The Kingston-upon-Hull wastewater scheme is Yorkshire Water‘s largest ever capital project and at US$286M (2001) represents 15% of the five year capital expenditure programme. The new combined storm and foul water scheme links the existing east and west pumping stations to a new sewage treatment works and outfall and is designed to serve a population equivalent of 1.1M. The upgrade commenced in October 1996 to meet the 2nd EC treatment deadline in December 2000.
At the outset Yorkshire Water decided that team working was essential to deliver best value and achieve the programme. Key principals included early contractor involvement and use of framework consultants and contractors. The scheme development Phase 1 commenced in January 1997 with a four-month programme to develop a detailed scheme design, detailed risk assessment and evaluation and a fully priced target cost.
Following this proposals were evaluated on: Rates for Phase 1; management fee and pain/gain share proposals for construction phase; creativity; and commitment to teamworking.
Subsequently, the Phase 1 team was established comprising: Client – Yorkshire Water; Designers – Arup, BWEL, Mott MacDonald and Montgomery Watson; Project Manager – CDGI; Contractors – Birse, Miller (now Morgan Est).
The original scheme for a near surface rising main was abandoned for a value engineering tunnelling alternative. The increased construction risk of the tunnel was offset by the increased storage capacity reducing pumping costs and avoiding construction in busy surface locations.
The initial out-turn cost was estimated to be US$78.7M and the Client had a US$18.6M risk budget giving a total client budget of US$97.3M. The target cost was monitored on a fully open book approach with cost breakdowns available to all the team. 90% of subcontract work was let to 3 subcontractors to ensure transparency of transactions.
A joint insurance was established for the contract that covered the Client/Contractor and Project Manager. The Contractor’s All Risks policy had some specific extensions covering; TBM breakdown, acceleration and building settlement.
Cost over-run insurance cover was also taken out. This provided a cap to the Client’s liability and comes into force when the project cost exceeds the construction cost + the costs of the clients most probable risk. This is new to civil engineering and kicked in at US$97.3M.
A cost over-run insurance of US$64.4M was also taken out for a premium of US$2.1M.
The 10.6km, 3.6m id segmental tunnel was to be driven at depths of 15-25m through water-bearing glacial and alluvial soils under the north bank of the River Humber. The project included 10 shafts spaced at up to 1.8km of 7.5-12.5m diameter and depths of up to 30m. The target cost was based on ground treatment followed by underpinning. A value-engineering proposal to sink the shafts as wet caissons was adopted using long reach excavators for the stiff clay that was considered too strong for the grab to work. Ground treatment in the form of soft piles/jet grouting was used at the tunnel eyes with Bullflex seals used to seal the breakouts/ins around the EPBM’s. The tunnel lining, designed by Millers and manufactured by Charcon, comprised a 6 segment, tapered trapezoidal, 250mm thick conventionally reinforced ring with EPDM gaskets.
Two US$2.6M EPBM’s were supplied by Lovat to construct the tunnels. All wearing faces of the machine and the EPBM screw were treated by TRI-Form plate. The machines incorporated a ground conditioning system and a Zed guidance system. Grouting was carried out through holes in the rings.
Jet grouting was used on the project to treat the ground for construction of a 40m long back shunt. This was only partially successful and additional de-watering was required. It was also used for the construction of pre-planned stops to allow safe access to the EPBM cutter head. This method was found to be expensive and de-watering down to 5m above the tunnel combined with low Pressure compressed air was used instead.
In November 1999 the eastern section had been completed early. The western section was 200m past shaft T3 with the EPBM located underneath a marina.
On the 16th November the tunnel started to leak and a substantial 60m diameter crater formed around shaft T3, up to 2m in depth. The tunnel was evacuated and compressed air put on within 12 hours to stabilise the tunnel.
Recovery Plan and collapse investigation
Arup’s Bill Grose reported that the collapse occurred in an alluvial valley which represented the worst ground conditions on the project.
The collapse initiated close to Shaft T3 in a completed section of the tunnel. No immediate cause for the collapse was apparent. The investigation methodology included intensive ground investigation of the collapse location focusing on: construction; the lifetime of the section of the tunnel prior to collapse; the collapse; verifying the design; confirming workmanship; analyses; centrifuge modelling; and numerical modelling
The best chronology of events that could be established from piecing together eyewitness accounts is given below.
At 19.00hrs on Monday the 15th November 1999 a leak of around 2 litres/minute began, jetting around 75mm at the right side knee joint between rings 2403 and 2404. At 01.00hrs on the 16th November the leak on the right side remained the same. A leak occurred in the crown when a train passed and a subsequent invert leak started on the left side bringing in sand, which bubbled like a lava flow with clear water and brown sand. At 02.20hrs a gang started to re-insert the bolts in ring 2399. Plates started moving, with invert rising up and left side Shoulder moving in. Spalling damage occurred on axis joints.
Geology
The tunnel was in alluvial conditions with peat/organic clay at the crown. A substantial body of mobile fine wind blown Aeolian (single sized) sand occurred adjacent to the knee that could have been disturbed during shaft and tunnel construction and could be transported by the 19m head of water acting at the tunnel invert.
Seven 1/75th scale centrifuge tests were carried out at Cambridge University using epoxy resin segments encased in a plaster grout. These enabled the post-failure behaviour and impact of leaks to be assessed, demonstrating that:
- Without a void occurring or soft soil, no collapse occurred
- The shaft provided fixity of the tunnel leading to differential deformation if the tunnel moved
- Once material started to wash in, the tunnel eventually collapses
- A small length of tunnel in soft ground is as bad as a long length
The main limitations of the testing where:
- The tunnel was unrestrained longitudinally
- Non-repeatability of the experiments
- Lining stresses not measured
- Strength of lining not modelled
Numerical modelling
A 3 dimensional model was developed of the soils to assess the possible modes of failure. Issues modelled included the circle joint and details of the circular shear pad between segments and the impact of the gaskets. The modelling allowed an assessment of how the tunnel might distort longitudinally and whether any of the circle joints would open when restrained by the shaft. It also included a study of the magnitude of the stresses generated on the lining by longitudinal distortion and a parametric study on the ground conditions that may lead to distortion.
A parametric study was undertaken on the following:
- No gap between tunnel rings (concrete to concrete contact)
- 5mm gap between rings
- Original soil properties
- Soft layer of disturbed material around the tunnel
- Buoyancy of tunnel
- Buoyancy of tunnel + dewatering of the peat in the crown
The modelling indicated that:
- Relatively short lengths of soft soil have almost the same effect as long lengths
- Gaps between rings would be necessary at the outset for a breach of the gasketting system to occur even in very soft/disturbed ground
- Stress concentrations are worse if the joints are tight
- Breakage of a plate due to the shear stresses at a shear pad is a realistic possibility
- Breakage of a plate due to direct stresses on the outer face of a circle joint is possible but less likely to cause a leak
To summarise, two primary factors in combination led to collapse, fine sand under considerable pressure adjacent to the tunnel combined with a leak through the lining large enough to allow fine sand to wash into the tunnel.
The leaks are most likely to have been caused by movement of the tunnel relative to shaft T3, which lead to opening up of the circle joints and shearing between adjacent rings causing local structural failure around the gasket.
The movement was most likely attributable to compression of the peat above the crown, caused by the upward buoyant pressure of the tunnel combined with loosening of the ground by shaft sinking/tunnelling and dewatering of the peat layer by a leak into the tunnel or shaft.
Factors considered and eliminated included, the design of the ring, tidal variations in ground water pressure and construction defects.
Recovery and Key objectives
The recovery requirements included determining the cause of collapse to reduce the risk of reoccurrence; deciding upon a safe method to complete construction in the highly disturbed ground conditions taking account of surface constraints; facilitate connection to the segmental tunnel already constructed and protecting the underlying aquifer.
The preferred options comprised either vertical ground freezing with segmental lining or horizontal ground freezing with sprayed shotcrete lining. The former was discounted because the large frozen section of ground required could lead to large differential movements at the tunnel/shaft connection.
The solution adopted was a liquid nitrogen ground freeze with sprayed concrete lining (SCL). The main advantages being its flexibility to deal with changing conditions, the reduced drilling lengths, a high integrity freeze structure, and it is less affected by salinity and moving ground water compared to brine.
The method also allowed the safe excavation around and connecting to the existing tunnel, rapid establishment of primary freeze phase and the SCL provided flexibility to construct underground drilling chambers.
The safety critical issues of the method were the use of liquid nitrogen on a construction site and in a confined space, the accuracy of drilling the horizontal holes from the tunnel to avoid the vertical holes carrying liquid nitrogen and the use of sprayed concrete lining on frozen ground.
Beton-und Monierbau provided the SCL expertise with Keller drilling the vertical freezeholes, Insond the horizontal holes and Linde the freezing expertise.
The recovery was constructed in stages:
Stage 1: freeze bulkheads developed using vertical drilling external to the shaft;
Stage 2: break out tunnel eyes;
Stage 3: drill 25m long horizontal holes from the shaft to construct an enlarging cone of ground freezing around the tunnel and vertical holes to establish a frozen zone ahead of the excavation;
Stage 4: excavate the ground and construct a shotcrete lining including the enlarged drilling chamber for the second phase of horizontal drilling;
Stage 5: drill second 25m length of horizontal holes and vertical end and freeze;
Stage 6: excavate and line this section with SCL;
Stage 7: drill final section of horizontal holes around existing tunnel;
Stage 8: excavate final section of tunnel;
Stage 9: install permanent SCL lining and membrane.
The design of the AGF was based on tests carried out on frozen samples undertaken at Cambridge and Jessberger @ Bochum University.
The SCL was designed with BS8110. The ground around the tunnel was very disturbed and pressuremeter tests gave a Ko of 0.27. Prior to commencing shotcreting spraying trials were undertaken on a mock up which had a nitrogen freezing circuit to enable mix design adjustment.
Conclusions
The form of contract and the problems encountered resulted in outstanding performance and commitment from all the companies and individuals involved.
Innovative solutions to problems were encouraged and good communication and understanding were fostered between parties.
For the client cost savings of 25% where realised. For the insurance industry this was a disaster.
Related Files
Figure 2: Shaft 2TA near the collapse was constructed to remove the Lovat TBMs used to bore the majority of the tunnel
Figure 1: Map of the tunnel alignment with associated shafts
