Search Documents

By P Holtham, R Dunne, D Stoll

Flash flotation is a process designed to remove coarse, liberated, fast floating minerals from hydrocyclone underflow. Irrespective of liberation, coarse particles would normally remain in a grinding circuit until they are reground fine enough to report to the hydrocyclone overflow. Additionally, minerals of high density also report to hydrocyclone underflow despite being fine in size and/or liberated because hydrocyclones separate on the characteristics of both particle size and density. Dense particles continually reporting to the hydrocyclone underflow can lead to over-grinding and the production of ultra fine particles. These particles exhibit slow flotation kinetics and thus are difficult to float in conventional flotation circuits. Therefore, removing coarse, liberated particles from the hydrocylcone underflow by flash flotation has the potential to improve recovery and can improve dewatering unit operations and reduce grinding costs.The impact on the overall performance of the flotation circuit by removing coarse, fast floating minerals prior to conventional flotation is not well understood. Ultimately, predicting the overall recovery improvement provides the justification for the installation of a flash flotation cell within the milling circuit. This paper will describe attempts to estimate flash flotation recovery improvements to the overall circuit using kinetic data from simplified batch cell flotation tests combined with limn-based modelling techniques.CITATION:Stoll, D, Dunne, R and Holtham, P, 2012. Modelling the impact of flash flotation from batch cell flotation tests using limn, in Proceedings 11th AusIMM Mill OperatorsÆ Conference, pp 179-184 (The Australasian Institute of Mining and Metallurgy: Melbourne).

Oct 29, 2012

By J. W. Perez

A simple exponential model has been developed (Perez and Aplan. Process Mineralogy, VI, 1986) to describe the fractional yield of a given species out of a locked assemblage: Fractional Yield of Species A = a e X/U where 'X' is particle size and 'a' and 'U' are constants. In the present study, the general validity of this model has been further demonstrated. experimentally, for the liberation of coal and pyrite out of various raw coals and beneficiated fractions using a variety of crushing devices. Using data from the literature, the model is also shown to be applicable to the liberation of sulfide minerals out of ores.

Jan 1, 1987

By Arash Habibi, Anna Cable, David Loring, Aleksandar D. Bugarski, Logan Ingalls, Calvin Rutter

The results of diesel particulate matter (DPM) monitoring were used by the underground mining industry to mitigate adverse effects of exposure to DPM. NIOSH Method 5040, currently used for ambient and personal exposure monitoring, does not provide needed real-time information. The objective of this study is to develop a methodology for continuous monitoring of DPM concentrations in an underground environment, founded on the application of the Dekati electrical particle sensor (DePS). The DePS, which measures concentrations of submicron aerosol in real time using a diffusion-charging sensor, was evaluated by comparing the results of the concurrent measurements with reference methods at several locations downstream of the diesel-powered equipment. The DePS results were found to correlate well to those obtained by reference methods. However, it was found that dependence of the response on the physical and chemical properties of aerosols necessitates sitespecific correlation factors for the targeted matrices. The site-specific factors for number and mass of diesel aerosols were derived using the results of concurrent measurements with TSI NanoScan, FLIR Airtec, and carbon analyses performed on the concurrently collected filter samples. The results showed that methodology based on compensated DePS measurements would be suitable for continuous monitoring of DPM in underground mining operations.

Dec 1, 2022

By Y. Luo

"It is generally known that spontaneous combustion potential is a function of coal rank, with low-rank coals having high susceptibility to self-ignition and high-rank coals having low potential. To establish this functional relationship and use it as an engineering tool to identify the propensity of coal to self-heat, both rank and spontaneous combustion potential should be quantified. ASTM D388 is normally used to classify coals by rank, but this classification system can only serve a qualitative analysis. To classify coals by their potential for spontaneous combustion, the self-heating temperature (SHT) method originally developed by the U.S. Bureau of Mines (USBM) was used by Litton and Page (1994) to correlate coal rank with spontaneous combustion potential. In that method, the SHTs of coal determined by an adiabatic experiment are linked with proximate analysis results through a proposed concept known as the “moist fuel ratio.” This ratio is a function of moisture, volatile matter and fixed carbon, all of which have widely different content percentages in different ranks of coal and have strong relationships with coal rank. However, the method is highly empirical, and there is large standard deviation between experimental data and calculated results. The equation correlating the coal quality parameters to determine SHT should be improved with more experimental data.This paper reports on the development of a new system for predicting the spontaneous combustion potential of U.S. coals, based on coal composition data from the U.S. Geological Survey’s coal quality database. First, the user is provided with a quantified classification of coal rank. Second, based on experimental SHT data covering a wide range of ranks from lignite to low-volatile bituminous coal, the original USBM correlation for predicting SHT is improved for the respective rank of coal. The relationship between the rank and SHT of U.S. coals was developed and plotted. The plot diagram can be used for a quick estimate of the SHT of a U.S. coal and as a cost-effective tool for the initial risk assessment of a new mine development."

Jan 1, 2015

By Robert J. Timko

This U.S. Bureau of Mines study examined spontaneous heating episodes in coal mine pillars in an active underground coal mine. The information obtained from these incidents was then analyzed to learn which sampling methods provided the earliest indication of pillar heating. The objective of this study was to discover if the location of future events of pillar spontaneous heating could be inferred from the available information. The spontaneous heating-prone area in this evaluation involved pillars located just inby the mine portals. Several detection methods were used to determine gas levels outside as well as inside the affected pillars. It was hoped that, by incorporating external and internal sampling methods into an organized pro-gram, locations undergoing spontaneous heating could be determined more readily. This study found that by drilling small-diameter boreholes into the pillars, then obtaining gas samples from the affected pillars, the ability to locate early spontaneous heating episodes was improved. However, the ability to accurately predict future spontaneous heating events remains in question.

Jan 1, 2010

By Mei Long, Tawum Juvert Mbah, Jianhua Zhang, Haiwang Ye

There have been many improvements and advancements in the application of neural networks in the mining industry. In this study, two advanced deep learning neural networks called recurrent neural network (RNN) and autoregressive integrated moving average (ARIMA) were implemented in the simulation and prediction of limestone price variation. The RNN uses long shortterm memory layers (LSTM), dropout regularization, activation functions, mean square error (MSE), and the Adam optimizer to simulate the predictions. The LSTM stores previous data over time and uses it in simulating future prices based on defined parameters and algorithms. The ARIMA model is a statistical method that captures different time series based on the level, trend, and seasonality of the data. The auto ARIMA function searches for the best parameters that fit the model. Different layers and parameters are added to the model to simulate the price prediction. The performance of both network models is remarkable in terms of trend variability and factors affecting limestone price. The ARIMA model has an accuracy of 95.7% while RNN has an accuracy of 91.8%. This shows that the ARIMA model outperforms the RNN model. In addition, the time required to train the ARIMA is than that of the RNN. Predicting limestone prices may help both investors and industries in making economical and technical decisions, for example, when to invest, buy, sell, increase, and decrease production.

Jan 6, 2021

By James R. McVey

Tests during pillar robbing, longwall operations, and other mining activities indicate that roof-floor convergence rates can be used as an indicator of imminent roof failure. A large number of electrical-mechanical devices have been built to measure convergence as well as convergence rates, but most obstruct mine traffic and are costly and time consuming to install. The Bureau of Mines has designed and is evaluating three types of small ultrasonic units to make these measurements. The new instruments provide nonobstructing measurements up to 35 ft away. The ultrasonic transducers can be attached to a roof bolt, placed in an unsupported area (by means of a toss basket), or handheld. Total distance to the target and rate of closure are displayed digitally to 0.001 ft and 0.01 ft/min, respectively.

Jan 1, 1984

By Donald W. Gentry

This paper presents some of the results of a rock mechanics in¬strumentation program designed for, and implemented at, longwall panel 4N at the York Canyon Mine near Raton, New Mexico. The objective of this particular instrumentation program was to determine the response of the rock mass to longwall mining of a thick coal seam utilizing shield-type supports. Measurements were made in the rock along the ground surface, above the coal seam, in the coal seam and immediately below the coal seam. Longwall test panel 411 had a total developed length of 762 m (2500 ft) and a face width of 168 m (550 ft). The seam mining height in the panel area was approximately 3 m (10 ft). The average depth of overburden above the panel is 107 m (350 ft) al¬though it ranges from 73-134 m (240 - 440 ft). Roof support at the face was provided by Hemscheidt shields. These shield-type supports weigh 11.4 metric tons each.

Jan 1, 1977

By Steven J. Schatzel, James Addis, K. Vaibhav Raj, Berk Tulu

The National Institute for Occupational Safety and Health (NIOSH) is conducting research to enhance methane control and mine ventilation practices at coal mines, as a method to mitigate hazardous work conditions. NIOSH researchers conducted case studies to investigate the effectiveness of existing systems and improve performance to decrease the potential for unplanned or excessive methane emissions. At one mine, drainage was influenced by hydrofracking inefficiencies where two of the six wells showed good frack development. Methane content was variable, from 300 ft3/t to 600 ft3/t. This paper also summarizes overburden movement during undermining and the estimated quantities of gas transported to the mine using FLAC3D and CFD modeling. The findings from this research effort highlight considerations based on specific mine conditions that may also be applicable to most longwall mines utilizing methane control systems.

Feb 1, 2025

By J. D. Miller, R. J. Smith

Earl C. Herkenhoff Publication of the article "The Flotation Behavior of Digested Asphalt Ridge Tar Sands" by R.J. Smith and J.D. Miller has confirmed and again spotlighted a most significant and important development in the field of tar sand and heavy oil mining. This is the ability to recover 95 % of the bitumen in dry tar sands by a hot water process. The initial investigation by Miller and Sepulveda in early 1978, "Separation of Bitumen from Dry Tar Sands," was the first blockbuter. It is disappointing that there has been so little scientific and industry ackowledgement that the authors laid to rest the long entrenched belief that the sand grains had to be "water wetted" for a hot water process to be applicable. Dr. Miller and his researchers at the University of Utah used slightly different mineral dressing terminology in describing their process but it is essentially the old, established technique of high solids attrition scrubbing at elevated temperatures with a conditioner or modifier, followed by dilution and froth flotation. Similar techniques were described in detail for other nonmetallic minerals in the mid-1940s by Stuart Falconer of American Cyanamid Co. The authors used a 15-minute "scrub" period (or in their terminology "digester step") at 73% solids and 95°C (203°F) in a 0.3 m solution of sodium carbonate -about 4 kg/t (8 lb per st) of tar sand. Flotation was at 10% solids. Obviously, the scrub is an expensive step with respect to energy input but the high recovery of bitumen appears to justify it. Two basic questions then arise: • What are the effects of shorter scrubbing time, lower temperature, and less sodium carbonate? • Would a two-stage approach be advantageous? That is, could you get the major proportion of the bitumen with a mild scrub and apply the high power and temperature only to the refractory portion? These variables should be evaluated. Apparently, all results were only for rougher flotation. What would the grade and recovery of bitumen be if multiple cleaner flotation steps were employed? We must concede that a 60-62% grade of bitumen leaves considerable room for upgrading. Finally, it should be noted by all executives who approve funds for heavy oil recovery schemes, most of which are variations of in-situ methods, that mining can yield a 95% recovery of bitumen. That is correct, 95%. Imagine what that does to the total barrels recovered from a heavy oil resource compared with the scrawny 30-35% maximum recovery we are getting now. The conclusion-let's mine heavy oil and tar sands. Reply by R.J. Smith and J.D. Miller We, too, share Mr. Herkenhoff s enthusiasm for the mining of tar sands-heavy oil and continue the development of related processing technology. Specifically, it will be noted that by appropriate control of bitumen viscosity in the feed (probably the most important variable for successful separation) the recovery can be accomplished at low digestion temperature and low addition of alkali. A 91 t/d (100 stpd) pilot plant based on our technology is now being operated by Enercor in Salt Lake City (Hatfield, Oblad, and Miller, 1982). In these same pilot facilities, rougher concentrates containing more than 80% bitumen on a dry basis are being produced from Asphalt Ridge Feed. However, the processing of certain other tar sands results in incomplete digestion of bitumen/sand aggregates as discussed elsewhere (Miller and Misra, 1982). Such undigested aggregates are shown in Fig. 1. In this regard, it is intended that a recycle stream will be installed in the pilot plant to provide for redigestion of middling products. Further details regarding surface chemistry aspects of this tar sand technology have recently been published (Misra, Aguilar, and Miller, 1981). References Hatfield, K.E., Oblad, A.G., and Miller, J.D., 1982, "Pilot Plant Recovery of Bitumen from Oil Wet Tar Sands," Processing, Second International Conference on Heavy Crude and Tar Sands, Caracas, Venezuela, Feb. 7-17. Hatfield, K.E., Oblad, A.G., and Miller, J.D., 1982, "Pilot Plant Processing of Oil-Wet Tar Sands," Symposium, Rocky Mountain Fuel Society, Salt Lake City, Utah, Feb. 18-19. Miller, J.D. and Misra, M. 1982, "Hot Water Process Development for Utah Tar Sands," Fuel Processing Technology, Vol. 6, No. 1, April, pp. 27-49. Misra, M., Aguilar, R., and Miller, J.D., 1981, "Surface Chemistry Features in the Hot Water Processing of Utah Tar Sand, Proceedings, "Symposium of Separation Science and Technology for Energy Applications," Gatlinburg, Tennessee, May 5-8.

Jan 11, 1982

By J. L. Gillson

The word "other" in paragraph 4, line 11, column 1, page 205, should read "over;" the correct version of reason 2, this paragraph, is: 'L. . . 2) to evaluate the usefulness of certain geo-chemical techniques, V. C. Kennedy and S. W. Hobbs of the USGS had already studied the distribution of copper, lead, and zinc in the soils (both in background areas and over veins) and in the water and plants, providing a criterion for measuring the magnitude and effect of soil contamination; ...9 On page 206, in the top left hand corner of Fig. 1 (in the area bounded by 47°35' and 116°10), the reference to Fig. 3 is incorrect. The area of Pine Creek, South Fork Coeur d'Alene River, and portion of Big Creek (the area which includes, from left to right, Kingston, Pinehurst, Smelterville, and Kellogg) should be identified as that of Fig. 2A. The headings for columns 1, 3, and 4, Tables I and 11, page 209, column 2, were reversed when the paper was printed. The corrected tables are:

Jan 1, 1961

By Raymond E. Healy

Mercurian e1ectrum (Au-Ag-Hg alloy) is the principal Au-bearing phase in the massive sulphide ores at Trout Lake, Manitoba. The average composition of this phase (in wt %) is 49.2 % Ag. 38.7 % Au and 11.0 % Hg, with ranges of 17.34 to 76.47 % Ag, 1.62 to 79.86 % Au and 1.29 to 30.87 % Hg. Compositional variation reflects solid solution in the a-phase of the Ag-Au-Hg system. Backscatter electron imagery reveals systematic zoning and suggests a general crystallization sequence from metamorphic pore fluids of progressively Ag-rich compositions. Investigation of the occurrence of gold in mill products demonstrates that mercurian electrum accounts for over 90 % of the Au in Trout Lake ore. I Invisible" gold in pyrite (x = 0.72 ppm Au) and arsenopyrite (x 30.2 ppm Au) is confirmed by SIMS analyses, and accounts for 7-10 % of the Au in the ore, and 35 % of the Au lost to the tailings. The remaining 65 % of the Au lost to the tailings consists principally of mercurian electrum locked with pyrite. No significant losses of mercurian electrum in the size range amenable to flotation and/or regrinding is indicated.

Jan 1, 1989

By F. W. Lane

Some months ago the Bureau of Mines began research of a fundamental character on sulphur in petroleum. In the course of this work, the usual qualitative test for sulphur in organic compounds1 was applied to certain oils that were believed to contain approximately 0.5 percent sulphur. No sulphur was found, and, as a check on the qualitative test, the oils were subjected to quantitative determination of this element by the oxygen-bomb method.2 The sulphur content of the oils ranged from 0.10 per cent to 0.50 per cent. This experience led to further trial of the qualitative test for organic sulphur, both on the oils mentioned above, and on common sulphur-containing organic substances. No difficulty was experienced in detecting the element in the latter, but with the petroleum oils the test, as usually performed, could not be relied upon to detect less than 0.60 per cent of sulphur consistently. The use of mechanically polished pure sheet copper for the detection of free sulphur, hydrogen sulphide, and perhaps other sulphur compounds has been recommended.3

Jan 1, 1927

By Ralph Husack

The Bureau of Mines investigated the effect of gamma radiation on reactions between Pennsylvania anthracite and fluorine and chlorine. In the tests with fluorine about 90 percent of the coal reacted at 350° C to form volatile products whether or not irradiation was employed. Up to 0.8 ml of liquid products per gram of anthracite was collected. Among the 25 com-pounds indicated by gas chromatography, fluoroform and l,l-difluoroethylene were tentatively identified from their relative retention times and perfluoro-cyclopentane from its infrared spectrum. In addition, a clear, hard, wax, probably a fluorocarbon, was formed at 45° C in both the presence and absence of radiation. With chlorine there was no significant difference due to radiation, and preliminary tests indicated that the chlorine reacted more with the organic matter of anthracite at 200° C than with inorganic matter. At 400° C the results were inconclusive because of excessive corrosion in the reaction vessel.

Jan 1, 1966

By Naeimeh Shabani, Hamid Mirzaie, Javid Sadeghi, Hoda Jafarizadeh-Malmiri, Afshin Javadi

Iron oxide nanoparticles (Fe3O4 NPs), despite their small size, are stable and have a large surface area that allows for absorption and removal of ions as well as turbidity in a short period of time. Thus, the present study addressed the synthesis and characterization of Fe3O4 NPs synthesized using the co-precipitation method and evaluated their effect on the coagulation process in water treatment. The Fe3O4 NPs characteristics were evaluated as follows: magnetic properties were measured using a vibrating sample magnetometer (VSM), surface area using Brunauer–Emmett–Teller (BET), crystal phase using X-ray diffraction (XRD), and morphology using field emission scanning electron microscopy (FESEM). In addition, the effects of the synthetic Fe3O4 NPs on the turbidity reduction of water in various pH levels were evaluated. The synthesized Fe3O4 NPs were without any impurity in the sizes of 22–32 nm, surface area of 99.6 m2/g, pores’ mean diameter size of approximately 10.6 nm, and maximum magnetic saturation of 66.21 emu/gram (emu/g). Using 0.2 g of synthetic Fe3O4 NPs with an initial turbidity of 17 NTU and pH of 6 could reduce all water turbidity (100%). Considering the cheap and facile production and separation, as well as the possibility to reuse them and reduce the sludge water content, iron oxide Fe3O4 NPs can be effectively used for turbidity reduction in the water treatment process.

Oct 26, 2020

By Junwei Han, Kai Xue, Wenqing Qin, Tan Xu, Fen Jiao, Lianbing Cai, Wei Liu

"The research aimed at solving the problem of the stacked spent magnesia-chrome refractories using in non-ferrous metal smelters, which contain various heavy metals and are considered as a hazardous waste. The purpose of the research is to recycle the spent refractories environmentally and economically using gravity separation and flotation to obtain metal concentrates for smelting and using chloridizing volatilization to remove the trace metals from tailings, providing raw material to reprocess magnesia-chrome refractories. In the course of the study, laboratory techniques for the preparation of experimental samples of spent magnesia-chrome refractories were developed including the sequence of operations of material grinding, gravity separation, tails regrinding, flotation, filtration and subsequent unification of concentrates, filtration and chloridizing volatilization of tails, and collection of volatiles. Laboratory tests carried out on several types of samples of spent magnesia-chrome refractories from different smelters showed that gravity separation provided increase in quality of concentrates including about 80% metals (including Ag, Pb, Bi) and flotation separation provided increase in recovery of concentrates under the condition of 200g/t of emulsified kerosene, and 500g/t of combined collector. Based on combined technology of gravity and flotation separation, about 90% metals were recovered from the refractories by a closed-circuit test and the total concentrates contained 40% metals, which can be returned to smelting process as a feeding. Because a small amount of metal in raw materials will seriously affect the performance of refractories, chloridizing volatilization is necessary for removing the remaining 10% metals from tails mixed 8% coke and 13% magnesium chloride at a temperature of 1200°C. Thus, the tailings can be reused as a raw material for new refractories.INTRODUCTION Magnesia-chrome refractory has been used extensively in various industries such as secondary metallurgy, non-ferrous furnaces and cement making due to their high temperature stability, low thermal expansion and outstanding erosion-corrosion performance at high temperatures (Gotod et al., 2005; Vezza et al., 1997; Gregurek et al., 2015). Generally, the magnesia-chrome refractories are produced from a mixture of magnesia and massive variety of chrome ores. However, massive chrome ores are being depleted throughout the world. Even more serious is that the heavy metal chromium (Cr) can transfer into soil and water, which will have an influence on the survival and development of human beings (Zhao et al., 2006; Jiang et al., 2004; Qintie et al., 2013). As a result, regulations treating spent magnesia-chrome refractories as potentially hazardous waste are being enforced, making their disposal difficult and expensive."

Jan 1, 2018

By E. Chin

The countries of Asia and the Pacific have large resources of diverse minerals. Australia and China have extensive mineralization that are commercially being exploited. The major minerals in Australia with known reserves adequate for both domestic demand and export markets include bauxite, clays, copper, diamond, gold, iron ore, lead, manganese, mineral sands (titanium minerals), nickel, salt, silver, tin, uranium, and zinc. China is the world leader in proven reserves of antimony, barite, magnesite, molybdenum, rare earths, titanium (ilmenite), tungsten, and vanadium. It also has about 55 billion tons of iron ore, albeit with an average grade of only 30 % to 35 % elemental iron content. With the exception of chromium and copper, China has significant resources of a wide array of minerals and metals, based on its production and/or export capability for these commodities. Additionally, India has significant resources of bauxite and mineral sands; Indonesia, bauxite, copper, nickel, and tin; Malaysia, tin and associated titanium; Mongolia, copper, fluorspar, and lead-zinc; New Caledonia, nickel; New Zealand, gold and iron sand; North Korea, magnesite; Papua New Guinea, copper, gold and mineral sands; Philippines, copper, chromite, and gold; and Sri Lanka, gemstones. Moreover, there appears to be extensive mineralization III Burma (Myanmar), Cambodia, Laos, Thailand, and Vietnam, but they have yet to be definitively delineated.

Jan 1, 1994

By M Haest

This study demonstrates the application of HyLogging™ technologies for mineralogical characterisation of different iron ore types throughout the Hamersley Province (Western Australia), including a channel iron deposit (Murchison Metals Ltd) and bedded iron deposits in the Brockman and Marra Mamba Iron Formations (BHP Billiton Ltd).Drill core from these deposits was analysed using CSIRO’s hyperspectral drill core logging systems (HyLogging systems), which are commercially available at the AMDEL and AMMTEC laboratories in Perth (Western Australia), through a leasing agreement. These HyLogging systems provide high resolution photographs at 0.1 mm pixel resolution as well as visible and infrared spectroscopic refl ectance data at 1 cm spatial resolution. The minerals identified include haematite, goethite and its subtypes (ie vitreous and ochreous goethite), AlOH-clays (kaolinite, Al-smectite and white mica), carbonates, gibbsite, chlorite, amphibole and hydrated silica as well as water content. The abundance and composition of these minerals was established from the spectral data and was validated against XRF analyses.The full potential of HyLogging systems in the iron ore industry is only starting to be unveiled, but some major benefits have already been demonstrated:• HyLogging systems can delineate different ore zones by mapping the vitreous and ochreous goethites. This has a potentially large impact in mine planning and operations, as hard, vitreous goethite and soft, powdery, ochreous goethite have different processing characteristics that can affect plant throughput if not carefully managed.• HyLogging mineralogy can also be used to support stratigraphic interpretations of the host, for example in subdividing the Dales Gorge 1 - 4 submembers of the Brockman Iron Formation.• HyLogging can have a potentially signifi cant impact on waste rock characterisation, as many of the waste materials can be identifi ed and quantifi ed with hyperspectral data (eg clay-type,• hydration state).• HyLogging can provide accurate three-dimensional (3D) mineralogy to iron ore companies that will empower them to improve future exploration and resource planning.• Application of high resolution photography to further improve ore deposit domaining is also to be considered.

Jul 11, 2011

By Y. Fu, C. Aldrich

This study investigates the use of dense convolutional neural networks to estimate bubble size distributions in flotation froths. 20 videographically captured images of froths from an industrial platinum flotation plant in South Africa were labelled semantically with an image annotation tool, LabelMe (https://github.com/wkentaro/labelme). As a result, pixels in the images could be categorized into three classes, namely ‘bubbles’ (class 0), ‘bubble boundaries’ (class 1) and ‘background’ (class 2). A U-net architecture with a ResNet34 backbone was trained to segment the images via classification of the pixels in the images. 5 of the 20 images were used to validate the performance of the neural network. Explicit inclusion of the bubble boundaries as a class proved to be significantly more effective than training the network with two classes only (i.e. bubbles and background). However, the thickness of the boundary layer regions were critical. If too thin, it was difficult to distinguish between some contiguous bubbles and if too thick, some fine bubbles may not be identified. The results were promising, although there is still significant scope for improvement, among other by making use of more images for training and validation. Keywords: Flotation Froths, Convolutional Neural Networks, Deep Learning, Bubble size Analysis

Jan 1, 2020