Search Documents

By B. T. Lusk, J. J. Silva Castro

Currently, there are several methodologies to assess ground vibrations from blasting. Of all available methodologies, one that is becoming more important is the waveform superposition or signature hole technique. This technique is based on a signature waveform assigned to each hole to build the complete waveform for the whole production blast, using the convolution between the time initiation sequence and the signature hole waveform. In the basic methodology of the signature hole technique, the waveform hole-to-hole is assumed to be the same. Several studies and field tests have proven that the waveform hole-to-hole, even in ideal rock mass conditions (rock mass with little or no joint system), changes. In addition to distance or confinement, other factors that could change the waveform hole-to-hole include different geologic paths from the source to the monitoring point and a different energy output of each hole, due to hole loading, explosive composition and/or contamination, priming, etc. These factors result in varying waveforms generated for each hole, making the waveform produced by each hole a somewhat random process. There is available in the literature some proposals to include variability in the waveform hole-to-hole, but the parameters involved are complex and difficult to estimate. In this paper, the variability in the signature waveform hole-to-hole in a blast event was recorded through a field test, and a proposal to include signature waveform variability to develop an equation based on the Fourier series is presented. This methodology is the basis for a future signature hole analysis using a Monte Carlo scheme, taking into account the variation of the amplitude and the frequency content of the waveform hole-to-hole in a production blast event.

Jan 1, 2012

By L. K. Bandyopadhyay

The expert system component of Knowledge Base System supports the analysis and interpretation phases of ventilation design with an algorithmic program used for the solution of alternate design. The paper describes the ?ES-VENT? knowledge based system developed at Central Mining Research Institute, Dhanbad, India for an on-line diagnosing of the ventilation problems using the ?IITMRULE? expert system development tool. In the ?IITMRULE? expert system tool, it has been decided to use a backward chaining reasoning process to reach the goal state i.e. the final conclusion. ES-VENT system is made of three modules: On-line Data, Diagnostics and KBS. Each module is being a collection of several programs. The paper mainly deals with the parameters affecting the ventilation of underground mines such as air-velocity, air pressure, temperature, dust, and humidity and gas contents. The graphical presentation of these parameters on the computer and the interactive part of the developed program is expected to help a ventilation officer to determine the changes in a ventilation circuit necessary to bring the mine ventilation to the desired value. Once such changes are physically effected, the display will indicate if the changes have brought the desired effect or not on ventilation. Keywords: Ventilation network, on-line data, expert system, intelligent system, knowledge base, inference engine.

Jan 1, 2002

By M. S. Shishvan, S. Klerkx, J. Sattarvand

The inaccurately drilled blastholes from the planned locations cause changes in the spatial distribution of explosive energy; some areas of the blast may receive a surplus of explosive energy, while others see a reduction. This results in less uniform fragmentation, reduces the efficiency of mining operations and increases costs. Additionally, it can lead to slope stability issues, flyrock, and uneven benches. To mitigate these problems, this study attempts to reduce differences in the explosive energy distribution (EED) through an optimization approach that adjusts the height of explosive columns in blastholes. Two optimization algorithms, the Genetic Algorithm (GA) and Tabu Search (TS), are tested. The latter is preferred for its local search neighbourhood strategy that gives superior computation times. Applications on different blast patterns result in improvements of 0.53–1.54%, or 2.14–3.94% when only the affected blocks are considered.

Sep 1, 2021

By S. K. Kawatra, T. C. Eisele

Production of highly-cleaned coal for advanced utilization schemes requires close control of the coal ash content at very low levels. For this purpose, a sensitive, low-cost, robust on-stream ash analyzer is needed. The development of such an instrument based on x-ray backscatter and fluorescence is described, including features to improve its sensitivity at low ash levels. The performance of the resulting device in testing a real coal sample is discussed.

Jan 1, 1990

By M. Pekguleryuz, S. A. Argyropoulos

"The recovery of additions in light metals is an important parameter influencing consistency of the finished product, which ultimately means cost. Recovery is a function of many parameters such as assimilation speed, addition density, addition practice, liquid metal processing conditions, etc. In liquid metal processing operations, it is desirable to have the maximum possible recovery at the shortest possible time.In this paper a new index will be presented, namely, the rate of recovery (RR). This index quantifies the addition recovery and introduces the time dimension in the recovery increase. As such, the RR equips the liquid metal practitioner with an accurate picture of addition assimilation in liquid metals. Results will be presented on how this index is applied to a variety of additions in liquid aluminum. IntroductionThe data on addition assimilation kinetics offer a valuable parameter describing how fast a given addition enters the metal bath. These data may be expressed as gr m-2 s-1 of additive for a set of experimental conditions which might include bath temperature and degree of bath agitation among others. However, the liquid metal processing practitioner is more interested in the recovery of additions. In other words, the point of interest is the percentage of the used addition entering the metal bath. Ideally, the addition recovery can reach the 100% level. However, in practice, the addition recovery never or rarely reaches that level. Many factors can contribute to this. Some of these factors can be slow assimilation kinetics, addition density, addition practice, turbulence of the melt, etc. In some cases, sloppy addition practices can contribute to low recoveries, or even worse, to erratic recoveries.In this paper a new parameter is introduced. This parameter relates the maximum addition recovery observed over the time needed to achieve this maximum recovery."

Jan 1, 2002

By Leon Lorenzen, Trevor J. Britz, Pierre J. Van Zyl, Ellis R. Els

Enhanced granulation of UASB granules was tested on pilot scale with two different test reactors (about 5 litres capacity each). Sludge from a local sewage works was used, with peach cannery effluent as carbon source. Cyclic motion of the reactor contents, and a daily replenishment of feed material create the necessary stress conditions for enhanced growth. Sizable granules, bigger than 0.5 mm, were produced within 20 days. Of the two reactors tested, the reactor equipped with a horizontal agitation shaft proved more effective.

Jan 1, 2003

By R. Els Ellis

Enhanced granulation of UASB granules was tested on pilot scale with two different test reactors (about 5 litres capacity each). Sludge from a local sewage works was used, with peach cannery effluent as carbon source. Cyclic motion of the reactor contents, and a daily replenishment of feed material create the necessary stress conditions for enhanced growth. Sizable granules, bigger than 0.5 mm, were produced within 20 days. Of the two reactors tested, the reactor equipped with a horizontal agitation shaft proved more effective.

Jan 1, 2003

By J. Ouellet

The authors participated in a study for a new mine shaft in Saskatchewan?s Athabasca Basin where ground freezing is desired to reduce water inflows. Constraints were imposed on the design in terms of permissible water leakage (less than 10 m3 / hour) inside the shaft. Since the shaft goes through highly permeable formations, water control measures had to be incorporated into the design. Ground freezing was selected as the most reliable and proven method for the sinking and a hydrostatic liner design was implemented. While exploring various technological options available for providing a fully hydrostatic liner, the authors came up with a new design concept that offers a more cost effective solution while meeting all the design constraints. The new liner concept is a composite design using concrete and a sprayed-on membrane system. A proof of concept program comprising laboratory testing and field testing was successfully implemented. This paper presents the innovative liner design, as well as the adapted sinking methodology, developed for this type of project. The new concept offers a new cost effective high performance hydrostatic mine shaft liner comparing favorably with traditional methods. RESUME: Les auteurs ont réalisé une étude pour un nouveau puits de mine dans le Bassin de l?Athabasca en Saskatchewan. Des limites sévères sont imposées sur le design en terme de volume d?infiltration d?eau acceptable (moins de 10 m3 / heure). Comme ce puits de mine traverse des formations très perméables la congélation des terrains a été sélectionnée, une méthode éprouvée et fiable. Un design hydrostatique a été choisi pour le revêtement du puits. Dans le processus de design, les différentes technologies disponibles ont été étudiées. Les auteurs ont développé un nouveau concept de revêtement composite offrant un rapport performance/prix plus avantageux que les technologies conventionnelles. Le nouveau design proposé est composé de béton et d?une membrane projetée. Un programme expérimental a été réalisé pour démontrer la faisabilité du concept. Cet article présente le nouveau concept ainsi que la méthodologie de fonçage de puits adaptée. Ce concept innovateur offre un revêtement hydrostatique performant à un cout avantageux par rapport aux techniques de construction utilisées jusqu'à maintenant.

Jun 1, 2012

By L. A. Alcorn, H. E. Camargo, G. Gwaltney

"One of the most pervasive diseases in the mining industry is noise-induced hearing loss. The exposure of miners to noise levels above the permissible exposure level results in hearing loss in approximately 80% of coal miners by retirement age. In this context, the U.S. National Institute for Occupational Safety and Health (NIOSH) is conducting research to develop noise controls for longwall mining systems, which are used to extract half of the national underground coal production.From field measurements, previous research determined that the dominant sound-radiating components at the shearer of a longwall system are the two cutting drums used to remove coal. Due to the dimensions and complexity of longwall mining systems, the NIOSH project developed a validated finite element model of the cutting drum to use in predicting the sound radiated by the drum due to the excitation forces experienced in underground operations. This same model will be used to develop engineering noise controls for the drum. The excitation forces arise from the interaction of the cutting bits and the coal and are transmitted to the cutting drum through the bit holders. Due to the adverse conditions at the face, and a lack of instrumentation approved for underground use, the operational coal cutting forces have not been assessed before. To this end, NIOSH, in collaboration with Michigan Technological University, developed a self-contained, intrinsically safe instrumented bit to measure these forces.This paper describes the development of the instrumented bit and the measuring of operational coal cutting forces at three working coal mines. The measurements revealed that the forces have a flat spectrum, representative of the impact nature of these forces, up to around 100 Hz. Above this cutoff frequency, the force magnitude decreases at a rate proportional to the inverse of frequency squared."

Jan 1, 2014

By L. A. Alcorn, H. E. Camargo, G. Gwaltney

"Una de las enfermedades más extendidas en la industria minera es la pérdida de audición inducida por el ruido. La exposición de los mineros a niveles de ruido superiores a los niveles de exposición permisibles resulta en la pérdida de audición de aproximadamente el 80% de los mineros del carbón a la edad de jubilarse. En este contexto, el Instituto Nacional para la Seguridad y Salud Ocupacional (NIOSH, por sus siglas en inglés) está realizando investigaciones para desarrollar controles de ruido aplicados a los sistemas de minado de tajeos largos. Este método de minado es usado para extraer la mitad de la producción nacional de carbón subterránea. A partir de las mediciones en campo, se determinó en investigaciones anteriores que los componentes dominantes que generan sonido en el cizallador de un sistema de tajeos largos son los dos tambores de corte utilizados para extraer el carbón. Debido a las dimensiones y complejidad de los sistemas de minado de tajeos largos, el proyecto NIOSH desarrolló un modelo de elementos finitos validado del tambor de corte para utilizarlo en la predicción del sonido generado por este debido a las fuerzas de excitación experimentadas en las operaciones subterráneas. Este mismo modelo se utilizará para desarrollar controles de ingeniería para el ruido del tambor. Las fuerzas de excitación surgen de la interacción de la broca de corte y el carbón, y se transmiten al tambor de corte a través de los sujetadores de brocas. Debido a las condiciones adversas en el frente y la falta de instrumentación aprobada para uso subterráneo, las fuerzas operativas de corte de carbón no han sido evaluadas anteriormente. Para este fin, NIOSH, en colaboración con la Universidad Tecnológica de Michigan, desarrolló una broca instrumentada autónoma con seguridad intrínseca para medir estas fuerzas. Este artículo describe el desarrollo de la broca instrumentada y la medición de las fuerzas operativas de corte de carbón en tres minas. Las mediciones revelaron que las fuerzas tienen un espectro plano, algo característico de la naturaleza del impacto de estas fuerzas, hasta alrededor de 100 Hz. Por encima de esta frecuencia "cutoff", la magnitud de la fuerza disminuye a una razón proporcional al cuadrado de la inversa de la frecuencia."

Jan 1, 2014

By G Pizarro, O Miranda, D Rosales

In panel/block caving mines, large amounts of information are collected at the operations. This information includes grades and extracted tonnage, fragmentation, lithology, design of stopes and main infrastructure, geotechnical information, overbreak progress, seismicity, numerical modelling and many other mine variables. To improve the reliability of the production plan it is required to integrate the mine data into one platform. In this paper, the logic of a new technological platform is presented for analysis, modelling and design in panel/block caving mines. The platform includes an acquisition system for mine data and geotechnical information, a back-analysis module for calibrating statistical models and a module of visualisation of the results. The application and benefits of the platform is illustrated for mining induced seismic risk for panel/block caving mining operation. The platform can be used in the mine design and planning processes and is useful for reducing uncertainty in the mining business of panel/block caving operations.CITATION:Miranda, O, Rosales, D, Pizarro, G and Vallejos, J, 2016. Development of an integrated platform for the analysis, modelling and design in panel caving mines, in Proceedings Seventh International Conference and Exhibition on Mass Mining (MassMin 2016), pp 237–240 (The Australasian Institute of Mining and Metallurgy: Melbourne).

May 9, 2016

By Y M. Lee, Y Hong, B C. Kim

Since sintered ore is the primary source for producing molten iron in the blast furnace process, maintaining stable quality despite the utilisation of low-grade iron ore is crucial. However, conventional analysis methods, such as chemical composition, ore size distribution and types of iron ore, have limitations in developing the utilisation of new low-grade iron ore and predicting sintered ore properties. Sintered ores are formed through a complex melting process of granule particles consisting of iron ore, flux and fuel. Consequently, it has been investigated that analysis methods which apply the reactive properties of iron ore to predict sintered ore characteristics. In this study, we estimated the properties of iron ore by simulating the assimilation process of coarse ore, which influences the bonding strength of sintered ores and quantifying the degree of assimilation reaction. Pseudo particles were formed by granulating coarse ore (1–3 mm) with adhering fines, which act as primary melts and sintered at 1623 K. The bonding strength of individual sintered ores, formed from 11 different types of iron ore, was measured using a simplified Tumbler Index (TI) test and each value was indexed as the Iron ore Reactivity Index (IRI). By investigating the relationship between the degree of assimilation depending on iron ore properties and the characteristics of sintered ore, we demonstrated that sintered ores formed from more reactive iron ore exhibited stronger bonding strength. Moreover, when blending conditions resulted in a higher Iron Blending ore Reactivity Index (IBRI), the physical properties of sintered ore improved. The development of this analysis method is expected to not only enable careful consideration of iron ore usability but also allow for the advance prediction of sintered ore qualities by deriving the IBRI.

Sep 18, 2023

By Lutz Markworth, Fernando Parra, Wolfram Jaspers

"The worldwide demand of iron ore and the depletion of the high-purity deposits make it necessary to recover ores in increasingly fine size fractions or low Fe content in the feed material. Impurities for Iron ore processing plants are mainly Silica, Al2O3 and Sulfides.There is a copper processing plant in Chile, which has a high content of Fe in its tailings. These tailings will be re-processed to recover the Fe and to produce a pellet feed concentrate. This challenging flotation process was developed by the customer in his own laboratory. Different flotation technologies were tested. The best results were achieved with the PNEUFLOT® (PNEUmatic FLOTation). For a feed capacity of approximately 600 t/h, a flow sheet of two flotation lines with a reverse two-stage Amine-Silica flotation followed by a single stage reverse Xanthate Pyrite flotation was developed.Because of the two impurities the process control is very critical. Not even the reagent dosing has to be flexible for changes in the feed material; also the flotation cell has to be adaptable. Continuous development of the PNEUFLOT® cell with their specific advantages and the elimination of dysfunctions in the PNEUFLOT® cell have led to a growing demand of this type of equipment worldwide for the iron ore industry within the past years.INTRODUCTIONThe worldwide increasing demand of iron ore is mainly caused by China. This has a major impact on all economics worldwide as the metal price indicates. However, not even the iron ore and steel price, but all other commodities prices and the precious metals prices are increasing by a high percentage. Figure 1 shows the iron ore (pellets) price increase for Fe and the import of iron ore by China. Figure 2 shows the imports of iron ore to China."

Jan 1, 2008

By Y Yamaguchi, S Nomura, M Matsumura, C Kamijo

In order to increase productivity and decrease reducing agent rate (RAR) of blast furnaces, slag content of sinter ore must be decreased. However, gangue content in iron ore tends to increase because of resource depletion. Therefore, amount of high-grade concentrate in sinter ore feeds should be increased, but this situation causes a decrease in permeability of the sintering bed so that the productivity of sinter process decreases. In order to increase the amount of concentrate, NSSMS has developed the P-type separate granulation method, RF-MEBIOS and SPEx II technologies. These technologies have been introduced to actual sintering machines. In this report, a new technology using microparticles of iron ore is proposed to increase the amount of concentrate in sinter blends. Basic experiments simulating the sinter making process using microparticles of iron ore were undertaken. Crushed pisolite, containing >60 wt per cent -10 µm was prepared for binder. The pisolite microparticle binder was mixed with another Australian sinter feed ore in a high-speed agitating mixer. They were mixed for about one minute, and then granulated using a pan pelletiser. After that, they were added to another raw material mixed by a drum mixer, and then charged in a pot for a sinter pot test. As a result, when the amount of concentrate increased from 13 wt per cent to 20 wt per cent, productivity was almost kept constant using pisolite microparticle binder.CITATION:Kamijo, C, Yamaguchi, Y, Matsumura, M, Kawaguchi, T and Nomura, S, 2017. Development of an iron ore microparticle binder for increasing the concentrate content of sinter blends, in Proceedings Iron Ore 2017, pp 69–74 (The Australasian Institute of Mining and Metallurgy: Melbourne).

Jul 24, 2017

By J. C. Vreugdenburg, B. A. Dippenaar, J. Muller, T. L. de Vries

Iron ore fines are agglomerated in the sintering process to produce sinter: an important feed material for blast furnaces worldwide. A model of the iron ore sintering process has been developed with the objective of being representative of the sinter pot test, the standard laboratory process in which the behaviour of specific sinter feed mixtures is evaluated. The model aims to predict sinter quality, including chemical quality and physical strength, as well as key sinter process performance parameters such as production rate and fuel consumption rate. The model uses the finite difference method (FDM) to solve heat and mass distributions within the sinter pot over the sinter bed height and time dimensions. This method can further be used for establishing empirical relationships with measured sinter properties. Inputs into the model include the feed material physical properties, chemical compositions, and boundary conditions that vary over time. Sub-models describe relationships between applied pressure differential and gas flow rate through the bed of granulated fine ore particles, combustion of carbonaceous material, calcination of fluxes, evaporation and condensation of water, and melting and solidification. Four sinter pot tests were performed and the model predictions validated against temperatures measured at depths of 100, 230, and 420 mm. Differences between predicted and measured temperatures indicated the need for optimization of some model parameters that are uncertain. Optimization resulted in an improvement in the correlation between the predicted and measured temperatures for all tests, although an acceptable level of agreement was not achieved in all cases. Using these parameter values, reasonable agreement was found in terms of typical performance parameters for three out of the four tests.

Jan 1, 2014

By G. T. Adel, J. D. Miller, C. L. Lin

Washability analysis determines the maximum separation efficiency that can be achieved in the operation of coal preparation plants. Unfortunately, there are no on-line techniques for determining this most fundamental information. It now appears possible to determine coal washability on-line through the use of x-ray computed tomography (CT) analysis. This paper presents the results for the development of a CT-based, on-line coal washability analyzer. The development of appropriate hardware and software necessary to adapt a CT analyzer for use in determining coal washability is discussed. Testing of the analyzer using well-characterized coal samples is presented.

Jan 1, 2000

By Markus Timusk

This paper describes a research project that focuses on developing a diagnostic routine to assist in decision support for the condition based maintenance of mobile mining equipment subject to unsteady duty. The objective is to develop a general method for initial identification of mechanical faults in machinery operating under non-steady state conditions. The platform to be used for the experimental development and validation of the system is the swing machinery in an electric mining shovel operating in an open pit mining environment. The strategic approach that is proposed for the diagnostic routine will involve using novelty detection augmented by temporal classification.

May 1, 2004

By D van der Wal, R Dobbe, K Moeller

"Ore mineralogy and texture are primary variables controlling mill recovery efficiency in flotation plants. However, tracking daily variations of key mineralogical characteristics such as species-by-species liberation is not possible with current on stream analysers and assay lab techniques. In addition, traditional plant mineralogical analysis limited by long data turn-around times from central analysis laboratories typically provides only a time-averaged view of mill performance, e.g. by means of monthly composite feed, con or tailings analysis.This paper presents the results of a development and first field test of an on-site Automated Mineralogy solution (MLA or QEMSCAN) with the specific goal of monitoring recovery variation, ore quality and flotation performance on a mineral species-by-species and day-by-day basis, from up to twelve million direct particle observations per month. The trial was performed at a Pb/Zn-Ag operation in the US from November 2012 to April 2013. The focus of the study was three-fold: to validate the robustness of equipment and operational procedures for daily use by mine site technicians at a remote mine site to develop a methodology to boil down hundred-thousands of Automated Mineralogy particle data from size-by-size con and tailings samples to a couple of numbers meaningful to chief metallurgists to monitor recovery trend changes and correlate these to changes in the feed properties and/or flotation efficiency.The study focused primarily on the Pb/Ag rougher-concentrator circuit. Ten polished blocks of two and three size fractions of con and tailings were prepared by mine site technicians during a daily twelve hour shift. Eight hours of daily measurement time by the instrument revealed an estimate of the dayÆs recovery performance, ore quality and flotation performance for Freibergite (Ag-species) and Galena (Pb-species), within 36 hours from initial sampling. Data quality was assessed by means of calculated versus measured assay reconciliation. Cumulative sum curves of recovery performance, ore quality and flotation performance reveal trend changes with a frequency of 1 to 4 weeks. These were correlated to cumulative sum curves of e.g. locked and slimed particle classes of the mineral species of interest, attempting to identify a possible root cause of recovery performance trend changes. Systematic over-grinding and flotation chemistry issues were the primary causes of below average plant operations during the trial.CITATION:van der Wal, D, Dobbe, R and Moeller, K, 2013. Development of an on-site mineralogy solution for metallurgists, in Proceedings MetPlant 2013, pp 126-132 (The Australasian Institute of Mining and Metallurgy: Melbourne)."

Jul 15, 2013

By Keith A. Heasley, Christopher R. Newman

"Originally developed in 1993, LaModel has aided mining engineers and researchers alike in improving the design and safety of underground single- and multiple-seam mining operations (Heasley, 1998). Through the years (Heasley, 2011), the use and capabilities of the LaModel program have grown immensely while adapting to ground control concerns and problems within the mining industry. Presently, MSHA specifically mentions the use of LaModel for the analysis of, “complex non-typical roof control plans,” defined as either room-and-pillar retreat mining greater than 1000 feet, bump- or bounce-prone mines, or other criteria considered unusual by the District Manager (Stricklin, 2013).With an ever increasing number of users and an increase in available solution options and complex analyses, there is now a significant demand for better education and training in the practical application and detailed analysis using LaModel. This paper presents the development of an electronic user’s manual and comprehensive online training course created in an open online learning environment in order to better inform and train industry professionals as well as engineering students and new users. The development of both the user’s manual and online training course will ultimately increase the ground control design effectiveness of mining engineers, leading to safer and more productive mine designs.INTRODUCTIONOver the past 20 years, mining engineers and researchers have used the LaModel program for improving the design and safety of single- and multiple-seam mining operations. LaModel uses a laminated, displacement-discontinuity, boundary-element method which simplifies the overburden into a stack of frictionless homogeneous laminations and limits the detailed analysis to the seam itself. This approach is fundamentally simpler than either finite-element, finite-difference, or discrete-element approaches; and, therefore, the program provides a significant reduction in computational time and efficiently calculates the displacements, stresses, pillar safety factors, multiple-seam interactions, subsidence, etc. for thin-bedded deposits such as coal, salt, potash, limestone, etc. (Heasley and Salamon, 1996; Heasley and Agioutantis, 2001). The laminated overburden model in LaModel is different than previous homogeneous elastic formulations and more realistically represents the natural flexibilities of the stratified sedimentary overburden typically associated with coal seams (Heasley, 2008)."

Jan 1, 2015

By E J. Y Koh, G J. McLachlan, E Amini, N Beaton

Preconcentration greatly reduces the ore mass input to the processing plant while upgrading the grade. One of the commonly used methods of preconcentration is optical sorting. These methods rely on optical sensors which classify the material stream into “accept-reject” streams. Currently, it is common to use near infrared (NIR) or colour cameras to classify streams based on reflectance or colour thresholds set by experts. The imaging is done through line-scan sensors where the material is scanned on a belt or a chute. Then, a separation apparatus like a diverter gate or air jet is controlled to separate the streams according to the classification made by the algorithm. Separation efficiency is dependent on the classification algorithm. Existing classification algorithms rely on low-level feature discrimination like observing individual pixel values for reflectance or colour hues. This limits discrimination between minerals with low contrast in colour. Human experts perform mineral segmentation using texture, colour distribution and shape, but it is difficult to translate these visual rules to mathematical thresholds. Deep learning methods like Convolutional Neural Networks (CNNs) provide high-level feature discrimination similar to a human expert. Instead of relying on pre-defined features, the CNNs learn complex features from the dataset. CNNs can construct a hierarchy of features based on pixel clusters to learn texture, shapes, and colour spectrums of the mineral grain surface to provide better separation. In this study, a state-of-the-art deep learning instance segmentation method is used to outline boundaries and classify grains from the background (mineral segmentation). The algorithm is trained on images collected from a flotation feed of a gold deposit, where the gold is mechanically locked inside pyrite. This method can process on-line video inputs at 5 frames per second to provide a live grade. The proposed method is inexpensive as it only uses a camera and a desktop computer.

Nov 10, 2020