The FEBEX benchmark test: case definition and comparison of modelling approaches

International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638 The FEBEX benchmark test: case definition and comparison of modelling approaches E.E. Alonso a, , J. Alcoverro a , F. Coste b , L. Malinsky b , V. Merrien-Soukatchoff c , I. Kadiri c , T. Nowak d , H. Shao d , T.S. Nguyen e , A.P.S. Selvadurai e,p , G. Armand e,p , S.R. Sobolik f , M. Itamura f , C.M. Stone f , S.W. Webb f , A. Rejeb g , M. Tijani h , Z. Maouche h , A. Kobayashi i , H. Kurikami j , A. Ito j , Y. Sugita j , M Chijimatsu k ,L.Bo ¨rgesson l , J. Hernelind m , J. Rutqvist n , C.-F. Tsang n , P. Jussila o a Department of Geotechnical Engineering and Geosciences, Universitat Polite `cnica de Catalunya, Edificio D-2, Campus Nord, UPC 08034, Barcelona, Spain b G3S-E ´cole Polytechnique, France c Laboratoire Environnement Ge ´ome ´canique et Ouvrages, E ´cole des Mines de Nancy, France d Federal Institute for Geosciences and Natural Resources, Germany e Canadian Nuclear Safety Commission, Canada f Sandia National Laboratories, USA g IRSN, France h E ´cole des Mines de Paris, France i Kyoto University, Japan j Japan Nuclear Cycle Development Institute, Japan k Hazama Corporation, Japan l Clay Technology AB, Sweden m FEM Tech AB, Sweden n Lawrence Berkeley National Laboratory, USA o Helsinki University of Technology, Finland p McGill University, Canada Accepted 2 March 2005 Available online 10 May 2005 Abstract The FEBEX (Full-scale Engineered Barriers Experiment in Crystalline Host Rock) ‘‘in situ’’ test was installed at the Grimsel Test Site underground laboratory (Switzerland) and is a near-to-real scale simulation of the Spanish reference concept of deep geological storage in crystalline host rock. A modelling exercise, aimed at predicting field behaviour, was divided in three parts. In Part A, predictions for both the total water inflow to the tunnel as well as the water pressure changes induced by the boring of the tunnel were required. In Part B, predictions for local field variables, such as temperature, relative humidity, stresses and displacements at selected points in the bentonite barrier, and global variables, such as the total input power to the heaters were required. In Part C, predictions for temperature, stresses, water pressures and displacements in selected points of the host rock were required. Ten Modelling Teams from Europe, North America and Japan were involved in the analysis of the test. Differences among approaches may be found in the constitutive models used, in the simplifications made to the balance equations and in the geometric symmetries considered. Several aspects are addressed in the paper: the basic THM physical phenomena which dominate the test response are ARTICLE IN PRESS www.elsevier.com/locate/ijrmms 1365-1609/$ - see front matter r2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijrmms.2005.03.004  Corresponding author. Tel.: +34 93 401 6866; 34 93 401 7256; fax: +34 93 401 7251. E-mail address: [email protected]s (E.E. Alonso). discussed, a comparison of different modelling results with actual measurements is presented and a discussion is given to explain the performance of the various predictions. r2005 Elsevier Ltd. All rights reserved. Keywords: Benchmark; Case history; Nuclear waste; Bentonite; Thermo-hydro-mechanical analysis; Numerical analysis 1. The FEBEX ‘‘In situ’’ test The Febex ‘‘in situ’’ test is currently in operation at the Grimsel Test Site, located in the granitic rocks of the Aare Massif in central Switzerland. In order to perform the FEBEX ‘‘in situ’’ test, it was decided to excavate a new drift. Prior to it, two pilot boreholes (FEBEX 95.001 and FEBEX 95.002) were drilled in the area. Afterwards, the FEBEX drift was excavated between these pilot boreholes. It was parallel to FEBEX 95.002 borehole. Fig. 1 shows a perspective of the FEBEX drift and associated boreholes. Borehole FEBEX 95.002, with a diameter of 86 mm and a length of 132.36 m, was equipped with packers to isolate sections in which pore water pressures could be monitored. During the period 25.09.95 to 30.10.95, the FEBEX drift was excavated using a TBM. The FEBEX drift has a diameter of 2.2870.01 m and a total length of 71.4 m. Discrete flow measurements of water flow towards the excavated FEBEX tunnel were conducted in November 1995. Bulk inflow measurements were also performed in January 1996 using a flow gauge. During the period 12.02.96 to 04.04.96, 19 in-drift boreholes (Ø 66 and Ø 146 mm) with a total length of 233 m were drilled from the test area of the FEBEX drift (tunnel meters 50–71 m). Initially, the borehole layout was planned to be strictly radial. However, the boreholes were re-oriented in order to intersect the most relevant geological features (Fig. 1). These boreholes were initially used to conduct hydraulic tests and later to monitor several variables during the test (water pressure, temperature, stresses and displacements). Extensive geological and hydraulic characterizations have been performed. Relevant documents are [1–4]. Shear zones are of considerable thickness in the area (5–20 m). At the intersection with tunnels, they display major outflows indicating their relevance as preferential flow paths. Lamprophyre dykes have also considerable dimensions (thickness in the order of several meters), although their relevance as preferential flow paths is not as important as shear zones. A common preferential flow path is the contact surface between the lamprophyre and the host rock. Fig. 2 provides a geological cross-section of the Grimsel area and shows the position of the FEBEX drift, bounded by two main shear zones. The shear zones of high transmissivity constrain regional groundwater flow and therefore, they constitute natural hydro-geological boundaries of the FEBEX test area. Hydraulic and Mechanical properties of granite and the Aare massif granite, in particular, have been reported by several authors [1,5–7].Table 1 provides a set of significant parameters of intact Aare granite essentially based on core testing. In a backanalysis of a ventilation experiment, the following expressions for the water retention curve and the relative permeability of the Grimsel granite were ARTICLE IN PRESS BOUS-85001 Tunel de acceso principal Tunel laboratorlo BOUS-85002 Galeria FEBEX FEBEX-95002 FEBEX-95001 SB13 SB12 SF13 Zona de ensayo SF23 SF12 SF22 SK2 SG2 Sl2 SB23 SB22 SJ5 SF24 SF14 SF11 SG1 Sl1 SK1 SF21 Fig. 1. Layout of FEBEX test and associated boreholes [2]. E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638612 obtained [8]: s¼1:74ðS1:68 r1Þ0:405, (1) kr¼S0:5 r½1ð1S1:68 rÞ0:5952, (2) where sis the suction (in MPa), S r is the degree of saturation and k r is the relative permeability. The vertical lithostatic stress at the level of the GTS test galleries is around 9–12 MPa. However, the horizontal stresses are substantially higher than the depth-related overburden pressure. Significant horizontal forces have been measured in the main compression direction NW–SE. Maximum and minimum horizontal stresses have been reported in the range 18/45 MPa and 15/32 MPa. More details may be found in [6] and [7]. The ‘‘in situ’’ test consists of a near to full-scale simulation of a high level waste disposal facility, following the scheme shown in Fig. 3. Two electrical heaters, of dimensions and weight equivalent to those of a nuclear waste canister were placed in a 17.4 m long test section of the drift just described. The entire space surrounding the heaters was filled with blocks of compacted bentonite in order to complete the 17.4 m length of barrier for the test section. This test zone was isolated with a concrete plug. Sensors (632) were installed to monitor the different thermo-hydro-mechanical processes that occur in both the clay barrier and the surrounding rock throughout the life of the test. A heating stage of more than 3 years was planned, followed by dismantling of the test. The present paper reports on the measurements recorded during the first 1000 days of operation. Compacted bentonite blocks were manufactured by compacting statically in appropriate moulds a crushed bentonitic rock under vertical stresses of 40–45 MPa. The average values of water content and dry density of all the manufactured blocks were 14.4% (by weight) and 1.69 g/cm 3 , respectively. Fig. 4 shows a view of the test during the process of installation. Heaters were located inside a perforated steel liner as shown in the figure. A typical cross-section of the test showing the relevant dimensions is shown in Fig. 5. Sensors located inside the barrier were arranged in the cross-sections shown in Fig. 6. Cables were carried radially to outer grooves in the bentonite and taken eventually, through the concrete plug, to the service area of the tunnel. ARTICLE IN PRESS Fig. 2. Geological cross-section of the Grimsel area, simplified from [3]. Table 1 Parameters of intact granite at the Grimsel Test Site [1,5] Parameter Central Aare granite Units Density 2660723.8 kg/m 3 Porosity 0.4–1.0 vol% Uniaxial comp. strength 169.1737.1 MPa Young’s modulus E 50 53.3711.0 GPa Poisson’s ratio 0.3770.12 — 0.3370.03 Tensile strength 9.0671.48 MPa Triaxial comp. strength (s3;s1) 5.0; 263.0729.9 MPa 10.0; 333.0720.6 20.0; 410.0763.8 Friction angle a 33 1 p-wave vel. (specimen) 31117278 m/s p-wave vel. (whole rock) 56007100 m/s Therm. conductivity (wet) b 3.3470.35 W/m K Therm. conductivity (dry) b 2.5870.19 W/m K Permeability 5 10 12 (10 MPa) m/s 3.5–4.5 10 12 (5–15 MPa) 510 12 (5–30 MPa) Coef. linear thermal expansion (5–12) 10 6 K 1 Coef. vol. thermal expansion (20–30) 10 6 K 1 Specific heat 800–1250 J/kg K a fractures; b the indicated values for dry and wet granite have been swapped if compared with the original reference, accepting that the thermal conductivity is larger for the wet material. E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638 613 A total volume of gaps of 5.53% of the emplacement volume was measured. Therefore, the average dry density of the barrier was established as 1.60 g/cm 3 . Typically, the dry density varied by 70.25 g/cm 3 among the different slices of assembled blocks. Heating started on February 27, 1997 a date identified as ‘‘day 0’’ on the time scale. The sequence of initiation was as follows: During an initial period of 20 days a constant power of 1200 W per heater was applied. Over the next 33 days the power was increased to 2000 W per heater and it was maintained constant to approximate the temperature of 100 1C desired at the surface of the steel liner. Finally, on 21st April 1997 (day 53) the system was switched to the constant temperature control mode. The necessary electric power was adjusted to maintain 100 1C at the hottest control point located at the canister–bentonite contact. 1.1. The FEBEX bentonite The properties of FEBEX bentonite have been extensively investigated over the past 8 years. Several recent publications describe the thermo-hydro-mechan- ical properties of the compacted material [9–17]. Extensive information is also available for the bentonite ‘‘S-2’’, a material of essentially the same origin (a quarry in Serrata de Nı´jar, Almerı´a, Spain) as the FEBEX one. A summary of relevant properties, taken from some of the mentioned references, is given here. The conditioning of the bentonite in the quarry, and later in the manufacturing factory, was strictly mechanical (homogenization, rock fragment removal, drying, crumbling of clods, and sieving) and aimed at obtaining a granulated material with the specified characteristics of grain-size distribution and water content: Fraction of particles of more than 5 mm less than 5%, and fraction of particles smaller than 74 mm greater than 85%; water content, after conditioning, between 12.5% and 15.5% in weight. Table 2 provides a set of identification parameters as determined by two laboratories: CIEMAT and UPC. Swelling pressure tests were performed using conventional oedometers on samples saturated with distilled water. A regression curve was developed for the swelling ARTICLE IN PRESS Heater (diameter 0.9) Bentonite blocks Steel liner Granite Heaters Bentonite barrier Concrete plug Service zone, control and data acquisition system 4.54 1.004.54 4.34 17.4 2.7 70.4 (Dimensions in meters) 2.28 Principal access tunnel to KWO Granite Fig. 3. General scheme of the FEBEX ‘‘in situ’’ test. Fig. 4. A view of FEBEX test construction. E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638614 pressure of the FEBEX bentonite, as shown on Fig. 7 (solid line), and it is expressed by the equation Ps¼expð6:77rd9:07Þ, (3) where P s is the swelling pressure in MPa and r d is the dry density in g/cm 3 . A series of swelling under load tests, performed in oedometer tests, provided additional information on the expansivity and mechanical behaviour of the material. The following equation provides a relationship between 1D swelling strain upon saturation, , applied vertical confining stress, s(MPa) and dry density r d (g/cm 3 ): ¼46:919:4 log sþ36:6rd. (4) ARTICLE IN PRESS 0.900 0.940 2.280 0.970 1.125 1.105 0.640 O' O'' O''' (Dimensions in meters) O': Center of the drift O'': Center of the steel liner O''': Center of the heater Center of the steel liner Center of the heater 0.030 Z Y Fig. 5. Typical cross-section of the clay barrier. Fig. 6. Arrangement of instrumented sections. Table 2 Identification properties of FEBEX bentonite Property CIEMAT UPC-DIT Water content in equilibrium with the air in the laboratory, in % 13.771.3 13.371.3 Liquid limit, in % 102749371 Plastic limit, in % 53734772 Plasticity index 49744672 Specific weight 2.7070.04 — Grain-size distribution, in % Fraction less than 74 mm, in % 927187 Fraction less than 2 mm, in % 687245 Specific surface, in m 2 /g Total 725747 — External, BET 3273— E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638 615 Many determinations of saturated hydraulic conductivity are available. Some of them are collected in Fig. 8. Infiltration tests were also performed to determine the relative permeability. It was found that the relationship kr¼Sn r, (5) where ‘‘n’’ is a parameter determined to fit experimental data and S r is the degree of saturation. A reference value for nis 3 although a significant variability was found. Water retention curves were determined under free and constant volume conditions. The latter are more precise. Curves relating suction and degree of saturation for wetting-drying paths at different constant dry densities are shown in Fig. 9. It was found that the concept of a constant intrinsic permeability does not hold, even approximately, for an expansive material such as the FEBEX bentonite. Fig. 10 shows values of ‘‘intrinsic’’ permeability determined in gas and water flow experiments. Extremely large differences were found, due to the change in microstructure of the compacted material when its water content changes. Thermal properties are also available. Specific heat has been determined for bentonite S-2. The relationship between specific heat and temperature fits the following equation, in a range of temperatures ARTICLE IN PRESS 0.0 5.0 10.0 15.0 20.0 Swelling pressure (MPa) Dr y densit y ( g /cm3) 1.10 1.30 1.50 1.70 1.90 Fig. 7. Swelling pressure of FEBEX bentonite (square dots). Crosses correspond to the ‘‘S-2’’ bentonite [9]. Dr y densit y ( g /cm3) 1.10 1.30 1.50 1.70 1.90 Hydraulic conductivity (m/s) 1.0 x 10-15 1.0 x 10-14 1.0 x 10-13 1.0 x 10-12 1.0 x 10-11 Distilled water Perpendicular to compaction Parallel to compaction Granitic water Saline water Fig. 8. Saturated hydraulic conductivity as a function of dry density of FEBEX bentonite [9]. 0.30.2 Degree of saturation, Sr Suction (MPa) 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1 10 100 1000 0.1 0.01 Dry density g/cm3 1.75-1.70 1.59-1.58 Fig. 9. Water retention characteristics of FEBEX bentonite. Bold symbols correspond to wetting paths [14]. E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638616 between 45 and 150 1C: cs¼1:38 Tþ732:5, (6) where c s is the specific heat, in J/kg1C, and Tis the temperature, in 1C. Measured thermal conductivities are given in Fig. 11, as a function of degree of saturation. Measured thermal expansions for heating and cooling paths are given in Fig. 12. More sophisticated suction controlled tests which explore the behaviour of the bentonite under specific stress and suction paths are described in the references given above. Tests on small scale cells involving simultaneous hydration and heating were also performed. They are boundary value problems and their analysis may provide a refined evaluation of constitutive parameters. Some of the research groups participating in the FEBEX benchmark test have used this information to their advantage. Their analyses are published elsewhere. 2. The benchmark The benchmark was divided into three parts, described as follows in general terms: Part A: Hydro-mechanical modelling of the rock. Based on the available geological, hydraulic and mechanical characterizations of the site as well as on results of hydraulic tests performed in boreholes, a hydro-mechanical model for the zone around the FEBEX tunnel was to be prepared. Using this model, changes in water pressure induced by the boring of the FEBEX tunnel in the near vicinity, as well as the total water flow rate to the excavated tunnel were required. Part B: Thermo-hydro-mechanical analysis of the bentonite behaviour. Based on the characterization of the bentonite and on the details of the process of test installation, a thermohydro-mechanical model for the bentonite barrier and the heaters was to be prepared. Using this model, the thermo-hydro-mechanical response of the bentonite barrier as a result of the heat released by the heaters and the hydration from the host rock was required. Local field variables such as temperature, relative humidity, stresses and displacements, as well as global variables such as total input power to the heaters were required. Part C: Thermo-hydro-mechanical analysis of the rock. Based on the characterization of the rock massif and on the details of the process of test installation and performance, the rock response in the immediate vicinity of the buffer was required. The rock is now subjected to the heat released by heaters and to swelling pressures resulting from bentonite hydration. The initial hydrological regime (Part A) is also modified by the presence of the impervious barrier. Temperature, stresses, water pressures and displacements in selected points of the rock were required. A maximum number of 10 Modelling Teams have participated in the different benchmark activities related to the FEBEX test. Their names, codes and symbols used in the presentation of results are given in Table 3. Details on the modelling approaches used by the different participants in the Benchmark are described in companion papers in this issue. The performance of the ARTICLE IN PRESS 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 10-22 10-20 10-18 10-16 10-14 10-12 Intrinsic permeability (m2) y = 5.103 · 10-12 x4.3 Gas flow y = 1.107 · 10-23 e7.744x Water flow Accessible void ratio Fig. 10. Intrinsic permeability of the compacted clay obtained from saturated water flow and from unsaturated gas flow tests. In gas flow tests, the accessible void ratio indicates the ratio between gas accessible pore volume and particle volume (e(1S r )) [10]. 0.00 0.4 0.6 0.8 1.0 1.2 1.4 0.25 0.50 0.75 1.00 Degree of saturation, Sr Thermal conductivity (W/m·K) FEBEX bentonite values FEBEX correlation Bentonite S-2 values Bentonite S-2 correlation Fig. 11. Thermal conductivity as a function of degree of saturation [14]. E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638 617 test and a comparison of calculated and most significant measured variables are described in the next three sections. The Benchmark was conceived as a blind prediction exercise although additional calculations and analyses were made once the field data was released. The graphs presented include the final calculations submitted ARTICLE IN PRESS Heating path First heating 20 1.0 x 10-3 8.0 x 10-4 6.0 x 10-4 4.0 x 10-4 2.0 x 10-4 0.0 x10+0 30 40 Temperature (°C) ∆ vertical strain /∆ temperature 50 60 70 Degree of saturation 56-64 % 81-83 % 91-98 % Cooling paths First heating paths Subsequent heating paths Cooling path Fig. 12. Linear thermal expansion as a function of temperature [14]. Table 3 Codes, symbols and colours assigned to participants and co-ordinator Funding Organisation Modelling Team/Coordinator Code Symbol Color National Radioactive Waste Management Agency (ANDRA) Laboratory ‘‘Sols, Solides, Structures’’ Grenoble. E ´cole Polytechnique ANG Red National Radioactive Waste Management Agency (ANDRA) Laboratoire Environnement, Ge ´ome ´canique et Ouvrages. Ecole des Mines de Nancy ANN Red Federal Institute for Geosciences and Natural Resources (BGR) Federal Institute for Geosciences and Natural Resources BGR Green Canadian Nuclear Safety Commission (CNSC) Canadian Nuclear Safety Commission McGill University CNS Green U.S. Department of Energy (DOE) Sandia National Laboratories DOE Blue Institute for Radiological Protection and Nuclear Safety (IRSN) Institute for Radiological Protection and Nuclear Safety Ecole des Mines de Paris IPS Blue Japan Nuclear Cycle Development Institute (JNC) Hazama Corporation Kyoto University Japan Nuclear Cycle Development Institute JNC Brown Swedish Nuclear Fuel and Waste Management Co (SKB) Clay Technology AB FEM-Tech AB SKB Black Swedish Nuclear Power Inspectorate (SKI) Lawrence Berkeley National Laboratory SKI Black Radiation and Nuclear Safety Authority of Finland (STUK) Helsinki University of Technology STU Orange Empresa Nacional de Residuos Radiactivos SA (ENRESA) Technical University of Catalonia (Coordinator) UPC Orange E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638618 by different teams. Reference is made, when appropriate, to a true ‘‘blind’’ prediction. Part A: Hydro-mechanical modelling of the rock 3. Introduction Some of the observations made during the excavation of the tunnel and immediately afterwards provided an interesting large scale experiment of a hydro-mechanical nature. Two types of measurements have been selected to develop the modelling exercise: the actual water inflow rates into the FEBEX tunnel and the water pressure response in the vicinity of the tunnel outer perimeter as a consequence of the tunnel excavation by the TBM. Flow measurements into the open tunnel provided an integrated variable, controlled by the problem geometry, rock mass fracture pattern, fracture anisotropy, rock permeability and boundary conditions. The hydrogeological characterization carried out [3,4] reveals that the rock mass surrounding the FEBEX tunnel has a rather involved distribution of cracks. This zone was explored by a number of instrumented boreholes, in which hydro-geological tests were carried out. The surface of the test zone of the FEBEX tunnel was carefully mapped. A hydro-geological model, in which several important fractures were specifically considered, whereas the rest were included in an equivalent continuum, has been described in [3,4]. The flow through fractures was found to be of the same order of magnitude as the flow through the rock matrix (Fig. 14). The second part of the exercise asked for a prediction of the transient changes in water pressure recorded at two borehole intervals in the close proximity of the advancing tunnel. Fig. 13 is a Plan view of the Febex tunnel and the Borehole FEBEX 95.002 where observations were made in the intervals P3 and P4. As the excavation of the FEBEX tunnel was taking place, a consistent relation between water pressure variations and TBM activity was observed in some packed-off segments of borehole FEBEX 95.002. They are shown in Fig. 15 where measured pore pressures correspond to the acronym ‘‘UPC’’. During periods of TBM activity water pressure increased, whereas in periods of inactivity water pressure decreased. The rate of the tunnel drilling and the permeability determine both the rate of increase and decrease of pore water and, consequently, the magnitude of the pressure peak. Whereas in one interval (P4) a marked pore water pressure variation was recorded, in a neighbouring interval (P3) this variation was much smaller, a fact attributed to the presence of a pervious fracture connecting the measuring interval and the tunnel. Only the behaviour in P4 and associated predictions have been included in Fig. 15. 4. Modelling approaches for Part A In the zone surrounding the FEBEX tunnel it is possible to identify macroscale features at a scale comparable to the scale of the problem, such as lamprophyre dykes and shear zones, and microscale features at a scale smaller than the scale of the problem, such as secondary fractures and microfractures. ARTICLE IN PRESS 159315 159320 159325 159330 159335 159340 667410 667415 667420 667425 667430 667435 667440 667445 667450 P4 P3 FEBEX 95.002 TEST ZONE Fig. 13. Plan view of the test zone of the FEBEX drift and the borehole FEBEX 95.002, showing the intervals P3 and P4. Grimsel Test Site coordinates (in m) are used (north is parallel to the y-axis). E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638 619 is indicated as ‘‘THM’’. Two of the Modelling Teams did not introduce mechanical effects (TH). Finally, two teams used two separate programs (thermo-hydraulic and hydro-mechanic, respectively), which were linked in the sense indicated: mechanical effects were derived as a result of thermal and hydraulic changes. The reverse effect was not introduced in those cases. Gas flow was considered by one team. All of them modelled liquid flow and permeability was made saturation dependent. Vapour flow and phase change (liq2gas) was considered by a subset of teams. Heat conduction was also a common feature of all models and the thermal conductivity in the buffer was in some cases made dependent on the bentonite degree of saturation. Model calibrations varied very significantly among participating teams. The best approaches involved the backanalysis of some of the suction controlled tests made on bentonite behaviour. 9. Comparison of field data with model calculations for Part B Fig. 17 shows the evolution of measured input electrical power on heater 1 and the set of predictions. Most of calculations reproduce the progressive decrease of required power but only in one case (SKB) the reverse trend experienced after some time is well reproduced. The variation of relative humidity with radial distance in Section E1 is shown, in Fig. 18a, for t¼1000 days. Measured data of four radial directions (two vertical, two horizontal) are lumped together in the figure. The ARTICLE IN PRESS 0 10 20 30 40 50 60 70 80 90 100 0.40 0.50 0.60 0.70 0.80 0.90 1.00 1.10 1.20 Radial distance (m) Relative Humidity (%) BGR RD1 BGR RD2 BGR RD3 BGR RD4 CNS RD1 CNS RD2 CNS RD3 CNS RD4 DOE RD1 DOE RD2 DOE RD3 DOE RD4 IPS RD1 IPS RD2 IPS RD3 IPS RD4 SKB RD1 SKB RD2 SKB RD3 SKB RD4 SKI RD1 SKI RD2 SKI RD3 SKI RD4 STU RD1 STU RD2 STU RD3 STU RD4 UPC RD1 UPC RD2 UPC RD3 UPC RD4 0 10 20 30 40 50 60 70 80 90 100 0.40 0.50 0.60 0.70 0.80 0.90 1.00 1.10 1.20 Relative Humidity (%) BGR RD1 BGR RD2 BGR RD3 BGR RD4 CNS RD1 CNS RD2 CNS RD3 CNS RD4 DOE RD1 DOE RD2 DOE RD3 DOE RD4 SKB RD1 SKB RD2 SKB RD3 SKB RD4 SKI RD1 SKI RD2 SKI RD3 SKI RD4 STU RD1 STU RD2 STU RD3 STU RD4 UPC RD1 UPC RD2 UPC RD3 UPC RD4 Radial distance (m) (a) (b) Fig. 18. (a) Variation of relative humidity with radial distance in Section E1 for t¼1000 days; (b) Variation of relative humidity with radial distance in Section H for t¼1000 days. E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638626 measured data does not show a clear preferential direction of hydration. The test response seems to have a cylindrical symmetry. This is an interesting observation since a lamprophyre dyke, which provided a nonuniform inflow of water into the tunnel, crossed Section E1. Some of the predictions capture well the distribution of Relative Humidity. Others have difficulties to handle the change of phase of water at the hot areas and predictions are far from reality. In one case (SKI) the 3D analysis performed leads to small variations of relative humidity with the different radial orientations, a result that is partially supported by the experimental evidence. A similar plot is given in Fig. 18b for Section H, centered between the two heaters and, therefore, not directly affected by the heat released by them. This section is obviously wetter. Predictions require now a 2D axisymmetric (or 3D) geometry. Two of the models predict quite accurately the experimental data. Evolution plots of relative humidity at some selected points (H, C and G) are given in Figs. 19 (Section E1) and 20 (Section H). Measurements exhibit a very distinctive pattern. The point close to the granite (G) becomes rapidly saturated. The point in the bentonite close to the heater (H) experiences an intense desiccation. The initial RH decreases to values as low as 10% (this is equivalent to a very high suction: more than 300 MPa) 200–300days after the beginning of heating. Later, the humidity increases at a slow, progressive rate. The point in the center (C) displays a more complex behavior: it becomes initially wetter, reaches a peak value of RH, but partially dries immediately afterwards. These transient phenomena take place during the first 80 ARTICLE IN PRESS 0 10 20 30 40 50 60 70 80 90 100 0 200 400 600 800 1000 Relative Humidity (%) BGR E1H BGR E1C BGR E1G CNS E1H CNS E1C CNS E1G DOE E1H DOE E1C DOE E1G IPS E1H IPS E1C IPS E1G UPC E1H UPC E1C UPC E1G 0 10 20 30 40 50 60 70 80 90 100 0 200 400 600 800 1000 Relative Humidity (%) SKB E1H SKB E1C SKB E1G SKI E1H SKI E1C SKI E1G STU E1H STU E1C STU E1G UPC E1H UPC E1C UPC E1G Time (day) Time (day) (a) (b) Fig. 19. Measured and predicted variation of RH at three radial points in Section E1 (H: heater; C: center; G: granite). E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638 627 ARTICLE IN PRESS 0 10 20 30 40 50 60 70 80 90 100 0 200 400 600 800 1000 Time (day) Relative Humidity (%) BGR HH BGR HC BGR HG CNS HH CNS HC CNS HG DOE HH DOE HC DOE HG UPC HH UPC HC UPC HG 0 10 20 30 40 50 60 70 80 90 100 0 200 400 600 800 1000 Relative Humidity (%) SKB HH SKB HC SKB HG SKI HH SKI HC SKI HG STU HH STU HC STU HG UPC HH UPC HC UPC HG Time (day) (a) (b) Fig. 20. Measured and predicted variation of RH at three radial points in Section H (H: heater; C: center; G: granite). 0 10 20 30 40 50 60 70 80 90 100 0.40 0.50 0.60 0.70 0.80 0.90 1.00 1.10 1.20 Radial distance (m) Temperature (°C) BGR 90 BGR 1000 CNS 90 CNS 1000 DOE 90 DOE 1000 IPS 90 IPS 1000 JNC 90 JNC 1000 SKB 90 SKB 1000 SKI 90 SKI 1000 STU 90 STU 1000 UPC 90 UPC 1000 Fig. 21. Measured and calculated temperature variation across the barrier, at Section D1, for two times: 90 and 1000 days. E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638628 days of heating. Later, a progressive hydration is recorded. Some of the model results reproduced in Fig. 19 capture correctly this phenomenon. The difficulties found by others are associated with incomplete formulations of the phase change of water and the transportation of water vapour. One thousand days after the beginning of heating the barrier is still far from saturation. ARTICLE IN PRESS -1 0 1 2 3 4 5 0 200 400 600 800 1000 Stress (MPa) CNS E2H1 CNS E2G1 CNS E2G2 DOE E2H1 DOE E2G1 DOE E2G2 IPS E2H1 IPS E2G2 UPC E2H1 UPC E2G1 UPC E2G2 -1 0 1 2 3 4 5 0 200 400 600 800 1000 Stress (MPa) SKB E2H1 SKB E2G1 SKI E2H1 SKI E2G1 SKI E2G2 UPC E2H1 UPC E2G1 UPC E2G2 Time (day) Time (day) (a) (b) Fig. 22. Measured and predicted variation of RH at three points in Section E2 (H: heater; C: center; G1 and G2: granite). Table 9 Main features of the models used for Part B Team Couplings Deform. Wat. flow Vap. flow Liq 2gas Gas flow Heat flow ANG — — — — — — — BGR TH-HM*  CNS THM     DOE TH-TM     IPS THM   JNC TH   SKB THM     SKI THM     STU TH   In the BGR model, M* indicates that the mechanical part reduces to consider a dependency on the swelling pressure. E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638 629 A more complex initial transient is measured at the inner point in Section H (Fig. 20). The central and outer points exhibit a progressive hydration from the early stages. One of the model predictions in Fig. 20 (SKI) is remarkably accurate. Only one figure regarding temperature distribution across the barrier is included here (Fig. 21). At the particular location shown (Section D1 at the edge of Heater 1), temperature decreases from a value above 80 1C at the heater to 30–35 1C at the granite contact. The lower value (30 1C) corresponds to 90 days of heating and the upper one (35 1C) to 1000 days of heating. All the predictions follow the measured decay with reasonable accuracy. However, when comparing extreme predictions for a common time (1000 days) a maximum difference of 27 1C is found. Measured radial stresses are given in Fig. 22a, b. Two of the selected measurement cells (G1 and G2) are at the contact between the bentonite blocks and the granite wall. The third one (H1) is closer to the heater. The outer cells develop consistently stronger radial stresses than the inner one during the considered time span. Stress build-up is faster in one of the G cells but the recorded radial stress in both cells is similar and close to 3.5 MPa at the end of the measuring period. The inner cell, however, shows a slower and steady development of the confining stress. The recorded value at the end of the period has not reached 1 MPa. It is clear that the buffer is still far from reaching the full swelling potential of the bentonite (about 6 MPa). One of the modelers (SKB in Fig. 22b) made a very good prediction although the predicted rate of increase of stress at the end of the period is slower than the actual value. Some models (IPS, DOE, SKI) predict a similar stress development in the inner and outer positions considered. Most of them underpredict, to a different degree, the actual measurements, especially at later times. 10. Discussion—Part B The heating power is a global variable of the barrier performance dominated by the heat conductivity of buffer and rock. The initial strong drying of the inner part of the bentonite leads to a reduction of the heat conduction coefficient and, therefore, a reduction in power is required to maintain a constant temperature at the heater–buffer contact. As the buffer becomes progressively saturated in the mid and long term, the average heat conductivity of the bentonite increases and the required heating power should increase. Models which do not include a dependence of lTon S r cannot reproduce this effect. Surprisingly, only one of the predictions (SKB) matched correctly the actual behaviour. The distribution and evolution of relative humidity offers a very good image of relevant thermal and hydraulic phenomena taking place in the barrier. Early stages of heating are dominated by phase changes in the inner part of the buffer. Evaporation close to the heater and condensation of the outwardly migrating vapour flow explain the observed changes in relative humidity. Models which do not include vapour generation and flow (BGR, IPS, JNC) cannot reproduce the measurements. One of the relevant features of the measurements is the transient drying–wetting–drying and final wetting of points located in the inner rings of the buffer. This behavior is explained by the role of water vapour and phase changes in water transfer inside the barrier. Water evaporates in areas of the bentonite close to the heater. Hot vapour at high concentration migrates outward and becomes eventually cooler thus leading to condensation. This explains the peak observed in RH. This is an expanding process. As the temperature increases inside the barrier, this phenomenon moves in a radial direction and the inner points of the barrier become drier again. At higher times the water inflow (in liquid phase) from the outer zones eventually dominates the barrier hydration. Models that include a correct TH coupling, as outlined above, are capable of reproducing the observed behaviour. This is specially the case for some predictions affecting cross-section through the center of heaters. The recorded behaviour of some points in the central section is more complex and it was not reproduced accurately by any model. Preferential vapour flow paths along open joints, which may have played a role in this case, were not considered in any of the models used. Relative humidity changes are weakly related to stress changes, provided the changes in void ratio (or joint openings among bentonite blocks) are not significant. It is expected that the joints in the inner part of the buffer remain open during the simulation period considered in the analysis. Joints of the outer buffer will certainly close and this change may affect, to a certain extent, the convective flow of water. However, in global terms, the porosity of the buffer remains constant due to the rigid confinement offered by the rigid granite. Therefore, changes in intrinsic permeability coefficients, induced by stress variations, are expected to be low. Accordingly, the correct prediction of relative humidity changes does not require a mechanical coupling. This is the case of STU which is able to reproduce in a satisfactory manner relative humidity changes in the buffer. Another issue refers to the dimensionality of the problem. One-dimensional coupled modelling (IPS) is capable of making good predictions in cross-sections centered in the heaters. Axial symmetry is a good simplification in this case. Obviously, the central section ARTICLE IN PRESS E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638630 of the test requires 2D or 3D analysis. The effect of discrete conducting features of the rock (shear zones, lamprophyre–granite contacts) on buffer hydration was, according to measurements of RH distributions along different axis of a given cross-section, of minor relevance. It is concluded that the low permeability of the saturated bentonite, (1 10 21 m 2 ) compared with the granite matrix permeability (5 10 19 m 2 ), leads to a fairly homogeneous hydration of the barrier. In other words, bentonite permeability controls its own rate of hydration and the rock matrix is capable of providing all the required flow of water. One of the 3D analyses (SKI) included in a simplified manner the actual position of the conductive rock features. The results obtained indicated a minor effect of the concentrated water inflows. Therefore, 2D axisymmetric models provide good results in this case. Temperature distributions and evolutions offer limited information to discriminate the physical phenomena taking place in the buffer. Somewhat unexpectedly, temperature predictions varied significantly among modelling teams. The actual recorded values tend to remain at the average of all calculations. Stress predictions require fully coupled THM models. Recorded radial stresses display the development of the swelling potential of the bentonite as it becomes wetter. However, the inner zones experience a strong drying which induces the shrinkage of blocks and favours the concentration of stresses at some contacts. Radial stresses recorded in the outer part of the buffer have shown a continuous increase during the observation period and their value is in excess of 3 MPa at the end of the first 1000 days of testing. However, the cell located in the inner part of the buffer remained essentially unloaded during the same period. Some models reproduce this trend (SKB, CNS, DOE) whereas others (SKI, IPS) predict a rather homogeneous development of stresses within the buffer. In general, models indicate a faster initial stress increase than the actually recorded evolutions, probably due to joint closure. At later dates, however, measured stresses maintain a constant rate of increase, unlike most models, which show a decreasing rate of increase. Model behaviour is a typical one for stress development associated with an underlying swelling mechanism induced by transient flow. The kinetics of swelling stress development in the bentonite is probably a more complex phenomenon. Despite all the difficulties mentioned, some of the fully coupled 2D and 3D analyses, and notably those presented by SKI and SKB, have achieved a good overall representation of the buffer response against the combined action of internal heating and external wetting. Part C: Thermo-hydro-mechanical analysis of the rock 11. Features of the experimental data Several instruments were located, at increasing depth, in the 19 auxiliary boreholes, perforated in radial directions from the FEBEX tunnel (Fig. 1). An enlarged view of the borings is given in Fig. 23. The length of these radial boreholes does not exceed 15 m. Typically, readings are available in three or four positions: close to the tunnel wall (say at 1–3 m distance from the origin of the boring, one or two intermediate distances and a distant position (13–14 m deep). The Cartesian coordinate system shown in Fig. 23 was used to orientate and locate the measuring devices. Data is available on the following variables: temperature, water pressure in rock, water pressure in packers, stress state and radial displacements. Water pressure was measured in some boring intervals (a few meters long) separated by packers. The rubber packers were also filled with water and their pressure was maintained and measured externally. Water pressure gauges are located in the measuring area of the FEBEX tunnel. They are connected with the measuring intervals by means of steel tubing. Three normal components of stress are measured by means of total pressure cells. Four sets of 3 cells were prepared and grouted ‘‘in situ’’ in boreholes SG1 and SG2. Each cell had five sensors oriented in different directions and fixed on a common support 2 m long. Each one of the sensors was a circular steel flat cell. An interpretation of the five readings provides the normal stress components in three directions: radial (with respect to the tunnel axis) (sr), axial (along the direction of tunnel axis) (sx) and circumferential (sy). Once the sets were in position the borehole was filled with a ARTICLE IN PRESS Fig. 23. Auxiliary instrumentation boreholes around FEBEX test section. E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638 631 slightly expansive mortar. Once the mortar was cured, cells were pressurized against the mass of surrounding mortar to guarantee a good initial contact. This procedure provides increments of stress over the stress state prevailing at the moment of cell installation. It will record therefore stress increments due to temperature effects, modification of pore water pressures and swelling of the bentonite. Radial displacements were measured by means of borehole extensometers installed in borings SI1 and SI2. They were located close to the position of the stress cells. Each extensometer consists of graphite rods with independent anchoring points located at a depth of 1.0, 3.0 and 7.0 m into the borehole. Plots showing the measured evolutions of temperature, pore water pressure, total stress and radial displacements are given in Figs. 24–28 (‘‘UPC’’ acronym). Temperature (Fig. 24) exhibits a monotonously increasing pattern. Temperature increases faster during the first two months after which they slowly approach asymptotic values. Measured pore water pressures near the tunnel wall increase slowly in a monotonous way from a value close to 0.5 MPa to a value around 0.7 MPa. No marked transients in pore water pressures were observed (Fig. 25). It is interesting at this point to observe the data of the packer pressures. The evolution of water pressure in the first packer located in borehole SF21 at a radial distance of 1.86 m is shown in Fig. 26. The initially recorded water pressure is related to the pressure applied to the packer to seal the borehole. Afterwards, the circuit is closed to maintain constant volume conditions. The substantial increase in pressure observed in Fig. 26 (maximum values close to 5 MPa are recorded) is most probably due to the temperatureinduced dilation of the water inside the packer system. Presumably, the deformation of the packer and measuring system and/or minor leaks led to a progressive decay afterwards. The relevant point is that this type of behaviour was not observed in the water pressure measuring intervals in direct connection with the granitic rock. Apparently, excess pore water pressures dissipate immediately after they are induced. In addition, recorded water pressures are average values over the measuring interval. This effect contributes also to reduce the excess pore water pressures. The different conditions of the packer system (a ‘‘high’’ porosity and a rigid confinement) explain the completely different behaviour recorded. The evolutions of all the three measured total stress components (Fig. 27) are characterised by an initial fast increase until a peak is reached at about 100 days, followed by a slow and slight decrease and an even slower increase, without, however, approaching any asymptotic value during the period of 1000 days. Measured total stresses decrease significantly with the distance. Measured maximum radial displacements were in the range of about 100 microns (Fig. 28). Their evolutions exhibit an initial fast increase followed by a very small subsequent increase. The magnitude of these displacements and their evolutions cast doubts on their accuracy. 12. Comparison of field data with model calculations for Part C The model developed for Part B was used for Part C by most of Modelling Teams. A summary of the modelling approaches is given in Table 10. The inclusion of thermal dilation properties of rock and water is now necessary to tackle Part C. A full THM coupling is also needed to make meaningful predictions. ARTICLE IN PRESS 0 5 10 15 20 25 30 35 40 45 50 0 200 400 600 800 1000 Time (day) Temperature (˚C) ANG BGR CNS DOE IPS JNC SKB SKI UPC Fig. 24. Measured evolutions of pressures at borehole SF21. Water pressure at the packed off interval closest to the tunnel (centre at r¼3:03 m) and inflation pressure of the packer closest to the tunnel (centre at r¼1:86 m). E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638632 Measured and computed evolutions of temperature in Boring SF 21, at a radial distance of 1.20 m are shown in Fig. 24. Most predictions are quite accurate. Two of them, however, depart in a significant way. Prediction of temperature depends essentially on a correct estimation of thermal conduction coefficients and boundary conditions. Water pressures are represented in Fig. 25a and b. Both figures show evolutions of water pressures recorded in Borehole SF 21 at radial distances of 3 m (Fig. 25a) and 13.6 m (Fig. 25b). These radial distances locate the centre point of the measuring intervals, which have lengths of 2.34 and 2.82 m, respectively. Some of the predictions made are very good and reproduce the absolute measured values and the increasing trend with time (IPS, DOE, and SKI). Measured water pressures are stable and close to 1 MPa at the more distant point (Fig. 25b). Recorded normal stresses in borehole SG2 (sx,sy,sr), are reproduced, together with predictions of several teams in Fig. 27a–c. Measuring cells were located at ARTICLE IN PRESS -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 0 200 400 600 800 1000 Water pressure (MPa) ANG 1 BGR 1 CNS 1 DOE 1 IPS 1 JNC 1 SKB 1 SKI 1 UPC 1 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 0 200 400 600 800 1000 Water pressure (MPa) ANG 3 BGR 3 CNS 3 DOE 3 IPS 3 JNC 3 SKB 3 SKI 3 UPC 3 Time (day) Time (day) (a) (b) Fig. 25. Measured and predicted evolution of temperature in borehole SF 21 at a radial distance of 1.20 m. 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 0 200 400 600 800 1000 Water pressure (MPa) UPC water pressure UPC packer pressure Time (day) Fig. 26. Measured and predicted evolution of water pressures in borehole SF 21 at radial distances of 3.03 and 13.6 m, respectively. E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638 633 radial distances of 2.59, 2.97 and 3.25 m, respectively. Recorded values at a larger radial distance (r¼7.10–7.80 m), not shown here, were significantly lower. Interestingly, a peak is recorded at early stages in all the three stress components, followed by a transient decay and a progressive increase at later dates. The stress cells measure total stress and the observed behavior is consistent with the expected change in pore water ARTICLE IN PRESS -2 0 2 4 6 8 10 0 200 400 600 800 1000 Total stress (MPa) ANG x CNS x DOE x IPS x SKB x SKI x UPC x -2 0 2 4 6 8 10 0 200 400 600 800 1000 Total stress (MPa) ANG theta BGR theta CNS theta DOE theta IPS theta SKB theta SKI theta UPC theta -2 0 2 4 6 8 10 0 200 400 600 800 1000 Total stress (MPa) ANG r BGR r CNS r DOE r IPS r SKB r SKI r UPC r Time (da y ) Time (day) Time (day) (a) (b) (c) Fig. 27. Evolution of normal stresses (sx , sy,sr), recorded in boring SG2 and predictions of four research teams. E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638634 pressures, explained before. The stress predictions shown in the figures do not reproduce the measured transient, however. Some of the calculated values (especially for SKI) are quite close to actual absolute values. Radial displacements measured in one of the extensometer rods installed are shown in Fig. 28. The extensometers measure the relative motion of the anchoring point with respect to the tunnel wall. The data in Fig. 28 refers to a 4 m long bar. Measured displacements (away from the tunnel wall) are very small, in the order of 0.1 mm. They increase rapidly at early stages due to the stressing and dilation of the rock and remain essentially constant for the rest of the measuring period (despite the increasing swelling pressure). Some of the predictions are quite accurate (IPS, SKI). Some uncertainties remain, however, because the measured values are close to the precision of the measuring system. 13. Discussion—Part C Temperature distributions are well reproduced in general terms. In fact, some of the calculations are very precise, unlike the case of temperature prediction for the buffer. In the case of the (saturated) rock heat flow is essentially controlled by constant parameters: the granite heat conduction and heat capacity coefficients. Convective heat transport is very small because of the negligible flow rate towards the bentonite. In the bentonite buffer, besides the effect of degree of saturation on conductivity and specific heat, other phenomena are at play: the heat transport associated with convective and diffusive flows of liquid water and vapour. This situation may explain the different performance of models in both cases. Rock water pressures development integrates several separate phenomena: the modification of the hydrogeological regime in the vicinity of the tunnel due to the presence of the barrier, temperature effects and the influence of the suction of the bentonite. Due to the different dilation properties of the pore water and the rock skeleton, a transient increase in temperature in a saturated porous medium induces a transient increase in pore water pressure. The magnitude of the increment of pore water pressure increases with the rate of temperature increase, the porosity and the rock stiffness. The rate of dissipation of the excess of pore water pressure increases with the rock permeability. As no clear transients of pore water pressure were recorded, this indicates comparatively high values of the rock permeability in the measuring zones. This view is supported by the observed evolution of the pressure of the water used to inflate the packers dividing the instrumented boreholes into intervals, exhibiting essentially an undrained behaviour. Some of the phenomena controlling the evolution of pore water pressures in the rock go in opposite directions. For instance, if temperature-induced pore water pressure development is a dominant phenomenon, points closer to the experiment axis should be expected to develop higher pore water pressures than points deeper in the rock, because the former will experience a larger temperature increase. On the other hand, if the natural gradient-induced flow dominates, points located at increasing radial distances will tend to show higher pore water pressures. Bentonite suction will tend to enhance this second effect and may even induce negative pore pressures in the granite under certain circumstances. In fact, the interaction between granite and buffer in the proximity of the interface is not a straightforward phenomenon. Important qualitative differences in behaviour may be explained by relatively minor changes of the bentonite and granite constitutive parameters. A sensitivity analysis of the behaviour of this interface is given in [19]. The measured pressure response at r¼13:06 m in boring SF21 (Fig. 26) is unfortunately not conclusive because of the abnormal fall of pressure recorded during a relatively long period (t¼80–480 days). Some ARTICLE IN PRESS -2.E-04 -1.E-04 0.E+00 1.E-04 2.E-04 3.E-04 4.E-04 5.E-04 0 200 400 600 800 1000 Time (day) Radial displacement (m) ANG BGR CNS DOE IPS SKB SKI UPC Fig. 28. Evolution of radial displacements recorded in boring SI1, extensometer fixed at a radial distance of 8 m and predictions of research teams. Table 10 Modelling approaches used for Part C Team Couplings Heat flow Rock th. dil. Wat. th. dil. Wat. flow ANG — — — — — BGR TH-(TM, TH)   CNS THM   DOE TH-TM   IPS THM   JNC TH  SKB THM   SKI THM   E.E. Alonso et al. / International Journal of Rock Mechanics & Mining Sciences 42 (2005) 611–638 635