You are on page 1of 13

Journal of Occupational Accidents, 3 (1981) 163-175 163

Elsevier Scientific Publishing Company, Amsterdam - Printed in Belgium

THE INVESTIGATION OF STRUCTURAL FAILURES IN STEEL


BUILDINGS, CRANES AND BRIDGES*

F.M. BURDEKIN
Department of Civil and ~true~ra~ engineering, BLAST, ~anehester ~6~ 1QD (Great
Britain)

ABSTRACT

Burdekin, F.M., 1981. The investigation of structural failures in steel buildings, cranes
and bridges. Journal of Occupational Accidents, 3: 163-175.

The various modes of failure relevant to structural engineering are discussed, and
brief summaries given of the methods of investigation and conclusions for some examples
of such failures in steelwork structures.
An examination of the contributory causes of structural failures in steel buildings,
bridges and cranes shows that in the vast majority of cases there have been several techni-
cal factors involved, but that in genera1 there has been a failure to communicate or use
existing information rather than a lack of research knowledge.

INTRODUCTION

Failures in structures occur from time to time and they occur perhaps
more often than the general public are aware. Whilst the occasional spectacu-
lar failures are reported by the media, there are many failures of smaller
structures which receive perhaps only a brief reference in the technical press,
and others which are not reported at all.
Almost invariably there are individual circumstances about each failure
which make it different in detail from any other failure. Nevertheless it is
possible to make broad classifications of the different types of failure in
different materials and types of structure, and to describe the general prin-
ciples of investigation of such failures. It is also important to recognise the
lessons to be learnt from failure investigations.
It must also be recognised that any structure can be made to fail if the
loading applied to it is increased sufficiently. The design process considers
first the loading condition likely to occur during the intended service life
of the structure (often specified in Codes, and often based on a probabilistic
approach). The structure is then designed to withstand these loading condi-

*Presented at the Symposium on the Science and Technology of Accident Investigation


held at the University of Manchester Institute of Science and Technology, April 7-9,
1981.

0376-6349/8~~0000~000/$02.50 0 Elsevier Scientific Publishing Company


164

tions; with appropriate safety factors applied against failure. Some failures
occur because the loading conditions exceed those allowed for in the de-
sign, and it must be emphasised that the choice of design loading condition
is affected to some extent by cost as well as by the primary objective of
safety. In other words, an element of risk of failure is considered acceptable,
provided the incidence of failure due to loads exceeding the design condi-
tion is itself kept to an acceptably low level. The order of level of probability
of failure considered acceptable for critical structures where the consequences
of failure are catastrophic is 1OP structure/events per year. This implies that
on average there would only be one failure per year for every million struc-
tures of such type. For structures where the consequences of failure are less
severe a higher probability might be acceptable.

MATERIALS AND MODES OF FAILURE

The materials most commonly used for structural applications are steel,
concrete, brickwork and masonry. Of course many other materials are also
used including aluminium, cast iron and other metals, plastics, timber and
composites, but it is not possible to cover all materials and their special modes
of failure in a short technical paper. Steel, aluminium and some other metals
are used to carry tension, compression and shear stresses, whereas cast iron,
concrete, bricks and masonry are used primarily for oompression loading
because of their brittle nature. Some modes of failure may occur irrespective
of the materials used, but others are more likely to occur only in particular
materials.
The most common material independent modes of failure are overall stabil-
ity and insufficient bearing support. The term overall stability is used here to
cover both the cases when a structure topples over because its centre of gravity
moves outside the line of support, and the cases when the structure collapses
as a mechanism through insufficient stiffness or bracing. Failure through in-
sufficient bearing occurs if a higher member loses its seating on a lower mem-
ber so that it is no longer supported.
The following modes of partial or complete failure are more likely to occur
in the particular materials shown:

Steelwork Reinforced or prestressed Brickwork or masonry


eoncre te

yielding crushing crushing


brittle fracture chemical deterioration cracking
shear fracture steel deterioration chemical deterioration
fatigue bond failure mortar joints
buckling shear failure frost damage
corrosion frost damage
creep
welded/bolted joints
165

Each of these modes of failure has its own characteristics and an experienced
investigator can generally identify the relevant mode without difficulty. The
full es~blishment of all the factors responsible for causing the failure will
generally require careful examination of all the evidence at site, followed by
selected laboratory checks and investigations.

CAUSES OF FAILURE

As with modes of failure, there are a number of general headings for causes
of failure which can be applied, although in most cases there will not be a
single cause, but a number of contributory causes. The headings selected for
the purpose of analysing structural failures are as follows:
Design
Materials
Workmanship
Overload
Inadequate erection
Deterioration in service
Foundation failure
In addition there are two general headings for factors which are ~por~t
from the point of view of learning from failures:
Failure to use or communicate existing information
Lack of knowledge.

INVESTIGATION OF STRUCTURAL FAILURES IN STEELWORK

General

Much can be learnt from careful examinations at site. Where there is a


total collapse it may be difficult to identify the first critical event, since a
number of secondary failures may have occurred as a consequence of an
initial failure. It is therefore important to record exactly which parts have
failed, and the final position of all parts after failure. Photographs of a col-
lapsed structure often play a vital role and certainly the relative positions
of failed parts will often help to establish the sequence of events. Checks
must be carried out on welded joints at major points of support, as well as
at the apparent points of collapse. Any undried regions similar to those
involved in the failure should be checked to establish their condition.
Patterns of deformation should be observed and recorded.
It is also important to obtain information about the loading applied to
the structure at the time of failure, and to record eyewitness reports of
the failure (although these should not be regarded as infallible). A design
check to establish the stresses and behaviour of the structure under actual
loading is usually necessary.
In most failure investigations, it should be regarded as an essential routine
166

requirement to check the material properties against those specified, including


hardness, tensile and fracture tests and chemical analysis.
Me~l~~c~ investigations on the quality of steel, and on the nature and
extent of any material defects present may be necessary. Examination of
the microstructure will show whether the heat treatment and manufacturing
processes for the steel have been correctly executed. Optical microscopy will
also show the non-metallic inclusion morphology, which may be particularly
important in some welded joint problems. Hardness tests on parent material,
weld metal, and heat affected zone will show whether the welding procedure
was correct, and may help in showing the causes of HA2 cracking.
Where fractures have occurred, the initial visual inspection should seek to
identify any primary origin. Transverse sections from the fracture will show
whether it has occurred by transgramdar, intergranular or shear modes. A
study of the fracture surfaces using binocular microscope, scanning electron
microscope, and/or replication techniques in the transmission electron
microscope may be necessary to confirm the type of fracture. Particular
attention will be paid to any original defects at the primary origin, to estab-
lish their nature and cause.

Yielding failures will almost always be due either to overloading or to in-


correct design. The overloading may arise as a consequence of failure else-
where, of inadequate erection procedures, or more directly from incorrect
assumption of loads, impact loads, earthquakes or explosions. Yielding will
be clearly evident from local stretching and thinning of the steel, or from
flaking of paint or rust markings.

Brittle fracture

Brittle fracture is a catastrophic form of failure which can occur under


normal design loading conditions, without the need for overload. This type
of failure requires a combination of circumstances, namely:
(1) tension stresses
(2) an initial defect or severe stress concentration
(3) material of poor fracture toughness.
The type of fracture is usually clearly evident from visual examination,
with little or no evidence of ductility or thinning, a square break, a bright
sparkling appearance on a new fracture and characteristic chevron markings
on the fracture surface. From visual examination of the markings on the
fracture surfaces, it is possible to identify the initial defect from which the
fracture started. An investigation needs to establish the nature and cause of
the initial defect, and to determine the loading, material fracture toughness
and other mechanical properties and chemical analysis of the steel. In
welded structures the initial defect is often a weld defect, although some-
167

times the initiating defect for brittle fracture may be a fatigue crack. Where
possible, at the time of first investigations at site, a protective oil coating
should be applied to fracture surfaces at the initiation region to prevent
corrosion destroying vital evidence which may not be revealed until laboratory
investigations can be completed later.
In brittle fracture failure investigations, the use of fracture mechanics ana-
lyses may be invaluable in explaining whether a particular defect could have
been the primary origin under the loading conditions reported. Such an ap-
proach can also be used to show whether some additional overloading must
have been present or whether additional welding residual stresses were im-
portant.
Brittle fracture failures have occurred in virtually every type of steel con-
struction, although with the precautions now taken they are fortunately less
common. Nevertheless the effects of a brittle fracture failure can be so cata-
strophic that careful attention to precautions is essential. A number of fail-
ures have occurred in bridges. The failure of Kings Bridge in 1962 in Mel-
bourne was one classic case. Figures 1 and 2 show general views of one of
the fractured steel plate girders, and of the welded coverplate detail from
which the fracture started. Investigations showed that heat affected zone
cracking had occurred immediately after welding, and that the steel plates

Fig. 1. Kings Bridge, Melbourne - general view of fractured girder.


Fig. 2. Kings Bridge, Melbourne - view of cover plate detail from which brittle fracture
initiated.

had very poor toughness at the temperature at which failure occurred (O°C).
The steel also had poor weldability, which contributed to the formation
of the initial cracks. The Royal Commission of Inquiry laid great stress on
failure of communications as a major contributory factor to this failure. Sub-
sequent to this, changes were introduced into the British Standard for steel
bridges, BS.153, to require notch ductile steels fur welded tension areas,
with maximum thickness limits for different grades of steel. These have re-
cently been revised in the new British Standard BS.5400, to give limits for
a range of different minimum temperatures.
An example of a building collapse as a result of brittle fracture failure was
the Tehran Airport building collapse in 1974. Three roof trusses failed by
brittle fracture of gusset plates connecting the two halves of the trusses to-
gether at mid span (Fig. 3). Examination of the fracture surfaces showed the
fractures to have initiated from the bottom edge of the gusset (Fig. 4). Labo-
ratory investigations showed the gusset plate to be mild steel with sheared
edges. The embrittled edges had high hardness and minute cracks and at the
temperature of failure (0--5°C) the toughness was low. Investigations at site,
showed that the roofing and ceiling materials were subs~ti~y thicker and
heavier than those intended in the original design. These dead loads combined
with a heavy snow load at the time of failure, were sufficient to cause the
PROJECTING COLUMN AND PARAPET

Fig. 3. Tehran airport building roof truss failures with fracture in central gusset plate.

Fig. 4. Fracture surface of gusset from Tehran airport showing brittle fracture initiation
at bottom edge.

gusset plate sheared edge to be stressed to yield, and brittle fracture failure
was the result.

Fatigue

Fatigue failures occur by the slow growth of a crack under fluctuating


stresses. Final failure occurs if the crack grows to a sufficient size to cause
unstable brittle or shear fracture, or to reduce the remaining cross-sectional
area to an extent that yielding of the ligaments occurs.
170

A fatigue crack surface has a distinctive smooth appearance, and can


usually be identified by visual examination. Examination in the scanning
electron microscope or by replication techniques may show the presence
of striation markings left by the crack front with each cycle of loading. If
present, such markings are conclusive proof of fatigue, and may also permit
the rate of crack propagation and fatigue life to be established. Striations
are not always present, however, and their absence does not mean that the
fracture cannot be fatigue.
In welded steelwork, fatigue failures commonly start from welded details
primarily because of the stress concentration effect of the geometric shape
of the weld. The presence of any weld defects may aggravate matters, and
fatigue is in fact the mode of failure most sensitive to defects. Research
work has identified the typical fatigue behaviour of different types of weld
detail, and allowed design rules to be formulated giving an allowed stress
range for the required fatigue life for each detail, These design rules are
included in the British Standards for steel bridges (now BS. 5400) and for
cranes (BS.2573).
The occurrence of fatigue cracking is not uncommon in cranes, and has
also been found in steel bridges. The cracking has usually (but not always)
been found at a stage before final failure, when repairs can be effected and
the structure allowed to continue in service. Figure 5 shows an example
of a fatigue crack which started at a weld, but has not reached final failure.

Fig. 5. Example of fatigue crack starting at welded joint but growing in parent steelplate.
171

In cranes and bridges it is difficult to give precise estimates of the loading


and number of cycles which will occur in service. In the codes an equivalent
design loading is specified, based on typical experience. The most common
cause of fatigue cracking is either the incorrect designation of weld detail
for design purposes, or the occurrence of additional secondary stresses not
taken into account by the designer.
An example of fatigue cracks in the roof trusses of a building was the
case of a brewery warehouse, where the cranes were slung from the roof
steelwork. One member fractured completely, and fatigue cracks were then
found at virtually all similar positions.
In a recent survey of twelve major steel bridges in this country, fatigue
cracks were found in four cases. The cracks were found at a sufficiently
early stage that repairs were possible without closing or restricting the bridges
concerned, albeit the cost of repairs was significant. The failure and collapse
of the offshore floating platform Alexander Kielland is believed to have
been due to a fatigue crack starting at a welded joint, and growing to a size
sufficient to cause fracture of a critical member.
Fracture mechanics analyses may again be a valuable aid in fatigue investi-
gations, as they may be used to calculate the predicted fatigue life for a
crack to grow from initial to final size under the reported loading.

Buckling

Buckling failures are also primarily a design matter, although there may
be contributions from overloading and from initial flatness or residual stress
aspects of workmanship.
In major structures fabricated from stiffened plate, several forms of
buckling can occur. These are summarised below:
Overall buckling of complete compression member
Lateral buckling of bending member
Torsional buckling
Shear buckling
Local plate panel buckling
Stiffener buckling
It is possible for progressive failure to occur by local buckling starting in
one area, and for the load shed by this local failure to cause buckling of
adjacent areas spreading across the whole section.
A series of four major failures of steel box girder bridges in various parts
of the world during construction led to the Merrison Inquiry in this country,
and the introduction of revised and more stringent design rules for stiffened
plate structures. The collapse of the Westgate Bridge, Melbourne in 1970
was the subject of another official inquiry. This established that although
the immediate direct cause of failure was the removal of a number of bolts
at critical splice joints, there were several other contributory factors. The
span concerned had been designed as a series of steel boxes, with each box
172

divided in half longitudinally. The method of erection devised was for the
two complete spans of half boxes to be assembled end to end, and lifted
separately onto the piers, with the longitudinal bolting together to be com-
pleted in position in the air. Despite precautions taken, problems were en-
countered in matching the two halves together, and the various actions taken
to correct this led to overstressing and progressive spread of buckling across
the top flange at midspan. The Royal Commission report again laid great
stress on communication problems between the various parties involved.

Fig. 6. Milford Haven Bridge - general view of collapsed span.

Fig. 7. Milford Haven Bridge - view at pier where buckling failure of internal diaphragm
and webs occurred.
173

The collapse of Milford Haven Bridge in 1970 (Figs. 6 and 7), was due
to buckling failure of a load bearing diaphragm over a pier. The initial buck-
ling took place due to the horizontal component of shear force from the
inclined webs of the trapezoidal box, Investigations of the failed diaphragm
at site showed from the folded shape of the diaphragm that it collapsed hori-
zontally first, and then vertically, leading to loss of bending strength of the
complete box and inevitable collapse. As in any failure investigation of a
major nature, routine material tests were carried out on the different plates
in the critical area. These tests showed one of the web plates to be of a lower
yield strength steel than that specified, even though it had been stamped at
the mills with the higher grade markings. This discrepancy was not a factor
contributing in any way to the failure. Examination of various failed welds
showed that their failure must have taken place at a later stage of the overall
failure sequence, and that their failure could not have occurred as the first
stage.

Other modes of failure

Corrosion of steelwork is of course a major problem requiring frequent


preventative attention. Actual collapses of steel structures due to corrosion
are rare, since deterioration is progressive and is generally found by main-
tenance inspections before the loss of metal is sufficient to cause failure.
The cost of remedial work because of corrosion can be very considerable
however.
Creep is a form of failure which occurs at high temperatures, and is not
relevant to bridges, buildings and cranes except for the cases of buildings
which may experience-high temperatures. Even for buildings containing
furnaces it is not usually necessary to use special steels for the framework
of the building, appropriate precautions being taken with insulating materials.
A number of failures occur because of trouble at welded or bolted joints.
This sometimes arises when the joint is incorrectly designed for the combina-
tions of forces to be applied, or sometimes when the joint is incorrectly made.
Occasionally problems arise with the weldability of the steel, and cracking
defects occur which reduce the load bearing capacity of the joint. A number
of collapses have occurred because a fillet weld connection has been made
when a T-butt weld was intended, through mis-reading of the drawing, and
others when welds of the wrong size or containing significant defects have
been made. Examination of such cases by an experienced investigator can
show readily whether the weld has failed by shear at its ultimate strength
or whether any defects were present to cause failure at lower strength. Metal-
lurgical investigations will usually be necessary to confirm the nature and
cause of any defects present, and chemical and mechanical tests of the mate-
rials involved should be carried out as a routine check.
Failures of bolted connections in service are not common, although some
have occurred due to hydrogen embrittlement of high strength steels, some-
174

times due to effects of pickling and plating. The cause of such failures has
been established by metallurgical examinations of the quality of the bolt,
together with fractographic investigations. In these cases the problem was
found by inspection at a stage when only a few had failed and replacement
was carried out without a major overall failure. Similar problems have oc-
curred occasionally with high strength steel shot fired pins used as roof
fixings. It is not uncommon for problems to arise with breaking of HSFG
bolts during tightening, and this is almost invariably due to combined tension
and torsion due to dry and/or dirty threads rather than any fault with the
steel quality of the bolt itself. Problems have been encountered on one
major bridge with nuts for HSFG bolts, manufactured from a free machining
steel with a very highinclusion content, which led to cracking and splitting
of some of the nuts on tightening. This was established by metallurgical
investigations of the quality of the steel nuts, chemical analysis, and laboratory
trial tightening experiments.
One failure has occurred due to mixed use of metric and imperial nuts
and bolts. The bolted connections concerned were at mid-span of the roof
trusses of a new concert theatre, and the whole building collapsed during
rehearsals when the trumpets sounded. It was found that 20mm diameter
bolts had been used with 3/4 inch nuts, and although the nuts appeared to
screw satisfactorily they rested only on the crest of threads. The nuts stripped
off along the bolts leading to complete failure of the roof trusses and destruc-
tion of the building.

CONCLUDING REMARKS

With appropriate combinations of site and laboratory investigations of


structural failures, the causes can usually be clearly identified from a techni-
cal point of view, and will usually involve more than one factor.

TABLE 1

Approximate estimates of percentage cause of different modes of failure in structural


steelwork
_~____~ ~~ ~~
Causes Modes

Yielding Brittle Fatigue Buckling Joints Stability Corrosion


fracture
(%I m__ (%)__ (%) (%) (%I (%)

Design 10 5 90 90 20 20 5
Materials 3 75 10 10
Workmanship 10 10 5 60 5 25
Overload 75 5 5 5 25
Inadequate erection 12 5 50
Deterioration in
service 5 60
175

On the basis of investigations of many failures over a twenty year period,


and a study of reports of many other failures, an attempt has been made
in Table 1 to show approximate estimates for the percentage cause responsible
for the different modes of failure.
From this Table the principal causes of the major structural modes of fail-
ure are as follows:
Yielding - overload
Brittle fracture - wrong material selection
Fatigue - design errors
Buckling - design errors
Joint failure - workmanship
In nearly all failures a major factor is that existing technical knowledge
was not applied or communicated adequately between those involved. Struc-
tural failures due to insufficient research knowledge do occur occasionally
but are rare. Further research should be aimed at improved economy of con-
struction and establishment of balanced probability of failure between the
various relevant modes.

You might also like