The text ONLY of the Paper is given below; but it will probably be more useful to read the .pdf file that includes illustrations and can be found in the Library.  The text is given below simply so that Google and other "crawl engines" can find the text.

This Paper was presented at the International Conference on TUNNELLING UNDER DIFFICULT CONDITIONS on 28 October 1997 in Basel, Switzerland.

Anticipating the Unexpected

Flood, Fire, Overbreak, Collapse

by   Robin B. Clay   and   A. Paul Takacs

September, 1997

Abstract

Driving a car is perhaps the most risky of every-day activities; driving a tunnel is perhaps the most risky form of construction. For the driver or the engineer, no matter how good his training and how broad his experience, the unexpected may still occur.

In the design phase of a tunnelling project, the engineer must make some estimate as to what challenges the ground ahead is likely to hold, and for this he has to carry out site investigations.  The more he probes, the more it costs, but the lower the risk. It is a matter of engineering judgement as to how much time and money to expend in investigation;  too much, and there may be accusations of extravagance; too little, and there may be a disaster. During the construction phase, possibly one of the most significant “management” factors is the amount and quality of supervision, both by the Contractor and by the Engineer. In the same way that Site Investigation is a trade-off, so, too, is the matter of site supervision. On the practical side, the question of what support to provide at any given time and place is again a matter of judgement, and a balance must be struck between too much and too little.  This may be entirely a site matter - shotcrete support, for instance - or entirely a design matter - strength of segments, say - or a bit of both;  but it remains a question of judgement, an art, undefineable by conclusive technical science. The other major factor is the technique.  In any given circumstance, there is a choice of methods available, and the choice, very largely a matter of opinion and experience, is often left to the Contractor who must carry out the work. Every tunnel is different; new ground is broken in the most literal of senses.

It is self-evident, then, that there always remains an unquantifiable risk.  The risk itself may be known or guessed and the cost of failure may be calculated;  but it is not possible to estimate how likely it is to happen.  All involved will hope for the best; the prudent will prepare for the worst. After something has gone wrong, an inquiry will try to find the reason, either with the practical aim of trying to prevent a recurrence, or to establish who should pay the very large sums of money involved, which tends to take precedence. Unfortunately, when culpability is at stake, there are vested interests, and truth takes a second place.

The Authors have personal experience of several different tunnelling disasters - fortunately, none involving any injury or loss of life.  However, the costs have been very high - sometimes of the same order as the Contract value. Some case studies are presented, looking not so much at the direct causes, but at the possible under-lying factors. 

The conclusion is that occasional disasters are unavoidable, because the costs of trying to avoid all risks is unrealistically high.  ‘Post mortem’ studies can highlight failures in the procedures being followed before the disaster, and are therefore valuable aids to  safeguarding the remainder of the work.  But Nature - and sometimes our forebears - can still spring surprises.

1.   Introduction

Of the risks associated with the driving of tunnels, the most serious are those given in the title of the Paper:

• Flood, caused by water flowing into the tunnel during construction;
• Overbreak, where the ground being excavated (usually rock) breaks away from the intended profile (overbreak may also be used to describe extra excavation of ground due to survey errors, bad drilling technique, temporary works, etc.);
• Inrush, where material (usually sand or mud from above the strata being excavated) flows into the tunnel, in most cases accompanied by much water, such that the inflow material interferes with construction work; and
• Collapse, where the actual structure of the tunnel is damaged or destroyed.

            Fortunately, in the incidents referred to in this Paper, there was no loss of life or injury, and in some cases not even any significant damage to plant or equipment; but in all cases there was considerable delay. The Paper is concerned with construction risks to the Works, not with the safety of the workers - although these may go hand-in-hand.

            After something has gone wrong, an inquiry will try to find the reason, either with the practical aim of trying to prevent a recurrence, or to establish who should pay the very large sums of money involved, which tends to take precedence. Unfortunately, when culpability is at stake, there are vested interests, and truth may take second place. In what follows, deductions have been made (possibly erroneously) as to what is believed to have happened. This Paper attempts merely to record some of the pertinent facts, without bias; any views expressed herein are purely those of the Authors alone.  The purpose is to try to help prevent similar occurrences elsewhere.

2.Risk

2.1The Concept of Risk

Before he takes a car out onto the road, a driver must undergo training. The more experience he gains, the less chance there will be of an accident, but every time he drives, he takes a risk. The faster he drives, the greater the risk, but the greater, too, the benefit of time saved. The driver weighs both risk and benefit, although usually entirely subconsciously.

            With most other forms of construction, the engineer knows what he is facing, but with tunnelling, the next step is always into the unknown. Tunnelling is dangerous, and collapses occur [1]. Tunnelling is an art rather than a science; science can be used to reduce the risks [2] - but even such use is a risk, as to whether it gives ‘value for money’.

            An example of ‘insurance’, and whether the ‘premium’ was worth paying, occurred on the Storebaelt Railway Tunnel [3]. Two under-sea tunnels were driven from one coffer-dam. Two risks were easy to identify - first, a tunnel might penetrate the sea-bed, resulting in a catastrophic flood; second, a flood in one might flood the other. The premium required to insure against the second risk was the cost, delay and inconvenience of having a dividing wall to separate the two tunnels. This had to be weighed against the remoteness of the possibility of the first risk, sea-bed penetration. No doubt those involved considered that the precautions taken - the premium paid - to insure against the first risk were sufficient to render unnecessary any separate insurance against the second. It is easy to point the finger afterwards; not so easy to make the decision beforehand.

            It is self-evident that in any tunnel project there always remains an unquantifiable risk.  The risk itself may be known or guessed and the cost of failure may be calculated;  but it is not possible to estimate how likely it is to happen, only to guess.  All involved will hope for the best; the prudent will prepare for the worst.

2.2  Site Investigation

In the design phase of a project, the engineer must make some estimate as to what challenges the ground ahead is likely to hold, and for this he has to carry out site investigations.  The more he probes, the more it costs;  but at the same time, the lower the risk of anything going wrong. It is a matter of engineering judgement as to how much time and money he expends in this investigation; too much, and he is accused of extravagance; too little, and he runs the risk of disaster. Those involved in a tunnel project must be pragmatic, and accept that occasional collapses may be part of the price to be paid for the job [4].

2.3   Site Supervision

During the construction phase, possibly one very significant factor is the amount and quality of supervision, both by the Contractor and by the Engineer. Supervision, like Site Investigation, requires a balance to be struck between quantity and cost.

            It is often the case where several tunnels are being driven contemporaneously that supervision, both by the Contractor and by the Engineer, is spread, so constant close watch cannot be kept on every face.  Increased supervision reduces the risk, but more men will not necessarily prevent an incident.

            Where major incidents occur, it is frequently the case that the situation develops slowly, with each progression seeming not to be particularly significant as it happens, until it is too late to try to prevent escalation. Frequently, each stage in the escalation process is ‘something we have dealt with before’, and the overall picture is lost, possibly because attention is concentrated on the matter in hand. A classic case, in a totally different field, of this collective state of mind, was the crash of Eastern Airlines Tri-Star Flight 401 into the Florida Everglades on 29 December 1972 - the crew was concentrating upon the failure of the under-carriage locking light to come on, and ignored the accidental disengagement of the automatic pilot which allowed the aircraft to descend slowly into the ground [9 and 10].

2.4   Technique

A third major factor is technique.  In any given circumstance, a choice of methods is available, and the choice, very largely a matter of opinion and experience, is left to the Contractor who must carry out the work. Every tunnel is different; new ground is broken in the most literal of senses. The Engineer is in a difficult position if the Contractor is determined to use a technique which the Engineer believes to be inappropriate, particularly where, as an example, the Contractor has experience of using that technique in other circumstances, while the Engineer has experience of similar circumstances but not of that technique.

3.   Case Studies

The following incidents are considered, in greater or lesser detail :

• Victoria Line Underground Railway, London
• Inrush, Green Park, 1964
• Inrush, Somer’s Town, 1965
• Southend-on-Sea Main Drainage, Inrush, 1966
• Orange-Fish Irrigation Tunnel, South Africa :
• Flood, Shaft 2 South, 1969
• Fire, Shaft 5 South, 1970
• Flood, Shaft 2 South, 1971
• Overbreak, Shaft 2 North, 1972
• Hong Kong Mass Transit Railway :
• Inrush, Modified Initial System, 1982
• Inrush, Island Line, 1983
• Gibei Railway Tunnel, Romania, Inrush, 1985
• Istanbul Sewerage Scheme, Inrush, 1989
• Motorway Tunnels currently under construction,
• Analysis of Overbreaks, Tunnel 4, 1993/4
• Overbreak, Tunnel 4 South Tube, 1993
• Collapse, Tunnel 3A South Tube, 1995/6

            In some cases, just a sketch is enough to show at once what happened, and long explanation is not required.

3.1 Victoria Line Underground Railway, London

3.1.1 Inrush, Green Park, 1964

This was one of the two segmentally lined running tunnels driven by drum-digger shield through London Clay from Green Park to Victoria, with low cover under water-bearing sands and gravels close above. The crown of the shield penetrated through the top of the clay, probably at a buried stream valley, and an inrush of sand and gravel occurred, burying most of the shield.  A shaft was sunk from the surface to enable this material to be staunched and treated. Little physical damage was caused, but there was much delay while the shaft was sunk and the loose material dug out from the machinery by hand.

3.1.2   Victoria Line Underground Railway, London. Inrush, Somer’s Town Goods Depot, 1965

This tunnel was a running-tunnel re-alignment, some 300 metres long, driven with a hand-shield and lined with SG Iron segments 12'-2½" (3721mm) internal diameter, through London Clay under a disused railway marshalling yard.  It was known that the sands and gravels were not far above the crown, and great care was urged upon the miners. Upon arrival at work one Monday morning, it was found that the crown of the tunnel face had given way, and the sands and gravels had poured in, borne by water.

            A lengthy grouting operation ensued, to stabilise the ground in the vicinity, before the face could be recovered and tunnel driving resumed.  No significant damage was caused, but the drive was delayed by about six months.

            Some five years or more later, on another job, the Author met a miner who said he had been on the last shift before the inrush, and the true story was revealed - there had been a dispute about rates of pay. The men had felt hard done by, and after their shift, the leading miner had sent the men away, and then had attacked the crown of the tunnel with a clay-spade until the inrush occurred.  Or so he said.

3.2  Southend-on-Sea Main Drainage, Inrush, 1966

This tunnel was one of a network of similar sewerage tunnels, driven mostly by hand, through London Clay, again overlain by sands and gravels. The tunnels were lined with PCC segments with a finished diameter of 4[RBC1] '-6" (1350mm), and this tunnel had reached a length of some 30-40 metres from the shaft. A report was received in the office early one morning that “water had been struck” in this tunnel - a most unlikely occurrence in London Clay. The men had been digging for the next ring, when the miner had noticed needle-like jets of water issuing from the centre of the face, and across to one side. The crew immediately vacated the tunnel at speed, and went off for breakfast. When the Author arrived at the shaft, there was a slight trickle of water issuing from the tunnel eye, and a few bricks were lying at the bottom of the shaft. Half-way to the face was a lump of clay of about 100 kg, lying in the tunnel invert.

            At the face, the cause of the incident was apparent.. The tunnel had intersected the bottom of a well, at tunnel axis level, the well centre-line being in line with the side of the tunnel. The well was under a car park, and there was no record of it whatsoever on any of the plans. It had been capped probably some fifty years or more before, and forgotten.

            The well was about 600mm in diameter, and lined with open-jointed un-mortared brickwork; it had been sunk probably at least a hundred years before, through running sand and water-bearing gravels. It is interesting to speculate as to how it had been sunk, how the muck was removed, and how the brickwork was laid in such a small shaft.

            The tunnel drive was continued in a timbered box heading. Two steel plates were fabricated such that when placed in position, overlapping, they closed off the bottom of the well. Two pipes were included, one short, and one long for ventilation. Once these were in position, a mix of one bag of Ordinary Portland Cement (OPC) and one of Ciment Fondu (High Alumina Cement) was blown dry through the injection pipe. Upon contact with the water in the well, a rapid ‘flash’ set occurred, and the well was sealed off. Conventional grout was subsequently injected, and then the segmental lining was laid through the box heading and grouted up.

3.3    Orange-Fish Tunnel, South Africa

The Orange-Fish Tunnel in South Africa was constructed in the late 1960s and early 1970s.  It is 80 km long, some 1200 m above sea level and about 800 km from the sea, and carries irrigation water from the Orange River, stored behind the Gariep Dam, through the tunnel due South under the 200m high watershed divide to the headwaters of the Fish River in a different catchment.  The water starts in Lesotho, and would otherwise run Westwards through semi-desert to discharge into the South Atlantic at Oranjemund.  Instead, it is diverted to provide much-needed irrigation to the fertile Fish River valley, before discharging Southwards into the Indian Ocean [12]. The tunnel was divided into three Contracts:

• The Inlet Contract had an Inlet heading, and two inclined access adits (Shafts 1 and 2) at right angles to the tunnel, which was approx. 150 and 200 metres below ground, from each of which the tunnel was driven in both directions. Ancillary works included an intake structure and an emergency gate shaft / surge shaft.
• The Plateau Contract had three vertical shafts (Shafts 3, 4 and 5) about 400 metres deep, with a heading driven in each direction from each shaft.
• The Outlet Contract had two 150m deep shafts (Shafts 6 and 7), each with two headings, an Outlet heading, and outlet works, with an underground valve chamber / power station.

            The circular tunnel was excavated to about six metres diameter, by rail-mounted drill+blast equipment - seven-boom jumbos and Conway muckers. It was lined with 225mm of in-situ concrete to a finished diameter of 17'-6" (5.334m). The tunnel curved to follow the North-South meridian on the map, on a uniform down grade of 1:2000, with extra allowance for the earth’s curvature. The strata were sandstones, siltstones and mudstones, generally horizontally bedded, with occasional dolerite dykes. Normal roof support comprised a grid of 1.5 m long tensioned resin-grouted bolts 1.5m apart. Shotcrete, sometimes reinforced with steel mesh, was occasionally used, applied in increments of 25mm, and also, but rarely, steel ribs. Round length was generally 3 m, with progress normally steady at six rounds every 24 hours, six days a week, on each face.

            This Paper is concerned with risks to the Works, not with the personal safety of the workers. It is interesting to note that, on the Inlet Contract, there were 17 deaths, precisely the number predicted beforehand for each of the three Contracts. Of these 17, only one could fairly be described as a tunnelling accident - the others were more in the nature of “railway” accidents. No deaths were related to the serious incidents described below. The Plateau Contract had 34 deaths; again, none was related to any such incident.

3.3.1   Orange-Fish Tunnel, South Africa, Flood, Shaft 2 South, 1970

Shaft Two access works had been fully developed, and progress had built up well in both headings.  About 1500 metres had been driven in total, when the quantities of water met in the North heading started to become significant, and grouting operations were introduced. Then the quantities of water encountered in the South drive began to become significant, so the same system was adopted. The water in both headings was issuing through fissures in good rock. 

            The quantities increased; the Contractor introduced more pumps. The flows in the North heading were increasing to such an extent that radical re-thinking appeared to be indicated.  The inflows in the South heading also started to increase significantly, but were considerably less than in the North heading.

            One day, after a blast, the men re-entered the South face, to find that the tunnel had intersected a fissure about 75 mm wide crossing the tunnel almost perpendicularly, and extending all round the perimeter of the tunnel, issuing an estimated 3,300 cu.m. of water per hour at about 14 bars.  The entire 1.6 km of tunnel and 200m of shaft were flooded within 24 hours.

            Subsequent investigation indicated that the fissure was ‘unlimited’ in extent, and the incident affected farmers’ boreholes up to 100 km away. The ground water in the boreholes was found to be tidal, despite the elevation and distance from the sea.  Subsequent close examination of the geological records both before and after the event showed that the tunnel had passed through a shallow anticline, and the associated bending of the strata had formed the cracks, which increased in width as the tunnel approached the apogee - at the point where the flooding occurred. Hindsight also showed up the pattern of steadily increasing water inflows and increased quantities of injected grout.  At the time, it had been the North heading which had been the cause of anxiety.

            Grouting was carried out from the surface, and eventually the tunnel was pumped dry.  A concrete bulkhead was built, the void beyond, and the fissures, were then grouted up. A series of holes was drilled, at 20 degrees to the tunnel centreline, and arranged around the tunnel at 45 degree intervals, and these were then grouted. A new series of holes was drilled, rotated by 22½ degrees, and also grouted. The process was repeated with a pattern at 10 degrees to the tunnel centreline, and repeated until every hole drilled was dry.

            After this the bulkhead was removed, and tunnel driving was resumed. Excavation stages of 40 metres were phased with ground treatment stages of 60 metres, the face being always protected by at least 20 metres of treated ground.

            Initially, enormous quantities of cement were injected, without limit, in an attempt to staunch the flow. Subsequently, a method of ‘blob-grouting’ was developed, by which a ring of ‘blobs’ of very thick grout, rich in unhydrated bentonite, was built up in the fissures a short distance away from the tunnel, and the space between that ring and the tunnel was grouted tight.

3.3.2   Orange-Fish Tunnel, South Africa, Fire, Shaft 4 South, 1970

The fire at Shaft 4 South was very similar, in many ways, to the flooding of Shaft 2 described above.  Methane had been predicted from the initial pre-Contract site investigation, and all the equipment was ‘flame-proofed’ as a precaution. Methane testing was carried out after every blast, but no significant traces had been found previously.

             One day, after a blast, the men re-entered the face, to find it ablaze.  The drive had intersected a fissure, this one carrying methane, which fortunately had been ignited by the blast before it had reached an explosive concentration.  The fire burned for about six months, until a wall was built across the tunnel, and the void beyond, including the methane-bearing fissure, was grouted up with cement, in a method very similar to that used in Shaft 2 South described above.

3.3.3    Orange-Fish Tunnel, South Africa, Flood, Shaft 2, 1971

This incident is rather different - during very heavy rains, a farm dam collapsed. The flood waters spread out from the stream bed, and much found its way down the inclined shaft into the tunnel, carrying with it large quantities of silt. The tunnels by this time were sufficiently long that there was ample storage space for the flood water, but had it occurred when the tunnels were short, it might have presented a threat to the workers inside the tunnel.

            There was some slight damage to the inclined muck conveyor, and the muck chute and its access passage were completely filled with mud, which also extended along the main tunnels for some distance, being about half a metre deep at the foot of the inclined adit.

            This mud was simply dug out by hand; the major effect was delay.

3.3.4   Orange-Fish Tunnel, South Africa, Overbreak, Shaft 2 North, 1972

            This heading encountered a strata of weak mudstone which gradually crept up to the crown of the tunnel, where it caused some problems and no little concern. It tended to break along the joints, and could carry very little corbel effect, resulting in a very high arch, compared with the circular shape intended. Further, this material softened with exposure. The stratum eventually moved away from the crown, but there remained a length of tunnel very considerably higher than intended. The section was lined with circular ribs, and the ground propped by steel cribbing.

3.4    Hong Kong Mass Transit Railway

3.4.1    Inrush, Modified Initial System, 1977

The Contractor was driving by drill+blast through sound granite towards the reclamation. At some stage, he was required to construct a chamber for the installation of a shield and airlock, for the continuation of the drive through soft ground.  Unfortunately, the Contractor’s interpretation of the geological information led him to decide to continue with drill+blast just one round too far, and the blast from that round penetrated the rock cover, allowing the water-bearing fill above to flow into the tunnel, opening a hole in the road over the tunnel.

3.4.2    Hong Kong Mass Transit Railway, Inrush, Island Line, 1983

The Contractor was to drive a station tunnel through weathered granite, with fill close above.  He was required to treat the ground above the station tunnel from the pilot tunnel.  He was also required to treat the ground above the running tunnel where it emerged from the station tunnel.  Unfortunately, although both types of ground treatment were carried out, there was a gap between them, through which the ground flowed, again opening a hole in the road above.

3.5    Gibei Railway Tunnel, Romania, Inrush, 1985

The Gibei Tunnel lies on the railway between Vilcele and Rimnicu Vilcea in the Romanian Sub-Carpats - the southern foothills of the Carpathian Mountains. The tunnel is 2210m long, and passes under the Ciofringeni hill; it was the longest tunnel driven by shield in Romania during the 1980s.

            The geological investigation comprised the drilling of cores only in the portal zones and these revealed only compact clay. The same type of clay was found in trial pits which were excavated at the top of the hill, and it was therefore assumed that the same geology would extend the full length of the tunnel. Additional   information was available dating from 1939 when German engineers had studied the possibility of a railway on almost the same alignment. At that time construction had been considered to be technically impossible, the idea was abandoned, and the records disappeared. By the 1980s, however, it was considered that the use of a shield would enable the tunnel to be driven in safety.

            The shield was a hooded mechanical shield 9.05 m in diameter, fitted with a hydraulic bucket and a bottom-mounted conveyor belt. During the first few hundred metres of advance, the crown of the face became unstable, and a fixed platform was installed which enabled the top of the face to be supported, but this imposed manual excavation.

            From the start of driving, the progress rate was only half what had been expected. After the platform was installed, the rate fell to less than a quarter, but the over-break was virtually eliminated. Water occurred only as damp patches or occasional drops. The clay was compact without defined layers, but with fissures that allowed large chunks to fall away from the face.

            After nearly 600m of advance the water seepage slowly start to increase, but was less than a third of a litre per second, and was not considered to be a risk. Unbeknownst to the tunnellers, however, there was a lens of waterlogged fine-grained sand just above the crown of the tunnel, which was being driven on a rising grade into this lens. The clay between the top of the shield and the sand lens became more and more soft as the tunnel advanced, and eventually lost stability due to the dragging stresses created during shoving of the shield. The clay layer broke suddenly on the right hand side of the crown, allowing an ingress of water at more than ten litres per second accompanied by running sand, which came into the tunnel, covering the machine and coming to rest at a natural slope more than 30m in length. Every attempt to recover the face and to free the machine resulted in renewed water-sand inflow.

3.6    Moda Collector, Istanbul Sewerage Scheme, Collapse, 1989

The Moda collector tunnel crosses the Kurbaglidere (Frog Stream) near the Fenerbahce Stadium by cut and cover.  On the Northern side, a cofferdam was excavated, through very fine and unstable mud on the stream shore, but through blasted rock on the Northern side away from the stream.  The articulated TBM was installed in this shaft, with about three metres of rock cover over the tunnel eye.  While waiting for the TBM to complete a previous drive, it was suggested to the Contractor that a horizontal probe might be useful. In the light of subsequent events, the response was memorable: “You must be joking! We’ve had to blast out all that rock!” The TBM broke out of the cofferdam on its way North, and had gone but eight metres or so, when it intersected a hidden valley of fine mud which flowed through into the shield, opening a hole in the road some five metres above.  Broken rock also came in, jamming the shield and preventing it from turning.  A shaft was sunk down onto the TBM to release it.

3.7  Motorway Tunnels currently under construction [1997]

3.7.1General Background

The Motorway Contract runs from West to East for about 260 kilometres across relatively flat countryside, but in the middle, the route is obstructed by a range of hills. The motorway crosses the steeply-undulating divide by a combination of tunnels, cuttings, embankments and viaducts. There are nominally four tunnels; they are three-lane, twin-tube, and thirty metres apart between centre-lines, with a finished internal cross-section of about 103 sq.m.:

   T1 - 376 & 376 metres long      T3 - 2717 & 2760 metres long

   T2 - 536 & 562 metres long       T4 - 1225 & 1230 metres long

Tunnel T3 is divided into two sections, T3A to the West and T3B to the East, by an access adit. The tunnels are being driven by conventional drill+blast, heading+bench methods; the total area of excavation varies from one type of rock to another and is around 68 to 70sq.m. for the top heading, from 119 to 162 sq.m. for the complete tunnel.

            The strata through which the tunnels are driven are principally sandstone and shale, with zones of crushed and weathered material. The ground was classified into six Rock Support Classes, with RSC1 being the best ground and RSC6 the worst, according to a system based on Austrian Standard ÖNORM B2203 [5] and Rabcewicz-Pacher.

            The tunnels are driven using wheeled equipment, though trimming for underbreak is performed by a hydraulic breaker mounted on a traxcavator. Support is provided as appropriate by shotcrete reinforced with wire mesh, and rockbolts, and, where appropriate, roof-ties fabricated from re-bar or I-section steel ribs, similar to the New Austrian Tunnelling Method .

Excavation is a multi-stage process, see Figures 9 and 10. In the best ground, there is a top heading, followed by bench excavation. In worse ground, an invert may be provided to close the ring, and in the worst ground of all, the top heading may be divided into two side-drifts.

           In order to provide support to the face in RSC4 or worse, the Contractor chose to have a support core, as shown in Figures 9 and 10. This reduced the amount of work required to advance the face, as advance of the support core could follow later. However, it did restrict access, and also precluded any rockbolts being placed in the front two or three rounds.

            For driving through the rather poor ground where the collapses occurred, the Contractor installed about thirty grouted forepoles (32 mm diameter re-bar) four metres long at 400 - 500 mm spacing around the top of the perimeter of the ground to be excavated, the ends were supported on the top of the rib of the previous Round, itself buried in shotcrete to its full depth.

            The face was drilled, then blasted, and the final half-metre or so around the crown was trimmed by hydraulic breaker. A thin ‘sealing’ layer of shotcrete was applied from the muck pile / support core / loader bucket to the roof and to the top of the face, after which the muck was excavated. Mesh was tied to the mesh of the previous round which oversailed the rib by about 200 mm, and the mesh was also fixed by tying wire wrapped round the exposed forepoling or embedded in the sealing shotcrete.

            The rib was erected in three parts, bolted together when the rib was up, and the rib was located longitudinally by tie bars clipped to it and to the previous rib. Surveyors checked the location and orientation of the rib by theodolite, and the rib was adjusted as required. The rib at this stage was very flexible, needing to be fixed into position by diagonal ties welded to it and the previous rib.

            Once the rib had been fixed into position, the ground profile was measured from it. This clearance between rib and ground was used for calculations of overbreak volume; it was therefore in the Contractor’s interest that the quantity of sealing shotcrete be small.

            The bottom three metres on each side, between the ribs only, was shotcreted, followed by shotcreting of both shoulders and then the crown. The half-metre or so length of excavation in front of the rib was not shotcreted, so as to leave the mesh clear for attachment of the mesh of the next round. This part remained unsupported until the full shotcrete was applied as part of the next round (except for the initial thin sealing layer).

            The ventilation in all tunnels was poor. In some faces, the Subontractor blasted with hand-mixed ANFO explosives, which generated more fumes than normal, and he also used un-filtered and poorly maintained diesel equipment, resulting in very long re-entry time after each blast, and further re-entry time after mucking. The effect of this was that the structural shotcrete support to most of the round was sometimes not placed until five or six hours after the blast.

            In late autumn to spring, the shotcrete materials were cold and damp. Considerable problems were experienced in trying to produce and place in the tunnels shotcrete which could achieve the specified strengths [7]. The footing of the rib was not thought to be important in ‘normal’ ground, following the theory that the shell of mesh-reinforced shotcrete enables the ground itself to support the loads. The purpose of the rib is to strengthen the shell, rather than to prop the ground;  a secondary purpose is to assist with maintaining the shape of the excavation.

There is a subsequent secondary lining of in-situ concrete (Figure 11), and the two are separated by a polythene membrane, with a drainage layer of artificial felt behind [Ref. 6, Figure 5.8], [Ref. 2, Figure 14.9].

3.7.2   Motorway Tunnels, Overbreaks, Tunnel 4, 1993

During construction of  the two tubes of Tunnel 4, 131 overbreak incidents were recorded, with a total volume of 1461cu.m. There were also cases when no record was made, either because this was impossible or because they were caused by construction errors. Where overbreak was considered to be the fault of the Contractor, no agreed record was kept, since agreement implied agreement that the Client would pay. The maximum deformation measured in this tunnel was 120mm.

            The number of cases (131) and their relatively small volumes (11 cu.m. average) do mean that statistical interpretation may not be accurate. However, by interpretation the data, it is possible to draw these tentative conclusions:

1. The daily distribution shows a  marked increase in failures at the beginning and the end of week, see Figure 2. This suggests a failure to apply the support in time, which may be attributable to delay in shotcrete supply, or simply to the method of working.
2. The monthly distribution shows two peaks. The peak at the end of the year may perhaps be due to less favourable weather conditions and to the New Year holiday season. The peak in the middle of the following year is more difficult to explain. It may be that in that period the Subcontractor had an interest in prolonging the execution time, this being shown not only by the number of overbreaks, but also by the slower rate of progress in the case of faces without special problems. This seems to be the only explanation for a rate of progress of only 11m/month on one face that had no overbreak problems.The difference in the volume of overbreaks between the pairs of tubes driven from each portal is hard to explain, since the rock conditions were obviously the same over the greater part of each pair of the drives. The difference in water flows is also not significant.

4. It is interesting to analyse the effect of  the overbreaks on progress. In practice there is a psychological pre-disposition towards avoiding overbreak, because it delays progress. The instinctive reaction is to reduce the round length to one that ensures safety, overlooking a simple thing: what was not dug today, will have to be dug tomorrow. By reducing the round length, the only things that are reduced are:

• the amount of rock that has to be excavated in each cycle; and
• the shotcrete quantity which has to be applied in each cycle.

The time needed for the other operations of the cycle, including dead time and preparation time, remains unchanged. Only when the amount of time saved is less than or equal to the time needed to fill the void created by the overbreak, is the method to be recommended. Otherwise systematic use of this method leads to the prolonging of the execution time rather than a reduction. That may have been the intention.

5. In case of Tunnel 4 about half is in sound rock (Rock Support Class 4). The round length should have been between 1.5-2.5m with an average of 2.0m. In practice, because of the attempts to avoid overbreaks by reducing round length, there were 32 extra rounds in the South Tube and 5 extra in the North Tube. The Contractor achieved an advance of one round per day, virtually regardless of round length, and thus wasted 37 days.

3.7.3    Motorway Tunnels, Collapse, Tunnel 4, South Tube, West Portal, 1993

During the early hours of Sunday 18 April 1993, at chainage 699m, a 200cu.m. collapse occurred following the blast. The support in the last round was affected and the right side of the face caved in totally. The peculiarity of this occurrence is not the volume (which is not greater than in some other cases), but the water inflow.

            Water infiltration had occurred in several locations, but none was greater than 0.3l/sec. Two days before the collapse, the decision had been made to change the rock support class from RSC4 to RSC5 at that same chainage, due to a deterioration of the jointing pattern, the size, number and width, and the amount of clay in the joints. The geological prediction did not indicate any unusual situation, especially regarding the amount of water.

            Immediately after the collapse, the water infiltration increased, reaching a value of 3.5 litres/sec on 25 April, with a maximum of 7.5 litres/sec. on 8 May, afterwards declining to less than 0.2 litres/sec. in the whole zone.

            The attempt to continue excavation under the protection of the usual umbrella of forepoling (32mm diameter rebar, 4-6m long) failed, because any removal of the collapsed rock mass brought down a new mass of fractured rock in its place. The Contractor decided to use, for the first time, an umbrella, not of solid steel bars but of 50mm perforated pipes 6m long. Ideally, these should have been 9-12m long, which would have allowed more advances under a single umbrella, but they could not be installed to this length because the drillholes tended to collapse. These pipes were grouted after insertion.

            This had the following effects:

• an increase in the bearing capacity of each forepole;
• injecting grout gave some filling of the voids and consolidation of the crushed material;
• drainage ahead of the face - for this reason, some of the forepoles were not grouted..

After a period of testing and as experience was accumulated, the solution proved satisfactory.

            In addition to the use of these forepoles, an ample central core was kept, protected with a shotcrete layer 50 to 100mm thick, and excavation was carried out in small sections, each sealed with 30 to 50mm of shotcrete. This combination proved successful.

            Unfortunately it took a total of five weeks to pass this critical zone only 6m long, although the first two of those weeks were spent in developing the system; thereafter a progress of two metres a week was achieved, which, in these conditions, may be considered as acceptable. Finally, by adopting these same measures - keeping a central core, and replacing the rebar forepoling with pipes - the same zone in the North Tube was crossed uneventfully.

3.7.4    Motorway Tunnels, Collapses, Tunnel 3A, South Tube, West Portal, 1995/6

The South Tube had advanced without incident for nearly three hundred metres, before a collapse led to the loss of about ten metres of tunnel. The original face was recovered after six months, during which time the face of the North Tube advanced past the face of the South Tube. Within a few days of this recovery, however, a second collapse of the same size brought work to a halt again, and caused distress in the adjacent North Tube. Work continues. The Authors have presented the collapses in detail [11], but the outline and some additional detail is included here for completeness.

3.7.4.1    Geology

The detailed geology in the vicinity of the collapses is very contorted, with the bedding planes constantly changing in inclination, accompanied by fault zones. The stratum at and above tunnel level at the collapse is thinly-bedded dark grey to black shale, which could be excavated by pick and shovel, and can be crumbled in the hand into pieces the size of a large thick biscuit and smaller. The shale which caused the problems may be a lens in the sandstone which generally predominates in the area. Although saturated, the shale appears to have acted as an impermeable barrier to the natural flow of water through the ground, resulting in a perched water-table which the tunnel pierced. The tunnels are generally comparatively dry.

3.7.4.2    Lead in

The ground appeared to be deteriorating, and the water flow was increasing, so the round length was reduced to 1.7 metres for Rounds 158 and 159, and then reduced again to 1.5 metres for Rounds 160 - 163, after which the Rock Support Class was changed to 5 and the round length was reduced still further to 1.2 metres.

            About a week later, when the face was at Round 163, settlement was observed in Rounds 153 to 159; extra bolts were to be provided along the right side of the tunnel from Round 152 to 160. Water flow declined, with only drips from the roof from Round 158 onwards, and Round 165 was almost dry.

3.7.4.3    First Collapse

On the evening of 11th October 1995, after the mesh and rib had been erected, shotcrete had been placed on the left hand side of the face, but before any had been placed on the right or in the crown, the ground fell away to form an overbreak chimney of some ten to twelve cubic metres from the right-hand side of the crown of Round 165, and extending forwards past the rib to the face which had ravelled to be over a metre in advance of the rib, see Figure 6. This stabilised some three metres above the tunnel crown, and an attempt was made to fill the cavity with  shotcrete. This was not easy to attempt, because there were two layers of mesh at that point, resting on the rib, and the shotcrete nozzle was pushed through the mesh. The shotcrete was thus being applied from some distance away, with not much control, nor was it possible for the nozzleman to see clearly what was happening [8].

            Due to the awkwardness of the position, much of the shotcrete fell back as rebound; shotcrete could be most easily placed on the face side of the cavity, but little was probably placed on the back face of the cavity, due to problems of access.  More than one full day was spent spraying this cavity. Unfortunately, too, there were considerable delays in the shotcrete supply, occasioned by a break-down of the local shotcrete batching plant, which had necessitated the use of a batching plant some distance away. This situation had been operating for over a week, and had been sufficient for normal requirements.

At about 22:00 the following evening, 12th October, 1995, work was continuing on filling the cavity, which had been slow, partly because the access was difficult on account of all the rebound material. Bundles of mesh had been pushed up into the cavity, to reinforce the shotcrete, and to try to help stabilise the free surface of the ground, but this obviously prevented the shotcrete from reaching the rock surface. It was felt that a shotcrete shell could be formed which would suffice to stop the flow of ground into the tunnel.

            The miners were removing the rebound material to improve access, when the small but steady trickle of loose material from the right shoulder of the tunnel, from the back face of the cavity, started to increase rapidly. This flow continued to increase, until it became a flood, and eventually the ribs and shotcrete slowly collapsed over a length of some six rounds, allowing about 650 cubic metres of this loose material to flow into the tunnel, accompanied by considerable quantities of water, estimated at about twenty-five litres per second. When water-logged, this material had the consistency of wet concrete.

            As is usually the case, the collapse occurred on the night shift. Fortunately, the inflowing ground and water soon eased off, as the pressure of both was reduced, and the passage was blocked, by the material which had flowed in reaching its natural angle of repose. There was, of course, no immediate action which could have been taken to prevent or reduce the inflow.

            The water flow tailed off over about a month, back to the steady flow of about two litres per second which had existed before the collapse occurred. This indicated that the tunnel had driven into a water-bearing strata below the water table, and that the water table had been drawn down. The total volume of water abstracted was estimated at some twenty to twenty-five thousand cubic metres.

            Considerable work was carried out, and the position of the old face was recovered on 5 March 1996. Five months after the collapse, excavation through new ground started slowly; the round length was increased from 0.8 metres through the collapsed material to one metre, and then to 1.2 metres for Round 173 and onwards.

3.7.4.4    Second Collapse

One month after the old face had been recovered, after driving a distance of seven metres through collapsed material and seven metres of virgin ground, the face was at Round 179 when a second, virtually identical, collapse occurred at 09:30 on 3rd April 1995.

            In this instance, the mesh had been installed and rib 179 had been erected, but not fixed. Sealing shotcrete had been applied to the left-hand side, but not to the right because of a delay in supply. At about four o’clock in the morning, while the surveyors were checking the rib, they became aware of noise and movement, and exercised their discretion by fleeing the face.  Radial movement of Rib 174 occurred, last measured at about 300 mm. Shotcrete in Rounds 173, 174 and 175 started to crack and spall away. Some pressure relief holes were drilled, but it soon became apparent that another collapse was occurring, and two loads of spoil were dumped near the face to act as a dam to limit the inflow. The collapse thus started on the right hand side, back from the face, but extended progressively forward to the face, backwards to Round 172 and across the full width of the tunnel.

            Water followed the ground, starting on the left hand side, and reached an estimated flow of about 25 litres per second, reducing to 3 litres per second within four days, and to less than half a litre per second after ten days.

4.    Discussion

Obviously, the incidents described above could almost certainly have been prevented or avoided, in the same way that road accidents could be almost totally eliminated if no-one drove faster than five k.p.h. In both cases, the extreme is unacceptably ‘expensive’ in terms of money or time, and some risk is taken for some extra benefit, knowing that the gamble may be lost. In the last example, work has progressed with extreme caution since the second collapse, in an attempt to avoid the remotest possibility of a third collapse. Cost and time are seen as irrelevant. It is difficult to counter the argument (adapted from Oscar Wilde’s “The Importance of Being Ernest”) that the first collapse could be seen as an accident, the second might appear to be due to carelessness, but a third could be interpreted as deliberate.

            There is thus a relationship between the foreseen risk and the precautions to be taken against that risk. Water is one such risk; even a small amount may seriously disrupt activity, if it is more than anticipated. In every tunnel provision will be made to cope with a certain amount of water.

            The case of the Gibei Tunnel is curious, for if the water-bearing stratum had been discovered during the prior geological investigation, then either the tunnel would not have been started (as with the first attempt by the Germans) or else a more expensive construction solution would have been chosen right from the design stage - dewatering by wells, perhaps, or a pressurised shield (not available in Romania), or very expensive ground-freezing, all for just a few metres of tunnel, and it is unlikely that the overall construction time would have been any shorter. An inventive and flexible mining crew, with experience and skill in several construction methods, solved the problems at probably less cost and less time than a “properly engineered” solution would have taken, had all the problems been known beforehand. The case also highlights the fact that, in tunnels above the natural water table, and driven on an up-grade, natural dewatering by gravity will be achieved, at almost no direct cost, even if progress has to be suspended for several months.

            The most dangerous risks in tunnelling are generally those associated with inflows of ground, of water or of gas, from the zone around the tunnel or ahead of the face. The more knowledge that there is available regarding this zone, the lower the risk, and geological prediction is therefore the most powerful - possibly the only - weapon in the tunneller’s armoury that can enable him to avoid or to prepare for problems which he may encounter during the tunnel drive.

            Well before construction starts, a geological prediction has to be made, in order to establish the viability of any tunnelling project. In most tunnels, the advanced prediction made by the Tunnel Designer is found to give a reasonably accurate assessment of the ground. The Gibei Tunnel contrasts with the case of the Motorway Tunnel 3A; here the Tunnel Designer’s prediction was accurate at the start of tunnel driving, but the ground deteriorated unexpectedly after some two hundred metres. The only prior incident was an overbreak of some twenty-two cubic metres which occurred at a fault which had been predicted. Prediction does not automatically mean that there will be no problems; it means that precautionary measures are taken which reduce the effects.

            Drilling of probe holes ahead of the face may frequently be justified, in reducing the risk; but these have a cost in both money and time. Percussion drilling is cheaper than core drilling, but gives only very limited information, except where water is encountered, in which case the information can be very useful.

            Of course there are many ways of overcoming overbreaks, but the solution adopted must be appropriate for the Contractor, his experience and equipment. A possible solution is to ensure safety in the zone and traverse it quickly, with the profile later brought to its final shape by filling the hole either by shotcrete or concrete.

            Apart from the geological conditions, which obviously are the most important factor, the following other causes favour the occurrence of overbreaks:

• A large expanse of exposed and unsupported rock, including the tunnel face;
• Overestimation of the rock quality;
• Failure to install initial support within the stand-up time of the rock.

            The effectiveness of a central core in increasing face stability in unusual situations led the Contractor of the Motorway tunnels to use this whenever rock conditions deteriorated, or there was significant water infiltration, and eventually he adopted this central core and forepoling, together with the application immediately after blasting of a 10 to 20mm thick layer of sealing shotcrete to the rock surface, as his standard procedure for the remainder of the tunnel, even in RSC4, where the drawings did not require this degree of caution.

5.   Summary of specific possible aggravating factors

In an attempt to reduce the risks of a collapse, we present here a list of factors which may possibly aggravate a situation, and which may profitably be addressed for future projects :

1. Inadequate site investigation
2. Failure to probe ahead of the face
3. Insufficient immediate supervision
4. Delay in producing deformation measurement reports
5. Inaccurate analysis of deformation data
6. Delay in applying full thickness of shotcrete
7. Delay in delivery of shotcrete
8. Failure to apply shotcrete ahead of the last rib
9. Weakness of shotcrete
10. The use of forepoling
11. Failure to grout rockbolts bolts fully
12. Failure of grout to penetrate the ground
13. Use of inappropriate grout

6.    Conclusion

Tunnelling is still very much an art rather than a science. Each interpreter’s perception of the ground will differ from every other’s, to a far greater extent than is generally the case in other forms of construction. At the same time, tunnelling is very much a team endeavour; but a ‘team’ concept does not lend itself to ‘Specification’, and is very greatly affected by the varying personalities of the individuals involved, each of whose own previous experience, knowledge and understanding will be quite different; and then the ground actually encountered differs again. Every tunnel project will be, to a greater or lesser extent, subject to a ‘learning curve’, as the Engineers gain in experience of that particular series of ground formations - and even then, the ground may spring a surprise. It is inevitable, then, that in every tunnel that is driven, there are likely to be instances where the conditions lie outside the collective understanding of those involved, and a collapse becomes more likely, and this is one of the risks that must be taken and accepted, although every endeavour is made to reduce it.

            The title of the Paper refers to “Anticipating the Unexpected”. An archetypal incident is that of the Moda Tunnel described in 3.6 above. Few tunnellers would have expected what happened, but it could have been anticipated by probe drilling. It is hoped that this Paper may help to increase awareness, and to persuade tunnellers and their masters to “think the unthinkable”, to consider what could happen, the costs if it did, and the cost of the ‘premium’ required to avoid it - the proper meaning of ‘risk assessment’.

References

1. Safety of New Austrian Tunnelling Method (NATM) Tunnels.  (1996) The Health and Safety Executive, London
2. Whittaker, B.N. and Frith, R.C. Tunnelling - Design, Stability and Construction.  The Institution of Mining and Metallurgy, London
3. Biggart, A.R., and R. Sternath (1996). Storebaelt Eastern Railway Tunnel: construction. Proc. Instn. Civ. Engineers, Storebaelt Eastern Railway Tunnel 1996, pp.20-39.
4. Takacs, A.P. (1996) May we prevent a collapse? - searching for answers.  Proceedings of the Second National Conference for Underground Structures, Romanian Tunnelling Association, Bucharest.
5. ÖNORM B2203  Österreichisches Normungsinstitut, Vienna
6. The Institution of Civil Engineers, London - Design & Practice Guide (1996) Sprayed Concrete linings (NATM) for tunnels in soft ground.  Thomas Telford Publishing, London
7. Clay, R.B. and Takacs, A.P. (1996) Contractual aspects of testing shotcrete and rockbolts.  International SCA/ACI Conference Proceedings, E&FN Spon, London.
8. Code of Good Practice.  Sprayed Concrete Association, Aldershot
9. David Grayson (1988) Terror in the Skies - The Inside Story of the World's Worst Air Crashes.  Citadel Press
10. Frank Borman with Robert Serling (1988) Countdown.  William Morrow
11. Clay, R.B., and A.P. Takacs, 1997. Tunnel Collapses - case study. Tunnelling ‘97 p.481. Institution of Mining and Metallurgy / British Tunnelling Society, Olympia, 2-4 Sept. 1997,
12. Boniface, A., J.G. McKelvey, S. Nthako and N.F.B. Schmidt, 1997. Tunnelling projects in South Africa. Tunnelling ‘97, p.379. Institution of Mining and Metallurgy / British Tunnelling Society, Olympia, 2-4 Sept. 1997.