In previous articles discussing the Muir-Wood/ Curtis methods and elastic-elastoplastic closed-form analytical methods, the mathematical models associated with ground movement and associated load development have been discussed. The purpose of this article is therefore to review the results of available instrumentation and compare them to the predictions from these models.

LOAD DEVELOPMENT IN CHALK

As outlined in the article discussing Muir-Wood and Curtis methods, the visco-elastic extension to Curtis’s method was developed to analyse the tunnel linings proposed for the 1974/1975 Channel Tunnel. The alignment of this tunnel crossed that of the old Duke of Beaumont’s 1880 tunnel beneath Shakespeare Cliff, Dover. As part of the 1974/1975 works, various extensometers were drilled from Beaumont’s tunnel in advance of the service tunnel TBM and sixteen lining rings were equipped with a total of 180 vibrating wire strain gauges (VWSGs).

When the 1974/1975 scheme was cancelled by the Wilson government, the service tunnel TBM was ready to launch and a special grant was made by the Transport and Road Research Laboratory (TRRL) to drive the TBM through the instrumented section and install the instrumented rings. The data collected was interpreted to establish the rheological characteristics of the ground and predict the loading behaviour onto the lining. This data was used as a basis for the design of the subsequent Eurotunnel scheme constructed between 1988 and 1994.

Curtis (1996) reports that as part of this work additional instrumentation was provided in both the running tunnel and service tunnel linings in the form of:

  • Vibrating wire strain gauges
  • Photo-elastic stressmeters
  • Piezometers
  • Convergence monitoring lines and
  • Environmental data in terms of lining temperature at the gauge position, air temperature and humidity.

In addition separate tests were carried out at the University of Bristol to measure the creep behaviour of the concrete linings.

A typical instrumented ring is shown in figure 1. VWSGs were the main form of instrumentation, and in instrumented rings these were installed in pairs in the crown, axis, invert and knees. This enabled both the bending moment and axial load to be determined at each location.

In addition, a limited number of photoelastic stressmeters (PESMs) were glued into preformed holes in the segments. These record stress by interpretation of the stress patterns viewed through a special optical reader. The stressmeter itself was in the form of a cylinder of optical glass. These instruments were used to corroborate the VWSG data.

Each location had a piezometer; however these recorded low and haphazard results and failed to give meaningful data. This was possibly because the selected instrument was capable of measuring heads up to 200m and was insufficiently sensitive to record the low pressures encountered. Alternatively, the instruments may have been sealed off by the tunnel backfill grout.

The convergence monitoring lines recorded a squatting profile of the rings as they came under load and this is consistent with VWSG data which showed that for crown and invert stations all gauges recorded an increasing compressive strain for the first 8-14 days after which the intrados gauges tended to lose compressive strain. Similarly, gauges at axis stations showed increasing compressive strain but diverged to indicate bending with tension on the extrados face.

Creep tests on concrete linings were carried out at the University of Bristol. This data was necessary to calibrate the VWSG data. The magnitude of creep strain was found to be highly dependent upon age of loading and moisture content. Samples loaded at an early age (28 days) experienced a high degree of creep with strain at 50 days being double the instantaneous strain. Samples loaded at 28 days and kept saturated experienced 15% less creep than samples loaded at 28 days and kept dry. When load was applied at 180 days, the creep was very much less and there was little difference between wet and dry samples. For this reason, in the interpretation of the VWSG data, it was necessary to be cognisant of the casting and curing history of the instrumented rings. In general, it was considered that some 40% of the measured strain was due to creep.

The correlation between the PESMs and the VWSGs was described as “reasonable to good.” In relation to the VWSGs, the PESMs both overestimated and underestimated the stress in the linings by up to 25%. On the other hand, the PESMs did offer the advantage of directly measuring the stress in the lining, thus negating the need to predetermine strain. The problem of determining creep and shrinkage strains is therefore avoided with the PESMs.

As illustrated in Figure 3, the results showed good agreement with predicted loads. In this figure:

  • TSS is the total expected service stress inclusive of bending effects.
  • EUS is the expected uniform service stress.
  • Z maximum stress is the highest monitored bending plus axial stress calculated for bending about the Z axis, and
  • Average stress is the highest monitored average stress. This was calculated from the four gauges at the station recording the highest Z maximum stress.

The monitoring of the Channel Tunnel Linings was considered a success. Lining performance was very much as predicted, with average stress for concrete linings ranging from 49.4% to 96.4% of expected values. No changes to the design were instigated by the monitoring results.

GROUND MOVEMENTS AROUND AN ADVANCING TUNNEL FACE

During the construction of the Brixton Extension of the Victoria Line, a series of observations on the motion of London Clay around an advancing open-face tunnel shield were performed. The tunnel was 24.4m deep and of 4.1m external diameter. A few undrained compression tests on small borehole samples showed the undrained shear strength of the clay was 7,000lbs/ft2 (335kPa) at the level of the tunnel. Assuming an average bulk density for the overburden of 20kN/m3, this would indicate a stability number Ns=1.46. Slow plastic squeezing of the ground would therefore be predicted.

Firstly, as illustrated in figure 4, a set of extensometers was installed from an existing tunnel, at the axis level of the approaching tunnel at points a, b and c which were respectively 488mm, 1.98m and 3.5m outside of the tunnel excavation.

Secondly, as illustrated in figure 5, a further set of extensometers was installed at points A, B and C at a point 7.62m before the shield entered the timbered end of a tunnel chamber. Point A is on the centre line of the tunnel, point B is at the axis of the tunnel at the periphery and point C is at the axis level but 305mm outside the periphery.

The shield had a front hood extending 518mm equipped with a cutting bead 3/8in thick and 229mm long which extended around the upper 300° of its periphery. The shield was advanced in steps of 508mm and a ring of bolted cast-iron lining was erected in the tail of the shield. A gap of about 38mm between the clay and the lining was filled with cement grout soon after the shield advance. The convergence measurements were considered accurate to a few thousandths of an inch.

The influence of the shield can be observed from point A at some 6.1-7.6m in advance of the tunnel face, i.e. some two tunnel diameters.

The lateral movement of point ‘a’ passing alongside the shield is instructive. As the shield approaches, the movement starts rather abruptly and remains almost linear until it comes to the bead. It then subsequently accelerates and then slows towards the tail of the shield, suggesting that after leaving the bead the clay converges and bears onto the tail of the shield. As it passes the tail of the shield there is the largest sudden movement which subsequently slows down as the grout hardens and the lining takes support.

The authors attempted to correlate these observations with surface settlement profiles and concluded that the “the loss of ground immediately at the tunnel boundary is of the same order as the loss at the ground surface.”

The observed movements correlate closely with the ground movements predicted by Panet’s curves – discussed in a previous article. The zone of influence of the tunnel extended approximately two tunnel diameters in advance of the face, approximately 50% of the radial ground movement occurred at the tunnel face, and slow plastic squeezing of the clay continued to occur until arrested initially by the shield body and subsequently by the permanent lining as the backfill grout started to set.

PORE PRESSURE CHANGES

The pore-pressure changes around a closed-face tunnelling machine are highly dependent upon the face pressure employed during excavation. This is illustrated by reference to two projects:

Singapore MRT Contract 301B

During the construction of the initial phase of the Singapore MRT the running tunnels between Lavender and Bugis were excavated by early Earth Pressure Balance TBMs. These early machines were not equipped with tail-void grouting systems and back grouting was performed through grout holes in the segmental lining. These tunnels were excavated through the soft, plastic Marine Clays of the Kallang Formation.

It was appreciated that the major cause of ground loss and surface settlement in these materials was ground squeezing into the tail void; without compressed air support, the soft clays would close almost immediately onto the segmental lining as soon as they left the shield. Although the theoretical volume of backfill grouting was injected through the grout holes in the rings, this grout tended to form bulbs and fractures in the ground and not to flow around the ring. This grouting was considered more a contractual obligation than an effective settlement control measure. Therefore, in order to minimise surface settlements, the contractor elected to over-pressurise the tunnel face and heave the ground in advance of the TBM, with the view that when the TBM had passed, the surface profile would return to approximately pre-existing conditions.

As part of a monitoring programme, pore-water pressure measurements were made by MRT staff from within the tunnel. BAT-type low-volume piezometer probes were jacked out through the axis level grout plug holes and the tips were located approximately 1m beyond the tunnel extrados. The probes were generally located in the last ring built as soon as it had cleared the tailskin and been grouted.

A single probe was installed in the eastbound tunnel at ring 665. At this point, the depth from ground to tunnel axis was about 18m with an estimated insitu total stress of 300kPa and water pressure of 165kPa. The average pressure during the shove was 350kPa with an at-rest pressure during ring building of 250kPa. The initial reading from the probe at ring 665, taken some six hours after excavation showed a pore pressure of 256kPa. The high excess pore pressure, some 91kPa over the hydrostatic, is clearly related to the high face pressure during shoving. Unfortunately, insufficient sealing thread was used to waterproof the probe tube which was observed to act as a local drain for the soil. As a result, it was not possible to accurately monitor the decay of the excess pore pressure with time.

The tactic of over-pressurising the face to limit settlement did not prove particularly effective: a heave of up to 6mm was measured in advance of the face and the initial settlements attributable to ground losses were typically 30-40mm. However, in the six to eight months following completion of the tunnels the additional consolidation settlement, as the excess pore pressures dissipated, measured between 18 – 81mm.

Crossrail Contract 300

The Crossrail running tunnels between the Royal Oak portal and Farringdon station pass beneath Hyde Park. London University was granted permission to establish a research site within the park to monitor ground responses associated with the tunnelling works. The results of this monitoring have been published in three companion papers by Wan et al (2019).

The EPBMs used to construct the Crossrail tunnels were 7.1m in diameter with tapered shields of length 11m, articulated into two sections. The precast concrete linings were 6.8m external diameter and the TBMs were equipped with a continuous tail void grouting system that filled the annular void between the tunnel lining and the excavated profile as the TBMs advanced.

The study area was divided into two main sections; an area designed to monitor the influence of the Crossrail tunnels on the Central Line tunnels (as the Crossrail tunnels pass underneath), and a second area where the Crossrail tunnels were in virgin ground. Figure 6 shows a cross section through the instrumented area in virgin ground.

At both instrumented zones, the axes of the TBMs were approximately 34.5m below ground level, the crown of the tunnels was in London Clay division B2, and the invert was in the more permeable A3ii unit. Hence, the total overburden stress was approximately 690kPa and the undrained shear strength measured from samples in Hyde Park ranged from 150kPa to 400kPa at the level of the Crossrail tunnels. The stability number Ns was therefore in the range 1.73 to 4.6. In the virgin ground area under Hyde Park, TBM face pressures during the excavation of the eastbound tunnel were reasonably consistent in the range 190kPa to 220kPa but tail void grout pressures were more variable in the range 60kPa to 210 kPa.

Multi-level piezometers, each containing six sensors, were installed in two boreholes; HP33 was installed directly over the crown of the eastbound tunnel and HP32 was installed at the side of the tunnel. Four spade cells consisting of flat rectangular spade-shaped oilfilled chambers connected to a VW pressure transducer were installed in boreholes and orientated parallel to the tunnel alignment. Each spade cell was also equipped with an independent piezometer connected to a separate VW pressure transducer. These spade cells indicated Ko values in the range 1.9 to 2.4 although, as discussed in the paper, these spade cells tend to over-estimate the true value.

The changes in pore water pressure measured in HP32 and HP33 are summarised in Figure 7. It can be clearly seen that the pore pressure increased in advance of the cutter head, an observation attributed by the authors to an increase in ground pressure due to longitudinal arching over the shield. These observations are consistent with the results of Barratt and Tyler (1976) as reported by Ward and Pender (1981):

Measurement of pore pressure changes in the London Clay at Regents Park were made around a tunnel driven with a 4.2m-diameter hand shield and lined with expanded concrete segments. A small rise in pore pressure was reported with the approach of the shield, then a substantial 75% fall in head abreast of the shield. This was followed by a slow increase with time, although the pre-tunnelling values were not re-established.

It can also be seen from figure 7 that a significantly greater change in pore pressure occurs in the soil above the crown and below the invert of the eastbound tunnel, compared to the pore pressure changes measured at the axis of the tunnel. As discussed in a previous article, this observation is consistent with Burger Schmitt’s load path analysis for excavations in over-consolidated clays. The resultant greater swelling pressure in the crown and invert explains why tunnels in overconsolidated clays squat rather than ovalize.

LONG-TERM LOAD DEVELOPMENT AND SQUATTING OF TUNNEL LININGS IN CLAYS Observations around London Transport tunnels

In the UK, a number of authors have attempted to monitor the long-term load development of tunnel linings installed in clays, a significant proportion of which relate to London Transport tunnels in London Clay.

During the period September 1954 to March 1956, the Building Research Station measured the existing circumferential stresses at a large number of points in the segmental cast iron linings of old underground train tunnels in the London Clay.

The work was carried out at four sites where two or more tunnels had existed for 50 years or more. At all sites, the construction of new access tunnels allowed the complete removal of several rings of the existing tunnel linings. At these locations, the rings to be removed were instrumented with vibrating wire strain gauges on both the skin and the web at several points around the ring. The change in stress as the segments were dismantled allowed the load that they had been carrying to be measured. It was concluded that the circumferential thrust in the linings corresponded to the full overburden pressure acting hydrostatically.

During the construction of the Victoria Line, at three points the stress distribution in the concrete linings was observed by incorporating photo-elastic stress gauges. Each gauge comprised a 1.25in (32mm)-diameter by 1.5in (38mm)-long glass cylinder cemented into preformed holes in the concrete segments. When lit by a probe light and viewed through an analyser, the gauge directly displays the magnitude and direction of the principal stresses in the surrounding concrete. About 200 gauges were installed. After one year, the loads were reported to have reached between 42% and 62% of full overburden pressure.

London Underground has collated the time-related load development on tunnels in the London Clay from various sources. These are presented in figure 8 taken from LU’s Manual of Good Practice. As can be seen from this figure, a progressive load increase is indicated in all instruments. It is also interesting to note that the horizontal loads monitored are typically in the range 55% to 80% of the vertical loads. This indicates an effective K in the range 0.55 to 0.8 and not the high Ko values typically associated with London Clay.

As discussed in previous articles, the long-term load development on linings installed in overconsolidated clays depends on swelling of the surrounding soil, as negative pore pressures created during construction are satisfied and a normal groundwater regime is established. The ability to re-establish the pre-existing groundwater regime is dependent upon the relative permeability of the lining and the surrounding ground. If the tunnel lining acts as a drain and dries the soil around the tunnel, swelling of the clay will be inhibited. In the case of the Victoria Line concrete segments, these were ungasketed and would have acted as a drain, locally drying the clay around the tunnel and limiting any build-up of pore pressure. The evidence for the extent of such drying effects is mixed. In measuring the pore pressure changes around the Crossrail tunnels, Wan reports that a piezometer line installed about 5m from the extrados of the existing LUL Central Line westbound tunnel showed a drawdown of approximately 3m to 5m head around the tunnel. The Central Line tunnels were constructed c.1900 and lined with cast-iron linings. Wan concluded that this drawdown was attributable to drainage from the Central Line tunnels. At this point, the Central Line tunnels are in a thick deposit of the low permeability B2 unit of the London Clay Formation.

Conversely Gourvenec et al studied pore water pressure around a section of the Northern Line tunnels at Kennington Park. These tunnels were constructed by shield methods in 1924 and are lined with bolted and grouted cast-iron segments. The tunnels lie close to the base of the London Clay and in contact with the Lambeth Group–Woolwich formation. At tunnel axis level the London Clay is described as:

“very stiff, dark greyish brown, closely fissured, fine sandy clay, with closely spaced partings of light greyish-brown silty fine sand”.

The authors measured only a slight reduction in pore pressure in piezometers installed just behind the lining and concluded that far field conditions were reached within about 1.5m of the tunnel extrados. The authors further considered that:

“the high pore pressures close to the lining were surprising as it is usually assumed that segmentally-lined tunnels in clay act as drains causing permanently reduced pore pressures close to the tunnel. The conditions observed at Kennington probably resulted from the sandiness of the clay”.

As a further consideration, considerable heat is generated within London Transport running tunnels and this is likely to bake the surrounding clay. The combined effects of drainage and heat are likely to create a thick cylinder of low induration strengthened soil around the lining. This strengthened zone is likely to limit load development onto the lining.

Thames Water Ring Main project

Long-term load analysis was also undertaken on concrete wedge-block linings installed as part of the Thames Water Ring Main Project, constructed during the 1950s and 1960s.

The original Donseg and subsequent wedge blocktype linings, as illustrated in figure 9, were expanded directly against the ground by the wedge key without backfill grout. The diameter of these rings varied between 2.54m to 2.73m depending upon hydraulic requirements. During construction of the Thames–Lee tunnel in 1956, nineteen rings were equipped with vibrating wire-type gauges to allow measurement of the circumferential thrust in the tunnel lining. Hemispherical reference studs were also fixed to allow hand measurement of the diameters.

Load development was monitored over a period of eight years, and the gradual changes in hoop loads are shown in figure 10. The tunnels were flooded periodically from 1960 onwards as shown, but only to low pressures over the instrumented section. At four sites it remained possible to still read the hoop load gauges in 1973 after a further nine years (i.e. 17 years after installation). All of these gauges showed a further increase of load up to a maximum of 89% of the overburden load.

Although the wedge block linings were provided with a caulking groove, they were ungasketed; the design philosophy being that the impervious London Clay stratum would limit losses from the main without the need for an internal steel liner. The wedge block liners would probably have drained the London Clay for the first four years post construction. Subsequently, when the tunnels were flooded, any unsatisfied pore pressures would have a ready supply of water that would allow the pre-existing groundwater regime to be re-established.

It is probably for this reason that the long-term hoop loads measured on the Thames-Lee tunnel are higher than those measured on the London Underground.

During the time the Thames-Lee tunnel was under construction in 1955 and 1956, a number of hemispherical studs were planted in the instrumented drives as diameter reference points. Figure 11 shows the difference between vertical and horizontal diameters. The tunnels showed a tendency to squat on average by about 7.5mm over periods up to three years. This represents approximately a 0.3% change in diameter under load.

CONCLUSIONS

Previous articles have discussed the mathematical models associated with various closed-form analysis methods. This article has compared the load and behaviour predictions to the results available from various instrumented tunnels. In this regard, the load development in chalk with regard to the Curtis’s viscoelastic method and the movement of soils around an advancing tunnel face with regard to the stability number and Panet’s curves are all considered to compare favourably to instrumented results.

With regard to the pore pressure changes around a TBM excavating in an overconsolidated clay, it can be seen from the instrumented Crossrail tunnels beneath Hyde Park and the results reported by Barratt and Tyler, that if face pressures are under-pressurised, then pore pressure changes closely follow Burger Schmitt’s load-path analysis.

With regard to load development on tunnels in clays, as previously discussed, Schmitt estimated that the long-term average pressure acting on a lining in a normally too lightly overconsolidated clay would be in the range:

pl = σ'v(1 – 0.5 sin ?') + uo and pl = σ'v + uo

i.e. full overburden pressure. Taking the Hyde Park profile as typical and assuming uo = 288kPa and ?’ = 25° this would indicate that an average lining load in the range 87% to 100% of full overburden pressure would develop with time. However, Schmitt’s analysis assumes that the pre-existing pore water pressure uo would be re-established around the lining. As discussed the LU tunnel lining will frequently act as a drain; if we therefore assume that uo is zero, then the average lining load would be predicted to be in the range 46% to 58% of full overburden pressure. These percentages will vary slightly depending upon the actual overburden pressure and degree of local drawdown in the London Clay.

Only in one instance, on the Thames Water Ring Main where the internal water pressure prevented the tunnel acting as a drain, did lining loads approach the upper predicted values. On all the instrumented LU tunnels, the monitored lining loads typically fell into the lower predicted range, indicating that an element of drainage was involved. The instrumented results discussed in this article all relate to ungasketed concrete linings or old cast iron that were, at best, caulked. It may be that modern segments, equipped with effective gaskets will attract greater loads.

In all instances, lining loads were observed to build progressively with time over many years and evidence from linings installed over 50 years ago indicates that loads may eventually approach full overburden. Even in soils with Ko greater than one, all linings were observed to squat.