You are on page 1of 8

World Tunnel Congress 2013 Geneva

Underground the way to the future!


G. Anagnostou & H. Ehrbar (eds)
2013 Taylor & Francis Group, London
ISBN 978-1-138-00094-0

The design of the Chillida Underground Sculpture


P. Varley(1), C. Merino(2), S. Macklin(3)
(1)

Rock Mechanics Consultant, Folkestone, UK. Formerly Scott Wilson, UK

(2)

Arup, Madrid, Spain

(3)

GHD, Melbourne, Australia. Formerly Arup, UK

ABSTRACT. The late Spanish artist Eduardo Chillida envisaged the creation of a large underground sculpture
within Tindaya Mountain, on the island of Fuerteventura in the Canary Islands. The structure enclosing the
sculptured space would consist of an entrance tunnel, a cavern of cathedral dimensions and two rectangular light
shafts. The shafts would capture the light from the sun and moon as they track across the peak; so providing the
only light in the space. The cavern would have a flat roof and sub-vertical walls, cut from the trachyte rock, with no
visible support. A series of possible solutions were tested for the design of the cavern. The eventual design
consists of a series of curved tunnels within the rock mass above the cavern roof. Cable anchors radiate from
these tunnels to create annular zones of supported rock: in effect rock arches within the rock mass. The cavern
roof will be pre-supported with fans of cable ties from these arches. The curved tunnels will be retained to
intersect any groundwater, to allow access for ground and support instrumentation, and to allow the cable ties to
be maintained and eventually replaced. The sculpture was devised by the artist in 1994. The investigation and
design of the project continued after his death in 2002. In 2008 the tender design for the construction of the
project was approved by the Spanish Chamber of Engineers and in 2011 the artists family signed a framework
agreement with the Canary Islands Government, ceding their intellectual rights, which now allows the
Government to procure the project directly. A previous paper (Macklin 2012 et al) described the geology, site
investigation and design philosophy of the project. This paper expands on the development of the support design
and the phasing of construction.

Introduction

A subterranean space is to be created within a mountain in the Canary Isles. The structure will
consist of a 13 m wide entrance tunnel, 49 m wide cavern and twin rectangular light shafts of 400 m2
and 600 m2 base area. The entrance tunnel and cavern will have flat roofs. The cavern walls and
lower shaft walls will be sub-vertical.
The structure will be excavated in igneous trachyte rock containing master joints and dykes. Some
local dental concrete is permissible where the dykes are weathered, otherwise the trachyte on the
finished internal surfaces must remain visible. External support; such as arches, are not permitted and
no rock bolts/rock anchors must be visible from within the space. Curved surfaces: such as a vaulted
roof, are not acceptable. A design life of 500 years means that the main support must remain
accessible and replaceable.
All roofs, walls and floor surfaces must be planar. Long term durability issues mean that the trachyte
has to be protected from the atmosphere.
The construction of the space must have minimal and recoverable environmental and zero
archaeological impact. General access to the mountain slopes is not permitted.

43

Geological setting

The mountain is composed of a series of inter-layered, west dipping basalt and trachyte sheets. The
trachyte is intruded by a series of thin, irregular and discontinuous basalt dykes, often having a stepwise outcrop controlled by the geometry of master joints. The dykes (and often the adjacent trachyte)
are highly fractured and hydrothermally altered. Geotechnical mapping over the cavern footprint and
digital terrain imaging were used to create a 3D model of the boreholes and dykes. This was used to
locate the dykes in the numerical modelling for the design of the excavation and support. As the dykes
are thin, they were included in the models as equivalent planes, with peak friction angle of 34, dilation
angle of 4 and cohesion of 0.05 MPa.
There are six principal joint sets (J1-J6): three highly persistent master joints (J1, J5 and J6), highly
persistent but shallow stress relief sheeting joints and less persistent minor joints (J2, J3 and J4).
The sheeting joints were not encountered in the boreholes at cavern level. Sets J1, J2 and J4 are
numerically dominant. Master joint sets J1 and J6 frequently control the distribution of the dykes. Most
of the joints have a thin coating of manganese oxide; of which some 5% is thick enough to be
considered to be infill.
Hydrofracture tests showed major and minor horizontal stresses of 0.8 to 3.0 MPa and 0.6 to 1.6 MPa
respectively at cavern roof level. These stresses are assumed to be a relic of the mountains previous
size and history. The vertical stress estimated from the weight of the overburden is 1.0 to 1.4 MPa.
The boreholes and perennial springs on the northern and western slopes suggest a seasonal
maximum water table slightly above the cavern roof and a slope reflecting the topography. The rock
-5
-7
mass permeability ranges from 1 x 10 to 5 x 10 m/s; including the fractured basalt dykes.

Design development

A 3D topographical model of the mountain slopes was created in SURFER and used to build an
elastic continuum model in FLAC3D, to try to determine an initial stress state compatible with the
current geometry of the mountain, the density of the rock mass, gravity and the measured in situ
ground stresses. This large model represented an area 1.4km by 2.0km and a height of 0.4km. Five
scenarios were considered, with stress ratios of 0.5 to 2.0 and original mountain heights of up to 3
times the current mountain height. The best fit with the measured in situ stresses was found by
eroding a thickness of ground above the mountain peak equal to the current height of the peak, with
an initial horizontal to vertical in situ stress ratio of 1. This is assumed to be analogous to the erosion
of the original Miocene volcano, remnants of which remain to the east of the mountain.
A 3D discontinuum model was made of the mountain, cavern, entrance tunnel and light shafts in
3DEC (version 4.00). This was eroded in the same way as determined from the FLAC3D model to
generate the in situ stresses. This 3DEC model was then used as the basis for the excavation and
support design.
The 3DEC model included the master joints and dykes anticipated to intersect the space from the site
investigation data. A Mohr-Coulomb model was used for the joint properties. Rigid blocks were
assumed due to the high strength of the rock in relation to the in situ stress. As the space is unlined, it
was assumed that the water table would drain to cavern floor level.
Initial models were run without support to investigate the failure mechanisms. These showed that;
apart from a single column in one light shaft, the ground was stable. Sensitivity analyses suggested
that roof stability was influenced by the cohesion along the steep angled joints and by the presence of
the low angled east-dipping (J5) joint set; which acted as release planes to the blocks bounded by the
steeper joints. The J5 joints were probably under-represented in the data, due to the difficulty in
mapping the steep western mountain slopes. The cohesive strength of the joints and dykes was
reduced to 0 and 0.05 MPa respectively and the number of J5 joints increased to a 5m spacing. This
resulted in large failures in the unsupported model; which typically started as columns in the light shaft
3
walls and tetrahedral blocks in the cavern roof (Figure 1 total volume of 10030m in the cavern roof
3
and 4522m in the walls).

44

Initial support concepts

The design proceeded in three phases, in parallel with the site investigation and increasing complexity
of the numerical analyses. The first phase considered six support and excavation concepts. Numerical
analyses of three concepts considered to be potentially viable, were made in the second phase.
Further detailed analyses, construction programming and costing of the chosen support and
excavation methodology were completed in the third phase. Design validation will be made as data is
collected during construction.

Figure 1. Ground instability in the unsupported cavern roof, cavern and light shaft walls

The six support concepts considered in Phase 1 were:


Concept C1: Rock bolts and cable anchors installed from within the cavern during the staged
excavation of the flat roof. The cable anchors would need to be replaceable and installing heavy
(+12 m long) cables in the cavern roof would be difficult. Grouting would be needed to reduce
groundwater infiltration, but could also induce locally high pressures, affecting block stability.
Concept C2A: Fans of cable anchors installed from the floor of a single horizontal tunnel, excavated
parallel to and above the cavern long axis from one of the light shafts. Multiple tunnels would be
excavated if more support was required (C2B). Alternatively (C2C) the cable anchors would radiate
downwards from points along six tunnels excavated parallel to the cavern short axis. An arch would
not be formed above the cavern roof in C1 and C2. The cables would not be anchored in stable
ground: their purpose would be to stiffen the rock mass by resisting shear and dilation along the
discontinuities, but not to otherwise resist the settlement of the stiffened plate above the roof.
Concept C3A: An arched roof with a cast in situ concrete vault and a false ceiling, suspended from the
vault on a steel and cable structure. This would allow support and drainage to be installed as the rock
mass condition dictated. Seismic loading was a problem and the concept was not consistent with the
aesthetic requirements of the space. In C3B the vault was moved up into the rock mass above the
cavern roof. This would be achieved by excavating a series of curved tunnels above the cavern roof,
from which cables would be installed to stiffen the rock above the roof. The curved tunnels would be
filled with reinforced concrete after the installation of each group of cable anchors, such that a series
of ribs would be formed within the rock mass. Alternatively, a series of parallel tunnels could be
excavated from the light shafts. These would be filled with concrete to form a single arch of
contiguous, horizontal piles. Filling the ribs or arch with concrete would make it more complicated to
replace the cable anchors, to access monitoring instruments and to intersect groundwater seepage.
Shearing along a continuous discontinuity could cut through the concrete ribs; causing the loss of the
arch. Wedge failure in the cavern walls could cause the ribs to rotate into the cavern.
Concept C4A: Horizontal cable ties on two levels: one parallel to the cavern long axis and the other
parallel to the short axis. C4B consisted of a single layer of cable ties aligned parallel to the cavern
short axis. In C4C the cables in C4B were replaced by steel tubes with 25 mm wall thickness; which
would subsequently be filled with cement grout. The pipes would extend to a little over half the cavern
width, such that they overlapped above the cavern roof. The cable ties and steel tubes would be
subjected to shear, as the main block movement would be sub-vertical. It would be impractical to drill

45

parallel holes and to install heavy cable anchors or steel tubes from small anchor galleries. Couplers
were potential weak points and would not be acceptable. Pre-stressing the horizontal cable ties would
not increase the resistance to shearing between the rock blocks as it would not be possible to pull the
blocks of rock above the roof together. Indeed the roof blocks would tend to sag, resulting in a
permanently unstable roof. Installing anchors in holes with a downwards curve, such that they
imposed an upward (support) force when tensioned, was not considered to be practical due to the
difficulty in drilling such holes and potential disturbance to the rock mass that the upward force could
cause. Holes that curved unintentionally upwards would impose a downward load; reducing the
stability of the roof.
In Concept C5 the curved tunnels in C3B were replaced with horizontal tunnels filled with concrete.
The cables in C4 would be installed in the tunnels with a downward curve and post tensioned; similar
to post tensioned concrete beams. The complexity of the build and the loss of an arching effect in the
rock mass were considered to be major limitations for C4 and C5.
In concept C6A the arched concrete ribs (C3B) were replaced with open tunnels, permanently
supported with rock bolts and reinforced shotcrete. In C6B, the permanent support included a zone of
cable anchors below the floor of curved tunnels excavated above the cavern roof. These zones would
form reinforced rock arches, which would replace the stiff concrete ribs. Longer cables would support
the ground between these rock arches and the cavern roof. The area between the light shafts would
be supported with cable anchors below further curved tunnels. The logical extension of C6B was to
install cable anchors in the walls and roofs of the curved tunnels (C6C), such that each tunnel became
the centre of an annular reinforced rock arch.
Three concepts were selected for the second phase analysis. These were considered to represent a
thick reinforced plate (C2C), a thin reinforced plate (C4A) and a reinforced rock arch (C6B).
Concept C2C initially consisted of six support installation tunnels with four arrays per tunnel of
5 x 20 m long cable anchors (four inclined at 600 and one vertical: 120 cable anchors in total). This
secured the roof but some of the cable anchors had exceptionally high loads (>200 tons). The loads
were reduced by doubling the number of anchor arrays and by adding 3 m long rock bolts in the
cavern roof, but the loads remained high. Installing the support from focal points limited the number of
anchor/joint intersections. It would be difficult to improve the distribution and to install enough arrays to
solve the excess loading problem. This was the least successful of the three models.
The 67 horizontal cable ties in C4A were located at 2 m centres, on two levels at 2 m and 3 m above
the cavern roof, and 1 m centres between the light shafts. Adding 3 m long rock bolts in the cavern
2
roof reduced the volume of unsupported blocks to a nominal 5 m , but some of the anchors had loads
in excess of 200 tons. While this concept was successful in the model, it was limited in reality.
The support in C6B consisted of pairs of 6m long cable anchors at 2 m centres, installed from
6 curved galleries above the cavern roof and 3 galleries between the light wells: 396 anchors in total.
The anchors were inclined at 700. The blocks within and above the arches were stable, but with no
rock bolts installed from the cavern roof, the blocks below the arches failed. Adding the rock bolts
reduced the volume of unstable blocks and reduced the anchor loads, but some still exceeded 200
tons.

Hybrid support solution

A final support solution (C7) was developed from C4A and C6B. This consists of a series of reinforced
rock arches above the cavern roof, formed by 7 m long cable anchors radiating from curved
installation galleries (Figures 2 and 3). The cable anchors create annular zones of reinforced rock,
which intersect to form a virtual vault. The ground between the arches and cavern roof is supported
with cable ties (cables between plates, rather than between a plate and anchorage) of up to 14 m. To
create a consistent support density of 1.5 x 2.0 m at the cavern roof, the inclination of the ties is
varied, depending on their location along the installation gallery. Any unanticipated structure that is
encountered during construction will be targeted with additional cable anchors above and below the
anchor galleries.
The cable ties forming the single arch between the light shafts are extended to the shaft walls. Subhorizontal cable ties at 2 m centres are also installed between the two light shafts below the arch to
resist the relaxation of the pillar above the cavern roof. The two rows of ties in the wall of the larger

46

shaft are inclined at +/- 3, such that they form 3 rows in the wall of the smaller shaft to maintain the
2 m spacing.

Figure 2. Cable anchor galleries above the cavern roof and between the light shafts

Figure 3. The support galleries and overlapping cable anchors above the cavern roof

In the final analyses (Figure 4), 3 m rock bolts were included in the cavern roof, 5 m rock bolts in the
cavern walls and 10 m rock bolts in the light shaft walls; all at 2 m spacing. In the actual excavation,
the longer rock bolts will be replaced with cable bolts.
The installation galleries above the cavern roof will remain open, but in the current design the gallery
between the light shafts will be back filled with concrete. It is probable that this gallery will be replaced
by replacing the 10 m cable bolts in the shaft walls with additional sub-horizontal cable ties.
In all of the models the support was installed and the excavation of the cavern, shafts and entrance
tunnel wished in place. The support was assumed to be elastic (infinite axial capacity), but in the final
model, the cable bolts and rock bolts were modelled with an elasto-plastic constitutive model, with no
shear capacity and with axial load capacities of 120 and 15 tons respectively. Axial capacity was
assumed to be lost at 1% strain. This resulted in just a single unstable block of <1 m2 in the cavern
roof. The cable anchors forming the reinforced rock arches had a maximum load of 44 tons: the
highest loads being in the centre of the roof and between the light shafts. The sub-horizontal ties
between the light shafts had a maximum load of 30 tons, with the majority <15 tons. The cable bolts in
the shaft walls had a maximum load above the cavern roof of 38 tons, with the majority <10 tons. The

47

majority of the rock bolts in the cavern roof and walls had a load of <10 tons. Some 53 of the 3140
installed rock bolts reached the yield load, though none failed in terms of the axial strain.

Figure 4. The 3DEC model for concept C7

Separate sensitivity analyses were run to test the robustness of the design. These included a
comparison of the Mohr-Coulomb and Continuously Yielding Models, sensitivity to joint orientation and
the dilation angle, the effect of a high water table and of the excavation sequence for the roof heading
and benches. This is discussed in detail in the previous paper (Macklin 2012 et al.).

Construction sequence

No construction personnel, materials or spoil will be permitted on the mountain slopes. The only road
access will be to elevation 242 m on the western side of the mountain, as above this the road gradient
would be >25%. The floor of the cavern entrance tunnel is at 282 m and there is no room to build a
working area at the portal. To overcome this, it is proposed to build a temporary steel platform at the
entrance tunnel portal. This will house the ventilation fans, pumps and workshops. Cranes will lift and
distribute the materials from the road to the platform. Rock spoil will be crushed and transported from
the platform by conveyor belt.
All excavation will be by drill and blast. A pilot tunnel will be excavated through the future cavern,
which will bifurcate to the footprint of each light shaft. This tunnel will be temporarily supported with
fibre-reinforced shotcrete and glass fibre dowels. Pilot shafts will be raised from the pilot tunnels to the
mountain slopes above (Figure 5). A minimum amount of surface excavation and support will be
undertaken at the shaft collars to allow the pilot shafts to break out. As with the site investigation, all
personnel and equipment for the surface work will be transported by helicopter. The initial excavation
and slope support will be by hand to minimise the environmental impact. The spoil will be stockpiled
and eventually mucked through the pilot shafts.
It was previously envisaged that a third branch tunnel would be excavated to the footprint of the larger
(south) light shaft to provide safe man access and ventilation. A 15 m wide trial excavation with a flat
roof would be excavated in this tunnel. It is probable that the entrance tunnel will now be excavated
full span and its flat roof used for the trial. This would also provide an additional working area.
Geotechnical data collected during the trial and from the excavation of the pilot tunnel and shafts will
be used as a first validation of the design parameters and models.

48

Figure 5. Pilot tunnel and pilot shafts

The light shafts will be excavated top down: the pilot shafts being used for mucking, personnel and
materials. Temporary covers will be built over the initial surface excavations to prevent dust and noise
pollution. These will be replaced with decking across the full shaft width. Space at the shaft collars will
be very restricted and it is therefore proposed to use small excavation and drilling equipment, which
will be housed on platforms fixed to the shaft walls when not in use (Figure 6).
Shaft excavation will stop at the level of the access tunnels to the support installation galleries and the
shaft floors will become temporary working areas for tunnel construction. The access tunnels and
installation galleries will then be excavated and supported, and the radiating cable anchors installed. A
second validation of the design will be made using the geotechnical data collected during access
tunnel and installation gallery excavation.

Figure 6. Cavern excavation

The light shafts will be extended to cavern roof level. Detailed mapping will allow the support needed
to secure the shaft/cavern roof edge to be determined. The horizontal cable ties through the shaft
pillar will be installed as the base of each shaft reaches the installation level.
The cavern roof will be excavated as a series of faces from the light shafts. Rock bolts will be installed
behind the face. The cable ties will be installed and tensioned from the installation galleries as the
cable holes are exposed in the cavern roof. On completion of the roof, the remaining cavern volume
will be excavated in benches and the walls supported.
The cavern surface will be initially under excavated to leave a thin shell of rock to be removed
mechanically just prior to the surface becoming inaccessible. The base of each cable anchor and rock
bolt hole will be enlarged by reaming. The anchor and a plate will then be installed at the top of the
reamed hole. The final rock surface will cut through the reamed hole. A disc of rock will then be
grouted into the hole to mask the presence of the support (Figure 7). It is currently envisaged that the
shell of rock will be cut off using a diamond cable; similar to that used in dimension stone quarries.

49

The cable will run on pulleys attached to the countersunk rock bolts by threaded extension bars. The
cable winches will be located in the light shafts.

Figure 7. Masking of the support

The particular micro-structure and porosity of the trachyte makes it prone to weathering. Trials have
been made of clear resins (similar to those used to preserve masonry) to determine their depth of
penetration into the trachyte. The resin will be applied to the rock surfaces and injected at low
pressure into the discontinuities to preserve the excavation surfaces.

Conclusion

An analysis and design has been undertaken for the construction of the Chillida Underground Space.
This is a unique project and has required an unusual support solution and construction methodology.
The authors believe that the project is technically viable and hope that it will eventually become a
reality.

Acknowledgements

The authors would like to acknowledge the following people: Pedro Varona and Montse Senis of
Itasca Consultores S.L who built and ran the 3D numerical models, Lorenzo Fernandez-Ordonez of
Estudio Guadiana who is the project architect and proponent of the artists concept, and Prof. Evert
Hoek for his encouragement and advice during all stages of the project.

References

Macklin S.R., Varley P.M., Varona P. and Merino C. 2012. The investigation and design for a unique architectural
space the Chillida Cavern, Mount Tindaya, Fuerteventura. Tunnelling and Underground Space Technology.
31 (2012) 9-19.

50

You might also like