The UX15 cavern is sandwiched between extensive existing and, until recently, fully functioning physics research facilities at Point One of the CERN Large Hadron Collider (LHC) Project. The cavern has an excavated span of 35.1m, by far the largest spanned underground excavation in the molasse in Switzerland and among the largest weak rock caverns in the world.

Construction at Point One is being carried out by a joint venture of Porr Asdag Tunnelbau (Austrian), C Baresel (German) and Zschokke Locher (Swiss). Design and construction supervision is by a JV of Electricité de France and Knight Piésold (UK). The LHC works are due for completion at the beginning of 2003, with the LHC coming on line in 2006.

The contract stipulated date restrictions on certain construction activities to co-ordinate the ongoing physics experiments with the new LHC works. One such constraint was that for fear of significant displacements affecting the existing Large Electron Positron (LEP) machine (T&TI May 1986, January 1985 and October 1982), no excavation could be carried out below crane beam level before LEP shutdown in October 2000. This left only 20 months after shutdown to excavate the 30m deep bench and concrete line the cavern, ready for the installation of the new 6,500t ATLAS detector and LHC machine. It was therefore necessary to line the vault prior to shutdown and minimise the amount of later work. This was a significant factor in the design for the UX15 cavern.

The 10,800t vault lining is therefore being temporarily suspended from 38 pre-tensioned cable ties installed between the vault and four overhead galleries, combined with rigid connections to the linings of the two access shafts that enter the vault (T&TI September 2001). When excavation is complete, the invert slab, up to 5m thick, will be cast and the cavern walls concreted up to the underside of the suspended vault structure. The ties will then be de-tensioned and the holes grouted.

Three-dimensional modelling has been undertaken to design the structures and determine the expected behaviour of this complex support system. The general layout and geotechnical conditions of the UX15 cavern have been described elsewhere (Parkin, T&TI 2000) and only features relevant to the tensioned tie support system will be described here.


The rock surrounding the cavern is supported by 25mm diameter grouted bars 5m and 7m long at 1m centres. A 20cm thick layer of shotcrete is used to protect the sensitive rock material against degradation through the action of air and water, and also to act as local support between the grouted bars (Laigle and Boymond, 2001). The vault concrete lining is therefore not designed to support the rock mass.

The suspended vault structure comprises a curved 1.3m thick concrete roof and 1.5m thick sidewalls. 1.45m thick crane beams run the length of the sidewalls and 1m thick endwalls are stiffened by an external beam 1m deep. The presence of a drainage blanket and waterproof membrane and no underlying supporting rock means that the vault is essentially freestanding.

The preliminary design of the vault support system was amended during the detailed design stage. Initially the vault concrete lining was planned to be supported entirely by 72 tensioned ties, with a movement joint in the concrete lining at the junction of the vault roof and the access shafts. In the final design the number of tensioned ties was reduced to 38, with the balance of the load carried by the concrete lining of the new shafts. This was achieved by replacing the joint at the shaft/vault junction with a rigid connection. It significantly increased the amount of vertical reinforcement in the concrete lining at the base of the shafts, up to a maximum steel content of 220kg/m3. The load in the shaft linings is then transferred to the surrounding rock via three shear keys formed in the concrete lining of the two shafts.

The vault support system has been designed taking into consideration two principal stages of construction. The first is the support of the self-weight of the vault concrete and the second stage comprises load increases due to ground displacements during bench excavation. To simulate these stages of construction, two models were created. The structural software program ANSYS was used to determine the distribution of the self weight of the lining between the ties and the shear keys following removal of the underlying rock. This allowed the positions of the ties to be optimised and an initial pre-tension value to be determined for holding the vault in place.

FLAC3D was then used to model the complex support system taking into account the geology, the existing and new concrete structures and the programmed sequence of excavation. Expected displacements of the vault were determined along with expected loads in the ties and stresses in the shafts at various stages of bench excavation. Because of cavern excavation, the rock mass settles up to the surface, 60m above. The differential displacement between the vault and the anchor galleries causes the increase in loads expected in the ties during bench excavation.

The initial analysis and design for the ties was completed before the start of excavation. Because of the innovative and complex nature of the support system, the FLAC3D model was re-run incorporating the encountered geology, as-constructed dimensions of the vault structure and the Contractor’s preferred sequence of bench excavation.

The as-built lining thickness varies considerably and in some areas is almost double its design thickness. This may be positive from a structural stiffness viewpoint, but it increased significantly the overall mass of the vault structure to be supported. The analysis showed that the increase in self weight was carried by the shaft concrete linings.

Construction of support system

Excavation of the new PX14 and PX16 access shafts was halted some 15m above the crown of the cavern and two small anchor galleries (cross sectional area of 27m2) were excavated from each shaft. The size of the galleries was dictated by the equipment required to drill the tie holes to the specified tolerance of 2% of the drill length. Being temporary, the galleries were supported with shotcrete and grouted bars and will remain unlined. Shaft excavation then continued while the concrete invert slab was poured and the tie holes, 16.5m to 22m long, were drilled. Because precise orientations were required for the ties, drilling was carried out by specialist subcontractor Stump of Switzerland, and the holes surveyed using an inclinometer. A number of holes were grouted and re-drilled to achieve the required accuracy. They were then cased with a metal tube.

The ties, supplied and installed by Freyssinet, are effectively the same as those used for normal pre-stressing, with a passive head, Type S13/15, embedded in the vault concrete and an active stressing head, Type C, in the anchor galleries. Between each of these heads are 13 low-relaxation multi-strand steel cables, Type T15S (cross section 150mm2), with an ultimate capacity of 345t. In line with relevant standards, a factor of safety of 1.6 on ultimate capacity of the ties has been used.

In the anchor galleries, the active head bears on a 1.3m thick reinforced concrete slab which, in turn, rests on a thick bed of strong massive sandstone. To accommodate the various tie installation angles, steel wedges were installed beneath the active head. This avoided the need for recesses or protrusions to form the bearing surface. Embedded in the slab were bearing plates with proprietary spiral confinement reinforcement fixed in the area immediately surrounding the active heads.

Slab reinforcement was supplemented locally with small diameter bars, closely spaced to resist spalling forces equivalent to 0.02 times the tie force.

Following vault excavation, each tie was installed from the vault up into the anchor galleries and the passive head and spiral reinforcement temporarily held in position between the inner and outer layers of the vault reinforcement. This ensured sufficient concrete cover to prevent ‘pull-out’ of the passive head. Modelling determined the expected loads in each of the tensioned ties at each stage of the future works. Sufficient length of cable has been left above the active head to allow for de-stressing or re-stressing to deal with the changing loads during bench excavation, and also to permit final de-stressing after cavern concreting works are completed.

After vault concreting, and before bench excavation, the ties were tensioned in a predefined two-stage sequence, up to their designed working load of 180t, using a C250 jack. After the walls are completed the sequence will be reversed for de-tensioning.


A design requirement was the installation of instrumentation to monitor the vault lining support system and ensure that the expected behaviour was followed and safety was not compromised (Saive and Parkin, 2001).

Each tie has a 250t capacity load cell, supplied by Geodata of Austria, beneath the active head which is connected to the site-wide automatic instrumentation data acquisition network. Also connected to this network are 36 vibrating wire strain gauges installed in the vault and shaft linings, and two multipoint extensometers have been installed between the vault and the anchor galleries to monitor the differential displacements. Optical targets have been placed on the inside of the vault to measure absolute displacements. The expected behaviour of each of these instruments has been determined from the modelling, and reading frequencies and alarm levels are set accordingly.

With 60% of the bench excavation completed to date, monitoring shows only minimal displacements of the vault structure and slight tension losses in the ties due to creep.

It appears that the vault is currently being supported from below by the two endwall thickenings bearing on the rock. These thickenings were not included as supports in the model, as the quality of the underlying rock was variable and the endwalls were expected to move during the subsequent excavation works. The sidewalls and endwalls of the bench, beneath the vault, are currently showing horizontal ground displacements of the order of 2cm, which is in line with those expected from the modelling.