First, a brief history. The invention and patenting of the use of compressed air to exclude groundwater from shafts and tunnels while under construction in permeable strata is credited to Sir Thomas Cochrane in 1830. This patent was concurrent with the construction of Marc Brunel’s Thames Tunnel, work on which had been started in 1824 and had suffered two inundations: in May 1827, causing work to be delayed for two months; and more seriously, in January 1828, which caused the work to be suspended.
Following Cochrane’s patent, the use of compressed air to exclude water from the tunnel was suggested to the directors of the Thames Tunnel but was resisted by Brunel who believed that the tunnelling shield alone was sufficient for the construction of the tunnel. Tunnelling on the Thames Tunnel restarted in 1835 with government finance and was successfully completed in 1843 without the use of compressed air.
The first recorded use of compressed air was for the construction of colliery shafts and bridge piers. Compressed air was first applied to tunnel construction in 1879 by H Hersent to construct a drainage gallery for a dry dock at the Port of Antwerp. In the same year, compressed air was used on the construction of the Hudson River tunnels in New York. On this major undertaking, the engineer De Witt Clinton Haskin chose to depend entirely on compressed air and not use a shield. The result was a blow out which occurred in July 1880 killing 20 men. After attempts to resume work, the tunnel was closed down in 1882.
In 1886-1887, the City and South London Railway tunnels were successfully constructed under the Thames by the combination of a Greathead Shield and compressed air; James Henry Greathead was engineer in charge of construction. Today these tunnels form part of the Bank branch of the Northern Line. After the benefits of using a shield had been demonstrated on the City and South London Railway, works on the Hudson River tunnels were reopened in 1889 under English supervision. Greathead himself acted as consultant. Using a Greathead-type shield and working in compressed air, the tunnels were successfully completed in 1905.
From 1886 until the recent development of closedface TBMs, the standard method of tunnelling in alluvial deposits below groundwater level has remained the combination of a Greathead Shield, compressed air working, cast iron linings and the grout pan to fill the annulus. Post 1945, the use of concrete segments became increasingly commonplace but without efficient gaskets their use was typically confined to impermeable soils. Such was the state of the art in 1971 when the author began his tunnelling career.
AIR PRESSURE FOR STABILITY IN CLAYS
In low permeability soils, it is possible for compressed air to act against the soil mass and provide support. The stability of a clay tunnel face in an open-face Greathead type shield is related to the stability number Ns where:
σvt is the vertical overburden pressure at the tunnel axis
pi is the internal air pressure and
cu is the undrained shear strength of the clay.
Broms and Bennermark (1967) demonstrated that for clay, a vertical face becomes unstable for Ns>6. Hence, it can be seen that for shallow tunnels, if sufficient internal air pressure can be applied, the stability number can be reduced below this value even for soft clays.
In the author’s experience of excavations during Phase 1 tunnelling of the Singapore MRT system, it proved possible to drive open-face Greathead-type shields through the very soft marine and estuarine clay formations using compressed air to stabilise the face. At tunnel depths of 15m to 20m, these materials typically had a cu in the range of 30-40kPa, which typically gave an Ns approximately equal to 9. However, with the application of compressed air in the range of 1.5-1.8bar, the stability number was reduced to between 4 and 5. The greatest air pressure employed was 2.7bar. The use of compressed air in this manner, plus the incorporation of suitable face support in the shields, proved effective to limit ground losses to acceptable levels.
AIR PRESSURE FOR STABILITY IN SANDS AND GRAVELS
A degree of cohesion is required for any soil to hold a vertical face. Coarse-grained soils lack cohesion, although fine-grained silts and sands will often display a useful degree of apparent cohesion due to capillary action if they can be maintained in a moist condition.
Below the water table there obviously will be a hydraulic gradient into the tunnel face, the magnitude of which will vary with the hydrostatic pressure. This water pressure can be balanced by compressed air pressure, but only at one level in the face.
As can be seen from figure 1, when excavating in an open-face shield, it is necessary to establish a balance point at some level in the face. Around the balance point, if sufficient progress can be achieved, moist conditions can often be maintained and this area of the face may be stable. Below the balance point, in fine grain soils, water ingress into the face will transport soil particles, creating flowing ground conditions – the soil behaves effectively as a slurry. Face support in this condition requires a close-boarded face with copious amounts of straw to prevent ground losses. As the face is generally mucked out from the invert, maintaining a close boarded face in this area is problematic. One method of limiting this problem is to drive a well point into the face close to the invert. If this well-point is then connected to a ’snorer drain’ that discharges to the atmosphere beyond the bulkhead, the air pressure will push the water towards the low pressure well-point head before it enters the face. Without such well-points, to facilitate excavation, it is often necessary to maintain a balance point close to the invert of the tunnel.
Above the balance point, air will seep out into the soil. Darcy’s Law states that fluid will seep through a permeable medium with a velocity of flow proportional to the pressure gradient and inversely proportional to the viscosity of the fluid. As the viscosity of air is typically 55 times lower than that of water, the excess air pressure above the balance point will begin to drive water from the soil creating dry running sand conditions. Again, close-boarded face support is required when moisture is lost.
Above the balance point, initially the fluid passing through the soil mass is water, but as the process continues the water is continuously displaced by the air. The seepage velocity increases as the air water interface advances through the soil mass and as the predominant characteristic of the fluid changes from that of water to air. The effect of this increase in velocity is twofold: first, the soil along the seepage paths may become loosened, particularly at points of exit; and second, the volume of air lost from the tunnel increases. As the tunnel advances, these seepage paths are progressively closed off and water will slowly seep back into the soil behind the shield, as the influence of the compressed air recedes. For most shallow tunnels it is probable that the air water interface will reach the ground surface.
AIR PRESSURE FOR STABILITY IN SILTS
Silts by definition are transitional materials between sands and clays and arguably display the worst properties of both materials. BS6164 (1990) states at section 7.3.5: “Silt is particularly dangerous because cohesion can be suddenly lost in the presence of water or when dried out. Close timbering of all faces becomes essential, unless the time of exposure is brief, the exposed area is small, or cohesion can be ensured. When a tunnel in silt is below the water table, the use of compressed air, freezing or other stabilisation measures is essential”. Further, at section 10.2.4 with regard to compressed air working: “Silt is much less permeable and has a useful measure of cohesion when damp, but the moisture content is so critical that dry silt will crumble away whilst wet silt will become fluid. The control of water is of vital importance, but is no substitute for close support which is essential where water is present.”
For many silts, the difference between the natural moisture content and the liquid limit is only a few percentage points. As described in BS6164, moist silt can display a useful measure of cohesion which is derived principally from capillary action between the soil particles. If under-pressurised, the hydraulic gradient into a tunnel face will progressively raise the moisture content; when the soil has become saturated all capillary action is lost and the silt effectively becomes a fluid. This change in character from a cohesive material to a flowing slurry can occur with great rapidity; unless suitable face support is already in place, it is probable that significant ground losses will occur. For this reason, silts have rightly earned a reputation as a treacherous tunnelling material.
PREVENTION OF BLOW-OUTS
There are essentially two circumstances in which a compressed air blow-out can occur. The first, termed a ‘seepage blow-out’, occurs when the pressure within the excavation suffers a sudden reduction on account of escape of air to the surface. The resultant air losses exceed the available compressor capacity and it is no longer possible to maintain the air pressure required to stabilise the face.
This type of blow-out can occur when the tunnel intersects high-permeability features; faults and gravel pockets represent two particular types of such hazards. In London Clay, a number of buried scour hollow features and relic ice-cored mounds which formed within the permafrost termed ‘Pingos’ have been observed. These are normally partially infilled with sands and sandy gravels. Such coarse-grained granular sands and gravels lack any cohesion and offer very little resistance to the escape of air, hence it is very difficult to maintain a suitable balance point in the face and to establish suitable working conditions.
A sudden loss of air pressure can also be associated with the tunnel encountering a man-made feature, such as piling that provides a low-resistance preferential escape route for air to the surface. Muir Wood (1975) reported that the Clyde Tunnel passed immediately below a timber jetty built for the Whiteinch vehicle ferry. The toes of the timber sheet piling to the jetty extended to approximately 1m above the crown of the shield. When the shield approached the jetty, a major blow of air to the surface occurred. The loss of pressure resulted in a slip of ground behind the jetty and about 150m3 of silt and silty sand entering the tunnel face. A significant cavity was also formed ahead of, and above, the tunnel face, exposing the toe of the sheet piles. Subsequent investigations indicated that a contributory factor arose from the buffeting received by the jetty over time from the ferry, leading to loosening of the piles in the ground.
Another incident in the author’s experience occurred during the construction of the Phase 1 MRT tunnels in Singapore, when the C103 tunnel shield encountered an ungrouted borehole in the tunnel face. Air pressure was suddenly lost which resulted in some loss of ground and the need to rapidly box the face. The situation at the surface was also hazardous; the borehole had originally been drilled next to a footpath, and this resulted in a mass of stones and earth being suddenly ejected at high velocity into the air. This obviously represented a hazard to third parties.
A further cause of sudden air pressure loss can arise from mechanical failure. It is always necessary to have at least one additional compressor on standby above that required. The power supply should be available from two separate sections of the grid and/or suitable diesel generators should be available. In establishing the size and number of compressors, the simplest method to overcome the problem of blow-out by seepage of air is to provide air pumpage capacity greater than the maximum possible rate of air loss.
Any air losses through the face can be mitigated by clay pocketing gravel layers, spraying the face with bentonite/shotcrete or partially covering the face with plastic sheets.
A second type of blow-out which is termed an ‘upheaval blow-out’ can occur when the pressure of the air on the soil mass in the roof of the tunnel lifts the soil mass as a block, permitting the air suddenly to escape as a large bubble.
Consider the stability of a subaqueous tunnel as shown in figure 2. One method of increasing the stability of the tunnel against an upheaval blow-out has been to place a blanket of soil on the bed of the water body.
The use of clay blankets over compressed air workings has been in use from at least the early 1940s, when they were used in the construction of New York City tunnels. Ideally, the blanket should be impervious, however the key question with respect to this method is how to achieve imperviousness in the blanket? A fine-grained material, such as clay, would appear to be suitable, however, if pulverised clays were used it would turn immediately to a slurry on contact with the water.
Alternatively, if clay was placed in large chunks, crevices between the chunks would provide easy escape paths for the air. For these reasons, such blankets cannot be regarded as impervious and they are not useful for reducing air losses. For design purposes, the advantage of such blankets is limited to simply adding additional weight to the soil above the tunnel.
For tunnels with mainly clay soil cover, if we consider the block of soil between the crown of the tunnel and the bottom of the blanket as a free body, and if the strength of the soil along the sides of this body is neglected, then the forces expressed as pressures acting on the projected area of the block of soil can be determined by:
Pressure of water: Pw = Zw.ρw
where ρw is the density of water
Pressure of the blanket: Pb = Zb.(ρb-ρw)
where ρb is the bulk density of the blanket
Pressure of the soil: Ps = Zs. Ρs
where ρs is the bulk density of the soil.
For equilibrium with respect to heave, the sum of these pressures must be greater than the air pressure Pa within the tunnel, hence:
ϒ(Pw+Pb+Ps) ≥ Pa
where ϒ is a safety factor (1)
The maximum air pressure commonly used in tunnels is equal to the hydrostatic head of water at the invert of the tunnel, hence:
Pa= (Zw+Zs+D). ρw (2)
Combining equations (1) and (2) yields an expression for the soil mass with respect to upheaval:
ϒ.[Zb.(ρb-ρw) + Zs.(ρs-ρw)] ≥ D. ρw (3)
If we ignore the terms for the clay blanket and assume that for many soils the bulk density is approximately 17.5 kN/m3 and water is 10 kN/m3, then adopting a safety factor of 0.9, an upheaval blow out should not occur if:
Zs ≥ 1.5.D
In practice, a more rigorous analysis which includes the shear resistance along the sides of the failure block, which have been neglected in this calculation should be made. Also in sub-aqueous tidal situations, the applied air pressure should be linked to the tidal head and the air pressure adjusted accordingly.
For tunnels with predominantly a sand cover, if we substitute Ps = Zs.(ρs-ρw) into the above analysis, it becomes apparent that a considerable overburden depth will be required to provide sufficient weight to balance the air pressure. For tunnels with a shallow cover of sand, air seepage from the tunnel face will destabilise the overlying sand cover and create quicksand conditions. The air losses through the disturbed material are likely to increase rapidly and so lead to a seepage-type blow-out.
There are numerous instances of upheaval-type blow outs in the literature, perhaps the most recent and dramatic occurred during the construction of the Docklands Light Rail tunnels beneath the River Thames
In this instance, assuming an average bulk density of 18kN/m3, the 7m overburden would have provided a safety factor of approximately unity against an upheaval blow out, provided the air pressure in the tunnel had been limited to balancing the hydrostatic air pressure at the invert of the tunnel immediately behind the bulkhead. Unfortunately, the air pressure used inside the tunnels for the excavation of the cross passage and sumps was considerably greater. Even at the minimum internal air pressure used of 2.1bar, this would have created a disturbing pressure of 210kPa at the tunnel crown. Assuming an average bulk density of 18kN/m3, the restoring pressure of the soil mass would only have been approximately 126kPa i.e. only 60% of the uplifting air pressure. The resulting unbalance was sufficient enough to shear through both the segmental tunnel lining and the overlying soils, and this caused a catastrophic upheaval blow-out.
In the author’s long experience of constructing inclined tunnels, it was always good practice to periodically include additional steel anchor plates between the segments. This enabled the locks to be advanced as the tunnel progressed deeper and the required air pressure increased.
SETTLEMENT CAUSED BY REMOVAL OF AIR PRESSURE
As illustrated in figure 1, if the balance point of the air pressure is below the crown of the tunnel, air from the tunnel will flow out into the soil mass and displace water from the capillary pores in the soil. If the air pressure is suddenly lowered on the completion of the excavation, then the air in the soil within the capillary fringe will disperse faster than the hydrostatic water pressure can be re-established. In this condition, the effective stress between the soil particles will increase, and consequently there is a risk of consolidation.
The magnitude of consolidation settlement that occurs after ‘air off’ can be significant. Shirlaw (1988) reported that the consolidation settlement over the C301A tunnels after ‘air off’ was typically only 10mm- 15mm. These tunnels were constructed using manual excavation in open-face shields equipped with a ‘wine rack’ and face ram-support system. They were segmentally lined and fitted with hydrophilic gaskets. Air pressure in this section of the drive was up to 2.7bar.
By contrast, settlement of over 100mm was recorded in the C105 tunnels following air off in the first tunnel. The final settlement after both tunnels had been driven, predominantly consolidation after air off, exceeded 350mm in places. The C105 tunnels were constructed using an open-face drum digger shield in air pressures up to 1.5bar. The tunnels were temporarily lined with steel ribs and timber lagging which, even after extensive back grouting, proved impossible to waterproof. A considerable proportion of the observed settlement could therefore be attributed to seepage through the temporary lining.
DE-OXYGENATED AIR
Oxygen may be removed from air travelling through ground containing certain organic or inorganic agents. Morgan and Bartlett (1970) in their Victoria Line paper note at Sections 3.51 and 3.52 that:
- 3.51 At Euston, while tunnelling in the sands of the Woolwich and Reading beds beneath the London Clay, some drives required the use of compressed air. On one occasion, the compressed air displaced the atmosphere from the voids of the sand beds into a neighbouring tunnel being driven in free air. This atmosphere had a greatly reduced oxygen content and it was fortunate that no fatalities occurred in the free-air tunnel.
- 3.52 It is understood that similar phenomena have been noted elsewhere in London, Melbourne and Seattle, and the Authors wish to draw attention to this danger which is not apparently widely recognised. The possibility must be considered of foul air being driven into basements of buildings by the application of compressed air for neighbouring tunnel or caisson construction.
Muir Wood (1975) noted that the incident in Melbourne involved loss of life when rapid depressurisation of a tunnel caused re-entry of air that was deficient in oxygen into the workings.
THE PROBLEM OF ENCLOSED SAND LENSES
In discussions of the Victoria Line paper, Bubbers (1970) reported that: ‘Compressed air was not always successful in stabilising lenses of silt and sand in the Euston area. These lenses, although extensive, were completely surrounded by clay and were probably supplied with water through small fissures in the clay. Under these conditions, the compressed air applied in the tunnel eventually permeated the whole lens, which thus became charged with air at the same pressure as that in the tunnel. When this balance is achieved the water is able to return through the face unimpeded.’
In the author’s long experience of similar conditions, this problem can be successfully treated by the combination of well pointing and compressed air. A well-point driven into the sand body and linked to a snorer drain that discharges to the atmosphere behind the bulkhead will effectively create a low-pressure sink within the sand body, thus allowing the compressed air to displace the water.
RECENT DEVELOPMENTS
The development of closed-face Earth Pressure Balance (EPB) and slurry shields has largely negated the need for compressed air working of the type described in this article where the whole workings are pressurised. These machines have allowed increasingly larger tunnels to be constructed in evermore difficult ground conditions and at significantly higher pressures.
However, these machines require picks and cutters to be changed at frequent intervals, and to facilitate this task it is generally necessary to make manned interventions into the plenum chamber under compressed air. Many of the issues discussed in this article remain relevant to such interventions and in the second article on this subject a number of illustrative projects will be discussed.
