In recent years there has been increased awareness of the risk of structural damage to tunnel linings due to exposure to fire. The risk seems particularly severe with regard to precast segmental linings where high density, low permeability concrete has been used in the construction. This risk has been highlighted by a number of high-profile incidents particularly those that occurred operationally in the Channel Tunnel and during the construction of the Storebælt Eastern Rail Tunnels.

Published reports of these incidents indicate that explosive spalling of the high-density concrete was a significant contributory factor to the high levels of damage sustained.

The phenomenon of explosive spalling has generally been attributed to the formation of vapour pressures within the segment as the heat front advances. Due to the low permeability of the high density concrete employed in precast construction, these vapour pressures cannot escape and create tensile stresses within the segment that lead to explosive spalling of the concrete’s surface.

Research has shown that the inclusion of a small quantity of very fine polypropylene fibres into the mix will, under certain test conditions, reduce or eliminate explosive spalling.

Polypropylene melts at low temperatures creating a continuous fine void structure in the heat-affected zone that allows vapour pressures, formed within the concrete pores, to be dissipated. Inclusion of polypropylene fibre into the mix has become an industry norm and seems to have been accepted as a panacea for the problem of explosive spalling of tunnel segments.

However, it is the author’s contention that vapour pressure within the segment, although a contributory factor, is not the single cause of explosive spalling. Internal stresses caused by the particular restraint conditions within the tunnel arch can also be a contributory factor.

Thermally-Induced Stresses

Concrete is a surprisingly good insulation material; hence, if a concrete structural element is subject to a sudden increase in temperature on one face, it will take a considerable period for the element to achieve a uniform temperature. So, for a segmental tunnel lining exposed to fire, very high thermal gradients are created between the heated internal face of the segment and the cooler core.

As the heated inner face of a flat slab or beam structure expands, it will sag (figure 2) in order to achieve strain compatibility. In a conventional fire-test arrangement, the slab under test is not usually loaded (apart from its self-weight) and consequently has the opportunity to deflect (sag) and achieve strain compatibility between the hot and cold elements of the structure.

Comparative fire tests conducted for the Channel Tunnel Rail Link (CTRL) in 1997 to evaluate the effectiveness of polypropylene fibres were conducted in the conventional manner. Sample panels 1240mm long x 250mm deep were typically observed to sag between 8-15mm over the course of the test – a deflection of 1:150.

However arch and vaulted structures do not have the same opportunity to relieve internal stresses by sagging as, in order to continue to carry applied soil and water loadings, they must maintain a hogging arch profile throughout the fire. Consequently, high shear stresses are created between the hot surface exposed to the fire and the cooler core.

Few fire tests have been carried out on arch structures under load conditions. In 1995, the Eureka–Project EU499 studied six arch profile test specimens of construction materials typically used in road-tunnel construction, four of which are relevant to segmental linings.

The geometry of the test specimens is shown in Figure 1 and each jack loaded a 1m width of the panel. The specimens were all subject to fire test conditions described by an ISO-834 temperature- time curve. This curve has a relatively slow temperature build-up with the peak 1100ºC reached after 200 minutes.

In describing the internal forces and temperatures, the report states: “One of the primary objectives of testing was to restrain the deformations in order to be able to determine the compressive forces. In respect of horizontal strains, this could be achieved, but vault crown lifting could not be prevented because of the limited capacity of the vertical jacks. In connection with vaults resting on hinged though fixed bearings at the sides, crown lifting generally results from an increase in temperature in the vault centre. A temperature increase of 200ºC produces in concrete segments 240mm-thick a rise of about 12mm, while vertical loads of P=200kN in the third points produces a sag of about 1.5mm”.

As the maximum jack capacity was 350kN, this would produce a sag of 2.63mm in a 240mm-thick segment, i.e. only 22% of the observed 12mm rise at 200ºC. The under capacity of the vertical jacks would become even greater at higher temperatures and for thicker segments. Despite the limitations of the test rig to adequately restrain the samples, extensive spalling was observed in all the four segments tested.

Mechanisms of Explosive Spalling

Explosive spalling may be attributed to two independent and separate mechanisms as described in the following sections:

Internal vapour pressures The mechanism of explosive spalling due to trapped water vapour has been described above. Most commonly, this has been observed associated with the heating of high-performance concretes that typically contain large doses of silica fume – 5-15% by weight of the cementitious materials.

In general the risk of explosive spalling due to vapour pressure increases as the permeability of the concrete reduces and the rate of temperature rise increases. An associated observation is that wet and saturated concrete is also likely to be more susceptible to explosive spalling. The mechanisms of explosive spalling due to vapour pressure, by their nature, are largely independent of the state of stress in the concrete.

Stresses Induced by Thermal Gradients

In an arch structure, strain compatibility between the hot and cold elements can only be achieved by vault rising (as observed during the Eureka test programme) and/or by delamination.

Delamination occurs when the shear stress between the hot and cold elements is exceeded and the cover concrete suddenly separates from the cooler body of the segment in an explosive spall. In practice, very high stresses can be generated at quite moderate temperatures.

In a sustained fire, the loss of surface concrete by spalling exposes the previously cold internal surface of the segment to the hot combustion gases and a new heating process begins that will eventually lead to the failure of the newly exposed surface layer. Successive sequences of spalling and reheating can lead to deep erosion of the lining until a point is reached where the profile is so weakened by a combination of loss of section and material strength that it is no longer able to sustain the applied soil and water loading – and failure occurs.

Analysis of the storebalt fire

During the construction of the Storebælt eastern railway tunnel in Denmark, an uncontrollable fire broke out in the tunnel boring machine Dania at approximately 7.20am on Saturday 11 June 1994. At around 18.00 hours, the fire was observed to have extinguished itself, judging by the lack of smoke emanating from the portal. It is estimated that the overall duration of the fire was some 10-11 hours. At 16.00 on Monday 13 June, a team of experts from the contractor’s and engineer’s staff entered Dania and surveyed the damage.

Severe spalling in the crown (over an approximately 90º-degree arc) was observed in the six segments immediately behind the tailskin of the TBM. The most severely damaged segments were immediately in front of the leading segment on the overhead conveyor where it appeared that an area of turbulence had been created. In the worst of these segments, the original 400mm thickness had been eroded by the spalling mechanism to only 130mm, i.e. only one third of the original segment thickness remained.

Fire Exposure of Damaged Segments

Subsequent investigations indicated that the primary fuel for the fire was hydraulic oil from the TBM’s power system, thought to be leaking and atomized at high pressure through a ram hose connection in the knee of the shield. Although a standard petroleum fire reaches 1200ºC, it was clear that the intensity of the fire was limited by the rate of leakage of oil through the ram hose connection. From various indications, it was estimated that the temperature during the fire was in the range of 300°C – 500°C.

The importance of the loading/ restraint conditions on explosive spalling can be clearly seen from figures 3 and 4.

The unloaded segment on the overhead gantry was obviously subjected to the same intensity of fire as the segment that had been erected and grouted in the tunnel. Indeed, it was the presence of this segment that created the obstruction and the turbulent zone.

However, unlike the erected segments, which were now carrying a full soil and water loading, this segment was free to deflect and achieve strain compatibility between the hot surface and cold core.

No spalling was observed from any of the unloaded segments on the overhead conveyor and they suffered negligible damage.

As all segments were taken from the same segment stockpile within an approximate period of 24 hours, their moisture contents would have been similar. Hence, if explosive spalling is purely a function of moisture content and vapour pressure build-up, similar damage would have to be expected to occur in the unloaded segment. This was clearly not the case and the difference in behaviour of the loaded and unloaded segments can only be attributed to their loading and restraint conditions.

Indications of the spalling mechanisms can be obtained from examination of the failure surface of a spalled element of cover concrete that was recovered from the TBM backup. Figure 4 shows an approximately A4-size element of cover concrete, approximately 35mm-thick which was recovered from the TBM backup.

Examination of the failure surface indicated that it was extremely rough. Areas where tensile bond failure had occurred around aggregate particles could be observed, however the greater percentage of the surface was represented by direct shear through the aggregate particles and an irregular rough concrete surface. As argued above, the observation of sheared granite aggregate particles in the failure surface is more consistent with delamination due to internal thermal stresses, rather than vapour pressures.

Analysis of The Channel Tunnel Fire

The fire in the Channel Tunnel occurred on a 29-car heavy goods vehicle (HGV) shuttle travelling from France to the United Kingdom on Monday 18 November, 1996 at 8.45pm UK time.

When the fire was noticed, the driver brought the train to a controlled stop approximately 19km (12 miles) into the tunnel from the French terminal. The segmental lining at this point was of French design.

The fire developed rapidly, fuelled by combustible material on the HGV vehicles. The heat developed was later calculated at 500MW and temperatures were estimated to have reached 1,000°C. Large amounts of concrete spalled-off explosively from the tunnel lining due to its exposure to the fire. The spalled material was described variously as “thin slices”, “onion peel” and “fish scaling”. It was noted that “you could actually see the aggregates in the material in these thin slices.”

Control of the fire was reportedly achieved at 5.00am on the following day, and it was reported to have been extinguished at 11.15am, having burnt for some 14.5 hours. Significant damage occurred to the tunnel lining for approximately 200m (656 feet), with serious damage to an additional 200m (656 feet). According to reports from the Kent Fire Brigade, in some areas approximately 406mm (16 inches) of concrete was destroyed, leaving only 25-50mm (1-2 inches) of concrete remaining. Fortunately, the incident occurred in an area of stable chalk marl which was able to stand unsupported for a considerable period.

Observation of the damaged lining, together with eye witness descriptions, indicate that the spalling had advanced in thin layers between the lateral ties within the segments.

Conclusions

It is generally recognised that explosive spalling of precast segmental tunnel linings can arise from two mechanisms, which are:

  • The formation of vapour pressures within the pores of the concrete.
  • Thermal stress gradients formed within the segment.

In this article, the author has discussed the nature of the spalling observed in both the Storebælt and Channel Tunnel fires. A common factor in these fires is that the segments were embedded in strong soils. In the case of the Storebælt linings, they were located in stiff to hard lower clay till, while the Channel Tunnel linings were embed in competent chalk marl. Both these materials would have exerted a high passive pressure resisting any upward rise of the crown vault during the fire.

A further common factor in these cases is the extended period to which the linings were exposed to fire (10–14 hours). As argued above, the extended duration of the fire can contribute to progressive spalling and erosion of the lining. It is appreciated that in both these cases the segments were made from high-performance, low-permeability concrete, and in neither case incorporated polypropylene fibres. Hence, arguably both segments were susceptible to explosive spalling due to increased vapour pressures within the pores of the concrete. Notwithstanding, it is argued that thermal stresses due to the combined effects of restraint and thermal gradient were also a contributory factor to the observed spalling.

This conclusion is based primarily on two observations:

  • First, in the case of the Storebælt fire, considerable differences were observed in the spalling behaviour of the unloaded segment on the overhead conveyor and those that had been installed and grouted into the lining. The only explanation for the difference in spalling behaviour between the segments is the restraint conditions: the unloaded segment on the overhead conveyor was free to deflect and achieve strain compatibility under the induced thermal gradient, while the segments installed in the lining were rigidly restrained and continued to support the applied soil and water loading throughout the fire.
  • Second, in both the Storebælt and Channel Tunnel fires, observation of the spalled particles indicated that they were generally thin, laminar in nature, with the failure surface sheared through and exposing aggregate particles. The thin onion peel layers, together with the observed “sheared aggregate in these thin layers” are, in the author’s opinion, consistent with laminar shear failure caused by high thermal stress gradients within the segment.

Obviously, the process of explosive spalling due to vapour pressures is limited to the cement matrix; it cannot explain the observed failure surfaces through the granite aggregate.

It is therefore concluded that, although neither the Storebælt nor the Channel Tunnel linings contained polypropylene fibres, the observed level and nature of the spalling cannot be solely attributed to vapour pressures. In both instances, thermal stresses induced by the combined effects of restraint and thermal gradient were, at least to some degree, also a contributory factor to the observed spalling.

Recommended test procedures

In specifying fire-test procedure for a tunnel lining, it is necessary that the test conditions are capable of modelling the spalling resulting from both vapour pressures and thermal effects. It is therefore considered essential that the lining is tested under load and subjected to restraint conditions analogous to the real world.

In this regard, it is recommended that a segmental lining is tested in a rig similar to that employed for the Eureka project as illustrated in Figure 1, with the provision that the loading jacks should have sufficient capacity to limit the arch rise to a level that might be permitted by the passive restraint of the surrounding soil.