Search Documents

By J Ledgerwood

Namakwa Sands is a mineral processing company based on the West Coast of South Africa and is solely owned by Exxaro Resources Limited. Mineral ore is mined and processed, concentrated and both chemical and mechanically upgraded before being sent for electrostatic and magnetic separation units. The final high value mineral products report as Zircon, Rutile and Ilmenite.The recovery of non-magnetic material is largely dependent upon a number of factors, including but not limited to dew point, ambient temperature, feed rate, mineral size distribution, specific gravity and mineral content in head feed. One important limiting factor to recovery in typical dry mill operations is the particle size distribution of the feed. Variance in the particle size has catastrophic consequences for recovery. This is the reason for this study. Until recently large variances in non-magnetic production were viewed to be a function of machine operation. This view was later rejected with the feed size distribution determined as the major driver of low recoveries, for Namakwa Sands orebody 50 per cent variance in particle size distribution can be noticed within one day. This is because of the successive mining blending techniques employed.The predicament is that the two modes were so closely placed together that recovery of a single mode is almost impossible without the consequence of greatly reducing recovery. The second mode was found at 125 µm while the first was at 90 µm. Interestingly, for the mentioned feed type a cumulative size distribution of a unimodal feed type is almost exactly the same as for a bimodal size distribution. Thus a plot of cumulative size distribution will show no difference due to the modes being so close together. This is the reason why this phenomenon was not noticed before.The study also found that, using an high tension roll (HTR) or rare earth roll, a fine stream with a single mode at 90 µm versus a coarse stream with a single mode at 150 µm would yield completely different results interms of yields. Larger conductive particles are thrown from the roll while small particles loose charge faster (entrainment). Results in terms of the bimodal size distribution show that even though 25 per cent of the feed type is greater than 150 µm the yields followed a similar pattern to that of the coarse particles rather than that of the majority (75 per cent) fine particles. This confirms what was published by M Ziemski and P N Holtham in 2005 on charge decay rates; they mentioned that particle bed effects played a major role in recovery.

Oct 5, 2011

By J Ortiz

In almost all copper mining companies, mine planning is done considering different time frames, leading to long-term (life-of-mine) (LOM), medium-term (first three years) and short-term (first year, monthly, weekly, and up to daily plans). Shorter term plans are usually associated with more abundant information, hence models with lower uncertainty. However, they capture more of the variability found on the geological attributes. In long-term models, usually estimation methods tend to hide this variability, providing smoothed interpretations.Medium and long-term plans consider average values for the parameters and optimisation techniques are applied aimed at maximising the economic profit. Short-term plans, on the other hand, are made to comply with the medium-term promises, maximising the use of the available resources (equipment and processing capacities) and dealing with the variability not characterised in the previous planning stages. This generates an important gap between the short-term and the medium and long-term plans. A case study was made using an important oxide copper mine, where the variability found in the short-term plan has a large impact in the productive process, thus reducing the daily ore grade variability is an important goal. In this paper, this planning issue is addressed by creating medium-term mine plans using mixed integer programming which are compared using the output of a simulation of the daily extraction that is aimed at satisfying the daily constraints of the processing plant, the availability of mining equipment and the variability of ore grade. The result shows the expected gap between these mine planning processes, which suggests correcting some of the parameters used in future medium and long-term plans.

Sep 26, 2011

By D T. Ribeiro, J F. C L Costa, J C. Koppe

Drilling and sampling are responsible for most of the costs during mineral exploration, especially in multi-elements and polymetallic deposits, such as iron, manganese and bauxite deposits. On these deposits, often it is necessary to estimate multiple attributes at several granulometric fractions, ie distinct particle sizes. This paper investigates the effect of sampling at a close spacing grid only the head grades for the multiple chemical species in iron ore deposits. The remaining required data at the various grain sizes are sampled at a sparse grid. The derived geological model is checked for precision and accuracy comparing the output obtained against a model obtained using a complete isotopic data set including head grades and chemical analysis at each granulometric fraction analysed at all sampled positions. Among the several tested alternatives, the best procedure for dealing with the posed problem involved utilising the correlation coefficients between the grades at distinct fractions to obtain an isotopic database prior to interpolation. Cokriging was the method used, based on the intrinsic coregionalisation model. Estimates were obtained with the quality needed for the proposal of the study. Results showed that at least 50 per cent of the blocks were estimated with errors less than ten per cent relative to the grades obtained at the reference model, using only 15 per cent of the granulochemical data employed to obtain the reference model. The methodology also preserved the correlations between the attributes in their several granulometric fractions on the estimated model.

Sep 26, 2011

By E Ben-Awuah

Mine planning is the backbone of all mining enterprises and determines the profitability or otherwise of any mining investment. Providing high quality ore and tailings containment at the right time are the main drivers in oil sands mining. Recent environmental and regulatory requirements also makes waste management an integral part of mine planning in the oil sands industry (Directive 074). Due to the limitation of lease area, the pit phase advancement in oil sands mining is carried out simultaneously with the construction of tailings dykes in the mined out areas of the pit. These in-pit dykes are constructed to hold tailings that are produced during the processing of the oil sands ore after an initial external tailings facility. Most of the materials used in constructing these in-pit and ex-pit dykes come from the oil sands mining operation, making it necessary to have a plan for supplying the dyke material. The research problem here is categorised as: determining the order of extraction of ore, dyke material and waste to be removed from a predefined ultimate pit limit over the mine life that maximises the net present value of the operation û a strategic schedule; anddetermining the destination of dyke material that minimises construction cost depending on the construction requirements of the various dykes as per their designs û an operational schedule. There has been developed, implemented and verified a theoretical optimisation framework based on mixed integer goal programming (MIGP) to address the oil sands long-term production scheduling and waste management problem. The research introduced the development of multiple material types and multiple destination optimisation techniques using MIGP in the context of mine planning. This study presents an integration of mixed integer programming and goal programming in solving large-scale mine planning optimisation problems using clustering and pushback techniques.

Sep 26, 2011

By B Muller

The disposal of tailings has become a greater issue due to the changing economics of mineral extraction. The processing of lower grade deposits and tougher environmental regulations has led to the selection process for tailings disposal options becoming more complicated. Current selection methods are too time consuming or too easily manipulated for inexperienced users to use successfully. This paper examines a different method for the initial selection of a tailings disposal system, using a numerical system that was introduced into the mining method selection field. The numerical method system was compiled by reviewing previous work done on the positive and negative aspects of the tailings disposal methods used by the coal and metallurgical mining industries. Assessing the common trends and design aspects of each of the methods was undertaken and a ranking was designed to explain how well suited the tailings disposal methods are to a situation. The tool successfully analysed sites using common data to indicate methods for disposal that would have provided minimal technical problems, low environmental risk and been practically possible. The tool also proved successful in identifying critical areas that would need further investigation. This new system was tested on some disaster case studies from around the world in order to show that it could have successfully selected a different method for tailings disposal that would have prevented the disaster.

Sep 26, 2011

By A T. Bernardes

Using the cellular automata computational technique it was possible to simulate the generation and discharge of an iron ore pellet pile inside a direct iron reduction oven, as well as to calculate the reduction degree of the pellets as they leave the oven. It allows simulation of the variation of operational parameters and verify the results of those variations in the pellet reduction degree. The simulated data was validated in two instances: for the dynamics of the pellet flow inside the oven it was used in an acrylic bi-dimensional silo; and data from direct reduction iron literature was used for validation of the pellet reduction. The simulated results were considered acceptable to describe the movement of the iron pellet inside the direct reduction iron oven and to describe the iron pellet reduction direct in the oven.

Sep 26, 2011

By B J. Wilde

Evaluating the resources and reserves in a mineral deposit requires many interdependent subjective decisions. The decision of stationarity is arguably the most important. This decision includes: the choice of domains that subset the entire deposit, modelling the boundaries of each domain, and understanding and considering the nature of the contacts between domains. Reasonable choices are made in consideration of the geological context, the available data, the goals of the study and the available resources for the study. This paper reviews aspects of the decision of stationarity in modern resource assessment with a focus on quantifying uncertainty and using numerical/statistical tools to help the geomodeller. The use of volume/distance functions to quantify uncertainty in domain limits is particularly important and is presented in detail. The use of contact plots provides help with the continuity of grades across domain boundaries. Realisations are not assembled as a branching combinatorial of all possibilities, but as a set of realisations that provide a measure of uncertainty. This paper amounts to a definition of stationarity for modern resource assessment.

Sep 26, 2011

By J F. C L Costa, J C. Koppe, C Diedrich

Currently mineral resources evaluation is firstly expressed in terms of volume. Next the resources are transformed into tonnes by using an average bulk density applied to the whole deposit or, in some other circumstances, various different densities to distinct geological domains. Previous work in uranium, copper and nickel deposits used linear correlation between grades and bulk density to derive the density. Geophysics methods are also another good way to obtain density indirectly.These approaches do not take into account that two samples with the same grade but with different bulk densities can characterise different sample supports, leading to severe bias in grade and tonne estimates. It is important that the systematic measurements of bulk density are included into the estimation process, particularly when density is not correlated with the grades. Additionally, the extra time and costs involved in obtaining the bulk density data have made the mining industry disregard it.This study reports the impact in resource estimates by using a methodology that includes the bulk density variability. It involves the determination of a variability model for the bulk density and another model for the composite that was weighted by density as well as by its length. Next, these variables are estimated by kriging and the final grade estimates are obtained. The tonnages are obtained with the bulk density estimates. The conventional methodology was also applied and both estimates are compared to the reconciliation data from a major copper deposit. The results show a significant difference between the tonnage and grade estimates obtained from these two methodologies.The results show that considering density variability when compositing leads to a more accurate methodology resulting in better grade and tonnage estimates locally and globally.

Sep 26, 2011

This paper describes modern long-term mine scheduling and gives an example of its application. It then discusses the obstacles to the implementation of such schedules.

Sep 26, 2011

By E Nettleton, A Melkumyan

Identification of geology and mineralogy in situ on the mine face is essential if mining processes are to be automated for tasks such as exploration, grading and reconciliation. Hyperspectral data acquired from field based platforms are ideally suited to these tasks, and are set to revolutionise the automated mining industry. Changing conditions of solar illumination introduces variation into the hyperspectral signal, which is unrelated to geology, complicating analysis of data. Thus, algorithms must be able to deal with changing illumination conditions and operate in a consistent and probabilistic framework, which is a prerequisite for data fusion with other sensors. To produce accurate classification maps of geology, algorithms must be developed that are effective under different conditions of illumination.The efficacy of three algorithms to classify hyperspectral data for geology mapping on vertical mine faces was assessed. These algorithms are the spectral angle mapper (SAM), and two machine learning methods operating within a fully probabilistic Gaussian process (GP) framework û the squared exponential (SE) and the observation angle dependent (OAD) covariance functions (kernel). Two experiments were done to assess the classification performances. The first determined the performance of the algorithms by classifying hyperspectral data of known rock samples acquired using a field spectrometer. The GP-based methods had a similar performance, with average accuracies and F-scores of 97 and 88 per cent (GP-OAD) and 95 and 81 per cent (GP-SE), respectively. Average accuracies and F-scores for SAM were significantly smaller (80 and 40 per cent respectively). The algorithms were then applied to hyperspectral imagery acquired from a vertical mine face at different times of the day, under very different conditions of illumination. In contrast with experiment 1, SAM and GP-OAD gave results which were superior to GP-SE and were entirely consistent with geology mapped in the field. Compared with SAM, GP-OAD was less sensitive to the effects of shadow and produced classified images that were less noisy. The GP-SE method failed to delineate even gross changes in geology on the mine face. These results show conclusively that in order to use independent training data (spectral library) to classify imagery acquired under a different set of conditions, algorithms must take into account variability in incident illumination. Measures based on spectral distance are inferior to those based on spectral angle.The OAD kernel within the GP framework (GP-OAD) is superior to the other two methods because:it provides a probabilistic framework;it is insensitive to variability in illumination within and between imagery acquired under different conditions; andit can identify and map geological zones on a vertical mine face using an independent spectral library, without a priori knowledge.

Sep 26, 2011

The JORC Code contains a mandatory classification system for Mineral Resource and Ore Reserve estimates for public reporting. The classification is a statement of the Competent PersonÆs confidence in the estimate and an attempt to convey that confidence to the various users of the resource estimate. In general the classification process is either largely subjective, relying on the Competent PersonÆs experience, or relies on statistical measures derived from the estimation process, or is derived from an estimated level of confidence in the outcome. However, the statistical measures may not necessarily relate to the overall confidence in the resource or reserve which must take account of the confidence in the geological inputs to the estimation.In addition to the traditional deterministic interpretation and estimations adopted by geologists, semi-quantitative techniques are now available to improve the estimation of confidence while continuing to involve the Competent PersonÆs experience in the process. Methods for quantifying confidence in mineral resource estimates for advanced projects and operating mines have also been developed.Classification is undertaken primarily as a communication tool between the estimators and the other stakeholders in the process. A brief examination of the stakeholdersÆ requirements is included.

Sep 26, 2011

By R J. Hardy

The aim of this study is to examine the volumetric and mass movement characteristics of load and haul operations using statistical modelling of relatively large samples of bucket loads that are transported as small samples in the form of truck payloads. Generally, the factors influencing bucket load variability are random, difficult to predict and to control. In the process of loading trucks the underlying variability of individual bucket loads will determine the variability of truck payloads. Bucket loads of material prepared by drilling and blasting for loading into trucks are fundamental quantities in open pit mine production. A study of payload data from a large surface mining operation indicates that bucket loads and truck payloads that are a small sample of bucket loads are normally distributed. Truck payloads have a reduced dispersion compared with bucket loads that is an inverse 0.5 power function of the number of passes per truck payload. Comparison of statistical modelling results with the æ10/10/20Æ policy promoted by Caterpillar Global Mining Inc, indicate that compliance with the æ10/10Æ part of the policy, applying to truck overloading is facilitated, as the coefficient of variation of truck payloads decreases, as a result of the increasing number of loading passes. Compliance with the more stringent æ20Æ part of the policy, relating to safe steering and braking is also facilitated, albeit at lower coefficient of variation for similar reasons.

Sep 26, 2011

By D M. Silva, J C. Alves

According to the literature, material transportation represents 50 per cent of the operating costs for an open pit mine. In this context, this paper proposes a multistage approach for dynamic truck dispatching in real open pit mine environments, implemented in the commercial package SmartMine®, marketed by Devex SA. The first stage of this approach defines the optimal number of trucks that maximises the tonnage production by means of a robust linear programming model, which considers the operational constraints of a real mining process. The second stage uses a dynamic dispatching heuristic joining computational simulation and multicriterion optimisation techniques for decision making for truck dispatching. Another relevant contribution of this work is the publication of realistic instances collected from a dispatching system managing an iron mine, containing information about the hauling and loading fleet, the equipment situation when the working shift starts, cycle times, the time line of special occurrences (such as equipment stoppages and loading equipment reallocation in order to satisfy the mine planning) and the tonnage produced. These instances were generated with the purpose of submitting the method to the test of a real situation that considers the uncertainties existing in the process studied. The proposed approach was validated by means of a discrete event simulation model applied to the previously designated instances. Using this multistage approach as a mechanism for dispatch decision making, it is shown that the haulage fleet productivity increased when compared with the real values reported from the instances. These results are important because they show that the proposed approach generates efficient dispatch decisions for trucks operating in a real environment, reducing operating costs, increasing production and making the mine more competitive.

Sep 26, 2011

By W Assibey-Bonsu

The paper provides a brief review of the multivariate uniform conditioning and the localised multivariate uniform conditioning techniques and presents a case study based on a porphyry copper gold deposit in Peru. Two other important scenarios have been investigated and will be particularly emphasised: The recovered selective mining unit (SMU)s depend on a combination of copper and gold content, which can be expressed by means of ænet smelter returnsÆ (NSR). Two approaches are compared. Firstly, the application of a composite cut-off on NSR on the joint distribution of each grade and the application of uniform conditioning to three variables is investigated, ie for the first approach NSR is used as the main variable. The second approach considers copper as the main variable. In the case of multiple domains the procedure used to apply the technique to process mixed panels is clarified and recommendations are made.

Sep 26, 2011

By M R. Asef

Although generation of environmentally harmful hydrogen sulfide is considered to be related to sour gas fields, sweet gas reservoirs and waste fluids, travertine springs and geothermal fields may also produce acid gas (H2S and CO2) under certain conditions. The potential parameters influencing such a possibility need to be fully investigated to evaluate the extent of the consequences and predict hydrogen sulfide generation accordingly. Reports on hydrogen sulfide production at different regions and fields were gathered worldwide for the cases when some of them did not have such a problem previously. In order to inspect different aspects of these case studies, some other regions with no hydrogen sulfide content were selected as control. In all cases water samples were taken carefully for chemical analysis in the laboratory. Besides, mineral compositions were characterised in some cases for better evaluation. In order to give a thorough prospect of hydrogen sulfide production, a geochemical model was prepared. In situ engineering geological conditions, especially pressure and temperature, were assigned in the model to investigate the possibility of H2S generation using PHREEQC software. The results of modelling revealed that although many similarities existed between mineral compositions of the cases, certain formations were particularly prone to hydrogen sulfide generation. For better evaluation, important exchangeable cation ratios such as potassium/sodium, calcium/magnesium and ion proportions like sulfate and chloride were calculated as trends of the formations. As expected, the model could successfully monitor a higher ratio of SO42- and Cl- in potential H2S generator formations. Wide variation in percentage of HCO3- was considered as a secondary observatory parameter for better estimation. In addition, it was noticed that higher anion content created an environment ready to absorb any cations and this makes H2S generation easier. Finally, using correlations between predictors and estimators for low depth and deep samples defined the degree of effectiveness for every sample.

Sep 26, 2011

Production scheduling of open pit mining is a key area as it deals with the management of operating cash flows of mining companies. The efficiency of long-term schedules is usually assessed with economic criteria such as net present value (NPV), discounted cash flow rate of return (DCFROR) and payback period, which are not sufficient on their own for revealing the complex picture of the decision-making process.This paper offers a thorough analysis of the most important contemporary criteria of long-term production scheduling, as way to provide the mining practice with a new approach to the decision-making process. The list of criteria includes technological, economic and probabilistic ones. A detailed case study is provided aiming to illustrate the applicability of the criteria for comparing the scheduling approaches regarding the economic interests of a mineÆs owner and contractor. The results obtained can be used for revealing new trends in the current methodology of open-pit production scheduling optimisations.

Sep 26, 2011

Large deposits present computational challenges to todayÆs commercial orebody modelling systems that are based on conventional block models. We describe a system that is specifically engineered to model and visualise large deposits. Central to the system is linear octree encoding of conventional block models and real-time conversion of three-dimensional (3D) meshes to linear octrees. We provide both methodological treatment and practical experience in implementing such a system. A commercial implementation is also presented to demonstrate the validity of the approach.

Sep 26, 2011

By J Zhou

This paper discusses the implementation of a mixed integer linear programming model for a coal preparation plant. The plant considered consists of primary and secondary heavy media cyclones, hydrocyclones for intermediate fines and froth flotation for the very fine material. The model is designed to optimise the plant settings and product flows to maximise revenues for a given specification for two product outputs. The model can deal with bonus/penalty aspects of the product pricing for products that exceed or fail to meet specification. We report on a model with two separate coal feeds with parallel circuits for the oversize, intermediate and fines streams. The genesis of this model is more than 30 years old but recent improvements in mixed integer programming solvers make a full mixed integer formulation practical. The original model began as an enumeration of cut points with arbitrary rules for product allocation. The product allocation was improved in a linear programming formulation by Gunn (1982) and Sarker (1991 and 1995). Current mixed integer programming capabilities make it possible to replace the enumeration aspects and solve directly as a mixed integer programming problem. The current implementation uses a general purpose modelling language to formulate the model. Open source and commercial mixed integer programming solvers are compared

Sep 26, 2011

By D Misra, R Ganguli, S Bandopadhyay

The objective of this paper is to assess the accuracy and applicability of recent state-of-art methods such as the machine learning algorithms (MLA) for mineral resource estimation. Conventional methods such as geostatistics have been predominantly used in the mining industry for this purpose with varying success. Recent advances in the use of MLA have provided a fresh approach to obtain improved accuracy in the estimation of mineral resources. Two MLA methods: the neural network (NN) and the support vector machine (SVM) have been used to estimate a polymetallic ore deposit in Alaska.The general characteristic of SVM and NN emphasises the fact that they can approximate any multivariate non-linear relation among variables. Although neural network models are able to capture the nonlinear spatial relationships that may be present in the data, they are usually difficult to optimise under sparse data settings. Of the various NN alternatives, despite their effectiveness, the model selection and estimation process is typically difficult as it involves solving complex integration and optimisation of parameters. In support vector modelling (SVM), estimating of parameters involves optimisation of a convex cost function. The working principle of SVM makes it robust against noisy and extreme value data. At the same time, it can capture the spatial distribution of ore grade more effectively with careful modelling and selection of SVM parameters.The outcome of the MLA methods were compared to those of the geostatistical ordinary kriging (OK) application to study their generalisation ability. The comparison was made using the following goodness of fit criteria: mean-squared error; mean absolute error; mean error; and coefficient of determination. The results indicated that the MLA methods may improve the predictability of ore deposits and thereby reduce the inherent risk in mineral resource estimation.

Sep 26, 2011

By A Melkumyan, F Ramos

This paper addresses the problem of prediction of grade variables over volumes (block estimation) for an autonomous mine. Resource estimation is usually concerned with the prediction of the quality of material over volumes (blocks in resource models). This problem is tackled by developing Bayesian block estimation with the Gaussian processes (GPs) framework. GPs are commonly used for nonparametric regression and classification and offer an elegant solution to deal with incomplete knowledge and information. A GP is defined by its mean and covariance function. The parameters of the covariance function can be learnt by maximising the marginal likelihood of the data. This offers a way of automating the covariance fitting process, while at the same time allowing for an assessment of how well the covariance function fits the data. Within the GP framework the calculation of the average grade over a volume can be performed using Bayesian quadrature, which treats an integral as a random variable. First, it is demonstrated that the block kriging equations in geostatistics are the same as Bayesian quadrature. Then, it is determined when closed form analytical solutions for the integrals are possible and quadrature approaches for approximations are briefly discussed.

Sep 26, 2011