The UX15 cavern, which will house the 7000t ATLAS detector (T&TI April), has an excavated span of 35.1m, some 50% larger than any cavern previously excavated at CERN and one of the largest in this rock type in the world. The cavern will be 42.3m high and 56.1m long resulting in a cross-sectional area of 1,380m². The USA15 technical cavern, adjoining at right angles to UX15 has a span of 23m, is 65.2 m long and between 18.3 and 31.4 m high (area 338-550 m²).

In addition to the caverns, two 60m deep circular access shafts, PX14 (20.5m diameter) and PX16 (14.5m) connect the vault of the UX15 cavern to the surface. Other excavations include junction and cryogenic chambers, personnel access tunnels and survey galleries. All caverns, tunnels and shafts will have permanent concrete linings, a waterproof membrane and drainage blanket. All of this work must be carried out adjacent to existing structures, much of it at a time when the LEP and SPS are still in operation and must not be disturbed. This has made the design and construction truly three dimensional requiring a great deal of design and organisation.

Organising the project

Following an international call for tender, responded to by 19 groups, the 50/50 Electricité de France-Knight Piésold Joint Venture (EDF-KP) was awarded the design and construction supervision contract for Package 1 at Point 1 in 1996. The design and construction supervision teams are fully integrated, with French and British engineers working together to complete the project within the tight programme constraints.

Construction of the works is being carried out by the Austrian/German/Swiss JV of Porr Asdag Tunnelbau/C Baresel/Zschokke Locher.

Principal technical difficulties

For the now complete design, a number of significant problems had to be addressed and solved (4). Unlike many engineering projects, little scope was available to the designers to modify and optimise the shape, size and location of the new works.

The molasse is a relatively well-known rock formation by virtue of the amount of work already undertaken at the CERN site. The molasse is overlain at Point 1 by between 5 and 10m of glacial moraine which comprises dense to hard sandy silt and numerous strong cobble and gravel-sized fragments. The molasse consists of irregular, sub-horizontal bedded lenses of variable strength rock with lateral and vertical gradational sequences of marls and sandstones. Contacts are generally gradational but some sharp erosion contacts can be seen. Generally, bedding can only be observed as a change in colour or local strength variations. The water table sits in the moraine with the molasse being effectively impermeable. One of CERN’s principal requirements is that no water should be allowed to pass via the shafts, as a number of the existing structures have slight water see pages which would be unacceptable to the new detector.

From the LHC site investigation, parameter values were determined from laboratory and site index tests, to define the general characteristics of the ground, and specific strength and deformation parameter tests to provide input data for the numerical modelling. Two sets of best-estimate and worst credible (mean minus one standard deviation) parameters were defined for the modelling.

Geological and geotechnical problems

The molasse is a highly heterogeneous rock mass comprising continually variable sequences of very weak ductile marls (UCS = 3 MPa), stronger marls, sandy marls, marly sandstones and strong brittle sandstones (50 MPa) with very different mechanical behaviours. This vertical and horizontal heterogeneity serves to concentrate displacements and stresses.

Some simplifications had to be made. Early modelling showed that the presence of weak marl beds played a dominant role in the behaviour of the underground excavations. Due to the dip of the beds, the locations of these marl beds also varied along the length of the major structures.

For initial simplification, the molasse sequence was divided into three primary units and a fourth unit for the more sensitive grumeleuse marls. The upper Unit 1 consisted of mainly sandstones. Unit 2, located above and in the upper parts of the caverns, is more variable and included significantly more marl beds (44%). Unit 3 consisted of 60-70% marls. Marl beds in important locations were included separately.

In-situ stresses

The vertical in-situ stress at Point 1 is effectively the weight of the overburden plus a small surcharge . The horizontal stresses are between 1.4 to 2.2 times the vertical stress. This high k value is probably due to the residual stresses remaining from the time when the molasse in the Geneva basin was significantly thicker and covered by a glacier and also from horizontal Alpine tectonic forces. These high horizontal stresses cause problems for the design of the large walls of the caverns and significant horizontal displacements are being observed during the current excavations.

Anisotropy of the marl beds

The sandstone beds can be considered as isotropic. This is not the case for the more marly beds, however, which can be thinly laminated with weak lamination planes. Horizontal and inclined polished fracture surfaces may also be present in weaker plastic marl beds.

Swelling and creep

The marls are very sensitive to water and swelling has caused problems with some of the existing LEP caverns, cracking the inner concrete linings and necessitating remedial works. The swelling is caused by a combination of deconfinement and increased water content. Where marl beds form large areas of the invert and vault, the swelling pressures have a significant effect on the design of the linings. In the sidewalls, the effects are still significant but generally more localised. Both the marls and sandstone beds are susceptible to creep deformations in the medium and long term.

Layout and geometry of the excavations

The layout of the works is very congested. With layout and geometry of the new excavations was constrained by existing structures. The sidewalls of the main UX15 cavern had to be vertical which is undesirable in this area of high horizontal stresses. The new works induce significant changes in field stresses around the existing works which they were obviously not designed to withstand. Some sections of the existing works will have to be demolished and some are required to remain in service during the works and may need strengthening. The design was also required to predict displacements of the LEP during various stages of the works prior to shutdown in order that adjustment could be built in to the machine. The existing PM15 shaft, which contains an important lift and US15, contains sensitive operating equipment and computers, which must all be maintained in working order, despite the proximity of the UX15 excavation. The two access shafts enter the vault of UX15, significantly reducing the desirable arching effect.

Planning considerations

Planning constraints have had a strong influence on the general concept of the new works and the determination of the construction phasing. The LHC completion date has lead to the undesirable situation of having excavated and lined (finished) new structures adjacent to very large ongoing excavations. This has resulted in very high densities (up to 10%) of reinforcement in some areas. The two main shafts and the technical cavern USA15 will be lined prior to the excavation of UX15 which will induce very high stresses in the linings and cause large displacements. Some of these effects have been mitigated by installing compressible material behind the linings.

The concrete vault and crane beams of UX15 must also be installed prior to the shutdown of the LEP in order to complete the cavern within programme. To accomplish this, it was necessary to design a system to support the vault before bench excavation.

Initially, it was envisaged that the vault could be supported from the crane beam which would be anchored to the rock using traditional "dead-end" ground anchors. However, the capacity of the anchors was limited by the strength of the molasse. The presence of the existing and new structures also limited the number of possible anchor locations. It was therefore decided to adopt a system of tensioned ties (38 no.), with an ultimate capacity of 3500kN, installed between purpose-built galleries and the vault. Instrumented re-stressable anchor heads will be located in the galleries to accommodate expected movements during bench excavation and dead-end anchors will be included in the vault concrete. After LEP shutdown, excavation and lining of the bench of UX15 and excavation and lining of numerous other galleries can begin. All of these works must then be completed before November 2002 when installation of the ATLAS services will begin.

Modelling

The design of the support for the principal excavations has been based on precedent from the LEP works, standard rock mass classification/support schemes (RMR and Q) and 2D and 3D modelling. During the initial design phases, modelling was carried out on 200MHz computers. Early 3D models took a week to run. The latest models had the use of 500 MHz computers and ran significantly faster (12 hours).

The modelling was necessarily interactive and iterative between the excavation and lining design teams in order to dimension the linings to the size of the excavation and vice-versa. It was also necessary to use a number of approaches to the modelling of the molasse as, although the molasse is generally considered to be a continuum, the inclined fractures and occasional bedding planes introduce important discontinuities into the rock mass.

2D modelling

The advantage of using a 2D model over a 3D model is that it is possible to include significantly more geological detail and the analysis is relatively fast. Both continuum (FLAC) and discontinum (UDEC) analyses were undertaken to investigate the effects of local stratigraphy, multi-staged excavation and support sequences, different support layouts and refined mechanical properties of the rock mass.

However, modelling in 2D had obvious restrictions with regard to the intersection of shafts with the caverns, and the presence of adjacent excavations.

3D modelling

After the initial 2D modelling, 3D models were analysed using the finite element program GEFDYN, but the main design effort was then switched to FLAC3D and all the main structures were included. The 3D models were considerably more time-consuming to establish and run, but were deemed necessary as the structures are so three dimensional that 2D modelling was not a sufficiently valid approximation for the displacements and load effects. ANSYS3D was also used for the design of the linings, vault and anchors.

The models were intended to analyse the global behaviour of the excavations and the existing works. Simplifications were necessary, such as reducing the detail of the geology and excavation sequences and the inclusion of only part of the foreseen rock support, as modelling was time-consuming.

Towards the end of the studies, more detailed models were created to investigate specific areas of concern, such as the effects of local geological variations on the excavations. These models were also used to compare the anticipated with the actual behaviour of USA15 cavern, in order to refine the predicted displacements and loads during the excavation and support of UX15.

Excavation, support and current progress

The design determined the dimensions of the headings for the caverns and also the principal support mechanisms. Support is installed in stages to provide both initial safety support and medium term support prior to concrete lining.

Due to the large difference in cross sections to be excavated (4m² to 1,380m²), a variety of machines have to be available. For the large excavations, Liebherr 954, 944 and 932 machines equipped with hydraulic hammers (1.6-3.9T) are used. The final excavation line is obtained in some excavations using Eickhoff ETH30 and ETH50 roadheader attachments. Mucking out is by gantry crane through the shafts using Secatol 13m³ skips.

To date, the existing PX15 shaft (10m diameter) has been refurbished and concrete lined. A thickened lining protected by 100 mm of compressible material was used for the base of PX15, due to the high expected displacements during UX15 excavation. Shear keys were installed to support the lining.

Now complete, the excavation of USA15 comprised two top headings and two benches. Waterproofing and concrete lining are ongoing with the first vault pours expected before the end of 2000. Roof rock support was by 6m long 25mm diameter grouted bars at 1m centres. The bars are pushed into a hole filled with Belcem mortar, at a water:cement ratio of 1:4.5. Drilling is by a three boom Atlas Copco 145 HD. Support in the walls was by 5m long grouted bars at 1.25m (V) x 2.5m (H) centres.

In all the excavations, wet mix fibre-reinforced shotcrete with 35kg/m³ Dramix ZP 305 fibres is used to prevent deterioration of the rock mass due to atmospheric and construction conditions and to provide local support. A final layer of mesh reinforced shotcrete without fibre completes the support. During bench excavation, the shotcrete was observed to crack along the vault centreline for most of the cavern length. The overall stability of the cavern was not compromised, but a layer of mesh was placed over the crack for safety purposes.

PX16 and PX14 shafts have been excavated and are supported by 4m and 5m long grouted bars at 1.5 m centres. Based on the results of instrumentation, the lower parts of the shaft had considerably reduced support. Heavily reinforced 1.6m and 2.0m thick collars of 100 MPa concrete are present at the base of these shafts to help support the vault of UX15. Both shafts have now been waterproofed and concrete lined.

The 13m high 35m span vault of UX15 began in May and was excavated in five headings and one bench down to crane beam level. Excavation of the vault was complete by the end of October. Support is by 5 and 7m long grouted bars at 1m centres and 200mm of shotcrete with layering as per USA15. The crane beam and concrete vault (1.3m thick) will then be constructed and hung from the tensioned ties. Excavation of the 30m high bench will be in 5m stages with support as per the vault. Support pressures for the two main caverns are between 200 to 300 kN/m². Note that due to the poor shape of the principal excavations, these values do not include the shotcrete.

Naturally occurring total hydrocarbon contents (THC) of 3g/kg have been measured in some poorly cemented sandstone beds which is considerably higher than acceptable limits for normal spoil disposal in France or Switzerland (0.25g/kg). A special containment area for up to 1,000m³ of polluted spoil has being constructed on site.

Instrumentation

To monitor excavations and compare model predictions, geotechnical instrumentation has been installed in most excavations. Some of the instrumentation cabling will be extended through the waterproof membrane and lining and optical targets have been installed on concrete linings to monitor displacements and loads in completed structures during later adjacent excavation works.

By the end, some 500 instruments will have been installed. During the excavation of UX15, most of them are moving and the results have to be examined on a daily basis! A well-designed, reliable and robust instrumentation system is therefore crucial.

Subcontractor specialist Geodata of Austria installed and reads the instrumentation and the engineer analyses the results. Optical targets are read manually using a total station theodolite. Other instruments, such as multi-point extensometers and pressure cells, are read remotely by a data acquisition system linked directly to the contractor’s computer, allowing real-time monitoring and reducing interruptions to the ongoing works.

Critical sections for monitoring were identified during the design. In USA15, settlements reached 50mm and horizontal displacements up to 30mm by the end of excavation. The results have been within 50-70% of the predicted displacements.

Based on the results of the instruments, significant reductions in the designed support have been possible in some structures.

Related Files
Three dimensional diagram