Search Documents

By A. T. van Zyl, M. Mthenjane, M. J. Cole

Growing global concern over the impacts of climate change, attributable largely to fossil fuel energy sources, has led to the widely shared goal for a ‘just transition’ to cleaner energy and reduced dependence on carbon-based fuels. As the world’s 14th biggest CO2 emitter and being particularly vulnerable to the impacts of climate change, South Africa must embark on a just transition pathway. This paper reviews expected coal mine closures and associated community vulnerabilities and local governance challenges in South Africa. Decommissioning schedules for all coal-fired power stations and operating coal mines are plotted, and 69 mining host communities and 21 municipalities are mapped, classified, and described. Community socio-economic profiles are measured using a set of SDG indicators and census data and municipalities assessed through financial audits. Our research shows that five coal-fired power plants (8.9 GW) and 15 coal mines (29.5 Mt/a) will probably close by 2030, and a further four plants (14 GW) and 23 mines (106 Mt/a) by 2040. Thus, the shift to cleaner energy will likely occur without the premature closures implied by the just transition. The impact of mine closure on the 2.5 million residents of host communities will be significant, particularly as levels of income, employment, and education are already very low and many municipalities are in financial distress. The South African approach to the just transition needs to take local realities into account and the narrative must support an effective transition that does not undermine energy security and economic growth.

Jun 5, 2023

By K. Woodward, Y. Potvin, B. McFadyen, M. Grenon

Open stoping has become a popular mining method in hard rock mines, not only due to the safety of the method as a non-entry approach, but also because of the high extraction rate and low costs. At mine sites, stope performance is evaluated by calculating stope overbreak using the stability chart. However, limitations of the stability chart regarding the precision of the predictions, nonconsideration of factors such as the influence of blasting, and the exclusion of underbreak have led to non-optimal designs. The capabilities of today's computers have increased the amount of data being collected and the power of models being built. This article presents a step towards a new stope design approach where stope overbreak and underbreak are measured and georeferenced using octrees at an approximately cubic metre resolution and predicted using multivariate statistical models (partial least square, linear discriminant analysis, and random forest). Results show that overbreak and underbreak location along the design surface and their magnitude are predicted with good precision using a random forest model. These predictions are used to build the expected geometry of the open stope. The resolution of the data and the use of multivariate analysis has enabled the prediction of variation in stope performance along the design surface, going well beyond the simple qualitative per stope face prediction provided by a traditional stability chart approach.

Jun 5, 2023

By Wei Victor Liu, Na Zhang, Chengkai Fan, Bei Jiang

This study used artificial neural networks (ANNs) coupled with Gaussian mixture modeling (GMM) to analyze truck haulage data from openpit mines. The proposed weighted ensemble models performed better than benchmark models in predicting truck productivity, providing a new means of solving complex data and improving mine planning and decision-making.

Jun 1, 2023

By Juan J. Monsalve, AMAN SONI, Nino Ripepi, Richard Bishop

The design and strength of pillars are important parameters when implementing the room-and-pillar mining method in a limestone mine. The process becomes complex with the presence of karst voids, which are frequent phenomena in carbonate rock formations. This research continues a previous study aimed at studying the effect of karst voids on the strength of a particular limestone pillar. Using calculated pillar strength and estimated redistributed stresses, it was assessed that the pillar can counter more stress than currently imparted on it. Numerical modeling studies combining in situ stress determination and excavation sequencing were used to estimate the existing stresses on the pillar. Different possibilities of modification to existing pillar design were simulated using distinct element method (DEM) modeling to determine a better design.

Jun 1, 2023

By Irina Chernyshova, Erik Larsen, Hanumantha Rao Kota, Vladislav Slabov, Garima Jain

The mining industry faces pressure to provide metals for the green digital shift using sustainable technologies. This study evaluates biosurfactants as collectors for ultrafine and coarser particles of hematite and malachite, aiming to improve the flotation of ultrafine particles.

Jun 1, 2023

By Yueyue Ding, Yuanhui Li, Zhengrong Li, Pengfei Yu, Kunmeng Li, Zhipeng Xiong

The overburden responses in gently inclined orebody mining are different from those in horizontal orebody affected by the orebody dip angle. In particular, the mining-induced stress presents asymmetrical distribution characteristics spatially, resulting in stress concentration and large-scale pressure relief phenomenon in the surrounding rock. As a result, frequent ground pressure disasters such as roof collapse and barrier splitting occur frequently. Moreover, the ground pressure evolution law and the interaction between stopes in multistope mining are complex. The effective ground pressure control measures cannot be taken due to unclear understanding of the evolution of stress and displacement fields, justifying the investigation of the distribution and evolution of underground pressure for safe and efficient mining.

Jun 1, 2023

By Ihsan Berk Tulu, Haochen Zhao, Deniz Tuncay

The process of designing pillars and selecting support measures in underground coal mines relies highly on the accurate assessment of mining-induced loads and the load-bearing capacities of the supports. This study proposes a new method to estimate the percentage of load carried by the gob in underground coal mines using a new parameter that introduces the site-specific overburden geology.

Jun 1, 2023

By M J. Nimmo

extremely powerful tools for building predictive and generative models. ML can be used for building highly accurate regression and classification models. But without careful data science and statistical analysis, a regression or classification model could be highly biased and completely wrong – and we may not even realise. An example rock density estimation task will be used to illustrate the potential gains and possible pitfalls in using ML and AI in Mineral Resource estimation. What could possibly go wrong? Nothing! For most simple regression or classification problems, very little. For density estimation, using the global mean may be all that is possible, but the resource classification would need to reflect the uncertainty in global estimation of tonnage. Everything! From data collection and measurement, from selecting ML and AI techniques and designing the analysis workflow to our own cognitive bias. The things that could go wrong are vast and may include but not limited to asking the wrong question, not asking questions, data bias, measurement errors, insufficient data coverage, not collecting the right variables, not collecting enough observations, collecting clustered samples, unbalanced data, filtering out data without reason (outliers), clipping data (outliers), relying on automatic feature selection, including irrelevant variables, excluding relevant variables, geological interpretation, changing context, blindly following best practice, splitting small data sets, subsetting large data sets, using the wrong tool for the task, insufficient budget to complete the analysis, not allowing enough time for testing and experimentation (the science), focusing too much on building the model (the engineering), not learning from the data, our assumptions, our preconceptions, and our skill. This paper will explore the question of how training regression models to predict rock density are affected by gaps in the data (missing observations and missing variables). What could possibly go wrong?

May 24, 2023

By W Patton

Investment decisions in the mineral resources sector are made on the basis of an assessment of the economic potential for a mineral deposit. Given the supporting information for the location, scale, and quantity of the commodity of interest, miners and investors aim to make risk-informed decisions when considering further investment. The objective of resource classification is the synthesis of various types and quantities of geologic information used in resource model construction to assign confidence categories to the estimated mineral resources, which can then be used to support informed economic decisions. In this paper a risk-based classification scheme for iron ore mineral resources using conditional simulations is presented. The method involves categorisation of each block according to a confidence level, considering expected recoveries from a kriging model over mining production volumes. The realisations of the simulated model are used to assign risk-based confidence categories to zones of the ore deposit. One motivation for this work is to assess the sensitivity of the scheme to geologic uncertainties. Specifically, that which uncertainty in the layout of mineralising alterations and weathering overprints influence variabilities of associated rock properties and the likelihood of obtaining expected ore recoveries. It is shown these risks have material implications for a project’s economic potential. How the scheme can be extended to the grade control context and the value subsequent grade control activities is also quantified.

May 24, 2023

By L Bertoss, D Carvalho

In the last decade, the use of advanced implicit modelling software/algorithms in mineral resource workflows has become a best practice in the mining industry. In most cases, these workflows can easily replicate the level of detail of explicitly modelled geometries, even for very complex deposits. These techniques are flexible and can produce fast results since software is responsible for generating 3D surfaces and solids based on geological and interpretation data. Nevertheless, depending on how unique and complex some deposits are, intense manual editing and input are required to honour the geological context and produce the desired outcome. The current geological modelling workflow at George Fisher Mine (GFM) is based on the semi-automatic built-in tool that generates stratigraphic sequences (stacked lithology layers) based on logging and mapping data. This tool (stratigraphic sequence from Leapfrog Geo™) works for the majority of areas in the mine but fails to capture observed mineralisation pinch-outs in operational areas and at extrapolated boundaries of the deposit. These pinch-outs have the potential to negatively impact grade forecasts if not resolved manually. Using built-in software tools only, the alternative for a more accurate model is to generate each individual layer and adding tens of thousands of manual interpretation points to honour pinch-outs (based on the amount of existing drilling in the mine and number of layers). This approach would require an estimated minimum of two to four years of a full-time employee (FTE) to be exhaustively completed. This study presents an automated approach to generate pinch-out points using a set of Python scripts that only take a few minutes to run. The newly generated points are then combined into the commercial software workflow as interpretation points, honouring observed pinch-outs. The remarkable time difference between approaches removes chance of manual errors and allows mine geologists to spend valuable time in validation and testing of scenarios, ensuring they are more representative of the deposit’s geology.

May 24, 2023

By H Arvidson, V Chamberlain, R Marinho, B Parsons, T Rowl, M Noppé, M Mattera

With the introduction of CRIRSCO (2019) based mineral asset disclosure, Regulation S-K part 1300, 2019 (S-K1300) by the United States (US) Securities and Exchange Commission (SEC), applicable to companies listed on the New York Stock Exchange from 2021, Mineral Resource is now required to be reported Exclusive of Mineral Reserve, referred to here as ‘Exclusive Mineral Resource’ (EMR). EMR is equivalent to reporting Mineral Resource additional to Mineral Reserve, which has been permitted in jurisdictions outside the US for many years. Despite this, there are no practical guidelines in the literature or relevant reporting codes and rules for Competent/Qualified Persons to refer to, which creates a risk that disclosure of EMR estimates across the industry will be inconsistent and potentially ambiguous for analysts, investors, potential investors, and other stakeholders.With the introduction of CRIRSCO (2019) based mineral asset disclosure, Regulation S-K part 1300, 2019 (S-K1300) by the United States (US) Securities and Exchange Commission (SEC), applicable to companies listed on the New York Stock Exchange from 2021, Mineral Resource is now required to be reported Exclusive of Mineral Reserve, referred to here as ‘Exclusive Mineral Resource’ (EMR). EMR is equivalent to reporting Mineral Resource additional to Mineral Reserve, which has been permitted in jurisdictions outside the US for many years. Despite this, there are no practical guidelines in the literature or relevant reporting codes and rules for Competent/Qualified Persons to refer to, which creates a risk that disclosure of EMR estimates across the industry will be inconsistent and potentially ambiguous for analysts, investors, potential investors, and other stakeholders.

May 24, 2023

By I Sucholutsky, D M. First, D Mogilny, F Yusufali

Machine learning is creating value in all facets of the mining industry, from exploration to production. The authors provide an accessible, high-level introduction to artificial intelligence (AI), machine learning (ML), and deep learning (DL); the latter being widely recognised as one of the most powerful forms of ML. In particular, the authors will introduce deep learning models known as convolutional neural networks (CNNs), how they are applied, and the economic considerations necessary for determining when DL may be the right solution to de-risking complex block modelling problems. The authors present preliminary results from the mineral resource modelling study of the Jundee orogenic gold deposit, Yandal belt/gold province, Western Australia. The primary goal was to identify the direction and location of the narrow gold veins more accurately and demonstrate how non-linearly correlated elements are used as direct inputs to the resource model to assist with the target element’s grade prediction. This demonstrates that: (1) existing techniques for finding correlations between assayed elements do not adequately reflect the complex geology of the asset, (2) nonlinear correlations that are difficult to model as simple mathematical functions are representative of geological patterns in a deposit, and (3) non-linearly correlated assayed data, fed as inputs, increase the performance of the resource model as reconciled through blind tests. To conclude, the authors hypothesise that the patterns represented by the DL block models may be revealing the results of overprinting geological processes that generated mineral deposits. For example, the primary hydrothermal processes that deposited metals created depositional patterns that become particularly complex as a result of being overprinted, in part or in whole, by secondary physicochemical processes. This may explain why non-linear geochemical relationships have the capacity to generate more accurate block models of mineral deposits.

May 24, 2023

By A A. Latscha, D O’Connor

Increasing orebody complexity, restrictions in ground access, and longer lead times for disturbance approvals, have generated the need for the Resource Development Team within Rio Tinto Iron Ore (RTIO), to explore optimised drill design techniques that deliver resource conversion value while maximising program outcomes. In response to this challenge, the Team initiated a case study to produce an AI assisted/designed drill program using a non-machine learning AI. The case study involved the assessment and implementation of an AI generated drill program at one of RTIO’s Pilbara iron ore deposits, located in the Hamersley Province of Western Australia. The Mineral Resource is currently classified as 100 per cent Inferred. Access constraints inherent to topography, and the presence of exclusions zones have prevented adequate drilling coverage, resulting in lower resource confidence and a poor definition of the orebody extents and geometry. The drill campaign, planned in Q1 2023, needs to consider access constraints, complex topology and the limitations related to additional ground clearance, while addressing the requirement of improving orebody knowledge. The drill plan had to address two objectives. The first and main aim of drilling was to upgrade the Resource Classification to allow for possible conversion to Ore Reserve. The input for this objective was a volumetric representation of the areas currently not meeting pre-defined quantitative Resource Classification upgrade criteria. The second objective was to better define the orebody extents and geometry. For the second objective, the input to the drilling optimiser was a volumetric representation (blocks) of the areas of high uncertainty in the mineralisation/waste boundary, which were defined using an indicator conditional simulation technique. The drilling optimiser output was a range of compliant drill plans. A set of KPIs based on economic and orebody knowledge criteria were used to assess the plans and to select the most optimal drill design to be drilled in 2023. The geology and management teams were able to make better informed, objective decisions using the trade-off between cost and orebody knowledge gained. As part of the assessment, the AI generated plans were compared to our current planning practises. The AI assisted plans, when gauged against the project KPIs, showed clear and measurable improvements in both meeting project objectives and in managing operational risk.

May 24, 2023

By D Clarke, I Minniakhmetov

This paper describes the application of the SBRE framework to model the Olympic Dam deposit, one of the largest copper and uranium deposits in the world. Geostatistical simulations are the best practice for quantifying uncertainty in the mineral value chain and estimating resources. However, these simulations are currently undertaken on an ad hoc basis in the mining industry due to the high requirements in computing resources and capabilities in advanced geostatistics. The SBRE project provides a standard framework to test, build, and run conditional simulation workflows, which has been applied to the Olympic Dam deposit. Conventional geostatistical methods face challenges when modelling the Olympic Dam deposit, such as non-stationarity, non-linear correlations, and a large number of grid blocks to simulate. To address these challenges, a combination of methods was used, including projection pursuit multivariate transform, trend-residual decorrelation, pluri-Gaussian simulations, and turning-bands simulations. Various tools and applications were used to build the model, including in-house developed applications, Autovariogram software (Varify), and the RMSP python library. All computations were accomplished on the AWS cloud using the EC2-Batch service. The final set of simulations was validated using various methods, such as grade-tonnage curves, histograms, swath plots, and variogram reproductions, as well as validation using historical samples back-testing. The study highlights the importance of an automated, modular, and scalable framework in the development of conditional simulation workflows and describes the learning gained from trials of different combinations of methods and software

May 24, 2023

By C Artica

Two quantitative methods for Mineral Resource classification have been applied to a copper skarn deposit beneath a large open pit that is mining a world-class porphyry complex. A drill hole spacing study (DHSS) using the single block kriging method (SBK) and a conditional simulation method (CS) were applied to the skarn deposit and the results compared. Inputs to both processes included the proposed mining rate and stope designs to estimate the likelihood of achieving grade and metal production targets. Mineral Resource classification of the deposit considers the relative precision of tons, grade and metal predictions over quarterly and annual production periods. These relative precision estimates are calculated at 90 per cent confidence interval (CI). The SBK method represented the production period as a single large block. Kriging estimates were performed in the large block for various synthetic drill hole grids, and the estimation errors derived. The error was assumed to be normally distributed due to the large block size and quantity of samples, allowing CIs to be deduced. The CS method followed a conventional sequential Gaussian simulation workflow where grades were simulated with a 5 ft × 5 ft × 5 ft grid node spacing inside mineralisation domains. Each realisation was then averaged up to monthly production volumes represented by a combination of designed stopes, and the coefficient of variation (CV) for the averaged grade was calculated. CI on the relative precision of grade predictions were then deduced from the CVs. The two methods produced similar results; despite being calculated independently. CS results were used to report uncertainty in the proposed life-of-mine plan and highlight areas for additional drilling. SBK results were used as the basis for Mineral Resource classification as this method has been previously applied to the skarn deposit.

May 24, 2023

By H Arvidson, V Chamberlain, J Joughin, N Pollock, F Cessford, T Flitton, T Rowl

Environmental, social and governance issues (ESG) have become a defining feature in the marketplace to differentiate preferred investments. With the sustainability commitment and reporting landscape anticipated to evolve exponentially, it is now imperative to rapidly develop the approach taken by mineral companies to adequately identify and address ESG risks and opportunities in the reporting of Mineral Resource and Mineral Reserve (MRMR). MRMR cannot be effectively reported as technically and financially assured and realistically executable, if ESG is not integrated into the strategy, business model, input assumptions and process that defines and drives robust MRMR public reporting. Regulatory reporting codes across the globe are presently in various stages of being updated and modified to ensure ESG factors are appropriately covered as a key part of diligent and transparent disclosure. Security exchanges are making it clear that companies must not omit ESG risks from reports to investors. This paper discusses associated challenges for MRMR reporting and suggests avenues for improvement. Key themes include: the importance of procedure rather than prescription and checklists; how ESG criteria impact Modifying Factors and determination of Reasonable Prospects for Economic Extraction (RPEE); the need for financial models to account for material ESG commitments; the uniqueness of each mineral asset; the need for education, upskilling and accreditation of both ESG experts about MRMR reporting and equally Competent Persons in ESG matters; and the need for clear and balanced reports that meet the needs of end users. (RPEE is adopted here for simplicity and consistency, recognising that some reporting codes still include ‘Eventual’ in the term (ie RPEEE). Whilst RPEE and RPEEE are subtly different the intent is to point to the process of assessing ESG factors in determining reasonable prospects.) The principles of transparency, materiality and competence remain paramount when considering the impact of ESG matters on MRMR reporting, and this necessitates an inclusive, cohesive, and integrated approach to assessing ESG issues to ensure the necessary expertise and diligence is applied.

May 24, 2023

By S Dunham

For decades we’ve been focusing on reconciliation as a tool to validate mineral resource estimates. And for decades we’ve been misleading ourselves. The concept of reconciliation is simple. Predict, measure, compare and assess the quality of the prediction. This four-step process reflects the ideas of statistical process control and the plan-do-check-act cycle of Shewhart and Deming fame and our intuitive understanding of using system feedback to improve performance. In an ideal world where data is robust and dense this type of feedback loop can be useful. In the mining industry however, there are some fundamental flaws with the conceptual framework. At almost every stage of the process from pre-estimation through to final production the mining system is plagued by material challenges with data and measurement quality and quantity, system delays, and losses from a supposedly closed system. These issues can and do impact on the benefits sought from reconciliation. Instead of providing feedback leading to improved estimation, reconciliation can result in poor decisions, distracting investigations, and exacerbate the noise in an already noisy system. Before we can rely on reconciliation as a management tool, we need to address the multiple challenges present in our everyday practices. Some of the most critical issues are addressed in this paper.

May 24, 2023

By M Bond, R R. Hargreaves, G W. Booth

There are numerous causes of speculative mineral resource and reserve inflation. Traditionally, these are linked to issues of data accuracy or misinterpretation, the use of overly optimistic estimation parameters or assumptions, an absence of independent expert verification, or the simple pressure to increase company value, covert or otherwise. Owing to its subjective nature, classification of resources and reserves remains an opaque process. Of necessity, Competent Persons rely on personal heuristic interpretations, ideally underpinned by appropriate estimation and classification experience across commodity and deposit types. Whereas independent verification of individual resource and reserve statements remains impractical, publicly available data can be interrogated for potential category inflation, particularly when viewed with exploration expenditure statistics included in annual company reports. Abrupt changes in category tonnages and market capitalisation without concomitant expenditure can aid in outlier detection. Consequently, a comparison of common exploration and resource reporting parameters offers a basis for risk assessing classification robustness, over time. In this introductory study, an analysis of globally listed companies was undertaken over a multi-year period, examining base and precious metal projects in both preproduction and operational asset classes. From these, a series of exploratory decision tree trigger points were generated against which target companies were anonymously identified and risk-ranked.

May 24, 2023

By D Corley, J Moore, M Grant, A, W R

Resource estimates are the corner stone of technical and investment decision-making. Prior to mining, resource estimation uncertainty has the greatest potential to lead to poor investment decisions, despite a significant component of resource estimation uncertainty being unknowable at this critical stage in the mining cycle. This presents a conundrum to the resource geologist in terms of risk evaluation and resource classification. Ahead of mining, there is normally considerable focus on drill hole spacing analysis to determine the ‘optimal’ drill hole spacing, taking grade and geological continuity into account; cost-benefit analyses balance improved resource definition against cost; the law of diminishing returns and the exponential cost increases acting together as spacing is reduced. Typically, a lower limit to the drill hole spacing is identified. After completing the resource drilling to an agreed spacing, resource estimation is then undertaken. At this point it is common to attempt to bracket the resource estimation uncertainty. In many cases conditional simulation is used, based on a modelled, but largely assumed, variogram model. It is also assumed that the histogram of the available drill hole sample data is representative of the in-ground mineralisation. The purpose of this paper is not to diminish the importance of drill hole spacing and simulation studies, but rather to illuminate an important aspect of estimation uncertainty that, although previously recognised, is typically overlooked – the question of how representative the available data set is in characterising the true (but unknown) distribution of mineralisation. To do this, we lift the lid on one of OceanaGold’s former operations, the mined-out Globe Progress Mine, by resurrecting a high quality, close-spaced, reverse circulation (RC) grade control data set. The Globe Progress Mine is within the West Coast Region of New Zealand, it was closed and transitioned to rehabilitation in 2016 and is now known as the Reefton Restoration Project. The exhaustive Globe Progress grade control data was used to repeatedly ‘redrill’ the deposit by extracting 35 m × 35 m spaced subsets from the original 5 m × 5 m spaced grade control data. Utilising closely spaced grade control data removed the need for assumptions regarding short range continuity, which are necessary with forward-looking analyses that are based upon broader spaced resource drilling. The extraction process used a nearest neighbourhood algorithm, repeatedly moving the origin in 5 mE, or 5 mN increments. Individual resource estimates were then completed for each of the extracted drill hole data sets (49 in total) as well as a grade control estimate based upon the exhaustive data set. Whilst the data unpinning each of the 49 estimates changed, the geological assumptions, variography, and modelling parameters remained constant. This approach was taken to isolate the impact of changing the input data. The 49 estimates were then compared against each other and the grade control estimate. The mean of all 49 sensitivity estimates was close to that of the grade control estimate in terms of contained gold, tonnes and grade. Whilst it is acknowledged that the grade control estimate itself is subject to some degree of estimation uncertainty, the close match between the average of the 49 estimates and the grade control estimate suggests that the resource estimation methodology is reasonable and appropriate. Whilst this comparison is important, the focus of this study is on the component of estimation uncertainty related to the underlying data, and this is reflected in the spread across the estimates. The spread across the 49 global estimates (highest to lowest) for this particular Mineral Resource Estimation Conference 2023 | Perth, Australia | 24–25 May 2023 142 case study was found to be significant (approximately 20 per cent in grade and metal) and that is attributable solely to the underlying data. This exercise quantifies a component of the estimation uncertainty that is inherent to all drill hole data and is distinct from the uncertainties associated with modelling methodology choices, sample and subsample quality and drill hole spacing-related interpolation uncertainty. Importantly, this uncertainty is unknowable prior to mining; we can only directly compare the resource drill hole data against grade control data after mining has taken place. For the case study, approximately 65 per cent of the estimates fell within 5 per cent of the grade control estimate for contained-gold, suggesting that for many projects the histogram of drill hole data is unlikely to differ noticeably from that of the in-ground resource. However, about 15 per cent of the estimates in this study differed by more than 7.5 per cent, suggesting that a not-insignificant proportion of estimates will be materially compromised. Whether or not the available drill hole data is representative comes down to the ‘luck of the draw’ and cannot be known at the time of resource estimation. Given the challenge of attempting to evaluate forward-looking estimation uncertainty, a component of which that can only be quantified retrospectively, what should we do as resource geologists? What are the implications for risk evaluation, resource classification and reconciliation? Furthermore, without rigorous post-mining data checks, resource geologists may conflate suboptimal modelling with the shortcomings of the underlying data. An example of this problem is discussed in the following section.

May 24, 2023

By P Dowd, A Adeli, C Xu, X Emery

Geometallurgical modelling is increasingly being incorporated into mineral resource modelling and estimation as a means of increasing efficiency, decreasing operating costs and reducing risk in mining operations. Traditional geostatistical workflows for geometallurgical modelling face several difficulties, in particular, related to the non-additivity of process response variables, to highly heterotopic data sets and/or to preferential sampling designs, all of which may explain why most attempts to estimate geometallurgical response variables from assay data and mineralogy are still based on statistics and machine learning. The aim of the work presented here is to demonstrate the applicability of geostatistics to estimate copper recovery from copper sulfide ores at the Prominent Hill Iron Oxide Copper-Gold (IOCG) deposit in Australia. As recovery is non-additive, the preferred workflow is the joint estimation of the masses of metal in the feed and in the concentrate, both of which are additive variables. While the mass of metal in the feed can be estimated from abundant online assay analyses, the mass of metal in the concentrate can only be estimated from a very limited number of laboratory-scale batch flotation tests. Traditional approaches to estimating the masses of metal in the feed and in the concentrate, such as simple and ordinary co-kriging, are described and discussed. A modified version of co-kriging is introduced, which incorporates linear relationships between the mean values of the input variables, in addition to a spatial correlation model. Co-kriging with related means is shown to outperform traditional simple and ordinary co-kriging in terms of precision, accuracy and consistency of the estimates, and offers a practical and efficient tool to deal with highly heterotopic and preferential sampling designs.

May 24, 2023