Search Documents

By F. Reyes, M. Hilden, G. Forbes, M. Yahyaei

Since AlphaGo defeated the best human player in the game of Go in 2016 - something thought to be an impossible achievement - machine learning, deep learning and reinforcement learning have caught the attention of scientists, the media and the general public alike. This recent development in machine learning (let us remember that neural networks have been available for decades now) opens new possibilities in the areas of modelling and process control. However, little has been reported in the minerals processing industry about these new possibilities and opportunities. This paper presents and introduces the topic of reinforcement learning and the different platforms that allow its implementation. The paper shows the consequences of using reinforcement learning for the specific area of process control and optimisation. As an example, the authors present the use of reinforcement learning for the control of a SAG mill closed with a pebble crusher using industrial data from an operation. A comparison with advanced model-based process control is provided, showing the benefits and shortcomings of reinforcement learning as a new control and optimisation technique. Keywords: Reinforcement learning, deep learning, process control, process optimisation

Jan 1, 2020

By C. T. O’Connor

Since the first IMPC took place in London in 1952 there have been many developments in our understanding of flotation, both in terms of chemical and physical factors influencing the process. Many of these developments have been captured in the proceedings of thirty congresses that have taken place since that first event, as well as in numerous journal papers and special volumes dedicated to flotation. This paper attempts to highlight the individual contributions which many eminent researchers and engineers have made over the past 70 years towards extending our understanding of, especially, the effects of chemical factors on flotation. Although the focus is mainly on more fundamental aspects, it seems that the ‘pull’ of practice has often preceded the ‘push’ from laboratory-based research. As the fundamental understanding of the role of chemical factors on flotation continues to be developed, the major challenge remains to transfer this knowledge effectively into the domain of operations of this profoundly complex process.

Jan 1, 2020

By O. A. Bascur

Ores are becoming extremely variable with mineralogy and hardness disturbing the grinding and flotation circuits. The current grinding and flotation sensors provide large amounts of data for process optimisation. Adding the right context and operational events augments operational knowledge for proactive actions to improve the performance of the grinding and flotation circuits. A data-driven strategy is needed to enable operations, maintenance, and business personnel to quickly and easily take corrective action when abnormal conditions occur. A digital plant template transforms data into operational insights in real-time. It uncovers the hidden, idle and downtime losses. By measuring and managing these unproductive times, people find new ways of avoiding them, improving the profitability of the plant. The information created by the real time analytics enables the calculation of recovery in real time and develops predictive analytics models to find the best operating condition based on the type of ore currently mined. The creation of new workflows and collaboration between mining and concentrator plant and the enterprise, including services providers, are enabled. A grade recovery model is used to identify the best operating conditions in real-time. An optimal Gaudin size distribution shape is used to find the grinding operating conditions that optimise the metal recovery in the rougher sections of the mineral processing plant. Sensors in the flotation circuits enable estimation of metal recovery and determine the optimal froth depth and aeration profiles, using an industrial air holdup flotation model. The implementation of a recovery/grind strategy with industrial examples in non-ferrous metal concentrator is presented in the paper. Keywords: Digital transformation, soft sensors, predictive analytics, grinding flotation optimisation, flotation bubbles, particle size, particle size distribution shape, froth aeration profiles, flotation air hold up model, recovery predictive analytics

Jan 1, 2020

By M. Le Roux, Q. P. Campbell, M. Fofana

The determination of densimetric information to predict coal washabilities is an important step in characterising the feed material, as well as the performance of operating plants. However, traditional methods like float-and-sink analysis are hazardous and time-consuming. This is especially true for finer particle size ranges, where very long settling times are required in a medium. De Beers Technologies recently developed a density analyser, known as the RhoVol, for use on diamondiferous kimberlite. Operating over size ranges from 25 mm down to 3 mm, it can determine the mass, volume, density, size and shape of all the particles in a given population and is thus able to provide information about the densimetric distribution on a single particle basis. A study was carried out to compare uncalibrated and calibrated RhoVol results to densimetric and size data obtained using traditional zinc chloride float-sink procedures, for three different coal samples in the size range 3 mm to 8 mm. The results indicate that the RhoVol can be used to provide simple, rapid and safe size-by-washability information for coal, provided proper pre-calibration is done, mainly to account for the effects of particle porosity in the density determination. Keywords: Coal washability determination, RhoVol, densimetric curve

Jan 1, 2020

By K. C. Corin, M. Tadie, R. M. Manenzhe

The near depletion of high-grade ores, among other factors, has necessitated the consideration of alternative PGM sources. Various factors in the metallurgical processing of ores result in valuable minerals being lost to the tailings, and hence, PGM tailings are a possible alternative PGM source. However, as the value residing in PGM tailings remains unquantified and uncharacterised, there is a challenge in knowing the recoverable fraction of value contained in PGM wastes. This paper outlines the in-situ techniques of oxygen uptake and EDTA extraction for their possible use for determining the recoverable value in PGM tailings. Both methods have shown efficiency in determining the weathering of sulphide minerals and sulphide-bearing ores. Keywords: PGM tailings, tailing retreatment, oxidation, oxygen decay, EDTA

Jan 1, 2020

By P. Ramsaywok, G. Legodi

Spiral concentrators are widely used in the coal industry treating fine (-1mm) coal. Their application was a limitation to the product quality until the introduction of low-cut spirals which allow a lower quality ash product to be produced, thus expanding the market. The Exxaro Grootegeluk 1 (GG1) plant has a two-stage traditional rougher-cleaner spiral circuit using conventional SX4 spirals to produce semi-soft coking coal product at 12.5% ash by mass. The overall plant product is 10.3% ash meaning that the coarser sized coal dense medium separation circuits must compensate. A low density cut-point spiral targetting 10.3% ash would thus be beneficial to the overall complex. The spirals were tested in the two-stage plant environment and in a pilot plant. This paper discusses the findings and compares a two stage four-turn low density cut and conventional spiral. The two spiral technologies were unable to sustain a 10.3% ash product. Pilot test work was also conducted on GG1 coal, demonstrating that an eight-turn low density cut-point spiral achieves the targetted 10.3% ash under controlled conditions. Keywords: Spiral, low-cut, coal

Jan 1, 2020

By D. Amariei, G. Bartolacci

Preconcentration of precious and base metal ore types is becoming more relevant with the discovery, the development and the operation of many lower grade deposits around the world. Ore preconcentration designates an operation consisting of sorting an ore to reduce the quantity of waste material sent to the concentrator. It can be seen as rejecting waste material as much as possible, without compromising the recovery, or as the recovery of as much valuable metal as possible into an optimum mass yield. Benefits of preconcentration are multiple: grinding energy savings, costs reduction, lower environmental footprint, extension of the reserves, etc. However, all of these benefits do not apply simultaneously and are not applicable to each ore type. Technologies employed for preconcentration are highly dependent on ore properties, especially on the particle size at which the gangue is liberated. Thus, from coarser to finer particle sizes, technologies such as screening, ore sorting, gravity separation (Dense Media Separation (DMS) and jigs) and coarse flotation could be used for the rejection of liberated gangue. The difficulty starts when a benchmark is needed to select a suitable technology, especially for low-grade gold deposits. This paper presents the challenges of comparing lab-scale technologies (DMS versus jig) for gold preconcentration. The experimental data will show to what extent comparing two lab-scale gravity technologies for gold preconcentration can be challenging. An attempt to evaluate the energy savings by assessing the hardness of the tailings after gravity separation will also be presented. Keywords: Challenges in gold preconcentration, gravity separation, jig, dense medium separation

Jan 1, 2020

By S. Polansky, J. Kolacz

The new sorting system has been developed to achieve efficient separation at low mineral content levels. The system employs high resolution X-ray sensors providing multi-energy analysis, thus giving maximal information about processed materials measurable by X-ray radiography. It makes it possible to analyse the internal structure of the particles and directly detect inclusion of minerals and metals like copper or gold. The paper describes sorting application examples for zinc, lead, copper and gold ore with low metal content below 0.1% of Cu. The sorting system still allows high efficiency analysis, quantification and removal of rock particles representing waste fraction with no metal or a negligible amount being uneconomic to process. This brings a huge advantage during pre-concentration, where almost no losses of valuable minerals in the waste fraction can be achieved. The presented sorting system has been developed by Comex and X-MINE EU Project Consortium. Keywords: X-ray sorting, ore sorting, XRF analysis, XRT analysis

Jan 1, 2020

By M. Khodadazadeh, K. G. van den Boogaart, T. Leißner, L. Tusa, J. Gutzmer, R. Gloaguen, M. Kern

Ore sorting is a technology that is increasingly used to process primary raw materials. Timeconsuming and expensive empirical state-of-the-art test work is carried out to assess whether the use of ore sorting for the enrichment of a particular ore makes technical and economic sense. With the innovative simulation-based approach presented here, it is possible to direct the selection of a suitable sensor based on quantitative mineralogical and textural data, thus avoiding much of the empirical studies. Required data can be collected quickly and cost-effectively using available methods of automated mineralogy. The obtained parameters such as mineral grain size distribution, modal mineralogy, mineral area and mineral density distribution have been utilised in this study to simulate the prospects of success of ore sorting, applying different types of sensors. Empirical tests with commercially available sensor systems have been conducted to experimentally validate the predictions of the simulations. The estimation of the target mineral grade can be further optimised by the use of machine learning algorithms for the integration of automated mineralogy data and sensor data. The approach can easily be adapted to other types of raw materials and thus has great potential to become a key technology for the optimisation of processing experiments. Keywords: Automated mineralogy, hyperspectral imaging, machine learning, ore sorting

Jan 1, 2020

By M. J. Latva-Kokko, CA Snyders, J. R. Peltola, M. K. Kostiainen

It is well known that carbon management is essential for minimising soluble gold losses in carbonin- leach (CIL) and carbon-in-pulp (CIP) circuits. Even so, carbon measurements remain predominantly a manual operation with a low amount of control and automation around CIL and CIP reactors. Carbon inventory, i.e. knowing the carbon concentration in each reactor, is crucial for good carbon management. CarbonSense, a novel in-situ measurement system, was therefore developed for continuous monitoring of the carbon concentration in each reactor and paves the way for more advanced control in gold operations. The carbon management and optimisation system is based on Outotec® HSC Sim dynamic process simulations where each process component is modelled in detail. Generated models are validated with real data from a particular operation. The simulator will allow for the carbon control sequence to be realistically tested and the efficiency optimised by controlling the feed and discharge pumps in a hazard- and risk-free process environment. Furthermore, by combining CarbonSense and Outotec® gold courier (online soluble gold measurements) continuous feed-forward and feedback information can be obtained. The optimiser will then define the best running parameters based on the cost function of gold production versus intrinsic factors such as utility costs and adsorption time. Eventually, the application can be installed parallel with a real process island and it can be continuously calibrated with real-time process data. The dynamic model will then create future predictions for the entire process island. The example case presents a CIL circuit, which is simulated using Outotec® HSC Chemistry Sim and operated through Outotec® ACT (Advanced Control Tool) application. Keywords: Carbon-in-leach, CIL, CarbonSense, carbon management, automated control, simulation

Jan 1, 2020

By E. G. Galvez, E. Mate, L. A. Cisternas, C. Cruz, L. Cortés, M. E. Osorio

Designing and optimising flotation circuits are critical for mineral beneficiation plants, and therefore several procedures have been published. The modelling of flotation circuits with the objective of design and optimisation is problematic in that different approaches are used for the modelling of flotation units and regrinding units. Usually, regrinding models are based on particles characterised by their size, while flotation models are based on species characterised by fast and slow classes. The relationship between particle size and liberation can be used, but it requires laboratory studies that are time-consuming and costs are high. It is therefore not surprising that few studies on flotation circuit design and optimisation include flotation units and regrinding units simultaneously. In this paper, it is proposed to model the regrinding stage using species characterised by fine-slow, medium-fast, and coarse-slow flotation species. In deep, the regrinding model is derived from a population balance approach based on flotation species. The fraction of these species at the feed and product of the grinding unit is determined by flotation tests. Then, based on these data, the model parameters can be determined by a regression estimation procedure. A local ore was used to validate the model. Keywords: Modelling, flotation circuit, grinding circuit, process design

Jan 1, 2020

By A. Chagnes

This paper presents a modelling tool that can be used to predict the behaviour of solvent extraction processes. In particular, this model allows calculating metal extraction efficiency as a function of pH, solvent flowrate, stripping flowrate and aqueous feed flowrate in the mixers-settlers. This model is applied to the recovery of cobalt from acidic chloride media by using CYANEX® 272 as extractant diluted in kerosene. Keywords: Cobalt, mixer-settler, modelling, solvent extraction

Jan 1, 2020

By T. Bui, T. Wallmach, S. B. Blancher, A. Back, M. Pierre, V. Delarue, A. Kanzari

This study investigates nickel laterite porosity and its impact on geometallurgical modelling. We performed a detailed mineralogical study on a set of Ni-laterite samples originating from New Caledonia. Samples were analysed after heavy liquid separation by automated SEM (QEMSCAN®). Laboratory and QEMSCAN® modelling results do not correlate well. Ideal physical mineral properties such as mineral densities used in QEMSCAN® modelling do not match the actual physical densities of minerals. The presence of empty or water-filled pores can represent important amounts in minerals and change the overall density of a particle. An adequate modelling of real density separation of minerals and particles should consider the water-filled pores. The first step towards a better simulation was to integrate pores larger than 10 μm in diameter filled with water (density of 1 g/cm3) in the QEMSCAN® database. The results approached, but did not match the physical laboratory results. The second step was to identify nanometer to micrometer sized pores, which could not be identified by QEMSCAN®. It was possible to identify nanometric pores in enstatite, Fe-oxides, serpentine, talc and quartz, and radial aggregates of fibrous Fe-oxyhydroxides with high resolution SEM (HR-SEM). The quantification of those micro-sized pores was performed on 2D HR-SEM images using image-processing techniques developed in Python to calculate the nanometer to micrometer scale pores in each mineral. Despite the limitation of converting 2D imaging into 3D objects, in our case pore sizes and porosities, we were able to calculate a mean value of porosity for each mineral. The integration of this porosity in the calculation of particle densities yielded a good match between the QEMSCAN® model and the laboratory experiment results. Despite the complexity of a strongly weathered material, such as Ni-laterites, a good knowledge of the ore makes it possible to build reliable processing models. Keywords: Mineralogy, geometallurgy, nickel laterites, SEM, QEMSCAN®, porosity, density

Jan 1, 2020

By K. -I. Higuchi, M. Mizutani, K. Haga, A. Shibayama, A. Batnasan

In recent years, the quality of iron ores has decreased due to the increase in the content of impurities, namely quartz, aluminum, phosphorus, and sulfur. It indicates that the removal of impurities from iron ore is an extremely challenging task. In particular, the high amount of phosphorus in iron ore could be a problem for energy consumption and CO2 emission in the steelmaking process. To reduce the impact of phosphorus, mainly two approaches: the utilisation of low-phosphorus iron ore, or dephosphorisation of high-phosphorus iron ore using mineral processing technology, have been suggested. In this paper, the removal of phosphorus from a low-grade high-phosphorus iron ore containing 0.1 mass% P and 57.5% Fe by alkaline roasting and wet grinding is discussed. The iron ore consists mainly of hematite, goethite, and quartz. Effects of sodium hydroxide concentration used as an alkaline source, roasting temperature, grinding time, and particle size on the dephosphorisation of highphosphorus iron ore were investigated. The dephosphorisation efficiency, iron loss, and mineralogical components of low-grade high-phosphorus iron ore under various experimental conditions were characterised by using XRD, SEM-EDS, and particle size distribution analysis. Results showed that the phosphorus content was lowered from 0.1 mass% to 0.034 mass%, while the iron grade reached 59.8 mass% from 57.5 mass% by alkaline roasting with 0.05 w/w NaOH at 300 °C for 2 h, followed by wet grinding in an attritor with water for 30 seconds at room temperature. Under the experimental conditions, the dephosphorisation rate was about 54.7%, and the solution obtained by wet grinding contained 22.7 mg/L of phosphorus and 21.5 mg/L of iron. The recovery rate of iron by this process reached 99.9%. Keywords: Iron ore, phosphorus, dephosphoriszation, alkaline roasting, wet grinding

Jan 1, 2020

By C. Bergmann, V. Ross

The concept of underground ore processing, in which valuable ore and waste are separated at a coarse particle size to improve the economics of operations and minimise the environmental impact, seems attractive from a sustainability perspective. Perceived benefits of such an operation include energy and cost savings due to a reduced hoisting load, a reduced footprint of surface operations, and using material for backfill that has not been contaminated by process chemicals. Progress in the South African mining industry in this area has however been below expectation, and opportunities for underground processing in deep, narrow reef operations were therefore explored by an Excel-based simulation with sensor-based sorting (SBS) as concentration technology. Various PGM mining scenarios incorporating factors such as the mining rate, depth of operation, and effects of dilution and particle size on the metallurgical efficiency, were compared. Whilst these initial results can be regarded as indicative only, they have shown that the economics of such an operation are driven by selective mining at significantly higher rates than achievable with conventional methods. Also, the cost of preparing, conveying and placing rock backfill will need to be lower than that of hoisting waste and depositing it on a surface dump. Another aspect to be addressed is whether ore sorting machines would be able to operate under harsh conditions underground and what provision would need to be made to ensure their efficient operation. Further investigations should focus on refining costs, incorporating environmental considerations in the modelling of economics and sustainability, and compare greenfields with brownfield scenarios. Keywords: Ore sorting, hoisting, backfill, mining rate, economics simulation

Jan 1, 2020

By J. Ojeda, R. Kuyvenhoven, M. Ziemski M

This paper focusses on the identification of operational opportunities in a 1100 tph pellet plant to reduce the specific energy consumption in its comminution circuit, which is designed to process magnetite ore and produce pellets to be used as raw material in the iron and steel industry. The assessment is carried out through a frequency analysis for three of the four processing lines of the plant, consisting of one ball mill each with an installed power of 4.5 MW and 280 tph of mill throughput design capacity. The methodology, as described in this paper, facilitates the visualisation and quantification of plant performance and provides a good first approach to prioritise alternatives for optimisation. In a pellet plant, grinding is a critical stage due to its high energy consumption and its major impact on the physical properties of the iron pellets, which quality depends to a large extent on grind size and Blaine index. From the evaluation that was performed it can be concluded that production line 2 has the lowest specific energy consumption of 15.1 kWh/t, while line 1 has the highest with an average of 16.6 kWh/t for the same type of ore. On the basis of the results and current operational practices, it is recommended to increase the average mill throughput from 240 tph to 265 tph, translating into specific energy reduction of 10%, without affecting the chemical and physical properties of the final product. The implementation of an HPGR for regrinding the magnetic concentrate is proposed to assure adequate specific surface area in case grinding coarser becomes an alternative, which would also be beneficial for the subsequent water recovery stages. Keywords: Iron ore pelletisation, grinding circuit, energy efficiency, particle size, frequency analysis

Jan 1, 2020

By S. Banisi, M. R. Yarahmadi, S. R. Mortazavi

Nowadays the development of automatic on-line sampling systems has enabled large amounts of data to be available to plant managers. These data are inevitably prone to random and sometimes systematic errors. Data reconciliation methods are used to deal with random errors. In these methods, the measured data are adjusted in accordance to their error, so that the mass conservation law is satisfied at the same time. Measured data may also contain systematic errors, using these data in mass balancing software will result in inaccurate or even irrational results. In this paper three methods for detecting systematic errors (global test, node constraint test and maximum power constraint test) were presented. A step by step method to locate the systematic errors was demonstrated. Then, the new version of Movazen©, a mass balancing software, was introduced and used to locate systematic errors in the measurements taken from the Sarcheshmeh pilot plant copper flotation circuit. Keywords: Mass balancing, data reconciliation, random error, systematic error

Jan 1, 2020

By P. M. Solozhenkin, O. I. Solozhenkin

Currently lead nitrate is used as the recommended activator in Au-Sb flotation. However, due to high levels of hazard, the application of lead nitrate constitutes a potential risk for human health and ecosystems. To mitigate the harmful impact of activators on the environment, further research on activation by alternative reagents has been launched to promote the efficiency of antimony ores. In the present study, adsorption of lead, copper and zinc cations has been investigated by nuclear quadrupole resonance (NQR), which has revealed copper and zinc cations penetration into the mineral lattice, although lead cations remain adsorbed on the mineral surface. However, the combination of zinc and copper cations leads to preferential volumetric adsorption of zinc in the mineral lattice and further surface adsorption of copper which positively affects the activation process. It has been established that activation of stibnite can be achieved effectively with the mixture of Cu2+:Zn2+=1:1 as much as with lead cations in antimony ore flotation. As a part of this work, the computational modelling has been used in prognosis of reagent activity evaluation (PRAE) for stibnite and aurostibite flotation by sulfhydryl collectors. The lower PRAE value the stronger the ability of a collector to form a stable complex with a mineral. PRAE has shown a strong negative trend to complex formation on the mineral surface of aurostibite and stibnite. Based on these studies, several collectors have been proposed for effective flotation. To verify the feasibility of proposed collectors tetramethyl (ethyl) thiuram disulfides (TMTD and TETD) have been used for flotation of stibnite, sphalerite, pyrite and arsenopyrite. Flotation performance has been improved significantly with TMTD that allows the avoidance of lead nitrate application and increases the recovery of antimony by 8,16%. Keywords: Stibnite, aurostibite, activation, lead, copper, flotation reagents

Jan 1, 2020

By P. Pięta, A. Surowiak, T. Niedoba

In a flotation process of copper ore three factors were measured: copper grade in concentrate (β), copper grade in tailings (J) and recovery of copper in concentrate (e). The experiment was conducted by means of the Jameson cell. The factors influencing the quality of the process were particle size (d), time of separation (t), type of the collector (k) and dosage of the collector (s). The considered vector function was then [β(d, t, k, s), J(d, t, k, s), e(d, t, k, s)]. Optimisation achieved was based on determining the values of the adjustable factors (d, t, k, s). The goal was to obtain the highest possible values of functions β and e (maximum) and the lowest possible values of function J (minimum). To this purpose the taxonomic methods were applied. Keywords: Flotation, copper ore, lithology, flotation agents, particle size distribution, taxonomic methods

Jan 1, 2020

By G. Mahieu, L. Perez-Barnuevo, P. Barbeau-Verreault, O. Gravel

Raglan mine is a copper-nickel mining operation, part of Glencore, and located at the northern border of the province of Quebec, in Canada. The concentrator is fed by a mix of four different underground mines, thus it must accommodate periodic variations in run-of-mine mineralisation and composition. In order to concentrate nickel and copper sulphides, the flotation circuit also includes a regrind step at the cleaner stage. This leads to challenges in maintaining optimal recoveries and in diagnosing potential disturbances. This paper advocates for the combination of automated mineralogical analyses, reconciliated mass balances from diagnostic sampling campaigns and innovative multidimensional visualisation to assist in tackling these challenges. Through this approach, it has been possible to quantify the regrind step influence on the liberation of valuable mineral particles in the final flotation concentrate. The methodology also allows for virtual sieving of the products in order to track unit loss in the flotation process. The proposed analysis scheme presents an opportunity to easily diagnose flotation circuit performance based on mineralogical information. Keywords: Flotation diagnostics, multi-dimensional visualisation, mineral liberation analysis

Jan 1, 2020