ANALYSE DES CAUSES REELLES DE LEFFONDREMENT DU TERMINAL DE ROISSY

1
ANALYSES OF THE REASONS OF ROISSY TERMINAL 2E
COLLAPSE IN FRANCE USING DETERMINISTIC AND
RELIABILITY ASSESSMENTS
Roy FEGHALY (Presenter) Wassim RAPHAEL Fouad
KADDAH
2
INTRODUCTION

• INAUGURATION JUNE 2003
• COLLAPSE MAY 2004
• LENGTH 650 m
• COST 750 MILLIONS OF EUROS
• NEGATIVE IMPACT ON THE IMAGE
• OF THE FRENCH CIVIL ENGINEERING.

STUDY
Examine the real reasons of the incident and
check if it were possible to predict the
structure deficiency from the beginning.
3
PLAN OF THE PRESENTATION
GENERAL DESCRIPTION OF THE STRUCTURE
FINITE ELEMENTS SIMULATION MODEL USING ST1
SOFTWARE
MECHANICAL RELIABILITY ASSESSMENT USING A
COUPLING PROCEDURE BETWEEN 2 SOFTWARES
DETERMINISTIC ANALYSES OF THE STRUCTURE
PROPOSITION OF SOME POSSIBLES SOLUTIONS
CONCLUSION ET RECOMMANDATIONS
4
DESCRIPTION

• MAIN BUILDING
• EMBARKING PLATFORM
• ISTHMUS

Main building
North
Isthmus
Footbridges
Sinister zone
Embarking platform
Pre-footbridges
South
5
2 particular shells zones 144 and 146
Opening of the Footbridge
Cross Section
6
EXTERNAL VIEW OF THE SHELL
Reinforced Concrete Shell
Glass
Connecting Rod
Tensional member
Stay
7
The shell of zone 146 is made of juxtaposed full
arcs and perforated arcs (Note the complexity of
the structure)
Tensional member

Full arc
Perforated arc
Horizontal beam
Keying
Opening of a ventilation duct
Stay
8
PROTOTYPE MADE OF 2 ADJACENT ARCS
14 m
31 m
Horizontal Beam
Elastomeric Bearing
9
INSTANT DEFORMATIONS OF ARCS AFTER THE ADJUSTMENT
PHASE
Deformation on the Keystone 11 cm instead of
5cm Predicted Rotation on the
Bearings 1.4 deg instead of 0.5 deg
Predicted And the deformations continued
increasing in the course of time
10
THE COLLAPSE
Collapse of a part of zone 146 causing 4 dead and
3 injured
11
THE MODEL
GOOD SIMULATION MODEL PERFORMED
1-Openings in the shell. 2-Keyings between the
segments. 3-Combined system of tensional members,
stays and concrete. 4-Irregular cross section
of the horizontal beams. 5-Openings in the
horizontal beams due to the crossing of
ventilation ducts. 6-Connection between the arcs
by fragile corners iron. 7-Simulation of the
elastomeric bearings by means of a complete
stiffness matrix.
FINE SIMULATION BY FINITE ELEMENTS
A-
TAKING INTO ACCOUNT THE LONG TERM EFFECTS OF
MATERIALS (CREEP AND SHRINKAGE) BY MEANS OF ST1
SOFTWARE
B-
12
1588 nodes 2320 bars
MODEL OF THE FIRST EIGHT ARCS STARTING FROM THE
ISTHMUS
East side
Elastomeric bearings
Isthmus side (West)
Horizontal beam
13
SIMULATION OF THE HORIZONTAL BEAMS
Horizontal Beams
HORIZONTAL BEAMS IN NORMAL ZONE
Neutral axis
Isolated cross section
Irregular cross section of the horizontal beams
14
SIMULATION OF THE FRAGILE CORNERS IRON
15
SIMULATION OF THE ELASTOMERIC BEARINGS
CARACTERISTICS
a 300 mm b 450 mm n 5 plates t 10 mm e
3 mm
All calculations being done using the technical
data of FREYSSINET INTERNATIONAL
16
RESULTS OF ST1 MODEL AND COMPARISONS
The expression  Short term  means after the
arcs adjustment phase (63 days after casting)
The expression  Long term  means at the
collapse date (915 days after the casting)

• VALIDITY AND ACCURACY OF OUR ST1 SIMULATION MODEL
• BAD SIMULATION MODEL PERFORMED BEFORE
  CONSTRUCTION
• IMPORTANCE OF TAKING INTO ACCOUNT THE LONG TERM
  EFFECTS OF MATERIALS

17
DETERMINISTIC INFLUENCE OF SOME PARAMETERS
This part aims at studying the effect of the
following parameters
Maximum relative variations of deformations due
to the different studied parameters

• The effect of creep is the most important among
  the studied factors.

• Little influence of shrinkage and thermal
  effects. The variation of
• temperature which preceded the sinister was not
  the main cause
• of the collapse.

• No practical influence of other parameters.

18
DETERMINISTIC JUSTIFICATION OF THE COLLAPSE
1. Punching shear of the shell by the stays
Maximum compression forces in the stays at the
serviceability limit state
The ultimate punching shear strength of our shell
using BAEL or the SIA 162
142.89 T
At short term Vd (Short term)125.37T lt
VRu142.89T ? Everything is fine At long term
Vd (Long term) 149.44T gt VRu142.89T ?
Strength exceeding
This explains the collapse after a period of time
19
The shell is shear punched by some stays.
20
2. Rotation of the bearings
The elastomeric bearings have an allowable angle
of rotation
aall
Comparison of the elastomeric bearings rotations
with the allowable rotation

• The rotation of the bearings is unacceptable at
  the short term and the long term.

• The rotation at the long term is more than double
  of what is allowable.

• This reveals the initial deficiency of the
  structure and explains the toppling over of some
  elastomeric bearing supports and the slipping of
  the horizontal beam out of its supports

21
The horizontal beam slipped out of the supports
22
RELIABILITY-BASED ASSESSMENT OF THE STRUCTURE
A software for the mechanical behavior
calculations (Finite Elements) which is ST1
This requires a mechanical-reliability
coupling procedure between 2 software
A software for probability calculations which is
Phimeca. Phimeca prepares the input file of the
studied structure to be analyzed by ST1.
Phimeca evaluates the failure probability
PHIMECA selects a realization of random variables
calls ST1
Yes
ST1 calculates the deformation for the given
realization of random variables
convergence is reached ?
A new realization of random variables
No
Iterative scheme
23
Coupling procedure in calculation progress
24
The compressive strength of concrete fc28
Creep is primarily sensitive to three parameters
that are considered random
Deterministic and reliability studies
The hygrometry ?h
The density ? of concrete
-Mean radius rm -Steel quantity ?s -Loading date
t0
Negligible influence
Deterministic
Random variables of the structure
25
No architectural elements likely to be damaged,
thus the permissible level of deflection is
calculated as follows
(where l is the span length).
The failure probability
Results of reliability calculation
Deficiency at the short term also
The reliability study
Predicting creep strains
The reliability study is a very interesting tool
in
Bringing complementary information concerning the
judgment of structures validity
26
ANALYSES AND POSSIBLE SOLUTIONS
Would the introduction of structural
reinforcements in the initial design have
prevented the catastrophe?
1 QUESTION COMES TO OUR MIND
So we have considered 2 possible solutions

• Use of tensional members that link the end of
  the arcs at the bottom level.
• Improve the compressive strength of concrete by
  taking fc28 55 MPa instead
• of 40 MPa.

In this case, the check of the structure shows
that it becomes acceptable
27
CONCLUSION

• The results have revealed the bad prediction done
  by the design office before construction for the
  following reasons
• A rough simulation of the structure.
• Not taking into account the concrete
  deformations
• at the long term (especially creep
  deformation).
• Not doing a reliability calculation.

This explains the unconformity of the design
values with the measured values on site.
Shear Punching Of the Shell
Excessive forces In the stays
Deterministic Deficiency at Long term
Long term deformation
Deterministic
Slipping Of the Horizontal Beam out Of
its Supports
Big rotations of the bearings
Our Study
High probability Of exceeding The
allowable Deflection
Probabilistic Deficiency at Short term
Probabilistic
28
Moreover, the results of our model ST1 are in
conformity with the measured values on site
This confirms the validity of our model and shows
that it was possible to predict the structure
deficiency from the beginning.
Recommendations for the future

• Perform calculations taking into consideration
  the long term effects of
• materials (creep, shrinkage, relaxation ).
• Make fine models of the structures in order to
  be faithful to reality,
• especially when it concerns non classic
  edifices from the structural or
• architectural point of view.
• Perform mechanical reliability calculations to
  assess the reliability of
• constructions especially when the failure
  causes serious consequences.

Whichever happened before, the punching of the
shell by the stays or the failure of the keyings ?
The keyings between the segments were also broken
It would be interesting to make a plastic
calculation of the keyings.
29
HOPE WE NEVER HAVE TRAGEDIES
THANK YOU
30
(No Transcript)