Rock mechanics and nuclear waste disposal 09 Jul 2020

With a long rock mechanics background, there has been the opportunity to gain some detailed insight into various international nuclear waste related studies. These have been studies with a geological disposal facility focus, and have involved strong crystalline and strong volcanic rocks, and specifically with higher strength rocks with reference to the UK-studies reported by Marsh, Williams and Lawrence in the recent TunnelTalk focus on nuclear waste disposal.

Many kilometres of cores add to the site investigation for suitable underground nuclear waste repositories

Personal involvement in studies started in 1980 and ended about ten years ago but with a degree of time extension due to the presently running, post-Fukushima ISRM commission that concerns siting of nuclear power plants underground. This is chaired by Professor Sakurai, a past-president of ISRM, and follows an interest in underground siting of nuclear power plants in large caverns of more than 50 years ago.⁽¹⁾

In the 1980s, the Office of Nuclear Waste Isolation (ONWI) in the USA funded a study by the geotechnical company TerraTek, to test several instruments in parallel to assess the in-situ performance of instrumentation when subjected to the heat generated by decaying nuclear waste in an underground repository. One task was to interpret the hydraulic behaviour of the first fully coupled hydro-thermo-mechanical in-situ block test of jointed rock (Fig 1). The test was performed on a heavily instrumented and flat-jack loaded, heated and flow-tested 8m³ of quartz monzonite in the Colorado School of Mines experimental mine in the USA.

Fig 1. Mean jointing trends of the heated block test⁽²⁾

Strain and deformation gauges used for the HTM in-situ block test

ONWI subsequently funded the completion of a joint constitutive model, which was finalized in 1982 and used in a two-volume report with Bakhtar for the Atomic Energy of Canada company and the Canada Centre for Mineral and Energy Technology. This concerned potential model application in fractured parts of the underground research laboratory in Manitoba granite. The model for joint behaviour, which became known as the BB (Barton-Bandis) model, was subsequently incorporated in the numerical code UDEC, UDEC-BB, by Itasca and NGI in 1985.

Yucca Mountain

At the Yucca Mountain test site in Nevada, planned to be developed in both the lithophysal and non-lithophysal jointed welded tuff, two reviews were performed of the extensive characterization studies. Much of this concerned the repository site characterization application of the rock mass rating (RMR) method of Bieniawski and of the Q-system.

Large diameter cores with planar and rough joint roughness coefficient values represent two of several joint sets at Yucca Mountain

Besides using core from numerous boreholes and a variety of down-hole testing, characterization at Yucca Mountain was performed after two TBM tunnels of several kilometres had been driven beneath a length of the mountain. The second tunnel angled into potential repository material, including the remarkable lithophysal or hole-bearing tuff, which was mostly without systematic jointing. Planned disposal of waste, in cannisters on rail cars, included options for future retrievability, providing the heated tunnel arches and inverts did not fracture too extensively along the parallel disposal tunnels.

Numerical modelling of gradual tunnel degradation as a result of earthquakes had surprisingly not distinguished between the properties of the different joint sets. In reality there were three distinctly different joint roughnesses. These differences can be described by means of the BB model (#3) for shear strength, and thereby lead to more realistic modelling results (Fig 2).

The illustrated non-linear peak strength equation was verified against tests on 130 joint samples. The joint roughness coefficient (JRC) has at least 50 additional equations to its name, developed by those who do not perform tilt or shear tests but instead analyse 3D joint roughness. The non-linear term, which includes JRC, incorporates the ratio of stress and joint wall-strength and a correction for scale or block size according to scaling equations suggested by Bandis et al.⁽³⁾

Site investigations at Sellafield in the UK

In the 1990s, a characterization and modelling study was carried out by NGI for the Nirex organisation at the Sellafield site in the UK in Borrowdale volcanics, which is basically a welded tuff called ignimbrite.⁽⁴⁾ The studies included the logging of 11km of deep core, using RMR and the Q-system. Hundreds of rock joint samples were selected for index testing and some for shear tests. Extensive UDEC-BB modelling was performed for the planned 700m deep disposal caverns. A planned TBM access spiral to 700m depth would first pass through the overlying St Bees sandstone and then penetrate the deeper ignimbrite. The UDEC-BB model illustrated several of the logged joint sets and a fracture zone (Fig 3).

Tilt test equipment was used to test the basic friction angle of the rocks, in this case the sandstone, using smooth, but not polished, core in line contact (Fig 4). Such tests generally represent input to UDEC-BB tunnel and cavern modelling. When performed on joints from the in-situ block test described earlier, they gave insight into permeability changes with joint closure, also as a function of temperature.

Fig 4. Mechanised or manual tilt test devices for characterizing basic friction angles (left) and the joint roughness coefficient value of an initially interlocked rock joint (right)

Extensive studies in Sweden

Studies at Stripa, Äspö, Simpevarp and Forsmark in Sweden included in-situ testing and occasional modelling of the excavation disturbed zone (EDZ) and extensive 1,000m deep Q-based core logging for the agency SKB. NGI had two years of rock mechanics research tasks in the site characterization and validation (SCV) project at Stripa for SKB (Fig 5).⁽⁵⁾ The BB model was an essential part of the core-logging, index testing, and coupled shear-flow testing at the different SCV scales. The BB shear-dilation-aperture-permeability behaviour from A to C was studied (Fig 5 left) as was the normal stiffness-closure cycling-aperture-permeability behaviour from D to E (Fig 5 right).

Fig 5. Successive scales of site characterization, prediction and validation in the site characterization and validation project at Stripa in Sweden

Site characterization in the Äspö pillar stability experiment (APSE) included Q-logging performed at 5m intervals to help locate a pillar-loading experiment (Fig 6). The Q-statistics indicate an extremely good quality rock mass, as also reflected in the frequent half-rounds following careful blasting. From such Q-value and Qc = Q x UCS/100 results, estimates can be made of the depth-dependent deformation moduli for modelling, and seismic P-wave velocities for correlation with, or interpretation from, geophysical testing.

The most frequent quality of Q = 200 means extremely good, and the typical maximum of 2,000 is basically a rock mass without jointing, which is an apparently ideal, but seldom realised goal for waste disposal. In reality it is not so ideal as there may be no hydration of cannister-surrounding bentonite if permeability is too low.

Fig 6. A set of Q-histograms representing overall rock mass quality in the APSE tunnel in Sweden

Four 1,000m deep cores drilled at the Simpevarp and Forsmark sites that were also Q-histogram logged generally gave similarly high-quality Q-value results except in fracture zones. Surface-exposure logging was performed above both sites, on each occasion using the same Q-parameter histograms. The logging was requested as an independent check of the results obtained by SKB and its geotechnical subcontractors.

The ideal for high-level waste repositories

Q as a measure of rock mass quality, with its simple inter-relationships to rock mass deformability and seismic velocity, and also to permeability via Q_H2O, is of some help in the choice and design of underground facilities. However, for nuclear waste disposal there will be chosen media, such as over-consolidated clays and shales, where the Q method is clearly inadequate or inappropriate.

There are other aspects of behaviour that rock quality by itself cannot solve, and this concerns the unconventional behaviour of local jointing in fracture zones. It is unfortunate that in the case of heat-generating high-level waste, smoother fractures need to be avoided when large-diameter drilling for cannister layout is occurring because of a phenomenon called thermal over-closure.⁽⁶⁾ We must ideally isolate high-level waste in the almost unjointed/unfractured parts of the rock mass at depth, but not too deep, to avoid extensional strain fracturing, which starts before possible unstable propagation in shear. Both must be avoided, possibly by siting at less than 500m depth and choosing sites where stress anisotropy is moderate.