Rockmass characterization is an essential component of geotechnical design for tunnelling infrastructure. Bid-stage design for tunnels relies heavily and often exclusively on borehole characterization for data inputs. The nature of data collected for conventional geotechnical characterization of drill core is often directed by inputs defined by empirical rockmass classification systems, such as Q (Barton et al., 1974) and Rock Mass Rating (RMR) (Bieniawski, 1976, 1989). When these systems were introduced, no practical numerical tools for routine use were available so design relied on an empirical process. Since then, numerical modelling has become a very powerful and ubiquitous design tool. Numerical methods have grown from their early elastic and homogenous continuum behaviour to having complex elasto-plastic and fully discrete capabilities. The advancement of modelling has allowed for more detailed analyses, including technically challenging underground excavations in more complex rockmasses, leading to significant improvements to design practice.

While significant improvements in the level of sophistication of numerical tools have been achieved, conventional core logging practices, whose procedural design predates numerical analysis, have not made similar advancements. Conventional core logging practices do not capture the sophisticated data required for numerical input parameters, especially in complex rockmasses such as nodular sedimentary rock or hydrothermally altered rock, where the geological features that exert significant controls on rockmass strength are ignored in common classification systems.

Geological settings with complex rockmasses contain healed structures such as nodules and veins that exist within jointbounded blocks of traditionally "intact" rock. Previous work by the authors suggests that these healed or partially healed structures (termed intrablock structure) within joint-bounded blocks (where the joints are the interblock structure) have a significant impact on rockmass strength. The authors have previously proposed methods to estimate the strength of a rockmass that contains both interblock and intrablock structure: at the structural scale using GSI (Day et al. 2012a, 2012b), and at the intact drill core scale using intact strength parameters (Day et al. 2013b).

In an effort to advance rockmass characterization practices for complex rockmasses, the authors have developed and tested several new core logging procedures that account for healed structures that are present in such rockmasses.

When compared to conventional methods, there are increasing amounts of detail in the recorded data for each method. Data from several drill core sections is used to discuss the implications of each method in terms of level of detail, logging time, and cost.

An example of a 15m section of core in nodular sedimentary limestone is used in this paper to illustrate the applicable core logging methods in the context of numerical models of a nuclear waste storage excavation.

Including intrablock structure in rockmass characterization
The method developed for GSI in complex rockmasses in Day et al. (2012b) is summarized here, as applied to the linearized and quantified GSI chart by Hoek et al. (2013). Conventional use of the GSI system dictates that when evaluating a rockmass that contains interblock and intrablock structure, the worst cases from each block size and joint condition ranking should be combined to give an overall GSI value for the rockmass. Using the hypothetical example shown in Figure 1, the combination of the interblock structure (square) and intrablock structure (diamond) would give a conventional worst case GSI value of 39 (white circle). This method has been shown in field observations to underestimate the rockmass strength and a new method is proposed here which calculates a more realistic GSI value, Composite GSI (GSI*), for a rockmass that contains both interblock and intrablock structure, of 54 as shown by the green circle in Figure 1.

The effect of GSI values on the Hoek-Brown strength envelopes for the hypothetical cases shown on the GSI chart in Figure 1 is shown in Figure 2, as well as the equations for the Hoek-Brown strength criterion.

In the updated GSI chart by Hoek et al. (2013), GSI values can be determined quantitatively by summing the two linear scales representing the discontinuity surface conditions (scale A) and the interlocking of rock blocks defined by these intersecting discontinuities (scale B) (see Figure 1).

Eq1 Common Core

According to Hoek et al. (2013), scale A can be estimated using the 4th Factor for Joint Condition (JCond76 or JCond89) in Bieniawski’s RMR classification (Bieniawski 1976; 1989) or Joint Roughness (Jr) and Joint Alteration (Ja) from Barton’s Q classification (Barton et al. 1974), as shown in Equation (2). Likewise, scale B can be estimated using the Rock Quality Designation (RQD) by Deere (1963) as in Equation (3). Scale B can also be estimated using the quantified GSI chart by Cai et al. (2004), as shown in Equation (4).

Common Core Eq 2,3,4

The new Composite GSI method for intrablock structure is described below with Equations (5) to (7). For a rockmass containing multiple, distinct suites of structure, an effective block size B* can be calculated from the individual spacings for each discontinuity set, while a weighted composite value for A* can be obtained considering the relative contribution of each structural set to rockmass integrity.

In Equations (5) and (6), A1, B1 apply to the first system (e.g. a clean, rough set of interblock structure), A2, B2 apply to the second system (e.g. an infilled, smooth set of interblock structure), An, Bn apply to the nth system (e.g. intrablock structure). This method becomes a powerful tool for rockmasses that contain multiple, distinct systems of structure, because each structure can be considered individually before being combined into a single value that describes the entire rockmass (Day et al. 2013a). In this scenario, any number of structures (e.g. third, fourth, nth) can be included in the calculations. For Composite GSI (GSI*):

Common Core Eq 5, 6, 7

Incorporating intrablock structure into lab testing
Intrablock structure is typically ignored by geotechnical loggers because they are generally trained only to assess fractures. Moreover, to follow conventional strength testing guidelines, intact samples containing intrablock structure are deliberately excluded as much as possible from laboratory strength testing. In a rockmass with pervasive healed structure, selecting the few intact samples without intrablock structure is not representative of the rockmass. Instead, intrablock structure should be included in geotechnical core logging and a variety of intact samples with and without it should be selected for laboratory testing in order to more accurately capture the rockmass behaviour. CODELCO-Chile El Teniente Division has addressed this issue by developing a methodology to include intrablock structure in laboratory testing (Marambio et al. 1999), shown in Figure 3. Here, intrablock structure is recognized as a main contributor to rock strength and geotechnical data for the different failure modes provides a more comprehensive understanding of rock properties for numerical design.

Incorporating intrablock structure into core logging
Four core logging procedures with increasing levels of data capture were developed to compare the effectiveness of the data collected in each for geotechnical design decisions (Day et al. 2014). The methods were tested using drill core in a hydrothermally altered porphyry deposit in northern Chile, as well as exploration boreholes in a magmatic Ni-Cu sulphide deposit at an active mine in Sudbury, Canada. Data for each logging method was collected from drill core sections each measuring 20m long.

The four logging methods and submethods are described as follows, and outlined detail in Table 1.

  1. Method 1: Traditional classification for empirical design (Q and RMR systems), for basic numerical models ((1a) unoriented and (1b) oriented core);
  2. Method 2: (2a) State of practice geotechnical logging (based on the method used by GeoBlast in Chile), plus (2b) basic information about intrablock structure (veins in unoriented core only);
  3. Method 3: Method 2a for interblock structure plus moderate detail of intrablock structure (veins in unoriented core only);
  4. Method 4: Method 2a for interblock structure plus high detail of intrablock structure for extremely detailed numerical models ((4a) unoriented and (4b) oriented core).

The block size measurement of intrablock structure is an assessment of spacing between veins. Only veins that are visible around the full circumference of the drill core are included. The measurement is systematically taken from the top inflection point of the vein around the outside of the drill core in order to avoid double counting veins.

The type of intrablock structure recorded refers to mineralogical composition. The types of veins are classified by strength ("Strength Class") based on a qualitative assessment of mineralogy and the quality of the bond between the vein and the wall rock. This becomes important for vein mineralogies that can be rockmass strengthening (e.g. quartz), but only if the bond to the wall rock is very good and the vein and wall rock appear fused together.

The thickness, alteration halo (if present), and Moh’s hardness rating each provide additional information about the intrablock structure.

The logging methods from 1 to 4 consider increasing amounts of data, but take increasing time to complete (see Table 2). Measuring orientation data (as in methods 1b and 4b) adds a significant amount of time to the logging process.

The relative data collected and the time required to log are significant decision making factors regarding schedule and budget for any project, and it is important to understand the added benefit (if any) of each piece of data that is included in the core logging procedure to the subsequent design process.

To put the importance of logging time into context, two examples of current industry core logging times and cost that are quoted for contracts are as follows: a North American consulting company quotes USD 120-160 per hour for core logging projects with an estimated rate of 40m per 10-hour day (~USD 30-40/m) for oriented core and 100m per 10-hour day (~USD 12-16/m) for unoriented core. A typical Chilean consulting company quotes USD 15 to 65 per meter, depending on contracts, with an estimated rate of 30-50m per day (mix of oriented and unoriented core). Based on these estimates, costs of core logging can range from USD 12,000 to USD 65,000/km. If the logging rate decreases due to collection of more detailed data, the core logging costs could easily exceed USD 100,000/km.

Implications of data input for numerical models
To apply the core logging methods and test the design implications in a sedimentary geological setting, a numerical analysis of a proposed tunnel section in an underground repository for high level nuclear waste (HLW) storage was conducted. Input data was based on a 15m drill core section of a nodular limestone. Numerical models of the tunnel with vertical canister storage were created using the finite element method (FEM) software, Phase2 (Rocscience, 2013b). The excavation was modelled in two different rockmasses defined by input data from the applicable core logging methods, 1 and 3, discussed above. While the intact rock properties and in-situ stresses remain constant, the types, geometries, and strength properties of the rockmass structure vary.

Model setup
This idealized repository excavation example was assumed to be approximately 500m deep and in an in-situ stress state with ratios of 2:1 ( H: v) and 1.6:1 ( h: v). The intact strength properties for the rock material represent a nodular argillaceous limestone and are based on published values (NWMO 2011) as shown in Table 3. The peak strength values were selected based on a linear regression fit from tensile, Unconfined Compressive Strength (UCS), and triaxial laboratory test data, as shown in Figure 4. Rockmass parameters are specified using the GSI and Hoek-Brown (H-B) approach (Hoek et al. 2002). Alternatively the strength envelopes can be simplified to equivalent linear Mohr-Coulomb envelopes (M-C).

A schematic of the general model geometry and boundary conditions for both models is shown in Figure 5.

The geometry and properties of structures in the discrete fracture network, and the corresponding GSI value in the homogenous equivalent continuum, vary between models. The continuum section of the model is included to lessen the computation time required for the models and was calculated using the Composite GSI approach for the applicable suites of interblock and intrablock structure.

Joint and vein properties
The model based on logging method 1 only considers interblock structure: joints and bedding. The selection of stiffness and strength properties was aided by information in Read and Stacey (2009) and Pitts and Diederichs (2011), and the values are shown in Table 4.

The model based on method 3 contains both interblock and intrablock structure. The intrablock stiffness properties were selected with help from Goodman (1968), and strength properties of the intrablock structure were selected as the lower bound regression fit of laboratory testing data, as shown in Figure 4.

The intrablock structure was modelled explicitly using Voronoi joint networks, which create polygonal shapes with high angularity and in an irregular pattern, and are widely accepted for modelling microstructures in rock (e.g. Lan et al. 2010).

FEM numerical model results
Several observations can be made from the models representing each of the core logging methods 1 and 3. Detailed views of the model results, in terms of the maximum principal stress ( 1), total displacement, and yielded mesh elements are shown in Figure 7.

There is a significant increase in total displacement around the excavation when intrablock structure is included in the models since the weaker planes of intrablock structure allows the rockmass to accommodate more displacement. Rockbolt design for this excavation example based on the conventional logging approach, method 1, may underestimate the length required for rockbolts to be effective, if the real rockmass behaviour is closer to that with intrablock structure.

Overall, these models show that different levels of structural detail, associated with different levels of core logging resolution in terms of data depth and breadth, produce significant variations in predicted rockmass response around an excavation.

Design implications
The numerical example of a deep underground repository excavation discussed in this paper shows that the input data based on different core logging methods has a significant influence on the results of rockmass behaviour.

Conventional logging methods which consider only interblock structure were compared to methods developed by the authors that consider both interblock and intrablock structure.

The detail of these logging methods, from traditional rockmass classification parameters to current state of practice conventional logging and detailed intrablock data in oriented core, increases significantly, but the time required for logging also increases.

While the models discussed in this paper show that greater detail in logging results in more geologically accurate numerical models, there are consequences for the cost and schedule of a project.

Further analysis of the core logging data that was collected by the authors is needed to make recommendations for a balance of detail gained and time spent that is necessary for accurate and useful geotechnical design.