Geotechnical Models and Data Confidence in Mining Geotechnical Design Michael Dunn Principal Consultant (Geotechnical Engineering)
Overview Geotechnical models Geotechnical model and design Data reliability and confidence Uncertainty Quantifying reliability / confidence tools available Design reliability Conclusions 2 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Geotechnical Model Geological Model Structural Model Hydrogeological Model Rock Mass Model Geotechnical Model NB: All components have uncertainties associated with them 3 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Geological model Uncertainties Lithological boundaries Alteration boundaries Major contacts Weathering profile Base of complete oxidation 4 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Structural model Uncertainties Major structures Existence Position Orientation (drillhole/mapping bias) Properties Fabric Orientation (drillhole/mapping bias) Properties Frequency 5 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Hydrogeological model Uncertainties Permeability of different rock units Groundwater level Seasonal fluctuations Pore pressure distribution 6 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Rock mass model Uncertainties Intact rock strength Rock mass strength Defect shear strength Rock mass classification Spatial variations Frequency 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 1400 1200 20 40 60 80 100 120 140 160 180 UCS (MPa) Frequency 1000 800 600 400 200 0 50 55 60 65 70 75 80 85 90 95 GSI Rating 7 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Geotechnical Model Used to define geotechnical domains with similar rock mass and structural characteristics. Forms the basis of geotechnical design sectors or areas. 8 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Geotechnical model and design Many design approaches imply a requirement for a geotechnical model and some level of confidence Peck (1969) Terzaghi s Observational approach Assessment of the most probable conditions and the most unfavourable conceivable deviations from these conditions. In this assessment geology often plays a major role. Establishment of the design based on a working hypothesis of behaviour anticipated under the most probable conditions. Bieniawski (1991) Minimum uncertainty of geological conditions. Stacey (2004) Minimization of uncertainty (collection of information e.g. site characterization, rock properties, groundwater, in situ stresses). Concept formulation (geotechnical model). 9 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Geotechnical model and design The geotechnical model is the cornerstone of any underground or open pit geotechnical design. Provides the basis for developing geotechnical domains and analyses inputs. Directly linked to the design confidence and reliability. Linked to the project life cycle: Conceptual to Operations. For mining projects there is a link with the declaration of resources and reserves. 10 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Data Reliability Data reliability is a state that exists when data is sufficiently complete and error free to be convincing for its purpose and context. This term is often used interchangeably with data confidence in my view this isn t strictly correct 11 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Data Confidence There are formal statistical definitions for confidence; In simple terms we want to know: How likely is it that a chosen value for a parameter falls within a certain interval; or How likely is a fault or geological boundary to fall within a specified distance of the interpreted position. Is often described in qualitative terms i.e. low, medium and high; Can be described using confidence levels e.g. 80% or 95%. 12 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Data Confidence Confidence interval can be determined using: Mean Standard deviation Number of samples Normal or Student-T distribution look up tables 13 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Data Confidence We are interested in confidence of: Rock mass and defect strength properties; Geological and domain boundaries; Major structural model; Hydrogeological model groundwater levels, pore water pressures; Geotechnical model; and Design domains. 14 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Confidence Levels What confidence level is needed for geotechnical models? Must be commensurate with the level of design/study. Basing complicated / high level analyses on crude / simple geotechnical model does not make sense. Design reliability is governed by confidence level of the geotechnical model and its components. Geotechnical data and design fall under Modifying Factors for JORC (2012) - requires: Confidence of the MF be considered in the conversion of Mineral Resources to Ore Reserves Increasing level of confidence is required 15 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Slope Design Confidence (Steffen, 1997) Category 1 - Proven slope angles Category 2 Probable slope angles Category 3 Possible slope angles Geotechnical investigations carried out to a feasibility study standard. In essence designs should have a minimum confidence level of 85% requiring: continuity of stratigraphy and lithological units confirmed in space through adequate intersections; detailed structural mapping of rock fabric is implied; strength characteristics of structural features and the rock mass through appropriate testing; ground water pressures have been measured. Equates to a design based on information that allows the following: reasonable assumptions on continuity of stratigraphy and lithological units; some structural mapping has been carried out and all major features and joint sets should be identified; limited rock testing for physical properties of the in-situ rock and defects has been carried out; preliminary groundwater analysis; enough information gained to conduct simplified design models with sensitivities. Equates to an inferred design using limited geotechnical investigations. Typical slope angles will be based on experience verified with rock mass ratings and some inference to geological conditions within the affected rock mass. 16 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Classifying Geotechnical Models Haile (2004) proposed a framework for classifying geotechnical models based on the structure of resource-reserve reporting codes. To avoid confusion with resource and reserve classifications, the following terms were introduced: Implied: i.e. with a low level of reliability, with only global estimates of geotechnical characteristics being available. Qualified: i.e. geotechnical model has a reasonable level of confidence. Justified: i.e. with a high level of confidence in the intrinsic spatial variability of geotechnical characteristics. Verified: i.e. based on in situ knowledge of the rock mass, which provides a reliable model of the intrinsic variability of geotechnical characteristics. 17 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Classification of mining projects (Haile, 2004) Data type Requirements Implied (inferred) Qualified (indicated) Justified (measured) Verified Geotechnical model has a low level of reliability. Based on global estimates of geotechnical characteristics. Will enable only a limited scope of analysis, and development of only conceptual level, mine-wide design parameters. Variability or uncertainty in the geotechnical model could have a significant impact on the project economic viability. Geotechnical model has a reasonable level of confidence. Provides a broad indication of the intrinsic spatial variability of the geotechnical characteristics. A reasonable scope of analysis could be applied, which broadly defines geotechnical domains, enabling the development of reasonably reliable, domain-specific design parameters. Variability or uncertainty in the geotechnical model could have a moderate impact on the economic viability of the project. Geotechnical model has a high level of confidence. Provides a good indication in the intrinsic spatial variability of the geotechnical characteristics. A comprehensive scope of analysis could be applied to well-defined geotechnical domains enabling the development of domain-specific mine design parameters. Variability or uncertainty in the geotechnical model would not significantly affect the economic viability of the project. Geotechnical model is based on in situ knowledge of the rock mass. Provides a reliable model of the intrinsic variability of geotechnical characteristics. Performance of the recommended design parameters have been verified through historical experience from neighbouring excavations and/or interim staged pit slopes. The design has been demonstrated to be practical and achievable. Variability or uncertainty in the geotechnical model would not adversely affect either the operational or economic viability of the project. 18 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Data confidence (Read & Stacey, 2009) Project Stage Project Level Status Conceptual Prefeasibility Feasibility Design and Construction Operations Geotechnical Level Status Level 1 Level 2 Level 3 Level 4 Level 5 Geotechnical characterisation Pertinent regional information Assessment and compilation of initial mine scale geotechnical data Ongoing assessment and compilation of all new mine scale geotechnical data Refinement of geotechnical database and 3D model Ongoing maintenance of geotechnical database and 3D model Target Levels of Data Confidence for each Model Geology >50% 50 70% 65 85% 80 90% >90% Structural >20% 40 50% 45 70% 60 75% >75% Hydrogeological >20% 30 50% 40 65% 60 75% >75% Rock Mass >30% 40 65% 60 75% 70 80% >80% Geotechnical >30% 40 60% 50 75% 65 85% >80% 19 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Uncertainty Natural uncertainty - properties vary Spatial uncertainty properties vary in space Temporal uncertainty properties vary over time Errors result in uncertainty Uncertainty results in errors Interpretation uncertainty Things we don t know lack of knowledge or data 20 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Geotechnical model and uncertainty When developing the GM the following should be considered: Is there sufficient data to capture natural variations in the rock mass? Is the spread of data sufficient to adequately define boundaries between different rock units? Is the data collection consistent using industry accepted practices? Quality of data collection - what QA/QC processes are used? Is there bias (direction of drilling, sampling stronger materials) in the data? How good is the laboratory testing programme (quantity and quality)? How is the data managed? How good is the data interpretation? 21 Geotechnical Models & Data Confidence in Mining Geotechnical Design
How much data? This question regularly arises during geotechnical studies. Geological model Generally determined by resource reporting requirements. Major structural model - unlikely to include the rock mass fabric. Hydrogeological model Development often driven by mining considerations - dewatering and process water Leads to the development of large scale hydrogeological models that do not adequately cater for near mine impacts e.g. the groundwater level behind a slope. The rock mass and structural fabric models are often developed as part of the geotechnical data collection programme. 22 Geotechnical Models & Data Confidence in Mining Geotechnical Design
DMP Guidelines Geotechnical guidelines for UG mines (1997) - suggests that the appropriate geotechnical data are collected from a representative number of cored bore holes, preferably oriented. Geotechnical guidelines for OP mines (1999) - outlines the need to collect geotechnical data that is consistent with type, size and life of the open pit. Stage of mine development Suggested percentage of geotechnically logged holes Prefeasibility study 25-50 % Feasibility study 50-100 % Operating mine 25-75 % 23 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Industry experience Haines et al (2006) considered geotechnical data in terms of geotechnical risks - need to reduce risk by improved understanding of the geotechnical environment. Summarised their experience of geotechnical logging conducted for various studies (Scoping FS), expressed as a proportion of resource holes drilled. Generally there is a doubling in the percentage of geotechnical holes as the study progresses. Stage of Study Geotechnical holes to total resource drilling Conceptual Engineering (Scoping) 1.6-4.9% (2.8%) Pre-feasibility (Advanced Scoping) 4.0-10.6% (6.6%) Feasibility 5.0-24.0% (11.9%) 24 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Industry experience (Dunn et al, 2011) Classification Geotechnical guidelines Newmont JORC Resource JORC Reserve Stage Logging requirements Laboratory testing (per major lithology) Potential Economic Mineralisation Non Reserve Mineralisation Reserve Conversion - - 1 Basic: 75-100% - Inferred - 2 Indicated Probable 3/4 Basic: 75-100% RMC: 30-50% Structural: 20% (min) Basic: 75-100% RMC: 30-50% Structural: 20% (min) Dedicated: as required 5*UCS, UTS & elastic properties Defect strength per major sets Additional testing depending on variability Triaxial if required Grade Control Measured Proven 5 & Operations As required As required Notes: Hydrogeology and specific testing (stress measurements, raise bore index, soil Atterberg limits etc) to be conducted as required. Dedicated geotechnical holes required to address data gaps and for infrastructure. Basic - RQD, fracture frequency, field strength estimates, weathering and alteration. RMC detailed descriptions of discontinuities and the collection of parameters for rock mass characterisation systems (Q and RMR). Structural - Orientated core and the collection of detailed data on individual discontinuities. 25 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Data quality and uncertainty Data collection to minimise uncertainty Critical step in the minimisation of uncertainty Includes the field collection as well as laboratory testing Often limited by the resources available motivate for additional resources or focus on critical aspects Sometimes you have to proceed with limited data document assumptions and limitations develop contingencies. Hadjigoergiou (2012) provides a good overview of shortcomings in data collection and how data can be more effectively used in solving geotechnical problems. 26 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Data quality and uncertainty Understand what data is needed and the goals of the program; Data density & distribution - define geotech zones & critical structures; Well defined data collection procedures & adequately trained staff; Implement data collection quality control procedures; Implement sound sampling procedures; Use accepted testing procedures & certified laboratories; Use statistical methods to define minimum number of samples; Develop statistical descriptions & distributions for all parameters; Ensure that data is managed and stored databases; Use visualization tools to view data in 3D Better understanding of spatial distribution of data; Improved interpretation. 27 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Data reliability tools available Read (2013) discussed methods that can be used to assess parameter uncertainty and model uncertainty. Similarly a number of other authors (Fillion and Hadjigeorgiou, 2013; Thomas, 2013) have recently evaluated the application various statistical methods in assessing geotechnical laboratory testing data and for comparing data sets respectyively. 28 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Parameter uncertainty Measured with statistical and probabilistic methods. Commonly use: Expected value (Mean) Standard deviation (SD) Coefficient of variation (CoV) SD/Mean Customary to accept the expected (average) value CoV is a subjective measure of performance (including assessment of the integrity of all data sources) CoV <10% low variability; Cov >30% - high variability 29 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Parameter uncertainty Read (2013): Coefficient of variation is a valuable screening mechanism when making decisions about the level of confidence in a selected design parameter; It is subjective and does not provide a numerical measure of the reliability of the data. Bayesian approach (Harr, 1996) could be applied to estimate the expected value of the reliability of a dataset. Uses a simple spreadsheet format - can be applied to any set of geotechnical data such as rock mass and hydrogeological parameters; Particularly useful for evaluating laboratory testing data sets. 30 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Parameter uncertainty Expected value of reliability E[R] E[R] = (S + 1) / (S + F + 2) Where: S = the number of successes and F is the number of failures in N trials. 31 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Parameter Uncertainty (UCS example) UCS (MPa) Strata No AVE STD CV S F E[R] Granitic 6 156 34 22% 3 3 50% Mafic 14 143 42 29% 8 6 56% Meta-Sedimentary 26 130 60 46% 17 9 64% Mineralisation 30 115 41 36% 15 15 50% Quartz 1 169 - - 1 0 67% Ultramafic 4 75 47 63% 2 2 50% UCS (MPa) - Used (UCS<50 for Meta-Sedimentary and Ultramafic were not considered) Strata No AVE STD CV S F E[R] Granitic 6 156 34 22% 3 3 50% Mafic 14 143 42 29% 8 6 56% Meta-Sedimentary 22 150 42 28% 11 11 50% Mineralisation 30 115 41 36% 15 15 50% Quartz 1 169 - - 1 0 67% Ultramafic 3 91 42 46% 1 2 40% UCS (MPa) - Used (UCS<50 for Meta-Sedimentary and Ultramafic were not considered) Strata No AVE STD CV S F E[R] Granitic 6 156 34 22% 3 3 50% Mafic 14 115 42 36% 10 4 69% Meta-Sedimentary 22 115 42 37% 18 4 79% Mineralisation 30 115 41 36% 15 15 50% Quartz 1 169 - - 1 0 67% Ultramafic 3 91 42 46% 1 2 40% 32 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Parameter Uncertainty (FES example) 33 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Parameter Uncertainty (RMR 89 example) 34 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Parameter Uncertainty Method can be applied to logging and rock mass classification data provided the data is arranged into equal length intervals. This approach has been applied to testing and logging data for various projects - found to be useful in determining properties for geotechnical domains and choosing design analyses input properties. Supplements the more common approach of using the mean, standard deviation and median for assigning values. Value in applying this approach to data sets for different study stages to assess whether the data reliability does in fact improve as the project progresses. 35 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Data reliability sample size Fillion and Hadjigeorgiou (2013) explored how small-sampling theory could be applied to assessing the results of laboratory testing from an operating mine. Showed that even if the number of specimens tested is higher than the minimum proposed by the ISRM suggested methods, the sample size was too small to obtain a reliable strength value for most of the rock domains. Minimum sample size obtained using the confidence interval approach is significantly influenced by the test results sequence. Can arise when samples from a relatively small zone are tested resulting in misleading statistics Highlights the need to have a reasonable spatial distribution of testing data so that the material variation is captured. 36 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Data reliability sample size 37 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Data reliability understanding data 38 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Data reliability sample size Understand how weathering and alteration influence rock strength and quality A representative number of samples is required for each weathering and alteration grade Sampling bias it is easier take samples of the good stuff but stability will be controlled by the weaker materials can be avoided by training and proper procedures Account for invalid tests over sample Account for anisotropy especially in folded strata 39 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Data reliability combining data Thomas (2013) provided an overview of various statistical tests that could be used to assess the similarities in properties between data sets from different areas. These methods can be applied to assess whether data can be combined or to assess if there are real differences between data sets or areas and is potentially a useful tool to assist in defining geotechnical domains. 40 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Data reliability combining data 41 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Spatial uncertainty Various statistical methods can be used to assess uncertainty and reliability of parameters. Read (2013) concluded that assessing model uncertainty in relation to the locations of through-going fault traces and the boundaries between lithologies and alteration units is significantly more complex and reviewed the following two solutions: Subjective assessments prepared by competent geologists, engineering geologists and geotechnical engineers, acting individually or as members of a review panel, as a means of quantifying the uncertainty associated with model geometries and boundaries. Generalised plurigaussian simulation to simulate lithologies and structures as a means of quantify the uncertainty associated with model geometries and boundaries. 42 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Spatial uncertainty 43 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Spatial Uncertainty Confidence Assessment 44 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Spatial uncertainty sensitivity analyses Sensitivity to boundary and fault positions can also be evaluated in design analyses 45 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Design Reliability (R) R = 1 POF This is different to the definition of Data Reliability However, they are related through Design Acceptance Criteria (FOS & POF) If you are able to reduce the POF you increase the Design Reliability Data Reliability and Model Confidence influence the POF, thus by reducing Uncertainty in your data and models you can reduce POF 46 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Acceptance criteria Higher FOS does not always mean a lower POF - it is important to reduce the spread 47 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Conclusions The geotechnical model forms the basis of geotechnical design. Complexity and confidence of geotechnical models should be matched to the project life cycle & resources / reserves reporting. There are useful qualitative guidelines relating the required level of geotechnical effort and data to the project lifecycle. In some cases confidence levels have been specified but there are difficulties in calculating confidence levels. There are a number of statistical tools that can be used to assess the uncertainty and reliability of laboratory testing and rock mass classification data. These methods can be used to determine the confidence level and reliability of some components of the geotechnical model such as intact and defect strength properties and the rock mass quality. 48 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Conclusions For assessing the confidence and reliability of major geological boundaries and fault traces we are reliant on the use of subjective assessments based on the judgment and opinion of experienced practitioners and this is likely to continue for some time. It is possible to reduce data uncertainty through well designed data collections programs. Data uncertainty and the associated confidence of the data and models can be dealt with in the analyses and design stage by: Appropriate acceptance criteria; Choice of appropriate input parameters; Sensitivity analyses; and Probabilistic analyses. 49 Geotechnical Models & Data Confidence in Mining Geotechnical Design
Acknowledgements Thank you to SRK for permission to publish this paper and attend this conference? 50 Geotechnical Models & Data Confidence in Mining Geotechnical Design