GREAT BELT TUNNEL REPAIRS AND REFURBISHMENT FOLLOWING A FIRE
Niels Peter Høj, COWI, Denmark
Chris Tait, Mott MacDonald, UK
ABSTRACT
On the Storebælt Tunnel an uncontrollable fire broke out in one of the tunnel boring machines on 11
June 1994. All personnel were able to evacuate the tunnel safely. This paper describes the immediate
findings, the investigations, the methods for recovery and finally the full-scale fire tests, which were
carried out on spare tunnel segments in order to provide better data to be used in the update of the
Operational Risk Analysis.
1. THE FIRE EVENT
For the Great Belt Eastern Tunnel, which opened to traffic in spring 1997, had its share of problems
during construction. One of the obstacles for the completion of the tunnel was a fire, which broke out
in the tunnel boring machine on the morning of 11 June 1994. The boring machine, called Dania, was
one of 4 TBMs. Dania was the north tunnel TBM boring west. At the time of the fire the two north
tunnel TBMs were only 50 m apart, and the south tunnel TBMs had already met.
The cause of the ignition is still unknown but the primary fuel for the fire was hydraulic oil from the
TBM hydraulic power system, thought to be leaking and atomizing at high pressure from a ram hose
connection. About 2000 l Shell Tellus 46 was contained in the storage tanks were consumed. A
methane alarm was tripped, probably by the product of combustion gases, and operated a backup
power system and emergency lighting. Black smoke seriously impaired visibility, but all personnel
were able to evacuate safely using the temporary walkways and crossing into the south tunnel. Fire
suppression systems consisted of hand held extinguishers, water hoses, a low density foam "plug".
Those systems with a chance of success were attempted by trained personnel wearing breathing
apparatus but were not able to extinguish the fire. The fire brigade soon arrived but after some
unsuccessful attempts to fight the fire, they had no choice but to allow the fire to burn itself out. The
duration of the fire was thought to be in the range of 4 - 8 hours.
2. INSPECTION AFTER THE FIRE
As soon as the fire had cooled sufficiently, a number of inspections of the damage took place.
Photographs were taken (Fig. 1) and the depths of the spalled concrete in the crown precast concrete
reinforced segments measured. At the worst affected point in one segment two thirds of the originally
400 mm thick segments had spalled away. The average spalling in the worst affected segment was 150
mm. The spalled zones of damaged segments occurred mainly in the crown region about 10 rings back
from the TBM tailskin.
As an emergency safeguarding measure the contractor infilled the spalled segments with reinforced
shotcrete to reinstate the original segment thickness with the eventual object of propping the linings,
but within a few days, concern over the increasing inflow of groundwater through the badly cracked
segments led to an evacuation of personnel without the propping being achieved. Arbejdstilsynet
(Danish Health and Safety Authority) prohibited further entry until a full investigation had been
undertaken with risks studied and minimized and which had been subject to adequate independent
checking.
Fig. 1 The tunnel lining damaged after the fire
Although exposed cabling on
the TBM had been badly
affected by the fire and there
was severe scorching and
blackening by soot in the
crown, there was no obvious
damage to the structural steel
forming the main body and
tailskin of the TBM. The
muck in the screw conveyor
trapped by the closed rear
door of the screw preventing
blow-out. The TBM rams
had non-return valves, so
unless the seals had melted,
the rams in the invert would
still contain oil and prevent
the TBM from being pushed
back.
It was decided to reduce the risk of inundation of the rest of the works if the linings were to fail, by
constructing a bulkhead at the east portal of the north and south tunnel. The bulkheads had flood doors
to allow a passage of service train and also accommodated pumping pipes and ventilation which was
ongoing. Having achieved this position of comparative safety, methods of recovery were worked up in
more detail.
3. DAMAGE TO THE LINING
The parameters affecting the remaining strength of the tunnel were
•
•
•
•
•
amount of segment remaining after spalling
temperatures reached during the fire and consequential concrete strength loss
possibility of rehydration of cement
position of damage within the ring relative to possible disturbed ground outside the lining
squatting of the ring
The longitudinal profile of the spalling damage of the worst affected rings is shown in Fig. 2. This was
the result of the rather hurried survey taken after the fire. It was also apparent that the circumferential
sides of the segments and the radial joints were reasonably intact, possibly due to the throughthickness reinforcement in those regions placed to prevent splitting of the segment.
Fig. 2 Longitudinal section in crown showing profile of spalling damage
It was known that the residual overall strength was affected by the peak temperatures that the concrete
would have reached at the various depths both in the body and the joints of the segment. The fire
temperature was uncertain and use was made of a hydrocarbon fire curve similar to that applied by
RABT for petroleum fires in Germany, with a constant temperature of 1200oC reached after 5 min.
Early tests of spalled pieces ascertained that they must have reached at least 600oC
The concrete is a very dense C50/60 design mix but which had been achieving average cylinder
strength of 76 MPa in production. The concrete contained OPC and granite aggregates and had a
water/cement ratio of 0.32. It also contained plasticizer and superplasticizer and had flyash and
microsilica in proportions high enough to prevent swelling of calcium oxide in the concrete and
consequential bursting the of the concrete matrix. Because previous tests within fire research has been
carried out on more ordinary structural mixes, it was decided to ascertain the behaviour of the concrete
when subjected to high temperatures. Cylinders cored out of spare segments were heated to temperatures of 300oC, 500oC and 700oC and later tested for compressive strength and elasticity. The tests
were undertaken by Danish Fire Institute and Danish Technical Institute in Copenhagen. The results of
the tests can be seen in Fig. 3 and Fig. 4. It was found that the shapes are similar to those found for
normal concrete. The relationships between strength and temperature was, respecting the strain
compatibility, applied together with calculated temperature profiles through the body of the segment in
order to find the residual strength. Because of the uncertainties in the assessment a "best estimate" and
a "worst credible" approach was used.
H eating
C ooling
T estresults
1,0
0,8
1,0
0,8
0,6
0,6
0,4
0,4
0,2
0,2
0,0
0,0
0
200
400
600
800
1000
Tem perature (C )
Fig. 3 Relative residual strength after cooling
0
200
400 600 800
Tem perature (C)
1000
Fig. 4 Relative residual modulus of elasticity
after cooling
4. GROUND AND WATER LOADS ON THE DAMAGED SEGMENTS
Whilst the remaining strength was being assessed a parallel exercise was taking place to ascertain the
likely loads on the tunnel at the fire location. The basic strata are shown in Fig. 5. The TBM was
capable of operating in closed mode and so could excavate in safety, should any pockets of running
sand or glacial melt water in the clay till be encountered suddenly.
However, what had proved impossible to control at this location was sudden over-excavation due in
part to the slow rate of advance through the underlying marl up to axis level with a weak horizon of
sand in the crown ready to "fall in". This collapsing of the crown had led to successive forming of
pipes or chimneys to the surface and thereby a direct hydraulic connection to the sea of Great Belt
above. The impact of the direct hydraulic connection meant that an external dewatering exercise that
was using wells drilled from the seabed could be quickly rendered ineffective. The dewatering had
been operating successfully to facilitate TBM operation and maintenance, and help stabilize the
ground during construction of cross passages. It reduced the full hydrostatic pressure of 4.5 bar at the
fire location down to 3.2 bar; therefore it was considered vital to keep it operating and not to allow
loads to increase on the damaged linings and so to decrease what safety margin may still have existed.
The sink holes had been plugged when observed by divers using clay from bottom-dump barges.
A two-dimensional finite element slice model of the ground and lining was analysed postulating
various scenarios regarding the collapsed ground and its eventual restoration. By comparing the
various load cases and ranges of strengths, a better picture of likely safety factors began to emerge. In
order to increase these still further the eventual recovery method was carried out in 0.75 bar of
compressed air in front of a forward bulkhead and this gave added confidence that safety margins must
have been increased. However, care was taken not to use excessive air pressure in case air escaped
through the ground, causing it to be disturbed and so increase loads.
5. RECOVERY, RISKS AND EVACUATION
Four basic approaches to recovery and repair were identified for consideration:
• Place a new lining inside the existing damaged lining
• Plug the damaged length of the tunnel with weak concrete and mine through it by hand
• Ground treatment from an island to be crated above the tunnel
• Ground treatment from the adjacent south tunnel 25 m away
The first mentioned seemed immediately to be the most suitable and was developed first; however, all
schemes had a cost-benefit study and safety investigations.
It was fortunate that the tunnel was designed to be axially vented with jet fans and could take the slight
local "throttling" of the cross section resulting from the use of a smaller diameter cast iron (SGI) liner.
The risks of collapse whilst working within the tunnel on repairs were subject of a HAZOP study.
Recovery methods were developed by introducing procedures in a step-by-step process, breaking the
remedial operations into a series of carefully scrutinized activities for all personnel involved both
inside and outside the tunnel, with maximizing of safety always paramount. Failure, if it occurred, was
going to be as a result of further weakening of the damaged concrete or an increase in ground and/or
water loading or a redistribution of it. Various ways that this could come about were postulated and
chance of its occurrence investigated. For example. the likelihood of failure of dewatering wells for a
sufficient length of time to increase water pressure and time to react was one aspect considered.
The times available for evacuation
were also studied; this was made up
from:
• warning time of distress in the
lining
• time for full inflow conditions to
develop
• time of inundation by water at
full flow to reach parts which
prevent escape
Attempts were therefore made to
maximize the warning time. An
important part of this was to develop
various methods of monitoring the
tunnel which included video cameras
installed initially.
This equipment was followed by
strain gauges and other devices of
various types which took longer to
install but which was not attempted
until confidence had begun to be
gained. The data from these devices
was fed to remote data loggers
which were set to trigger an alarm
at certain thresholds.
Fig. 5 Ground strata and subsidence
Once the monitoring was in place, to minimize risks still further by reducing the time taken to
reach a place of safety, flood prevention doors were fitted on each of the cross passages (CP2 to
7), so making the south tunnel a place of refuge. The doors were part of the required safety
measures for cross passage construction to prevent inundation of the main tunnel and so could
easily be modified, Fig. 6.
Fig. 6 Location of bulkheads in the eastern tunnels (schematic)
Once this position had been achieved, sufficient confidence was gained to enable construction of a
forward bulkhead between the end of the Dania backup train and the last cross passage (CP7).
This then enabled compressed air pressure of 0.75 bar to be applied and was the key factor in
giving the final measure of confidence needed to have personnel working for longer periods on
the recovery.
6. THE RECOVERY METHOD
The chosen scheme was to place a new lining where the structural integrity was in doubt. The
repair required 16 of the damaged concrete rings, each 1.65 m long, to be covered with 36 SGI
rings each 750 mm long. The diameter and annulus gap had to be chosen carefully to suit the
damage, squat and eventual wriggle for the track. An additional consideration was that the TBM
itself was so badly damaged that it could not drive forward and so this had to become the eventual
junction position.
The method of effecting the repair was to first place "passive" props off the backup train just clear
of the crown to catch the segments if there should be any movement, and then to hand build the
SGI one ring at the time, working towards the TBM without loading the lining. The undesirability
of disturbing the existing damaged linings had been the reason for ruling out ground improvement
schemes or freezing through the lining. The ring was grouted up inside the damaged PCC ring
with low pressure grout. Rings 1 to 11 could be covered with only minor dismantling of the
backup train. This was the front "bridge" part of the TBM train and the SGI rings could be built
underneath it. This then supported the major length of damage and enabled more dismantling of
the backup before building the remaining SGI. Once all rings had been built, secondary grouting
could be carried out through the precast lining to fill any voids and ensure tightness of the ground
around the "composite" tunnel. The fire resistance of the SGI reliner was enhanced using shotcrete to fill the pans, suitably anchored. When this had been completed the tunnel was considered
as safe as the original design.
7. INVESTIGATION OF LESSER DAMAGED SECTIONS
The fire had affected the lining beyond the point at which the SGI lining needed to be inserted. At
this interface, inspection of the segments led to the conclusion that, even if cover had to be
reinstated, there was sufficient strength to carry the loads. The effects varied from delamination of
the concrete cover, cracking, heat damage to the epoxy-coated reinforcing bars and leaking
through the joints, to simple smoke blackening. Therefore to complete the repair process tests
were performed:
•
•
•
•
•
•
back analysis of temperatures reached in concrete using thermo-luminescence methods
thin section micro analysis looking for tell-tale signs of temperature degradation
porosity determined by thin section analysis and fluorescent microscopy techniques
water absorptivity (confirming porosity)
chloride profile diffusion testing
accelerated salt spray corrosion tests on degraded epoxy coated reinforcement
The proposed methods of repair considered for the lesser damaged sections was:
•
•
•
•
delaminated concrete to be replaced with mortar and protected by a cover plate
injection of cracks greater than 0.2 mm
corrosion monitoring.
coating by a two component thixotropic cementitious modified polymer
8. CALCULATION OF FIRE RESISTANCE OF THE TUNNEL
Fire design and the Operational Risk Analysis conducted at the design stage was based on
theoretical modelling of fire-performance of the lining. The event of spalling was foreseen to
influence the fire-life of the structures and was modelled in the analyses. Based on international
guidelines, literature reviews and experts advice it was assumed that the inner concrete cover, of
size 40 mm, would spall away after 20 minutes of fire and that spalling would then cease. A
transient heat equation was assumed for estimating concrete temperatures and thus the current
load bearing capacity of the lining. Based on a study of fire scenarios a 1200oC hydrocarbon fire
was selected for the analyses. It was considered that the fire curve of RABT, which indicates a
rapid increase in temperature and a constant temperature after 5 minutes, would provide the best
representation of a hydrocarbon fire such as it might arise from a train fire in the tunnel..
Based on the described assumptions and considering the tunnel specific behaviour, taking account
of temperature-induced stresses and stress-relieving interaction with the ground as well as
deformations in the lining, it was assessed that the lining would sustain the fire for between 3-5
hours. With this estimate the risk of service disruption in excess of one month was well within the
acceptance criterion.
After the fire tests were conducted in order to supplement the experience gained, but in a scientific controlled way. The purpose was to update the fire resistance and operational risk.
The major part of the update was the performance and interpretation of full-scale fire tests of
segments under realistic loading conditions. In addition the updating also encompassed small
scale laboratory tests carried out in order to confirm models for the concrete strength and elasticity during elevated temperatures, as well as reviews of relevant documentation. A literature study
was also conducted and international experts consulted in order to be up to date with recent
concepts and results of research.
9. CONDITIONS FOR FULL SCALE FIRE TEST
The main purpose of the tests was to explore the spalling behaviour of the concrete segmental
lining when subjected to a severe hydrocarbon fire (the design fire). The temperatures reached in
the TBM fire were believed to be below 1200oC.
The literature studies and consultations with experts confirmed that the concrete mix, the moisture
content of concrete and the external mechanical pressure are important factors for the occurrence,
and possibly also for the severity of spalling. As to moist in the concrete it is widely held that the
more moist the more vulnerable the concrete will be towards elevated temperatures. Two theories
for the influence of compressive load are: 1) Presence of compressive normal load may increase
the risk of spalling and 2) the magnitude of the load will be a contributory factor for the risk and
severity of spalling.
In the planning of the tests these aspects had to be considered plus the requirement that test
conditions were to resemble as closely as possible those expected to prevail in the tunnel in its
operational phase. As tunnel segments from the general production were still available these were
obvious specimens for testing. The test programme comprised two full-scale tests, one with a
segment at low pressure level (2 MPa) and another with a segment at realistic in-situ pressure
level (5 MPa).
The effect of wet soil backing the segment could not be readily simulated. However, Based on the
temperatures measures on the segment extrados this effect was assumed to have limited influence
on the spalling and on the temperatures reached in the segment.
Overall, the moisture content of the segments subjected to testing equalled that of the segments
already in-place in the tunnel; about 4% by weight. For the latter segments the moisture content is
most likely to decline over the years why this test condition tend to be on the safe side compared
with future in-service conditions.
The RABT fire curve was applied as in the design calculations: 1200oC was reached in 5 minutes
and this temperature was kept constant for 4 hours and followed by 2 hours of linear cooling to
15oC. This fire curve results in high rates of heat flow and moisture movements most likely to
increase the chance of spalling and accelerate its occurrence in combination with a quite rapid loss
of concrete strength.
10. SET-UP OF THE FIRE TESTS
The tests, specified and interpreted by the consultants of A/S Storebæltsforbindelsen, COWI-Mott
Joint Venture, and undertaken by Swedish National Testing and Research Institute, took place
with the segment placed on top of a horizontal furnace as shown in fig. 7. During the testing a
purpose-designed loading frame provided the specified external load on the segment. For
simplicity this was done in the longitudinal direction rather than the hoop direction.
The destructive effects of the fire were visually observed and video recorded through dedicated
windows situated in the walls of the furnace. Temperatures inside the test specimen were recorded
by an array of 56 thermocouples which further provided information on the current depth of
spalling.
Hydraulic ram
Test
specimen
Furnace
Fig. 7. Test Set-up for full scale fire tests.
External pressures corresponded to forces of 342 and 870 tons in the respective tests.
11. TEST RESULTS
Spalling Depths
After 4 hours of fire and 2 hours of
cooling the segment was lifted off the
furnace and appeared as can be seen in
fig. 8. The amount of damage that the
concrete had suffered in the two tests
differed only slightly and in both tests
the remaining concrete appeared domeshaped as was the case in the TBM
tunnel fire.
The contour plot in fig. 9 is based on a
detailed mapping of the final damage.
The design thickness of the segment
was 400 mm.
Fig. 8. Concrete segment after fire test
The overall trend in the progression of spalling is illustrated in fig. 10. As can be seen the
majority of spalling took place within the first 20 minutes of the fire, and the final spalling was
more severe (90mm) than the original design prediction model (40mm) had suggested. The
spalling shown in fig. 10 is the spalling depth (cross sectionally averaged) in the worst affected
cross section of the test specimen. As it was only possible to carry out a detailed mapping of
damage upon completion of the tests, the curve has been established on the basis of visual
observations made during testing as well as readings of the thermo-couples
The fire resistance of the tunnel is determined by the weakest segment in the group of segments
exposed to fire. Due to the random nature of spalling an increase in the number of exposed
segments will, from a probabilistic point of view, increase the severity of spalling damage in the
worst affected segment. Conditional that 30-300m of tunnel are exposed to fire then the best
estimate for the worst affected segment is spalling damage of 100-110 mm.
100-125
75-100
50-75
25-50
0-25
Fig. 9. Spalling depth in mm. Result from test with 5 MPa pressure is shown.
Segment Temperatures
The segment temperatures did not rise as rapidly as was predicted by the transient differential heat
equation (fig. 11, Theory 1). As the temperature level determines the remaining concrete strength
the observation suggests that the decline of capacity of the remaining concrete will take place at a
lower rate than originally predicted.
The lack of detailed information on
the rate of progression of spalling
(in time and space) introduces
uncertainty
when
referring
temperature registrations to depths
below the exposed surface.
Spalling [m m ]
400
320
D esign
Test
240
160
80
0
0
40
80
120
160
200
240
Tim e [m in]
Fig. 10. Approximate rate of progression of spalling
observed in tests along with design prediction.
1200
1000
800
600
400
200
0
E xperim ent
T heory 1
T heory 2
0
50
However, For the deeper thermocouples this uncertainty plays a
minor role as regards the accuracy
of the temperature profile, and the
tests revealed that a more refined
heat flow model (fig. 11, Theory 2)
would be more accurate in estimating concrete temperatures than
the original design
prediction model (Theory 1). The
heat flow model in question takes
account of the energy consumption
of decomposing reactions in the
concrete during heating, and was,
due to its better performance for
the particular concrete, applied in
the subsequent update of fire
resistance
100 150 200 250 300 350 400
D istance to virgin surface [m m ]
Fig. 11. Temperatures in the remaining concrete.
12. UPDATE OF FIRE RESISTANCE AND RISK ANALYSES
The fire tests resulted in improved prediction models for spalling (depth and time of occurrence)
and for the heat flow in segments. Additionally, the smaller scale tests mentioned in section 3
verified the design prediction model concerning the concrete strength-temperature relation. In
addition other tests confirmed that the cylinder strength of segment concrete being achieved was
some 50% higher than the design value. Hence, a more refined calculation basis for determination
of fire resistance for the particular structure was made available.
Inevitably, an element of uncertainty is associated with updating and for that reason the reassessment of fire resistance involved a sensitivity study in which implications of combinations of worst
case and best estimate scenarios were investigated. The study embraced spalling scenarios ranging
from 110mm (test result); 150mm (the spalling recorded after the TBM fire) to 180mm, and fire
temperatures of 1200oC as well as 1350oC. Based on the study it was concluded that the fire
resistance is 4 hours or longer and as the design prediction was 3-5 hours the result gave no
changes to the existing operational risk account for the tunnel.
13. CONCLUSIONS
The time taken to recover the tunnel was about ten months. Each step of the recovery was studied
carefully and the procedures were made with the purpose of minimizing the risk to personnel and
to safeguarding the structure. In times gone by, similar circumstances might have called for
"heroic" deeds working virtually blind, but this can no longer be the case today.
It has been possible by careful back-analysis using a best- estimate / worst credible scenario
approach, to gain sufficient confidence to be able to reinstate the severely damaged tunnel without
further incidents, and without having to resort to extremely expensive and even more time
consuming methods.
The fire event and the full scale tests have confirmed that spalling is an important issue when
determining the structural behaviour of a concrete structure subjected to fire. Knowledge of
concrete and spalling is not yet sufficient to predict with accuracy how the spalling will progress.
For the Storebælt Tunnel the tests established a sufficient basis for estimation of the spalling
during severe fires. It is advantageous that the load carrying capacity of a circular tunnel lining
acting in hoop compression will decrease relatively slowly unlike a structure carrying the load in
bending, where the structure may collapse shortly after the spalling has reached the reinforcement
in the intrados, such as an immersed tube tunnel.
The end user can thus be confident that the tunnel after repair and refurbishment has been put
back to its original intended condition with respect to structural performance. After collection of
all experience from the fire and the fire tests it is also concluded that the intended condition with
respect to user risk and disruption risk is maintained.
14. ACKNOWLEDGEMENTS
The authors would like to thank their colleagues at COWI-Mott and A/S Storebæltsforbindelsen
for their assistance and comments.