Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Forensic Engineering Proceedings of the Institution of Civil Engineers

Forensic Engineering 165 May 2012 Issue FE2


Volume 165 Issue FE2 Pages 63–75 http://dx.doi.org/10.1680/feng.10.00001
Forensic engineering and collapse Paper 1000001
Received 11/06/2011 Accepted 20/10/2011
databases Keywords: failures/risk & probability analysis/safety and
Breysse hazards

ice | proceedings ICE Publishing: All rights reserved

Forensic engineering and


collapse databases
Denys Breysse PhD
Professor of Civil Engineering, Civil and Environmental Engineering
Department, Université de Bordeaux, Bordeaux, France

Collapses in civil engineering are not uncommon. Some cases become known worldwide but, in many cases, the
details only reach the limited audience of the client and companies directly involved. In these latter situations, after
in-depth analysis of the disaster, useful information could have been obtained and shared. Collapses of structures can
result from many factors (loading, material, organisation, lack of maintenance, etc.) but analysis of failures can always
identify lessons for improved practice. Forensic engineering based on post-mortem analysis aims to understand the
reason for failures in order to learn useful lessons. At the present time, however, forensic engineering is still an
unknown concept for many French engineers. This paper demonstrates from general considerations, with a few
examples, what can be gained from collating ‘collapse experiences’ on a more formal basis. Three famous French cases
of collapse are analysed. The use of organised feedback and databases for risk management in civil engineering
projects, for teaching and for improving the quality of work is explained.

1. Introduction is the name of a new field of engineering that consists of post-


The principal area in which construction engineers are learning mortem analysis in order to understand the reason for the
from experience is seismic collapse. After a significant earth- collapse and learn useful lessons (Delatte, 2000). This paper
quake, groups of experts analyse failures to understand what demonstrates from general considerations, with a few exam-
could have been improved (in building codes, building practice, ples, what can be gained from collating ‘collapse experiences’
etc.) to reduce the consequences of such hazards. By observing on a more formal basis. The main difficulties of such work are
damaged structures and grouping them according to various described. The knowledge drawn from these databases would
criteria, they also develop fragility/vulnerability curves be useful for risk management in civil engineering projects, for
(Rossetto and Elnashai, 2003) that can be used to help teaching and, more widely, for improving the quality of civil
decision-making when evaluating designs. However, earth- engineering work.
quakes are not the only field in which analyses of collapse can
provide useful information. Collapses or warning incidents in 2. Civil engineering structural failures as a
civil engineering are not uncommon. While some cases are source of knowledge
famous worldwide (such as the Tacoma and Palau Bridge In construction, explanations for failure often focus on
failures, Heathrow and Roissy airport collapses and the Hyatt material properties or loading: failure happens when the load
Regency footbridge disaster), most events are only known to a produces more severe effects than a material can bear, without
limited audience – the client and the company directly involved. forgetting possible deterioration over time (Breysse, 2009a).
This is unfortunate as useful information can be obtained and However, while this ‘mechanical logic’ can explain the fracture,
disseminated after in-depth analysis of what happened. human factors at any stage (design, building or service) also
need to be considered (Breysse, 2003). Human error can come
Collapses of structures can result from many factors (e.g. loading, from bad decisions, insufficient knowledge, or understanding
material, defective organisation or maintenance) but analysis can or communication problems. Accidents can also be the result
always bring useful lessons for improving practice. They cannot of risks that are consciously considered to be acceptably low
be considered just an ‘act of God’ since rational reasons can (in one case out of four), either because elimination of the risk
always be found. These often relate to the sociological, is technically impossible or economically not acceptable
contractual and economical context of when and where failures (Schneider, 1997).
occur. Public reaction to such collapses varies between countries
and, in any given country, can change with time. When a failure occurs, one wonders about its ‘acceptability’. Is
it acceptable that building collapses frequently happen in
Some efforts have recently been made to gather and analyse Egypt (partial data sampling led to the identification of 18 such
data on collapses on a more formal basis. Forensic engineering collapses between 2000 and 2009, resulting in 154 victims

63
Forensic Engineering Forensic engineering and
Volume 165 Issue FE2 collapse databases
Breysse

(fatalities))? Is it acceptable that urban tunnelling work can the Western world in modern times. The subsequent investiga-
induce so much damage in a neighbourhood? Is it acceptable tions (Roubault, 1970) led to a general improvement in dam
that within a time period of 2 years, two collapses with striking management strategies. The design could not be faulted on the
similarities occurred in Moscow (Transvaal Park with 28 then state-of-the-art as, at that time, there was a lack of
victims and Yasenovo market with 60 victims)? To answer knowledge of the risks induced by high pore pressures in the
these questions the technical, economical and social contexts fractured rocks below the dam. Because the failure and its
must all be considered. consequences were comprehensively documented, further
analysis is still possible, for instance for calibrating flooding
Case studies of past failures make it possible to identify causes models after a sudden dam collapse. Other significant collapses
(certain, probable or possible), analyse them and draw some include
conclusions (Bordes, 2005). These conclusions enrich collective
expertise. Feedback qualifies the data gathering process in & the Auguste Perret Meteor tunnel collapse, in Paris, on 14
order to learn lessons after an unforeseen (and often dramatic) February 2003
event. Feedback can be qualitative or quantitative, based on & the roof collapse at Roissy airport terminal E on 23 May
case studies or on statistics of similar failures. A detailed case 2004 (four fatalities)
study requires both a high level of expertise and extensive & the Saint-Etienne River Bridge collapse, on La Réunion
means (laboratory tests, numerical simulations, etc.). It is also Island, in the Indian Ocean, on 25 February 2007.
used when some judicial/legal dimensions appear (i.e. when the
question of responsibility arises). The development of this These three failures were technically analysed by experts
dimension, mainly in Anglo-Saxon countries, gave birth to nominated by the Ministry of Transport and the experts’
‘forensic engineering’ (Carper, 2007). Forensic engineering can reports were made public (albeit briefly and then withdrawn in
be seen as comparable to medico-legal analysis: it is a (post- the Roissy airport case). A summary of the context and
mortem) retro-analysis after a failure, going back to its root technical aspects of each case is now given, from which some
causes, in order to rebuild the logical chain of causes and general considerations are derived.
consequences that finally resulted in the collapse. This is
important for courts considering the allocation of responsi- 3.1 Auguste Perret Meteor tunnel collapse
bility, but is of greater importance for adding to engineering During the night, a sinkhole opened on the building site of the
knowledge. Meteor line in the centre of Paris. The sinkhole originated at a
site where a large underground station was being built. At the
A classic case is the Hyatt Regency footbridge collapse in surface, it caused the sudden collapse of ground between
Kansas City in 1981, resulting in 114 fatalities and more than several school buildings (Figure 1). Fortunately, the collapse
200 injuries when a suspended walkway suddenly fell on people occurred during the winter holidays when the school was
who were dancing in the hotel atrium. The investigation closed. The authorities, however, had to face ‘local psychosis’
showed that the cause was the connections between beams and about what possible consequences the collapse could have
the metallic bars that supported them. The drawing had been brought.
modified (without checking) during the building stage, leading
to double the stress value of that in the initial design. Lack of
coordination between the project partners was critical in the
failure (Moncarz and Taylor, 2000).

An engineer starting to investigate a failure is now able to


consult forensic academic journals and conferences (Rens
et al., 2000), but France has until recently remained outside the
field of development of forensic engineering: it was not until
April 2011 that a conference on the topic was held in Paris
(Breysse et al., 2011).

3. Three major French cases of structural


collapse and lessons learned
France is not devoid of structural collapses. The most famous
in France is the sudden failure of Malpasset dam on 2 Figure 1. Sinkhole at Auguste Perret school (after Dubois and Rat,
December 1959 in the town of Fréjus, causing 423 fatalities. It 2003)
is known worldwide as one of the most devastating collapses in

64
Forensic Engineering Forensic engineering and
Volume 165 Issue FE2 collapse databases
Breysse

Three investigations were simultaneously initiated by the was fractured with an orientation roughly along the tunnel
Ministry of Public Works, the Administrative Court and the axis, thus playing an unfavourable role. Post-failure on-site
client (RATP, the transportation authority in Paris). The tests revealed a high anisotropy with a 350% difference
information given in the following is mainly drawn from the between the modulus in orthogonal directions because of these
report written for the Ministry (Dubois and Rat, 2003). fractures.

The underground station was a 145 m long and 14?5 m wide During excavation, a light support was designed combining
hall, 9–10 m below the surface. This depth was necessary in shotcrete, bolts and trusses spaced at 1 m intervals. However,
order to avoid the plastic swelling Sparnacian clays. A typical during the work, the contractor suggested, on the basis of
cross-section (Figure 2) shows that the tunnel was dug into additional two-dimensional (2D) finite-element computations,
limestone, with a cover depth in that layer of about 6?40 m. It using only bolts and shotcrete. His view was that this was
is situated in a complex environment with many old limestone sufficient to ensure vault stability and to avoid the fall of rock
quarries that had been backfilled. The upper part of the blocks. The client accepted these changes, while requesting a
limestone comprises layers of highly contrasting properties. thicker shotcrete layer.

Mechanical property results from on-site and laboratory tests The procedure used was alternatively digging two half tubes,
revealed a high variability (modulus of 850 MPa on plate tests separated by a central pile. There was no full-time geotechni-
for the lower part of the limestone and 25–90 MPa in cian on site during the work. The forensic analysis was based
pressuremeter tests on borings), while laboratory tests were on pictures of the excavation wall, daily reports, reports of
performed only on the ‘good’ layers, thus introducing a weekly meetings between the client and the contractor, and
sampling bias. The post-failure report noted that it is difficult convergence and settlement measurements. The reports docu-
to get representative parameters in such a heterogeneous mented ‘bad soil conditions’ from September 2002 and many
formation, but that more relevant information could have been local rock falls that had been mitigated using a denser bolting
mesh. During excavation of the hall, surface settlements locally
derived after a careful analysis of the available data. For
reached 30 mm, compared with the calculated and accepted
instance, boring records showed that the thickness of high-
value of 10 mm; however, even if the stakeholders were aware
strength layers was regularly less than 50 cm. In addition, the
of the possible consequences for the existing buildings, these
limestone was considered homogeneous and uncracked, but
settlements had not been retro-analysed in terms of ground
properties.
Boring Pressuremeters
0.00 m MT122 SP3
The collapse happened in an area were the excavation had been
Fillings stopped for 2 months and were the settlements were stabilised.
E = 26 MPa

3.35 Some minor work was undertaken in the lower part of the
9.76 m

cross-section. This possibly induced vibrations that could have


been a trigger for the collapse. All observations confirmed that
the coarse limestone above the tunnel did not have the
Upper
properties it was assumed to have. It appears that, because
limestone
of its layered structure, the vaulting effect was not efficient,
preventing the transfer of loads on the sides of the cavity. The
E > 90 MPa

Maintenance dissymmetry of the excavation process could have induced


14.00
10.41 m

hall of the station bending and the existing cracks could have induced disconti-
nuity. The shotcrete layer was not rigid enough to create the
Lower
vault effect.
limestone
As a conclusion, while the under-designed support system and
21.70 the condition of the limestone layer can be considered the
causes of the collapse, human components must also be
Plastic
clays considered. Organisational deficiencies before and during the
work were identified, including

Figure 2. Cross-section in the area of collapse (after Dubois and & the fact that the limestone was considered homogeneous
Rat, 2003) when it actually consisted of highly heterogeneous and
cracked layers

65
Forensic Engineering Forensic engineering and
Volume 165 Issue FE2 collapse databases
Breysse

& the fact that the design of the support system was modified dissymmetry because it has three openings on one side
without careful checks (the report highlighted an ‘excessive (Figure 4).
confidence in inadequate calculations’)
& the fact that neither the bad soil conditions encountered Fortunately, the collapse occurred very early on a Sunday
during excavation nor the unpredicted settlements were morning, at a time when very few people were in the area, thus
considered to necessitate an update of the prior site model. explaining the limited number of fatalities. The external
temperature was 4?1 ˚C; 90 minutes before the collapse, a piece
3.2 Roissy airport terminal 2E roof collapse of concrete had fallen from the roof and a safety perimeter had
The fatal collapse, 30 months after completion, of the been installed. All possible causes for the final collapse were
concourse building at Roissy Charles de Gaulle international investigated by the committee and it was shown that
airport is well known. A substantial factor was the architec-
& the concrete and steel initial properties were as expected
tural character of the building. Even though only 4% of the
650 m long concourse structure was affected by the collapse & the support structure behaved satisfactorily (some problems

(Figure 3), the impact was not limited to the four fatalities: it encountered with the design of the head of piles during
also affected the client (development at Roissy airport was building had been correctly remedied)
& there was no significant settlement and that the soil
delayed), the architect (Paul Andreu), the main operator (Air
France, user of the concourse) and the whole profession of properties had been correctly anticipated.
French structural engineers. Many possible explanations arose
in the very first days, most being based on hypotheses about The technical explanation for the collapse pointed towards
deficiencies in structural behaviour, missing reinforcements or structural causes. Numerical simulations were performed that
defective foundations. confirmed the sudden character of collapse, which was also
confirmed by eye witnesses. The following main elements were
noted.
The French government set up an administrative committee to
investigate the technical causes of the collapse, without
& The overall structure was very flexible. High deformations
interfering in judicial investigations. The committee’s report
could be induced either by permanent loading or by climatic
was released on 15 February 2005 (Berthier, 2005).
and variable loading. This flexibility had even increased
because of cracking in some parts, which developed very
The bearing structure is made of arches (span 26?2 m)
early on, probably because of insufficient or badly located
assembling horizontal and vertical concrete panels, with steel
reinforcement.
stiffening struts anchored in concrete. At their feet, the arches
& The 2D analysis performed during the studies was not
are stabilised by a longitudinal beam and lie on a series of piles.
detailed enough to understand what could happen when the
The longitudinal regularity of the structure is disturbed by
real geometry was accounted for with high 3D effects due to
many local singularities (e.g. openings linking to pedestrian
singularities, for example those due to links with foot-
access tunnels and holes for ducts in the longitudinal beam).
bridges (see Figure 4).
The collapsed section is a very particular area with significant

Figure 4. 3D finite-element analysis of the shell, showing the


Figure 3. The collapsed structure influence of singularities (Berthier, 2005)

66
Forensic Engineering Forensic engineering and
Volume 165 Issue FE2 collapse databases
Breysse

& The structure lacked redundancy and was not able to guidelines and improving professional practice. This project is
transfer loads in the case of local deficiency. in progress and reports are expected soon.
& The longitudinal beam stabilising the arches was not strong
enough. 3.3 The Saint-Etienne River Bridge collapse, La
& Steel struts were anchored in concrete panels with metallic Réunion Island
plates, at a depth that induced high local tensile punching The road that connects the two cities of Saint-Louis and Saint-
shear stresses in the concrete. Pierre on the south-west coast of La Réunion Island is a crucial
axis of communication with daily traffic of 54 000 vehicles
The main initial cause of the sudden collapse was linked to without any possible detour. The crossing of the Saint-Etienne
punching of the concrete panels by the struts. It was probably River includes two 520 m long bridges (Figure 5). The
facilitated by prior and gradual weakening of the concrete upstream bridge is a prestressed concrete box girder bridge,
under permanent and varying loads. Concrete creep induced a built in 1977 on masonry piers erected in 1937–1938. The
progressive development of local internal stresses while thermal bridge located downstream was a prestressed concrete box
cycles increased the local cracking of panels. The overall girder bridge made of ten continuous spans, built in 1992–1993
strength of the structure thus progressively decreased until a on the old piers of a railway bridge dating from approximately
trigger event occurred. The spalling and fall of a small chunk of 1880.
concrete on the internal face of the arch was the first sign. The
lack of redundancy of the whole structure explains why this In February 2007, a flood affected the Saint-Etienne River
local event had a rapid and global consequence: the shell, during cyclone Gamede. The seventh pier of the downstream
already cracked, was fully penetrated by some struts, inducing bridge was swept away by the river flood, resulting in the
bending failure of an arch component. complete ruin of the bridge. The whole deck was found
fractured in many elements at the bottom of the river bed and
However, the committee considered that the event also all the piers were broken at variable heights except for only one
revealed ‘organisational failure’ and suggested some changes (pier 4), which was cracked but still upright. Fortunately, the
so as to prevent future collapses (Berthier, 2005). ‘twin’ bridge located further upstream (see Figure 5) did not
show any damage. At the time of the event, the flooding of the
& The structure, although complex, had been designed and river had not reached its maximum value: the flow was
controlled more like a ‘usual building’ than like a bridge or estimated to be 1000 m3/s at the time of the collapse, compared
a dam. The committee suggested that such structures merit with an estimated value of 2000 m3/s two days later. This flow
verification of the design by an independent body, not just is not very unusual, corresponding to a return period of about
reliance on contractor analysis and models. Regular 5 years. With the piers of the bridge having supported many
inspections of the structure should also be carried out to other cyclonic events since 1880, experts thus had to explain
check construction. why the collapse happened in such ‘mundane’ circumstances.
& Regarding the construction project itself, it was noted that
the designer was not independent from the client. The
committee recommended that a formal contract be agreed
between all stakeholders for all phases of a project.
& The design studies and structural checks suffered many
deficiencies. The committee pointed out, first, that such
studies must be detailed enough and with a sufficient budget
planned for them and, second, that sufficient means must be
dedicated to control in order to correctly manage the
complexity of such a project.

Finally, the committee suggested that all professional experts


meet in order to influence possible evolution of practice
(including regulation and contracts) in this field. The Roissy
collapse was a seminal event for French engineering compa-
nies, who decided to initiate collective research in the field of
project risk management in civil engineering. This field had
largely been ignored until Roissy, and a research project Figure 5. The two bridges on the Saint-Etienne river (Binet, 2007)
(ANR-GERMA) was founded in 2008 with the aim of editing

67
Forensic Engineering Forensic engineering and
Volume 165 Issue FE2 collapse databases
Breysse

Periodic inspections previously carried out on the bridge had In parallel with these general considerations, an analysis of the
not revealed any critical defect either on the deck or on the pier position and state of bridge debris, supplemented by calcula-
shafts (made of basalt masonry and covered by a concrete cap tions, was carried out to try to understand the mechanism of
beam). On the other hand, the piers (with heights ranging from the collapse and draw conclusions for the design of new
7 to 13 m) had long been recognised as vulnerable to scour bridges. Analysis of the debris showed that almost all the piers
because they were only based superficially on supporting soils collapsed by being drawn towards pier 7. During its fall to the
that had never been the subject of precise geotechnical ground, the part of the deck between piers 6 and 8 pulled the
investigation, thus leading to uncertainties concerning exact remainder of the deck horizontally. The sections of the deck
stability conditions of the piers (Binet, 2011). located between piers 6 and 8 moved towards pier 7, with a
displacement of about 1 m on the pier 6/8 supports. Such a
The Ministry of Public Works initiated a technical assessment displacement induced a significant horizontal force at the top
because of the strategic importance of the bridge and the of these piers by friction or blocking. Since the unreinforced
catastrophic collapse. The report stated that the collapse was masonry piers were not able to resist the horizontal force
due to scour at pier 7, followed by the collapse of the two exerted at their top by the displacement of the deck, they
adjacent spans, which then induced the progressive collapse of fractured and rocked (Binet et al., 2008).
all other spans in a series mechanism. The other piers were
unable to support the horizontal pulling force induced by the The report stated that structures with important social and
consecutive collapses due to their old masonry and/or defective economical stakes would merit a specific process management,
foundations (Binet et al., 2008), as illustrated in Figure 6. with regular updating of their vulnerability, particularly after
each important event. This reassessment should be the subject
of a pre-established procedure defining the nature of observa-
The vulnerability of pier 7 was not judged critical at the time of
tions to be made and should include the obligation of a formal
deck building (in 1993) but became so due to a combination of
decision by a person in charge at a suitable level on actions to
two phenomena: deepening of the river bed due to excess
be enforced. This decision process must also apply to the
gravel extraction and lateral movements of the river channels.
reassessment/rebuilding/reopening of structures that have been
After a previous flood in 2002 (cyclone Dina) certain piers were
damaged by a climatic event, because such events can recur.
considered for immediate protection, but pier 7, located rather
far away from the principal channel was not among them.
4. A continuous improvement process
Gravel extraction downstream had then been prohibited by the
The three presented cases provided the opportunity to develop
public authorities but the awareness that upstream extractions
a new approach for such problems. French public authorities
also had an impact on the bridge came later, and the
chose to make public the technical elements collected by
effectiveness of police measures against illicit exploitation
experts and conclusions, despite responsibility or confidenti-
was insufficient to restore the normal situation of a balanced
ality constraints. Even if each case remains specific, it is easy to
longitudinal average profile of the river. It was also clear that,
understand that lessons can be learnt from each case and from
at the time of the collapse, pier 7 was in the main channel
similar cases that have occurred around the world.
(because of fluctuations of the river channels), but the level of
priority of its protection had not been updated.
4.1 The forensic dance
The learning process after failures and catastrophic events has
been theorised by Duffey (2005). Figure 7 illustrates how
forensic engineering can be seen as participating in a whole
process: the ‘forensic dance’ (Wood, 2008). The four possible
uses of failure analysis are

& legal matters and responsibility


& research and improvement of knowledge
& education and learning
& improvement of standards and practices.

4.2 Improvement of knowledge


Case studies have a long history in the Anglo-Saxon world and
gave birth to rich specific literature. They are of value to
Figure 6. Collapse of pier 2 and adjacent spans (Binet, 2007) practitioners not only because they enrich technical expertise,
but also because they draw attention to ethical or juridical

68
Forensic Engineering Forensic engineering and
Volume 165 Issue FE2 collapse databases
Breysse

Better standards Design


and regulation Contract

Building process
Changing standards
Banning And practices
(asbestos,etc.)

Education Use
Maintenance

Research (cost) Investigation


needs
(cost)

Failure
Conflict

Figure 7. Forensic dance (after Wood, 2008)

matters. Many civil engineering students have seen movie failure had not been anticipated and the passage of many years
footage of the Tacoma Bridge collapse of 1942. It remains the was necessary before UK authorities accepted the use of post-
most famous case of a new phenomenon being discovered due tensioned concrete bridges, with additional requirements
to a failure: aero-elastic coupling (‘Von Karman whirls’) was regarding design and building.
only fully understood many years after the event. Other
failures that have led to significant gains in knowledge include Over the last 20 years, urban tunnels have provoked many
the following. failures with, in some cases, dramatic consequences at the
surface for people and existing structures (e.g. Heathrow and
& Several catastrophic steel bridge failures in the 1970s Munich in 1994, Taegu in South Korea in 2000, Meteor in
occurred before major research programmes enabled Paris in 2003, Nicoll highway in Singapore in 2004, Lane Cove
design codes to be developed to cover loading conditions in Sydney in 2005 and Köln in 2009) either because the boring/
and the buckling of slender steel elements in box girders in digging techniques were not well tested or because the risks
final design and progressive building stages. were not correctly considered (Breysse, 2010). These accidents
& London’s Millennium Bridge and its sister, Solferino Bridge led insurers to ask for safety improvements and risk mitigation,
in Paris, both with low lateral stiffness and inherent or insurance cover would be withdrawn. A collective change in
damping, experienced discomforting vibrations that design and construction practice was necessary and this led to
revealed hitherto little known mechanical coupling phe- the publication of several texts that better define ‘good
nomena between pedestrians and a structure. practice’ (e.g. ITIG, 2006).
& The sudden failure of a metallic bar in Uster swimming pool
(Switzerland, 1985) initiated the consideration of stress 4.4 Improvement in standards
corrosion when the atmosphere contains a high concentra- Some failure events, often because of their high impact on the
tion of chlorides and use of the correct grade of stainless general public or on political leaders, have resulted in substantial
steel. research programmes and improvements in standards and
regulations. This was the case for seismic standards (the first
4.3 Improvement in practice version of French standards followed the Orleansville–El
Failure analysis can also help to improve professional practice, Asnam earthquake in 1954), fire safety in European tunnels
as illustrated by the following examples. after dramatic accidents like that of the Mont Blanc tunnel in
1999 and, more recently, for pedestrian footbridges.
The sudden failure (fortunately without dramatic conse-
quences) of Ynys-y-Gwas Bridge in Wales in 1985 revealed Concepts such as the importance of ‘structural robustness’,
that the metallic cables of the prestressed structure were highly which gives a structure some safety against localised failure,
corroded due to water and road salt penetrating the joints. The were highlighted by the Ronan Point collapse in London in

69
Forensic Engineering Forensic engineering and
Volume 165 Issue FE2 collapse databases
Breysse

1967. At Ronan Point, a localised gas pipe burst on one of the useful data while keeping facts confidential and offers a very
highest floors of the residential building and caused the cascade attractive way to add value to feedback. The Cross
collapse of the lower floors, showing the poor design of (Confidential reporting on structural safety) process was
prefabricated connections. This was also the case in a famous developed to reduce the number of accidents (or near-
storm that devastated France in December 1999 when steel accidents) involving structures and to improve reliability
towers supporting electric cables collapsed in series because of throughout their service life. Cross involves getting feedback
the destabilising effect induced by the collapse of the first from professionals (through confidential reports) and writing
tower. reports when needed, in order to influence practice. All reports
are analysed by independent experts and, after deletion of
This concept was given more emphasis after the horrific information in order to guarantee anonymity, they are
collapse of the World Trade Center in 2001, which led to an published. If facts emerge because similarities appear between
additional requirement for engineers to prevent total collapse different reports, in-depth research programmes can be
even after an explosion. In all these cases, the priority is to developed.
reduce the vulnerability of the structure and to ensure that a
level of safety is maintained (avoiding total and sudden Another database – not limited to civil engineering fields – is
collapse) even in an accident situation. being developed in Japan by Hatamura et al. (2003) (http://
www.shippai.jst.go.jp). The main question Hatamura asks is ‘is
4.5 Knowledge build-up and education it possible to develop useful and consistent models for failure,
Teaching on the basis of lessons learned from past failures can from the analysis of unique cases, since they can be many but
be very efficient. Failures or accidents can be used also remain specific?’. The issue of bridging the gap between a
descriptive and naturalistic approach to case studies and a
& at a statistical level, since repeated occurrence generic approach has no easy answers. It is, however, the basis
sometimes enables identification of laws and of any scientific approach – being able to find invariants,
quantification of occurrence probability logical causalities, etc. upon which (deterministic or statistical)
& at a qualitative level when a case is strong enough to rules will be identified.
draw conclusions that cannot be forgotten
& at an intermediate level when apparently isolated Hatamura et al. (2003) has chosen to develop a three-step
cases reveal, after thorough analysis, some generality that logical model as in a fault-tree analysis – causes/actions/
makes it possible to go from particular to wider consequences – in the form of three mandalas (in Buddhism,
conclusions with practical applications. mandalas symbolise the world in its totality). Each mandala
corresponds to a two-level classification and makes it possible,
Delatte (2000) developed a case study database for learning for each individual case, to identify a pattern and thus a group
purposes. For each case, the (unique or multiple) cause(s) for of cases to which it belongs. For instance, the mandala of
failure were identified and discussed, and the cases were thus causes (Figure 8) contains ten classes at the first level of detail
grouped into categories according to their causes. This concept and 27 classes at the second level. These classes are defined
needs to be developed in many fields – roof collapse due to according to the fact that the cause mainly comes from an
snow overload, instability collapse (such as buckling), dis- individual, an organisation, the societal context or that nobody
orders induced by tunnel work and accidents in temporary can be considered responsible (unknown or unpredictable
work being a few examples. events).

5. Added value of failure databases and This approach is too general to have direct application, but it
possible developments does provide a framework that can be applied to specific types
5.1 What can be learnt from databases of structures (dams, bridges, tunnels, etc.) or situations. It will
thus be possible to consider more detail and obtain more
The development of failure databases facilitates organisation
of information, recording it for the future and disseminating it fundamental conclusions, both for statistical analysis of the
worldwide. Databases have particular value in identifying failure processes and for practical risk management.
factors that give rise to disproportionate risk that only
becomes clear from the analysis of larger scale data as distinct Several other examples of files or databases exist, such as those
from individual cases. developed by Bailey et al. (2001) focusing on safety analysis of
bridges, Xu and Zhang (2009) devoted to modelling breaches
The reporting system developed in the UK under the auspices and floods on dams and Delatte (2000), which contains a very
of Scoss (Standing committee for structural safety) spreads limited number of in-depth analysed cases for teaching

70
Forensic Engineering Forensic engineering and
Volume 165 Issue FE2 collapse databases
Breysse

Individual is responsible Lacked prior research No one to blame


Lacked environment study
Misread situation
Misrecognition Lacked virtual exercise
Lacked
Wrong understanding analysis Environment changed
Economy changed
Narrow view
Misjudgement Change of
Ignored documents condition Lacked rights research

Poor communication Ignored Poor organisation


procedure Root Poor
Bad health cause of planning Poor strategy
failure
Lacked precaution
Careless Different culture
Bad
Poor understanding value Bad corporate culture
Ignored tradition Ignorance Bad Poor safety awareness
operation
Lacked knowledge Inflexible operation
Unknown Poor management
Abnormal
Discovery Poor staff Organisation is responsible
No one is responsible

Figure 8. Mandala of causes (after Hatamuras et al. 2003)

purposes. Other groups are also gathering information taking 4. Since no common failure database exists at the moment, the
an historical perspective, or focussing on a type of structure. following considerations are based on a first initiative
promoted by the author.
5.2 The start of forensic engineering in France
Forensic engineering in France does not have a long history. A A prototype database is being developed from various sources
research group devoted to risk management in civil engineering such as technical reports on case studies, existing databases
(MR-GenCi) was founded by the author in 2003 (http://www. and information from the internet. The database presently
mrgenci.u-bordeaux1.fr/) and now more than 20 partners contains more than 650 cases. Tables 1 and 2 provide some
(authorities, research teams, insurers) are involved in risk and rough statistics about the cases for the period 2007–2010. The
safety issues. However, there is still no collective effort devoted high number of cases from Asia is mainly due to contributions
to the gathering and analysis of failure or near-failure data. from China and India, where 38 and 49 failures have caused
The civil engineering profession, as a whole, is just beginning to about 650 and 450 fatalities respectively. It can be seen that the
consider these issues and the reluctance to share and spread statistical distribution between geographical areas or types of
information about problems and failures remains a significant structures is relatively stable across the years. However, several
hurdle. Information is often said to be confidential, which is weaknesses due to the data gathering process must be noted.
sometimes true because of judicial matters.
& One important source of data is the internet and press
French engineers and civil engineering students are not used to releases. The problem here is not so much the reliability of
considering failure events. It therefore seems important to open data as the often limited amount of technical data. The
up a new perspective by showing the value of a failure quality of information in the database is thus very
database, with all possible improvements discussed in Section heterogeneous.

Europe America Asia Africa Oceania Total

2007 8 17 20 6 — 51
2008 11 15 20 7 — 53
2009 20 22 48 9 1 100
2010 17 34 49 12 3 115

Table 1. Number of cases in the prototype database according to


geographical origin

71
Forensic Engineering Forensic engineering and
Volume 165 Issue FE2 collapse databases
Breysse

Buildings Bridgesa Roofs, domes Damsb Tunnels Other Total

2007 13 17 2 3 3 13 51
2008 20 15 2 3 2 11 53
2009 28 39 8 7 8 10 100
2010 40 26 12 15 12 10 115

a
Including footbridges
b
Including dykes and tailing dams

Table 2. Number of cases in the prototype database according to


type of structure

& The collection process consists mostly of continuous demolition works. For the first category of failures in service,
monitoring of news events (plus contacts with experts and Figure 9 shows statistics regarding the primary cause of failure
specific attention paid to ‘historical cases’). The conse- so far as it can be identified from the available data. The
quence of this is that the sampling process does not offer the numbers are, however, only indicative because, in many cases,
guarantee of representative sampling; for example, the failure resulted from a combination of excessive load and a
higher number of events for 2009 and 2010 is probably due given material condition, and lack of maintenance or defective
to increased reporting and research effort rather than to a design can also contribute to such a collapse. The graph shows
multiplication of failures. Another bias is due to sources: that in more than half of the cases, failure was due to excessive
English and French language sources are noted whereas loading either because of a climatic origin (like excessive wind
others are disregarded or ignored. or river surge) or traffic consequences (overloading or impact).
& Other choices can have consequences on sample representa- The ‘climatic’ category covers mainly flooding while ‘material’
tiveness, since a given flood or earthquake can cause the mainly covers problems of fatigue and ageing (corrosion,
collapse of many structures. In such cases, the choice is to creep, decay, etc.). Even if the statistics are only indicative, the
select only a few of those structures for the database. large variety of causes can be noted and this is particularly
valuable for teaching purposes.
For these reasons, the prototype database cannot yet be
considered as a reliable research tool but rather a demonstra- Another illustrative result is the creation of F–N curves, which
tive tool of what could be achieved once the organisational represent the annual probability of death after a structural
problems of collecting comprehensive reliable data are collapse (also called fatal accident rate (FAR)) as a function of
identified and solved.
1 10 100 1000
The database contains 223 cases of bridge failures, of which
5 10-8
169 were failures during service, 47 were failures during
construction and 7 were unexpected failures during repair or

40
35
Number of cases

30
F 5 10-9
25
20
15 Buidings
10 Dams
5 Bridges
0 Tunnels
ic

ad

ct

cs

re

us

n
ia

5 10-10
ig

w
at

pa

is
Fi
er

ni

rio
rlo

no
es
lim

al
ch
Im

at

N
va
ve

nd
D

nk
M
C

te

s-
O

Va

U
eo

er
G

th
O

Figure 10. F–N curves for four types of structures, from 276 events
Figure 9. Primary causes of failure for bridges in service in the prototype database

72
Forensic Engineering Forensic engineering and
Volume 165 Issue FE2 collapse databases
Breysse

1 10 100 1000
5 10-7

F 5 10-8

Buidings
Dams
Bridges
Tunnels

5 10-9
N

Figure 11. F–N curves for four types of structures, from 276 events
in the prototype database assuming effective exposure ratios

the magnitude of the event (number of fatalities). The each year for a reference at-risk population of 65 million
probability value for each fatality number F is calculated by inhabitants such as in France.
dividing the cumulated number of victims due to events with at
least F victims by the number of years and the population at risk. Such curves can help regulators compare the real ‘safety level’
Such F–N curves are known to provide data (and guidelines for of different contexts (different countries or types of structures)
decisions) regarding the acceptability of risk levels. or compare performance (feedback) to safety objectives.
Further analyses are already possible, such as comparing the
Figures 10 and 11 are built from events documented in the level of risk between various types of countries or for certain
prototype database, in which 473 events that resulted in 276 situations (e.g. during erection). In this case, the issue of
fatalities were selected for the period 1991–2010. The events building safety overlaps that of building project risks, and a
were classified into building failures (including roof and dome database would be useful in understanding how collapses are
collapses), bridge and footbridge failures, dam and tailing dam related to procurement conditions, supervision arrangements
failures, and tunnel failures. For these four types of events, the or the local context (Morand et al., 2011).
number of documented events with fatalities is respectively
159, 74, 23 and 20. Figure 10 assumes that all the population is 6. Conclusions
equally at risk full-time. Figure 11 is a variant accounting for Despite recent changes initiated by the public authorities and
an average exposure rate (the population at risk is not the research teams, forensic engineering is still in its infancy in
whole population or it is not at risk full-time). In this case, one France. The interest of engineers and building companies in
must assume a multiplying factor, to consider the effective risk management issues is increasing, although some reluctance
exposure ratio, which is respectively equal to 1, 0?1, 0?05 and to share information remains. A first research project (Morand
0?01 for the four types of structures. When comparing the two et al., 2011) is devoted to preparing guidelines for better
figures, it can be seen that the result is very different since the managing risk in complex construction projects. The develop-
more dangerous structures in the first case (buildings) are the ment of forensic engineering practices and a failure database
safest in the second. Of course, assumptions on the effective will represent a wide field of progress.
exposure ratio have a large influence on the results and must be
improved in further analysis. Independent of cultural and psychological dimensions, one
main limit for the development of approaches based on
Because of the prototype character of the database, these feedback is the data gathering process. It is thus important
results are only illustrative. One can, however, check that the to consider issues such as the data collection process (the type
magnitude of the probability F, between 1?061028 and of data to gather, validation process, etc.), organisation of the
5?061028, is typical for such sources of risk (Breysse, database, access for specialists and/or the general public,
2009b): a value of 1027 corresponds to about 6–7 fatalities property and confidentiality (Breysse and Harouimi, 2007).

73
Forensic Engineering Forensic engineering and
Volume 165 Issue FE2 collapse databases
Breysse

The process developed by Scoss could serve as a model and Breysse D (2010) Les risques attachés aux tunnels urbains et
common decision-making and collective efforts between all l’amélioration des pratiques professionnelles. Coll. Génie
parties (authorities, professionals, teachers) are possible and Urbain et Infrastructures Souterraines, Marne la Vallée,
would be fruitful. France.
Breysse D and Harouimi R (2007) Comment mieux tirer parti des
A first step has been made with the creation of a committee of accidents de structure? Préventique Sécurité 95:
risk management in building projects under the common 36–40.
auspices of AFGC (French association of civil engineering) Breysse D, Duprat F, Melacca V, Motet G and Schoefs F (2011)
and GIS MRGenCi (scientific group for risk management in Ingénierie forensique: améliorer les performances, anticiper
civil engineering). This committee aims to prepare guidelines les défaillances. Journée Ingénierie Forensique, GIS
on methods and practices for better managing these risks. One MRGenci-FoncSI, Paris, France.
of its tasks has been, in relation with insurers, to discuss the Carper KL (2007) ASCE technical council on forensic
development of databases. engineering: enhancing the influence of forensic engineering
in the United States. Proceedings of International Forensic
Acknowledgements Engineering Conference, Mumbai, India.
Thanks go to Jennifer Henry for her improvements to the Delatte NJ (2000) Using failure case studies in civil engineering
manuscript. The author also acknowledges Prof. Jonathan education. Proceedings of 2nd ASCE Congress on Forensic
Wood for his continuous and useful support. Engineering, San Juan, Puerto Rico. ASCE, Reston, VA,
USA, pp. 430–440.
REFERENCES Dubois P and Rat M (2003) Effondrement sur le Chantier
Bailey SF, Antille S, Béguin P, Imhof D and Brühwiller E (2001) ‘Meteor’. Report for Ministry of Public Works, Report
Niveau de Sécurité Requis pour l’évaluation de Ponts-routes 2003-0054-01.
Existants. Ecole Polytechnique Fédérale, Lausanne, Duffey RB (2005) Predicting tragedies, accidents, errors and
Switzerland. failures using a learning environment. Science and
Berthier J (2005) Aéroport Charles de Gaulle – Jetée du Terminal Technology of Advanced Materials 6(8): 878–882.
2E. Présentation du rapport de la commission Hatamura Y, Ilno K, Tsuchlya K and Hamaguchi T (2003)
administrative, METATTM, 15 February. Structure of failure knowledge database and case
Binet C (2007) Effondrement du Pont de la Rivière Saint-Etienne expression. CIRP Annals Manufacturing Technology 52(1):
sur l’ı̂le de la Réunion. Conseil Général des Ponts et 97–100.
Chaussées, Paris, France, Report 005170-01. ITIG (2006) A Code of Practice for Management of Tunnel
Binet C (2011) Effondrement du pont de la rivière Saint- Works. International Tunnelling Insurance Group.
Etienne. Journée Ingénierie Forensique, GIS MRGenci- Moncarz PD and Taylor RK (2000) Hyatt failure from the
FoncSI, Paris, France. perspective of a forensic engineer. Proceedings of 2nd
Binet C, Lacombe JM and Martin I (2008) Collapse of the bridge
ASCE Congress on Forensic Engineering, San Juan, Puerto
over the Saint-Étienne river in the island of La Réunion. Rico. ASCE, Reston, VA, USA, pp. 28–36.
Proceedings of Institution of Civil Engineers 4th
Morand D, Perret P, Demilecamps L and Machu A (2011)
International Conference on Forensic Engineering, London.
Germa, a national research project on risk control and
Thomas Telford, London, UK, pp. 204–214.
management in complex construction projects.
Bordes JL (2005) A propos de l’histoire des barrages en
Proceedings of ICASP Conference, Zurich, Switzerland.
France, quelles leçons du passé? Annales du BTP Eska 2:
(Faber MH, Köhler J and Nishijima K (eds)). Taylor &
5–26.
Francis, London, UK, pp. 234–235.
Breysse D (2003) Dimensions humaines et sociales de la maı̂trise
Rens KL, Rendon-Herrero O and Bosela PA (2000) Forensic
des risques de défaillance d’ouvrages. Proceedings of
GC’2003, Environnement et Patrimoine: Les Nouvelles engineering. Proceedings of 2nd ASCE Congress on Forensic
Donnes, Paris, France. Annales du Bãtiment et des Travaux Engineering, San Juan, Puerto Rico. ASCE, Reston, VA,
publics, Eska, pp. 6–18. USA.
Breysse D (2009a) Ingénierie forensique: la défaillance Rossetto T and Elnashai A (2003) Derivation of vulnerability
comme base de savoirs. Rencontres AUGC, Saint-Malo, functions for European-type RC structures based on
France. See http://www.augc.asso.fr/dl/articles/AUGC_ observational data. Engineering Structures 25(10): 1241–
article_00127.pdf (accessed 27/03/2012). 1263.
Breysse D (2009b) Maı̂trise des Risques en Génie Civil, Volume Roubault M (1970) Peut-on prévoir les catastrophes naturelles?
1: Multiples Dimensions des Risques en Génie Civil. Hermes Presses Universitaires de France, Paris, France.
Lavoisier, Cachan, France. Schneider J (1997) Introduction to Safety and Reliability of

74
Forensic Engineering Forensic engineering and
Volume 165 Issue FE2 collapse databases
Breysse

Structures. International Association for Bridge and Xu Y and Zhang LM (2009) Breaching parameters for earth
Structural Engineering, Zurich, Switzerland. and rockfill dams. ASCE Journal of Geotechnical and
Wood JGM (2008) L’ingénierie Forensique – l’approche Geoenvironmental Engineering 135(12): 1957–
Anglaise. Colloque Le Pont, Toulouse, France. 1970.

WHAT DO YOU THINK?


To discuss this paper, please email up to 500 words to the
editor at journals@ice.org.uk. Your contribution will be
forwarded to the author(s) for a reply and, if considered
appropriate by the editorial panel, will be published as
discussion in a future issue of the journal.
Proceedings journals rely entirely on contributions sent in
by civil engineering professionals, academics and stu-
dents. Papers should be 2000–5000 words long (briefing
papers should be 1000–2000 words long), with adequate
illustrations and references. You can submit your paper
online via www.icevirtuallibrary.com/content/journals,
where you will also find detailed author guidelines.

75

You might also like