.
8) across the tee section for beams
with long throat length (Specimens A1 and A-2). For beams with relatively
narrow throat length (Series 4 specimens), the stress distribution is proportional to the distance from the
neutral axis as suggested by Boyer
(Ref. 10).
The average extreme fiber elastic
strain in the solid web section can be
predicted approximately ( 8% error)
by the relation = M c / E I . However,
the bending strain distribution across
the solid web section is not linear but
follows a pattern similar to that o b tained by Kolosowski (Ref. 2), Mandel
et al (Ref. 12) and Cheng et al (Ref. 7),
and is shown in Fig. 17(b & c). When
the intensity of loading is low, the
strain along the neutral axis appears
to be zero. However, at higher loads,
some strain seems to be present
along the centroidal axis of the beam.
An explanation for this behavior is
s u g g e s t e d in the f o l l o w i n g p a r a graph:
Referring to Fig. 17(a), the strain
distribution along section A-A should
be similar to that of two haunched
cantilever beams, as shown in Fig.
17(d). Along section B-B, the strain
distribution should be like that of an Isection as indicated in Fig. 17(e). The
length n/2 is the transition distance
within which the strain distribution
changes from the pattern shown in
Fig. 17(d) to that of Fig. 17(e). If n/2 is
sufficiently long the strain distribution pattern along section B-B would
approach that of an l-section. However, with the values of n/2 normally

Fig. 16 Web buckling due to shear

Fig. 17 Flexural strain across solid web section

used in castellated beams, complete

transition of the stress pattern does
not seem to be possible and the strain
distribution along section B-B resembles the pattern shown in Fig.
17(b).
At lower loads <ft equals So
and t h e r e f o r e the b e n d i n g strain
a l o n g t h e n e u t r a l a x i s is z e r o .
However, at higher loads, due to
yielding, the magnitude of \ and Sc
no longer remains equal and certain

unbalanced strain occurs at the neutral axis as shown in Fig. 17(c).

The o b s e r v e d elastic shearing
strain distribution along the web weld
is not parabolic as may be expected
on a rectangular prismatic crosssection. The strain is fairly constant at
the middle and drops a little at both
ends. The shear strain at the free end
of the weld line is far from being zero.
This is due to the fact that, because of

WELDING RESEARCH SUPPLEMENT!

339-s

R =
SPECIMEN

SPAN 3 0 7 5 in.
SINGLE

0.8BD
+ 0.94
L (1 in.)

MID-POINT

where B = flange width, in.

L = span, in.
D = expanded depth, in.

LOAD

ALTFILLISCH

Referring to Fig. 18, it seems that both

methods yield sufficiently accurate
results for longer span lengths although the U.S.S. method tends to be
a bit too conservative. However, none
of the methods is valid for short spans
(see Figs. 19 and 20).
A theoretical analysis, based on the
stiffness matrix method, was carried
out to calculate deflections of the
shorter beams which were assumed
as Vierendeel frames composed of
non-prismatic members. The effects
of shear and member length changes
were included in forming the stiffness matrix of the individual m e m ber. Good agreement was obtained
between the theoretical and the observed results as can be seen in Figs.
19 and 20.
H o w e v e r , in t h i s m e t h o d it is
necessary to make an assumption involving overlapping members. The
assumed shapes of the members,
used in the present analysis, are
shown in Fig. 2 1 . Cheng et al (Ref. 7)
used the finite element method to obtain the deflections in specimen A-2.
The calculated deflections were in
good agreement with the observed
values (see Fig. 19).

</

AA'

A*\
OBSERVED

AA

ALTFILLISCH

SPECIMEN
SPAN : 2 7 3
SINGLE

in

MID - POINT

LOAD

1
DEFLECTIONS (inch ) at

Fig. w

MID - POINT

Load-deflection curves: Series 4

the castellations, the weld line does

not coincide with one of the principal
axes at the free ends.
The experimental results revealed
that the points of inflection are not
located at the midspan of the throat
after yielding but shifts towards the
re-entrant corner which is critical with
respect to c o m b i n e d normal and
bending stresses. This confirms the
opinion expressed to this effect by
Hope and Sheikh (Ref. 16).
Deflection
The
vertical
deflections
were
measured in all 12 specimens and the

observed values were c o m p a r e d with

those calculated by the following two
approximate methods:
(a) Altfillisch's method (Ref. 8) with
modification as suggested by Hope
and Sheikh (Ref. 16).
(b) United Structural Steel C o m pany (U.S.S.) method (Ref. 23).
The first method is lengthy and inv o l v e s t h e use of a r e a - m o m e n t
theorem. The U.S.S. method, on the
other hand, is grossly empirical. The
deflection is equal to the calculated
deflection (based on the minimum
moment of inertia) multiplied by a
correction factor:

Fig. 19 Observed and calculated deflections:

Specimen A-2

ALTFILLISCH

OBSERVED

STIFFNESS

METHOD

Specimen B - 2

_nr^

0~TQ

4 <?I5.75 = 63

80
40

80
120
DEFLECTION IN 0.0OI inch

340-s I A U G U S T

1 973

IG0

200

I20
DEFLECTION

IN

I60
OOI INCH

200

240

Fig. 20 Observed and calculated deflections: Specimen B-2

The deflection for shorter beams

may be calculated approximately by a
slight modified form of the U.S.S. formula. The constant 0.94 in the correction factor, R, can be replaced by a
factor q which has different values for
different spans. The experimental
results showed that for beams with
L/D ratio around 12, q is approximately 0.94. The value of q tends to
increase as the span length d e creases. The correct value of q corresponding to a particular span may be
obtained from the q vs L/D curve presented in Fig. 22 which was prepared
from the experimental results of the
12 beams tested by the authors and
those reported by Altfillisch (Ref. 8),
Clark (Ref. 23), Hosain and Speirs
(Ref 17), and Galambos et al (Ref. 25).

Fig. 21 Assumed shapes of members

1 POINT

LOADING

LOADING

POINT

S u m m a r y and C o n c l u s i o n s
If local and lateral buckling are prevented in a castellated beam, failure
will probably be initiated either by the
formation of a panel mechanism or by
the yielding and fracture of the web
weld (Ref. 17) due to excessive shear
stress. Reduction in the length of the
throat, which makes the design less
susceptible to secondary bending
effects, reduces the possibility of
failure due to a panel mechanism but
increases the chance of failure due to
rupture of the web weld. The logical
design approach should therefore be
based on a minimum allowable web
weld length.
Table 8 lists the allowable loads for
flexural and the corresponding actual
factors of safety, with respect to experimentally determined yield loads
for the four specimens that exhibited
Vierendeel failure mechanism. These
calculations were based on an allowable stress of 0.6 Fy for flexure where
F y is the specified m i n i m u m yield
stress. The results indicate that the
actual factor of safety in elastic design
is very high and that a better use of
the reserve strength may be a c c o m plished by adopting the plastic d e sign method.
The test results also indicated that
in calculating the limit loads for
beams which fail by Vierendeel mechanisms, the effect of axial force on the
plastic moment of the throat section
m u s t be c o n s i d e r e d . H a l l e u x ' s
method (Ref. 13) which ignores this
effect, can therefore be used only in
cases where such effect is nominal.
However, as a basis for design, Eq. (1)
can also be used as first approximation to obtain the value of the normal
force, N', for use in a more rigorous
formula.
The interaction method (Ref. 16),
which takes into account the effect of
axial force on the plastic moment of
the tee section, makes valid predictions of the ultimate load.

.J_
15

20

SPAN/DEPTH, ( L / D )

25

RATIO

Fig. 22 Values of factor q

The test results revealed that

elastic optimization (more castellations in a given span) has little effect
on limit load and on elastic stiffness
but affects considerably the ductility
of the beam. The beam with the maxi m u m number of castellations r e c o r d ed the m a x i m u m rotation capacity.
The test results showed that by
guarding against the basic weakness
of a castellated beam, viz., the formation of hinges at the re-entrant corners, its performance may be greatly
improved. Failure in such a beam may
be caused by a 'flexure mechanism'
which is formed due to the yielding of
the flanges in the region of high m o ment.
The experimental results indicated
that the flexural strain distribution
across the tee section at the m i d length of the throat is uniform for
beams with long throat length. For
beams with narrow throat length, the
stress distribution is somewhat proportional to the distance from the c e n troidal axis. The bending strain distribution across the solid web section
is not linear but the average extreme
fiber elastic strains can be predicted
approximately ( 8% error) by the relation $ = M c / E I .

The test results also revealed that

the shearing strain distribution along
the web weld is not parabolic. The
shear strains at the free ends of the
weld line are not zero.
The modified Altfillisch method
suggested by Hope and Sheikh (Ref.
16) predicts accurately the elastic deflections of slender beams. The U.S.S.
method is simpler but is more conservative when applied to the same
beams. Both methods are grossly u n conservative for b e a m s of small
span/depth ratio. The U.S.S. method
can be extended to beams of small
span/depth ratio, if the constant 0.94
in the correction factor is replaced by
a factor q; the value of which may be
obtained from a curve prepared by
the a u t h o r s from e x p e r i m e n t a l
results. For a more accurate analysis
of d e f l e c t i o n , either the stiffness
method used by the authors or any
other method that includes the effect
of shear and axial force may be used
(Refs. 3, 7, 24).
The experimental results indicate
that the optimum hole geometry requires a minimum length of the throat
which makes the design less susceptible to secondary bending effects. It
was found that full depth stiffeners, at

WELDING RESEARCH SUPPLEMENT!

341-s

concentrated load points, are essential f o r e f f e c t i v e l o a d c a p a c i t y .

Acknowledgments
This investigation was carried out with
financial support from the National Research Council of Canada.
The authors are indebted to Mr. R. A.
Ritchie, Mr. Dan Stott and Mr. Metro
Hrabok for their assistance in conducting
the tests. The assistance of Mrs. Mabel
Wrigley, who typed the manuscript, is also
acknowledged.
References
1. Gibson, J. E., and Jenkins, W. M.,
"An Investigation of the Stresses and Deflections in Castellated Beams,"Structural
Engineer, Vol. 35, No. 12, December 1957.
2. Kolosowski, J., "Stresses and Deflections in Castellated Beams," Structural Engineer, Vol. 42, No. 1, December 1964.
3. Gardner, N. J., "An Investigation into
the Deflection Behaviour of Castellated
Beams," Transactions of the Engineering
Institute of Canada, Vol. 9, No. A-7, September 1966.
4. Shoukry, Z., "Elastic Flexural Stress
Distribution in Webs of Castellated Steel
B e a m s , " Welding
Journal,
44 ( 5 ) ,
Research Suppl., 231-s to 240-s, May
1965.
5. Faltus, F., "Calculation of Castellated Girders," Acier-Stahl-Steel,
No. 5,
1966.
6. Humphrey, A. T., and Sunley, V. K.,
"Finite Element Analysis of an Expanded
ISection Beam and an Axi-Symmetric
Flanged Cylinder," Conference Records,
A d v a n c e d Stress Analysis, 3-14 Joint
British Conference for Stress Analysis,
Vol. 3, No. 14, 1968.

7. Cheng, W. K., Hosain, M. U., and

Neis, V. V., "Analysis of Castellated Steel
Beams by the Finite Element M e t h o d , "
Proceedings of the Speciality Conference
on Finite Element Method in Civil Engineering, Montreal, Canada, June 1-2,
1972.
8. Altfillisch, M. D., Cooke, B. R., and
Toprac, A. A., "An Investigation of O p e n Web Expanded Beams," Welding
Research, Vol. 22, No. 2, February 1957.
9. Toprac, A. A., and Cooke, B.R., "An
Experimental Investigation of Open-Web
Beams," Welding Research Council Bulletin, Series No. 47, February 1959.
10. Boyer, J. P., "Castellated Beams
New Developments," Engineering
Journal,
American Institute of Steel Construction,
July 1964.
11. Douty, R. T., and Baldwin, J. W.,
"Measured and C o m p u t e d Stresses in
Three Castellated Beams,"
Engineering
Journal, American Institute of Steel Construction, January 1966.
12. Mandel, J. A., Brennan, P. J., Wasil,
B. A., and Antoni, C. M., "Stress Distribution in Castellated Beams," Journal
of
Structural
Division, ASCE, Vol. 97, No.
ST7, July 1971.
13. Halleux, P.,, " L i m i t Analysis of
Castellated Steel Beams,"
Acier-StahlSteel, No. 3, 1967.
14. Redwood, R. G., "Ultimate Strength
Design of Beams with Multiple Openings,"
Paper presented at the ASCE Annual
Meeting and National Meeting on Structural Engineering, Pittsburgh, Pa., September 1968.
15. Sherbourne, A. N., "The Plastic Behavior of Castellated Beams," Proceedings, 2nd Commonwealth Welding C o n ference, Inst. of Welding, L o n d o n , 1965.

16. Hope, B. B., and Sheikh, M. A., "The

Design of Castellated Beams," Transactions of Engineering Institute of Canada,
Vol. 12, No. A-8, September 1969.
17. Hosain, M. U., and Speirs, W. G.,
"Failure of Castellated Beams due to Rupture of Welded Joints,"
Acier-Stahl-Steel,
No. 1, 1971.
18. CSA, "General Purpose Structural
Steel," Canadian Standards Association
Standard G40.12, 1964.
19. A S T M , "Mechanical Testing of Steel
Products," American Society for Testing
Steel Products," American Society for
Testing Materials Standard A370-69.
20. Blodgett, 0 . W., Design of Welded
Structures,
The James F. Lincoln Arc
W e l d i n g F o u n d a t i o n , C l e v e l a n d , Ohio,
1968.
2 1 . C S A S t a n d a r d S 1 6 - 1 9 6 9 , Steel
Structures for Buildings, Canadian Standard Association, Canada.
22. Olander, H. C , ASCE Transaction,
Paper No. 2698, 1954.
23. Clark, P. R "Shear Stress and Web
Stability in Castellated Steel Beams,"
M Eng. Thesis, Nova Scotia Technical
College, 1964.
24. Cheng, W. K., Hosain, M. U., and
Neis, V. V., "Deflection Analysis of Expanded Open-Web Steel Beams." Presented at the National Symposium on
Computerized Structural Analysis and Design, Washington, D . C , March 27-29,
1972.
25. Galambos, A. R., Hrabok, M. M.,
and Hosain, M. U., " O p t i m u m Expansion
Ratios of Castellated Steel Beams," Structural Engineering Report No. 6, Civil Engineering Department, University of Saskatchewan, Saskatoon, Canada, August,
1972.

Elastic-Plastic Bending of a Constrained Circular Perforated

Under Uniform Pressure (Triangular Penetration Pattern)

Plate

by J. S. Porowski and W. J. O'Donnell

WRC
Bulletin
No. 180
Jan. 1973

The concept of an equivalent solid material is used to obtain elastic-plastic

solutions for circular plates under uniform pressure. Plates with elastically constrained outer edges are considered for the entire range of constraint stiffnesses.
Results are obtained for plates perforated in a regular triangular pattern over the
entire range of ligament efficiencies used in design. Since such plates may be supported by shells having limited yield moments, solutions are also given for an arbitrary edge moment support. In addition to the limit load results, solutions are given
for the pressure at initial edge yielding and initial central yielding, for the elasticplastic moment distributions, and for the plate deflections.
Publication of this paper was sponsored by the Pressure Vessel Research
Committee of the Welding Research Council. The price of WRC Bulletin 180 is $3.50
per copy. Orders should be sent to the Welding Research Council, 345 East 47th Street,
New York, N.Y. 10017.

Correction:
The article, "Electroslag Welding of Boiler Drums in India" (March, 1973, pp 125-s to 134-s) should
be corrected to read as follows:
Table 5, p. 128-s: The manganese content for 60mm plate should be 1.31 rather than 11.31. For
70mm plate, the wire-flux combination should be Union S.3 + Z 4 1 , rather than "Union."
Table 8, p. 133-s: The silicon content for 103mm plate should be 0.28 rather than 0.08.
342-s | A U G U S T 1 9 7 3