Abstract
Fractured rocks, specifically sedimentary sequences, present a particular challenge for grouting, as major connected fractures that must be sealed will rarely be intersected by vertical groutholes. The aperture controlled grouting (ACG) method can be readily applied using discrete fracture network (DFN) analysis approaches to understand the pattern of fractures that need to be grouted. Aperture controlled grouting then utilises a combination of the principles of fracture grouting and grouting intensity number (GIN) grouting for practical and cost-effective grouting of tunnel, dam, shaft, or environmental restoration projects. This paper illustrates the application of DFN techniques for fracture assessment for grouting design and then compares the ACG approach to more conventional grouting approaches based on recent grouting experience.
List of notations
is transmissivity: m2 s− 1 is Lugeon value is test stage length: m is hydraulic aperture: m is void filling aperture: m is cohesion: Pa is viscosity: mPa·s is grout flowrate: L s− 1 is accumulated volume injected: L is pressure at the hole collar: MPa is effective pressure down hole: MPa
Introduction
The injection of cement-based grouts to control groundwater flow through fractured rock is a long-standing technique in both civil and mining engineering. In the past 80 or so years, this type of grouting has evolved from a ‘black art’ to an engineered ground improvement method.
Groundwater flows through rock in networks of hydraulically conductive fractures and other discrete features, including brecciated layers, solution zones (karsts), and shear structures; these are collectively and generically referred to here as ‘fractures’. Grouting reduces groundwater flow by blocking key interconnecting fractures. In principle, grouting is a simple process of injecting a setting fluid to fill the void space of the fractures: a void-volume replacement. Grout injection pressures are used to control grout flow, but grouting only succeeds if a sufficient volume is injected to block the fracture network.
Aperture controlled grouting (ACG, Carter, Dershowitz, Shuttle and Jefferies 2012; Bonin, Rombough, Carter and Jefferies 2012) aims to match design grouting protocols for injection pressure, flow control, and grout mix sequencing to the hydro-geological properties of site-specific fracture networks connected to the groutholes. The method thereby establishes a basis for effective and lower cost ground improvement than could be achievable using more conventional grouting procedures and protocols. On the basis of geologic understanding of the fracture systems and characteristics, the ACG method can provide a quantitative assessment of the quantities, rates, and pressures required for overall control of a grouting program. Effectiveness is then assessed from computation of grout penetration distances into the rockmass surrounding each grouthole. Such distances are typically used as a basis for ensuring that overlapped, contiguous grouted zones are achieved. By means of ACG sensitivity studies, optimisation can be achieved to ensure effective grouting protocols (mix sequence, flow-rates and injection pressure) are employed to match identified fracture patterns.
Application of the ACG approach can be tailored to almost any degree of ground complexity, as well as the level of field and design effort involved in implementation. For small projects, the design approach can be simplified to spreadsheet calculations. For larger projects, particularly where geology strongly influences the chosen methodology, benefit is gained from the powerful insight provided by discrete fracture network (DFN) modelling of the rock formation to be grouted.
The ACG approach can be applied both with the aid of detailed DFN models (Fig. 1) and/or using a simplified ‘partial dimension’ fracture geometry (Barker 1988) approach. The partial dimension approach assumes an increase in the flow area within the fractures being injected with distance from the injection point by a fractal dimension, a simplified characterisation but one that allows flow simulations in a spreadsheet; the difficulty is primarily in identifying the fractal dimension, which is an abstract concept not immediately recognisable from a geological exposure. In contrast, DFN models offer far greater realism in replicating the fractured network, with concepts well-understood in geology, but now needing specialised software. Either way, explicit recognition of the fracture network has many advantages for optimising the grouting injection process, as how grout penetrates the fractures essentially controls the success (or otherwise) of the grouting. Estimating grout take requires data on interconnectivity and fracture size (i.e. areal extent, scaled by trace length or persistence), because, for the same fracture intensity, connectivity of groutable and ungroutable (i.e. non-connected) fractures can be significantly different, depending on fracture size.
Discrete fracture network (DFN) model with Bingham grout penetration
Fracture size is best assessed from exposure or outcrop mapping. Where such data are unavailable, fracture size can be assessed by extrapolating between large-scale fractures (faults) and small scale (core or borehole scale) features, using fractal methods. If hydraulic measurements of rock mass connectivity (including hydraulic interference tests) are available, fracture size can be increasingly well-constrained. Fracture size can also be estimated from borehole image log data, although less reliably (LaPointe, Wallmann and Dershowitz 1993, LaPointe 2002).
When used with the DFN approach, the effectiveness of the ACG approach can be determined by using a ‘numerical permeameter’ analysis method (Dershowitz, La Pointe and Doe 2004) to better estimate pre- and post-grouted rock mass permeability. This is one of the key advantages of generating a DFN model in the early stages of the ACG design approach. Once the model has been created, it can be used to directly aid the process of assessing the amount of grouting required.
The DFN approach makes it possible to forecast rather than estimate the change in rock mass permeability, resulting from grout placement in fractures. This is particularly valuable where the geometry of the natural fractures with respect to layout of any curtain configuration (i.e., hole spacing, orientation, depth and location) has a significant effect on grouting effectiveness. In such situations, industry practice has in the past been to carry out limited scale field grouting of trial panels, and in an idealised situation carrying out a field pumping test or at minimum a program of Lugeon testing of the block of rock mass to determine the rock mass permeability before and after grouting. While such testing is extremely expensive and only rarely undertaken in real geology conditions, it is readily possible to carry out such evaluations using ‘numerical permeameter’ simulations in a DFN model (Dershowitz et al. 1998). Once the fracture geometry is defined, and the appropriate boundary conditions and engineering requirements specified, multiple numerical permeameter simulations can be carried out quickly and efficiently to evaluate different grouting designs.
The ‘partial dimension’ penetration distance approach does not provide as direct a measure of the amount of flow that can occur post-grouting as can be achieved through more rigorous DFN simulation, but has enough robustness for use in the field for hands-on control of grouting operations. As a result, it is best that the ‘partial dimension’ approach is first ‘calibrated’ against DFN numerical permeameter simulations to help estimate travel distances and thereby define an appropriate safety factor for determining whether sufficient grout overlap/fracture blocking is being achieved by the grouting layouts and injection process.
Undertaking more rigorous DFN simulation of the whole ACG process rather than just relying on the DFN as a more sophisticated method for characterising the fracture fabric for application of the ‘partial dimension’ approach has several additional advantages. Not least of these advantages is that the DFN model can be used to directly simulate Bingham grout flow, (Chhabra and Richardson, 1999)considering the natural fracture network and the variability in fracture aperture and connectivity stochastically determined based on field conditions. The model can thus directly aid grouting optimisation by allowing direct visualisation of probable grout travel pathways. This, in turn, can help in delineation of possible potential ungrouted ‘windows’ in the design curtain.
Discrete fracture network simulation can also be used to assess both hydrojacking and hydroshearing potential on each fracture that may be subject to increased fluid pressure during the grouting process. This again improves insight of probable grout protocol deficiencies from the viewpoint of controlling formation damage throughout the field implementation of the grouting process.
The initial steps involved in carrying out ACG analysis including combining both ‘partial dimension’ and ‘DFN approaches’ as applied for fracture-specific grouting, are quite straightforward, and little different from fracture characterisation for any hydrogeologic situation. The grout protocol design steps then build on this acquired data and assessment, right through to the stage of implementation and control of the grouting works. The entire process from initial field characterisation, through design, and onwards to construction implementation, comprises five basic steps:
field assessment of the geologic fractures; quantitative description of fracture geometry and properties (fabric); optimised matching of grout rheology to these site-specific fracture characteristics; leading to grout protocol definition; and finally to field implementation of injection control.
Aperture controlled grouting method
Aperture controlled grouting is based on the concept that grouting is a void-filling process and thus optimising grout ‘take’ becomes an objective rather than just a measure of ‘injection performance’.
The implementation of ACG follows the general decision flowchart approach of the Australian Method (Houlsby 1990, as updated from earlier procedures outlined in Houlsby 1977), set within the framework of the pressure/volume control procedures embedded in the GIN method (Lombardi and Deere, 1993).
Figure 2 illustrates a typical ACG grout design flow chart, based on the approaches and methodology described in Carter et al. (2012). This type of design chart implements the measures in a rational sequence in order to most efficiently ensure optimum grout mixes are developed to provide the desired rheology based on the established range of Lugeon values for a given site.
Aperture controlled grouting (ACG) grout design flow chart
The ACG approach to design of a grouting program for achieving the most rapid injection possible of the required grout volume for each stage, with minimal ground disturbance and disruption, is based on defining and then controlling three factors throughout the injection sequence outlined in Fig. 2:
‘Penetrability’, which is an index of fracture storage aperture (
Grout injection pressure based on requirements to maximise grout injection rates, while keeping below critical threshold pressures that would induce formation damage (by, for example, hydrojacking), and also ensuring reasonable batch volumes (e.g., 200 L) and required mix quality control. Grout rheology adjusted to obtain ‘refusal’ (i.e., achieving negligible (very low) flowrate at maximum injection pressure) at the target take for the specific grout stage, thereby optimising grout-fracture matching.
If this ACG style of sequenced mix design is adopted, it becomes possible to formulate a suite of grouts with progressive change in apparent viscosity matched to the specific fracture characteristics of the site, thereby aiding the process for achieving effective injection for all of the identified fracture characteristics (Fig. 3). The correct field implementation of the required ACG mixes would then be verified through a program of simple field quality control testing aimed at checking and confirming in situ grout rheology (primarily with use of a mud balance for density according to ASTM D4380 and a Marsh funnel for apparent viscosity according to ASTM D6910.
Penetrability Based Optimisation of Grout Mixes with respect to Target Lugeon Range
Photograph of typical mobile monitoring equipment for effective ACG field injection control
While the details of grout mix optimisation are not the direct thrust of this paper, it must be appreciated that in parallel with developing a suite of ideal stable-balanced grout mixes, practical constraints must also be taken into account. As grout cannot be uniformly or reasonably thickened in the holding tank, once a batch has been properly mixed, it is imperative that each fresh batch of grout be injected as soon as practicable after it has been initially mixed. However, until a specific grout in the prescribed thickening sequence is in the process of being injected into the ground and a change in penetrability with injected volume actually measured by the monitoring computer control system (Plate 1), the adequacy of that particular mix rheology for the fractures characterising that particular stage being grouted will generally not be known.
Cumulative Curve Fit Matching of Transmissivity and Aperture Data
Consequently, for fractured rock, particularly sedimentary sequences, before the grouting program starts and such data can be obtained from ATV logging of groutholes, the need for mapping local outcrops, quarries, road cuts, lineaments and pavement that give more depth and areal visualisation cannot be over-emphasised. Such mapping is particularly valuable for gaining insight beyond the one-dimensional view available in borehole images and from rock core. It is also often helpful to focus on mechanically distinctive rock units of specific lithology so that correlations can be defined between fracture spacing and rock unit thickness or domain extent. The aim must be to develop a three-dimensional understanding of fracture fabric variability within the zone to be grouted.
This understanding is critical, as for the same fracture intensity, connectivity of groutable and ungroutable (i.e. non-connected) fractures can be significantly different, depending on fracture size. Although fracture size is best assessed from exposure or outcrop mapping, it is feasible, although of lower reliability, to interpret fracture size from borehole image log data. The techniques that are needed to undertake such intrepretation from image logs are somewhat difficult to apply, and depending on data quality, significantly less reliable than when estimates can be made from mapping data. However, various methods do exist (e.g. LaPointe et al. 1993) such that, with reasonable evaluation effort, they are sufficiently robust as to allow relatively reliable estimates of fracture size to be made solely from borehole image log data, although this approach is not recommended as being applied alone. Most other aspects of fracture geometric data for ACG grout design can, however, be acquired directly from interpretation of the ATV and 3-arm caliper surveys that are recommended be routinely carried out in primary order groutholes. However, commonly during the early stages of a project, limited downhole televiewer surveys are available and thus data deficiencies often exist. These problems diminish as primary and secondary groutholes are drilled, and with each sequence greater understanding is gained. However, such drillhole fracture fabric information is often myopic because of practical considerations of the grout curtain geometry, and as such, almost always needs to be supplemented with data from outcrops and from other boreholes outside the immediate grouting area.
Figure 5 shows in the lower three diagrams a typical set of ATV data from grouthole drilling in sedimentary rock units plotted as a stereonet and rosettes (0°–20° and 70°–90°).
Good reliability discrete fracture network (DFN) matching of field ATV fracture data acquired from primary grouthole surveys
Typically in sedimentary rock units, bedding is subhorizontal and typically one or two orthogonal subvertical joints are commonly present, with a more diffuse network of intermediate dip fractures. As will be evident from Fig. 5, only the mid-range structures have been plotted on the stereonet, as it is much easier to discriminate minor differences in flat or steep fabrics using rosettes – hence, the use of three different plots.
The other useful discriminator of zones of concern for grouting is fracture intensity. Typically for a given fracture fabric in a rockmass (i.e. with a similar distribution of fracture apertures), grout takes are higher in zones with higher fracture intensity. Accordingly, it is useful to try to categorise the groutholes (and indeed any previous exploratory holes also) in terms of rock quality designation (RQD) and fracture frequency. Figure 6 shows some typical fracture frequency data separated by rocktype for a grouting project. There is a clear variability shown in fracture intensity, varying from 1.7 fractures per metre to as low as one fracture every 2 m. The DFN matching to this fracture data in this case is reasonable, given all rocktypes are present on the site and not all have been equally well-sampled by intersecting drillholes.
Typical discrete fracture network (DFN) Calibration Checks, showing Fracture Frequency (P10) Histogram Comparisons for different lithological rock types
It will be noted that in the above figure the designation P10 has been applied to this fracture frequency data, consistent with the normal nomenclature generally applied in DFN modelling (Fig. 7), but rare in engineering or even geological circles.
Discrete fracture network (DFN) Nomenclature for Fracture Intensity (Dershowitz and herda, 1992)
For most engineering projects, details of fracture intensity rarely exceed a count of fractures per metre along the various drillholes. However, when undertaking DFN modelling for ACG assessment, there is a need for a more complete volumetric description of the fracture fabric to be gained. This typically necessitates drillholes be advanced in varying orientations at the site, or that other (e.g., outcrop) information is added so that more representative volumetric understanding can be gained of the variability that exists in fracture intensity across the site to be grouted.
Fracture hydraulic properties
In parallel with achieving a good understanding of fracture intensity across a given site, proper application of the ACG approach requires that a reasonable understanding also be gained of variability in fracture conductivity throughout the rockmass to be grouted [expressed as ‘transmissivity’, T (m2 s− 1), rather than permeability (or hydraulic conductivity)], as, for the purpose of ACG, it is generally necessary to consider the grouted volume, with flow through all of the individual discrete features rather than considering flow uniformly distributed through the rock volume. Unfortunately, availability of volumetric transmissivity data is not commonplace for most engineering projects. Therefore, reliance is generally placed on data from standard Lugeon water pressure tests (WPTs) conducted in the exploratory stages of grouting. Typically for grouting projects, these tests, which give conductivity values in Lugeons (UL), are carried out over a test stage of length Ls; thus, an estimate of the interval transmissivity can be derived as:
Two apertures require definition:
– the hydraulic aperture, eh; and – the storage aperture, es.
Both are a consequence of the rock surfaces that form the sides of a fracture, being rough. Roughness leads to the reality that apertures vary along the length of any fracture, thereby creating existence for preferential flow channels and constrictions. For both water and grout flows, head loss is controlled by approximately the aperture cubed. This means that the ‘choke’ points, or constrictions within the flow-paths, control head-loss along each fracture (and thus, the required grouting pressure). These constrictions define the hydraulic aperture, which controls the hydraulic resistance of the fracture to flow, (Long and Witherspoon, 1985). Grouting to stop water leakage requires that the majority of the void space within the fracture system be filled as the grout front moves into the rock mass (otherwise, hydraulic flowpaths will still exist after grouting). When taken into account, this volumetric component defines the storage aperture, which corresponds to something less than the fracture porosity because there are ‘dead ends’ within a rough fracture that grout does not flow into. Broadly, the hydraulic aperture determines which grout to use, while the storage aperture affects how much grout must be injected to achieve an effective grout curtain.
By convention, the hydraulic aperture is defined from the ‘cubic law’ for flow between two parallel plates spaced eh apart (for water as the injection fluid):
Some estimate of the distribution and variability of storage aperture for a given project site can be made in parallel using data acquired from fracture aperture measurements derived from ATV downhole trace data, as shown in Fig. 4. Figure 8 plots the typical inter-relationship seen between transmissivity and aperture for typical field distributions of storage aperture and hydraulic aperture. The general form of the relationship of transmissivity with storage aperture that is applicable for grouts is given by the ‘Doe Law’ (which is described in more detail in Dershowitz, Doe, Uchida and Hermanson 2003):
Fracture Transmissivity variation with Controlling Aperture (after Dershowitz et al. 2003)
DFN grout analysis
Methodology
Discrete fracture network grout analysis can simulate the whole grouting process by ‘drilling’ groutholes into the DFN model constructed from the site characterisation data. As illustrated in Fig. 9, the holes are laid out at the design spacing and sequence, and then grouting of each hole is undertaken in stages within the DFN model using the design sequence, stage spacing, and grout rheology.
Typical Grout Curtain configuration set within a layered sedimentary rock discrete fracture network (DFN) model
In the DFN model, grout flow is implemented as a viscous immiscible fluid that flows into the fracture network from the stage being injected, (Fig. 10). The interface between injected grout and displaced groundwater is assumed to involve no mixing/dilution. Grout is represented as a Bingham fluid, so allowing realistic proportioning between fractures of differing aperture. Grout is also not allowed to penetrate fractures with too small an aperture, reflecting that grout is a suspension and not a true fluid (here, a groutability limit has been set for T < 1 × 10− 9 m s− 1, roughly 0.1 Lugeons corresponding to the fine fraction of typical microfine grouts).
Idealised grout flow through a fractured rockmass, simulated within a discrete fracture network (DFN) model. Left diagram shows typical ATV fracture data; right diagram shows typical Grout Flow Realisations for various grout stages as replicated within the DFN (Grout Take colour coded by radial Penetration Distance from hole axis)
Because the purpose of grouting in the DFN model is to understand the grout travel ‘radius’ and thus the related extent of ungrouted fractures that could compromise the curtain, grouting in the DFN is carried out by injecting various volumes of grout at realistic flowrates. Several grouting intensities are simulated, with the effect of the grouting then assessed for each intensity. To date, for most projects where the method has been implemented, it has been found sufficient to use three intensities: 50 L/m, 100 L m− 1 and 150 L m− 1 in order to get an acceptable range in behaviour. Multiple realisations can then be run in the DFN to allow quantification of uncertainty and variability. The number of realisations necessary will depend on the geological heterogeneity and variability of site conditions. In the authors’ experience, 10–20 realisations of most DFN models has been found adequate to allow subsequent optimisation of each modelled grouthole spacing and stage sequence configuration.
Evaluation of grout effectiveness
The primary purpose of bedrock curtain grouting for typical civil and/or mining engineering purposes is to block groundwater flow by injecting a balanced, stable grout that will remain within the fracture system, without being washed out during its design life. If some fractures (or portions of them) remain ungrouted, these fractures could potentially create open conduits through the fracture network for groundwater flow to bypass the grout curtain. Altering the grouthole spacing and/or the grout stage lengths and/or varying the mix designs and/or the sequencing of injection can each change the balance between what gets effectively grouted and what remains ungrouted. It is this interplay between ungrouted and grouted fractures that then becomes the basis for assessing grouting in the DFN model.
The effectiveness of grouting for each grout injection simulation can be calculated by using the DFN as a ‘numeric permeameter’ (Dershowitz et al. 1998) to determine an equivalent ‘porous medium’ hydraulic conductivity through the block of grouted rock, KG. This equivalent grouted hydraulic conductivity is then compared to the formation's original (pre-grouting) block conductivity (KF) to develop a ‘grouting efficiency’ measure, defined as With this measure, perfect grouting has an efficiency of 1; ineffective grouting has an efficiency of zero. By computing the efficiency for each realisation of the DFN model an estimate of effective improvement per grout stage can be assessed and if required, time and cost estimates can also be made to examine the value of extra effort to achieve a higher grouted efficiency.
Example results
Figure 11 shows the generally poor computed efficiency achieved with a typical conventional grout mix after primary grouting of the outer row of a three row curtain planned to be constructed through one of the formations shown in the layered DFN model in Fig. 9. The simulation results are for grout injection termination at 50, 100, and 150 L m− 1.
It is important to understand the source of the scatter shown in Fig. 11 This paper was originally presented at the first I. This scatter occurs because, typically, knowledge of fractures in the formation to be grouted (as modelled within the DFN) is in terms of probability distributions; thus, multiple realisations have to be computed to allow proper sampling of this natural variability. However, while this natural variability and uncertainty are responsible for some of the perceived wide scatter in resulting grouting efficiency seen in the figure, this is not the primary cause of the spread.
Required grouting intensity
An underlying cause of the wide scatter in Fig. 11 is that the volume injected in any stage is not matched to the transmissivity of the fractures intercepted by that stage – recall that fixed grout volumes were injected in the simulations. The issue then becomes how to best-match the injected volume to the stage transmissivity, reflecting the ungrouted rockmass condition. Because grout volume to achieve a particular penetration is proportional to storage aperture (Carter et al. 2012), and because that aperture is related to transmissivity, (Fig. 8) grout volume must be related to stage transmissivity to maximise grouting effectiveness. This requirement is developed into a design grouting intensity by transforming the results shown in Fig. 11 in the manner shown in Fig. 12. This transformation is carried out by introducing the realised fracture aperture (transmissivity), a random value reflecting the distribution of fracture properties being simulated (and which is known for each simulation) and for which the grout volume to reach the desired penetration distance is simply computed, i.e., ‘Voptimum’. Dividing the actual grout volume by this stage-specific optimum gives the true trend. Randomness arises around the true trend simply because of the stochastic nature of the geology. The practical grouting engineer deals with this randomness by injecting more grout than the ‘optimum’; in the case of Fig. 12, about two to three times the optimum stage grout take (based on that stage's Lugeon value) will give satisfactory curtain closure (a high probability of more than 95% reduction in groundwater flow).
Simulation results normalised to infer optimum grouting intensity
Borehole logging and associated water pressure testing show a substantial correlation between observed fracture aperture and interval transmissivity (or the more commonly used Lugeon value). This field-scale correlation is supported by direct measurements on hockey-puck size samples taken from cores. As shown diagrammatically in the top part of Fig. 8, the basis of the correlation is that, although fractures are rough with spatially varying apertures, hydraulic head-loss is dominated by the ‘choke’ points along each fracture as head loss is proportional to the third power of the aperture. On the other hand, grout will tend to fully fill the void space (‘tend’ because there will be dead ends in which grout does not penetrate); thus, properly assessing grout take depends on the average void aperture. In the DFN model, fracture transmissivity has been independently realized, with the associated void aperture used to compute grout travel based on correlation between the field-derived ATV fracture data and measured conductivities, using the approach that was earlier illustrated in Fig. 4. In turn, this means that the effect of the three modelled grout takes (i.e. 50, 100, 150 L m− 1) plotted in Fig. 11 needs to be related to the transmissivities of the fractures grouted. Optimisation of the grouting protocol is now described.
The grouting engineer has a choice over design hole spacing, orientation and layout (1 row vs 3, for example) and the volume of grout to be injected. Various stage lengths can also be selected with longer grout stages requiring larger volumes. Accordingly, for simplicity, grout take is normalised by hole length to define a grouting intensity. Given that grout volume and intensity to achieve a particular penetration will be proportional to void filling aperture, it is essential that grouting intensity be carefully related to stage transmissivity in order to maximise grouting effectiveness.
As a starting point, based on the storage (void) aperture correlation established from the field specific data (as shown in Figs. 4 and 8), an anticipated best-estimate of required grouting intensity can be generated, with the resulting equation having the form:
What is known, is the actual grout injected in the model, say five 5 m long stages each taking 50 L m− 1 – or, 1250 L of grout. If one ‘guesses’ values for the coefficients c and d then, depending on the fracture transmissivities, Stage 1 might have needed 23 L m− 1, Stage 2 73 L m− 1, etc. This might then give an optimum injection volume for the hole of let us say 1876 L. The normalised grouting intensity applied to this hole would then be defined as 1250/1876. When this type of calculation is carried out for all holes in a simulation, it allows one to iteratively determine the normalised grouting intensity for the block.
The coefficients c and d are thus determined by using a ‘goal seeking’ optimisation procedure that minimises the bandwidth of the resulting normalised plot. This is readily done in Excel, as shown within Fig. 12, giving c = 4 × 104 and d = 0.5, for this example. This implied optimum grouting intensity is then applied in equation (4) and graphed within Fig. 13.
Example required grouting intensity chart, as developed from discrete fracture network (DFN) modelling
Notice in the transformed results of the DFN simulations as shown in Fig. 12, there always remains some scatter at any chosen grouting intensity; this is a real result of natural variability in the fracture network and something every grouting engineer is familiar with.
Even careful, diligent injection does not always result in a perfect curtain. Rather, effective grouting is a matter of probabilities.
Suppose that the groundwater control in the project, notionally a cutoff beneath a dam, requires a 20-fold reduction in groundwater flow compared to an ungrouted scenario. This is a design hydraulic efficiency across the curtain of 0.95. Looking at Fig. 12, it can be seen that 90% confidence in achieving this performance requires injecting about three times the ‘optimum’ grout volume given by the assessed coefficients and equation (4).
Neither grout rheology nor grout injection pressure is involved in assessing the best hole layout – it is strictly a geometric process matching void space to be filled with likely grout travel distance. The resulting grouting intensity governs the effectiveness of the grout curtain. Grouting does not have to be carried out to ‘refusal’, but it must achieve an intensity matched to the Lugeon value of the stage being injected. In this regard, ACG is similar to the grouting intensity number (GIN) method. Figure 13 illustrates a grouting intensity curve for a typical fractured rock site derived with the ACG approach; this required intensity is specific to this rock formation and chosen hole layout (i.e. this chart should not be directly used for other sites – it is simply an example of what was done at one particular site).
Grout injection protocol
The selection of grout mix, the sequence and decision points in changing to a thicker grout and the injection pressures to be used are collectively referred to as the grouting protocol. The protocol aims to meet the required grouting intensity delivered most efficiently – it is entirely an operational issue requiring (i) injection pressures that maximise grout flowrates without causing formation damage; and, (ii) adaptive grout rheology so that when a stage reaches its target ‘take’, the injection pressure has increased sufficiently to allow moving to the next stage without subsequent grout injection continuing to penetrate into the last completed stage.
Grout rheology (mix design)
Grouts have evolved over the past 80 years. The very high bleed (i.e., particles settling out of solution when grout is at rest), low viscosity mixes of the 1930s to 1950s became more stable bentonite-based mixes through the 1960s to 1980s. These, in turn, gave way to thixotropic control (i.e., requiring positive pressure to initiate movement) with fluidising additives in the 1990s and, by the early 2000s, reached the current balanced, stable, highly controllable mixes of today (DePaoli, Bosco, Granata and Bruce 1992a, 1992b). Current balanced, stable grouts, utilising a cocktail of additives are now the norm (Chuaqui and Bruce 2003) – we refer to these as ‘modern’ grouts. Modern grouts also show controllable, consistent rheology pre-gelation (Rombough, Bonin and Shuttle 2006) that approximate a Bingham fluid with a yield strength (cohesion, c, and a viscosity, μ), with both properties controllable by adjusting the additives and the water : cement ratio of the grout; Fig. 14 illustrates a range of achievable properties with modern grouts.
Range of grout rheology accessible with modern grouts (Bonin et al. 2012, after Rombough et al. 2006)
Mix behaviour during injection
Practical grouting requires either sufficient pressure build (with constant flowrate injection) or sufficient flowrate reduction (at constant pressure injection) so that the packer can be moved and the next stage injected without the grout moving into the just-completed stage and without waiting for the grout to gel. These two conditions can be expressed as a requirement that the penetrability must fall below a target value by the end of grout injection for the stage. Although site specific, and depending on the desired effectiveness for the constructed curtain, experience and DFN modelling suggest a < 1 UL standard curtain (a reasonable, practical value) requires penetrability < 0.1 L m− 1 min− 1 MPa− 1. The task for the grouting engineer is to select the mix sequence to achieve this penetrability while honouring the ACG grouting intensity criterion. A simple spreadsheet model of radial Bingham flow into rough fractures extending from the grouting stage suffices for evaluating different grout mixes. Figure 3, shown earlier in the paper, was computed with such a spreadsheet and illustrates the penetrability at specified take limit for five grout mixes.
Looking at Fig. 3, the uppermost trend is for a very penetrating modern grout (c = 3 Pa, μ = 12 mPa·s). With this penetrating grout, for a < 2 UL stage (as plotted on the bottom axis), the target penetrability range is achieved at the specified take limit. However, for larger aperture fractures, the low viscosity of this grout means it falls short by nearly a factor of 10 in terms of desired penetrability at the desired take in even 7 UL ground – the required pressure build/flowrate reduction simply will not develop with this grout. Conversely, the bottom trend line is for a rather ‘thick’ grout (c = 40 Pa, μ = 25 mPa·s) and the computed penetrability shows that this grout essentially will not penetrate ground with < 40 UL, although the chart shows this mix is well suited for 70–100 UL ground.
Thus, effective and efficient grouting requires a suite of mixes. For a simple grouting project, typically a suite of four to six design mixes might be contemplated, as shown in Table 1; each mix being designed to have a bleed of < 5% (at 2 h) and a pressure filtration coefficient of < 0.04 min− 1/2 (DePaoli et al. 1992a, 1992b; Chuaqui and Bruce 2003).
Example mix sequence
Once mix selection and design are completed, Bingham properties are then dropped in favour of the familiar field QA measures, such as the Marsh funnel time that is calibrated to each of the grouts (Table 1).
Mix sequencing
Grouting could proceed by carrying out a WPT before injection of each stage, with the appropriate mix then being identified from the project's specific version of Fig. 3. However, such an approach leads to an unnecessarily large number of WPTs and, more importantly, multiple smaller aperture conductive fractures in a stage can have the same response in a WPT as a single, larger aperture fracture (i.e. the Lugeon test is not a unique indicator of fracture conditions). This presents a problem as the appropriate grout for the multi-fine fracture case will differ from that needed for the single-fracture case. Thus, while a series of WPT are required during the site investigation, and in some (possibly, all) of the primary holes, it is much more efficient to conduct the remainder of the grouting using a step-wise sequence of thickening during injection that depends on measured ground response to actual grout injection.
Injection begins using Mix A (or an ultrafine cement grout for a very fine fracture case) and continues through to Mix F, as needed, to achieve the required reduction in penetrability. Table 1 illustrates a set of grout mixes with progressively changing rheology mapped to the simulation results presented earlier on Fig. 3.
The criteria for decisions on whether and when to change the grout mix are established before grouting, and typically best represented on a flowchart as shown in Fig. 15. Decision points on this chart are not only established in part using the spreadsheet-based grout flow model but also reflect equipment-specific issues (for example, the mix batch quantity in the holding tank).
Aperture controlled grouting (ACG) ‘Decision’ flowchart (Bonin et al. 2012)
It will be appreciated that Fig. 15 is a practical control method for implementing ACG despite the use of Bingham theory in its derivation.
Injection pressure criteria
The desired injection pressure is aimed to be as high as possible while avoiding hydro-jacking (dilation of existing fractures in the rock mass caused by excessive injection pressure) as this typically damages the formation and generally makes grouting ineffective. As established by Lugeon (1933) and subsequently confirmed by Cambefort (1977) and Houlsby (1977) the onset of hydro-jacking can be easily established through simple water pressure testing.
Too large an injection pressure causes a hysteresis loop on a flowrate-pressure plot, with the formation showing uncontrollable increase of flow, irrespective of pressure change.
This behaviour is particularly obvious with modern electronic data acquisition systems. Although effective grouting requires avoiding hydro-jacking, the ideal injection pressure is as high as can be achieved while avoiding hydro-jacking; somewhere between 75 and 90% of the determined hydro-jacking pressure is practical. This critical hydro-jacking pressure usually shows systematic trends with depth and with geologic formation, both of which need to be established in the site investigation and fracture characterisation stage. Once these trends have been established, and re-checked in the early stages of grouting operations, it becomes straightforward to grout efficiently and effectively to a defined pressure profile.
Injection control
Monitoring systems
Engineered grouting requires measurement of grout injection parameters. Today, electronic monitoring of grout flowrate (Q), accumulated volume injected (V), and injection pressure at the hole collar (P) is routine and there are readily available commercial systems to do it.
Use of modern electronic data acquisition systems recording and displaying grout injection parameters is required to implement ACG, but ACG requires a little more in terms of data display than standard practice as ACG requires that the grouting engineer can ‘toggle’ through alternative displays showing: (i) Q and P versus time; (ii) Q and P versus V; (iii) P versus Q; and, (iv) penetrability versus V. The reason for this requirement is now illustrated.
Field procedure
The methodology for injection control essentially duplicates experience from Lugeon-style water pressure testing in that pressure is slowly built up while controlling grout injection rate using the real-time graphical monitoring output to ensure that no damage occurs to the rock mass.
Following the steps shown in Fig. 15, grouting starts with ‘Mix A’ at 3 L min− 1 for 10 min then increasing the flowrate to 6 L min− 1. After it is observed on the computer screen that the pressure versus flowrate plotted response has stabilised (typically after 2–5 min of injection), it is possible to draw a ‘safe’ grouting line as shown by the mean trend line on Fig. 16. The back projection of this trend line to the pressure axis indicates the formation's hydrostatic pressure. With this ‘safe’ grouting line established, grout flowrate can be increased until one of the following two thresholds are reached – either the maximum pump flowrate (possibly limited by the grout mixing rate) is achieved or the maximum injection pressure is reached. As shown in Fig. 16, the expected ground response to increased grout injection rate is for the stage pressure to steadily increase (i.e. the vertical trending data on the chart) or for flowrate to decrease at constant pressure (i.e. the curve tracks to the left, still within the ‘safe’ zone above the trend line). If hydro-jacking is starting to develop, the curve will track downwards to the right.
Using electronically monitored grout injection data to define the ‘Safe Trend’ to avoid rock damage during grouting (Bonin et al. 2012)
With initial grouting established using Mix A, the next consideration is whether Mix A is the correct mix choice for the stage being injected, or whether a thicker design mix is needed. This decision is made using the plot of penetrability versus total volume injected. If the mix is appropriate for the ground conditions (i.e. properly aperture balanced), penetrability will decrease markedly with take, as illustrated by the descending plot of data on the left of Fig. 17. (This decrease rate can be predicted, based on analysis of injection parameters in real time, but in practice, it is generally easier to try a few stages until appropriate typical rates can be established for that rock formation).
Measured penetrability trends for assessing appropriate grout rheology for stage being injected
An inappropriate mix is readily apparent as the penetrability ‘flat lines’, as shown in the right hand behaviour illustrated on Fig. 17.
Grout thickening is indicated when penetrability results have ‘flat-lined’. Thickening must then be implemented in a controlled sequence to maintain formation aperture matching. Assuming that one was grouting with Mix A, then one would change to Mix B; if grouting with Mix B, again shows “flat-lining”, then one would change to Mix C etc. Each mix change requires review of the penetrability versus time plot to determine if the new grout is appropriate for the formation – as evidenced by a clear penetrability response (like the curve on the left side of Fig. 17). For practical reasons, it is usual to always inject at least one holding tank volume of each mix (typically 200 L) and this is normally sufficient to ascertain the adequacy of the aperture balance of the mix. Further mix changes would then be implemented as necessary to continue to maintain a declining penetrability trend.
Closure criteria
The ACG approach requires optimising stage grouting intensity with stage conductivity. Design work, using fracture data acquired from acoustic/optical televiewer logging and conceptualisation of the geology of the project site, leads to a site specific chart, such as Fig. 13. Grouting procedures are then based on efficiently meeting these chart criteria, adapting grout rheology to site conditions with a penetrability closure criterion. Verification ensures that the procedures ‘close the loop’ to meet project objectives.
In the early phases of the works, for those grout stages for which a WPT was carried out, the grout take should be check plotted against the stage Lugeon value on the design intensity chart (Fig. 13) for every stage grouted. Based on such check plotting, it will rapidly become apparent whether or not the computed ground response to grout injection is reasonably correct, or if further refinement to the grout mix rheology and/or mix change sequence is warranted.
With the dual aspects of grouting intensity and mix suitability verified, it is straightforward to complete the grouting works using the now readily available real-time graphical monitoring displays to ensure that each injected stage is grouted to target penetrability. In principle, that is a sufficient approach, but engineering prudence is based on the proverb ‘trust is wonderful…but distrust is better’. Proof-of-adequacy WPTs should always be used to verify grouting effectiveness. Grouting works should start with a test ‘panel’ in which WPTs are carried out in the secondary order holes and then, after grouting those secondary holes, yet further WPTs should be conducted in some tertiary order holes. Comparisons of those WPT results with the design expectations for conductivity reduction after primary and secondary injections allow rapid confirmation that the grouting to the specified grouting intensity is achieving closure, or not.
Discussion
The procedures outlined in this paper have been developed over three decades. Grouting experience with electronic monitoring includes more than 20 dams constructed into Canadian Shield rock through the 1980s and 1990s, the deep curtain construction at the Antamina Dam (Peru) over the period from 1999 to 2011, (Carter, Amaya, Jefferies and Eldridge 2003; Ritchie, Garcia, Amaya and Jefferies 2003), plus work undertaken on other dams, including two at Meadowbank (Nunavut) 2009–2011. The pioneering electronic monitoring of the Elliot Lake dams (Jefferies et al. 1982; Carter 1982) established the close correspondence between viscosity affected grout penetrability data and WPT results, suggesting that grout injectability trends parallelled those seen in classic Lugeon water testing (Fergusson and Lancaster-Jones 1964; Houlsby 1976; Carter 1982; Carter and Blair 1990).
Evaluation of the comprehensive data from the main grouting curtain construction at Antamina, initially using GIN protocols (Lombardi and Deere 1993) with only cement–superplasticizer-based mixes before changing to site specific thixotropically modified cement–bentonite–superplasticizer mixes and modified GIN protocols (Ritchie et al. 2003; Rombough et al. 2006; Shuttle, Rombough and Bonin 2007) shed further insight into grout–ground interaction under various grouting conditions, ranging from straightforward fracture filling to full karst injection (Ritchie 2003). Injection control in recent dam foundation grouting for Meadowbank, Little Bear Creek in Alabama and at other sites has further highlighted the applicability of utilising electronically monitored, grout penetrability behavioural patterns as primary control for adjusting flowrates and pressure build in response to ground behaviour during grout injection.
The typical reason more grout is often seen to be needed on a given grouting project than ‘optimum’ is that the grouting engineer has no control over which fractures the grout flows into. It is insufficient to grout simply to the hole spacing. One must go further to ensure sufficient overlap to block meandering pathways for groundwater flow. Depending on the project requirements in terms of hydraulic effectiveness and confidence in achieving that effectiveness, site specific charts such as shown in Fig. 12, can be used to judge the amount of extra volume needed over the ‘optimum’ amount – the ‘grouting safety factor’ referred to earlier. With the degree of over-grouting established, which is essentially multiplying up the coefficient ‘c’, Fig. 13 shows the end resulting plot that would then be carried forward into the execution of the actual site grouting as the underlying ‘rule’ for establishing the site ACG protocols.
Conclusions
Grouting of a given stage in a grout hole is most effective if the volume injected is based on that stage's fracture aperture. The procedure to do this, ACG builds on the ‘Australian Method’ for grouting (the seminal contribution by Houlsby 1977) where grout mixes were thickened in sequence using a decision chart. However, in ACG (i) modern near-colloidal grouts are used rather than the ‘thin’ 1980s mixes; (ii) the thickening sequence is based on the computed behaviour of these modern grouts flowing into rough fractures as opposed to being based solely on compiled experience; and, (iii) a penetrability target is used to match the volume injected in each stage with the fracture apertures intersected by that stage. Operationally, the ACG is a simple combination of aspects taken from the Australian and GIN (Lombardi and Deere 1993) methods with clear decision points.
Although some may view the use of DFNs as ‘much too sophisticated and esoteric’ for routine grout design, particularly for small projects, field experience is showing that their use within the overall ACG approach has wide applicability for aiding the optimisation of any grouting being planned within fractured rock. This is not unlike the reticence of the grouting industry to adopt the recommendations of Jefferies et al. (1982) to employ electronic monitoring as a cost-effective advance for controlling a grouting program, and the slow adoption also to utilise downhole OTV and ATV tools to collect in-hole fracture data to help understand grout takes and intensity variability along a curtain. It is to be hoped that with time and exposure to the DFN/ACG approach, the grouting industry will see the benefits and appreciate that DFN application as part of the ACG approach can significantly assist in generating cost-effective grout design solutions, which can then be readily verified by controlled monitoring. Grouting need not be a ‘black art’. It can be both a predictable and reliable ground improvement method if the nature of the fractured ground and the way cement grouts penetrate those fractures can be more accurately replicated.
Discrete fracture network usage within the ACG methodology simply helps to achieve this by aiding better characterisation of the ground, thereby moving typical grouting operations more towards a properly engineered ground treatment method.
For the Owner, ACG allows the grouting engineer to deliver the most effective grout curtain at least cost.
The first large scale application of ACG was the 200-m high Antamina Dam (Ritchie et al. 2003). However, for that project, the grouting intensity was established by test panels with WPTs, not DFN modelling. Interestingly, the required safety factor on injection volumes was found to be two; the probability of a factor of 2 in achieving the required hydraulic performance was uncertain from the panel tests, although post-impoundment measurements suggest that the curtain achieved what was required for the project.
Clearly, this is where DFN modelling can contribute much more than achieved to date. The similarity of the results that can be achieved when DFN control is exercised on a grouting project when compared to the baseline practical results achieved at Antamina are encouraging. More experience is, however, needed with different geologies to see if two to three times ‘optimum’ grouting intensity is a good general rule for application for all grouting projects, or if this factor, in fact, varies in a more geologic-specific manner.
What is not in doubt is that utilising DFN modelling provides a significantly more cost-effective approach than undertaking fully constructed panel trials to gain the same basic understanding.
Acknowledgements
The approaches outlined in this paper have evolved over three decades of application of early versions of these techniques since the first introduction in the 1980s of electronic grout monitoring for dams in Elliot Lake, Ont., Canada. Acknowledgement is due to the many individuals and companies who over the years have contributed to the successful implementation of these techniques, which have also evolved in step with the significant technological advances that have occurred in computing power, site investigation approaches and drilling methods. In addition, thanks is due to the wise counsel and guidance provided over the years by the various project review board members whose insight guided development of some of the key ideas behind the ACG concepts. This paper was originally presented at the first International Conference on Discrete Fracture Engineering (DFNE 2014) (19–22 October 2014, Vancouver, BC, Canada) and has subsequently been revised and extended before consideration by the Mining Technology journal.