« Back to Articles
This paper discusses three case studies conducted at Highland Valley Copper mine (HVC) that illustrate operational improvements and insights which resulted from using the McGill University gas dispersion sensors. In the first study, the superficial gas velocity (Jg) sensor was used to implement a gas profiling control strategy for the rougher, scavenger and first cleaner cells in the Cu/Mo separation circuit. The data showed a 40% reduction of down-the-bank Cu recovery at constant Mo recovery which resulted from operating the cells at the target profiles. In the second study, the bubble size analyzer was used to better understand two mechanisms of bubble formation in the two flotation columns that make up the final TM Mo cleaning stage: the conventional rubber-sleeve sparger and the Microcel system. The data showed a considerable reduction of bubble size in the Microcel column which resulted in a significant increase in concentrate quality (1-2% Mo absolute) and a dramatic (4-5 fold) increase in concentrate production rate compared to the typical operation. The estimated economic benefit resulting from the improved column flotation operation was an overall 4% increase in Mo circuit recovery. In the third study, real-time measurements obtained from the gas holdup and bulk density sensors were used to infer the pulp density (% solids) in a rougher cell of the bulk flotation circuit. Validation tests demonstrated that this approach provides reliable pulp density measurements. Although promising, more work is required before implementing this approach at HVC.
--------------------------------------- 2
INTRODUCTION It is generally accepted that the way bubbles disperse in a flotation cell plays an important role in metallurgical performance. However, carrying-out gas dispersion studies is frequently considered to be of secondary importance. This situation prevails because, on the practical side, there are only a few publications available providing evidence that significant metallurgical improvements were achieved in a plant environment resulting from gas dispersion studies [1-5]. On the fundamental side, quantitative interpretations describing the effect on metallurgy of changing the gas dispersion conditions in a full-scale operation are quite rare in the literature [6]. One challenge is to obtain reliable measurements. The research group led by Prof. Jim Finch - whose contributions are being honoured at this symposium - has developed a suite of sensors that measure three gas dispersion properties: (1) superficial gas rate (Jg), i.e., the rate at which gas is transported from the pulp phase to the froth layer, with units typically cm/s; (2) gas holdup (µg), i.e., the fraction of cell volume occupied by bubbles; and (3) bubble size (Db), with units typically mm. This paper will discuss three case studies conducted at Highland Valley Copper mine that illustrate operational improvements and insights resulting from using these sensors. Details regarding the working principles of these devices, referred here to as the McGill University gas dispersion sensors, can be found elsewhere [7-10]. HIGHLAND VALLEY COPPER Highland Valley Copper (HVC) is Canada’s largest base metals mine and one of the world’s largest producers of Cu-in-concentrate. The mine, located in the central interior of British Columbia, is owned and operated by Teck. In 2008, HVC produced about 254 million lb of Cu and 4.2 million lb of Mo. The grinding circuit at HVC consists of five lines: three SAG mills (A, B and C) each with two ball mills in parallel, and two AG mills (D and E) each with one ball mill. Each AG mill is equipped with one short head pebble crusher. All ball mill discharge streams are classified using clusters of cyclones, before advancing to the bulk flotation circuit. Bulk Flotation Depending on the grinding line, bulk flotation is carried out in rougher-scavenger banks of either 3333 36.1-m (1275-ft) or 17-m (600-ft) Denver cells. The reagents used are: potassium amyl xanthate and sodium isopropyl xanthate (Cu collectors), fuel oil (Mo collector), Tennafroth 350 and pine oil (frothers), and lime (pH modifier). The rougher concentrate reports to one of two cleaning circuits (one circuit for A, B, and C lines, and another circuit for D and E lines) which produce final bulk concentrate grading approx. 36% Cu and 0.7% Mo. The scavenger concentrate and cleaning circuit tails are recycled to the rougher feed. The final bulk concentrate is dewatered before proceeding to the Cu/Mo separation circuit. Copper/Molybdenum Separation 3 A bank of eight 10-m tank cells is split in half to perform roughing and scavenging duties. The rougher concentrate is reground and then sent to a first cleaning stage. The first cleaning stage is a bank of 3 four 5-m tank cells. The tailings from the first cleaning stage are combined with the scavenger concentrate and then recycled to the bulk concentrate thickener. The concentrate from the first cleaning stage can either 3 be pumped to a second cleaning stage (one 5-m tank cell) or by-passed to the final cleaning stage. When the second cleaning stage is in service, the tailings are recycled to the feed of the first cleaner bank and the concentrate is sent to the final cleaning stage. The final cleaning stage consists of two columns (0.9 m diameter by 10.4 m) in parallel. The tailings from the third cleaning stage are recycled to the feed of the first cleaner bank. The concentrate from the columns, containing typically 49% Mo and 2.5-3.0% Cu, is dewatered and sent to a ferric chloride leach plant where Cu content is reduced to less than 0.25%. Prior to feeding the first rougher cell, the pulp is conditioned with sodium hydrosulfide (to depress the Cu sulfides). To reduce oxidation and consumption of sodium hydrosulfide, N2 is used as the flotation gas throughout the Cu/Mo separation circuit. --------------------------------------- 3
CASE STUDY 1: GAS PROFILING Background Two survey campaigns have been performed thus far at HVC to test the effect of operating the cells in the Cu/Mo separation circuit (Figure 1) at a given gas profile, i.e., a down-the-bank fixed distribution of Jg. The first study focused on the rougher-scavenger bank (cells 1 to 8), whereas the second concentrated on the first cleaning stage (cells 9 to 12). Data from the former study has been previously reported [5]. Bulk Conc ConditioningBulk ConcentrateBulk Concentrate TanksStock TankThickener RoughersScavengers 12345678 Cu Conc 1st Cleaning 9101112 Regrind 2nd Cleaning Column #1Column #2 Final Cleaning Mo Conc Figure 1 - HVC’s copper/molybdenum separation circuit Experimental On a given day, with the circuit operating under steady conditions, the first step was to measure the as-found profile using a version of the McGill University Jg sensor that is capable of collecting multiple measurements at the same time. Then, a metallurgical survey was carried out to obtain the grade-recovery performance of the bank. Subsequently, a target profile was set by adjusting the volumetric gas flow rate (Qg) injected to the cells with the proviso that the total Qg in the bank had to be the same as in the as-found condition. Sufficient time was allowed before conducting the metallurgical survey for the target profile. Results Figure 2 shows the gas profiles tested in the rougher and scavenger cells (left) and in the first cleaning stage (right). The empty and filled symbols indicate surveys performed on different days. Figure 3 shows the metallurgical results expressed as Mo vs. Cu cumulative recovery. It is evident that the parabolic profile in the rougher-scavenger cells and the increasing profile in the cleaners gave better metallurgy than the as-found profiles, approx. a 40% reduction of Cu recovery at constant Mo recovery. --------------------------------------- 4
Figure 2 - Gas profiles as-found and tested Figure 3 - The effect of gas profile on down-the-bank metallurgical performance Implementation If all the bubbles in the pulp dispersion float (i.e., reach the froth layer), then the Jg can be calculated from: Jg = Qg / Ac, where Ac is the cell cross-sectional area. However, significant discrepancies were observed between the Jg measured with the McGill University sensor and the one calculated with the above equation. To illustrate this apparent inconsistency, the total Qg delivered to the first cleaning stage was evenly distributed among the four cells for the as-found case, however, the Jg measurements resulted in a decreasing profile (Figure 2, top-right). A similar situation occurred during the rougher-scavenger study [3]. The problem was traced to the presence of very small bubbles (< 300 ¼m) that were entrained with the tailings. Therefore, the McGill University sensor was chosen as the standard method since it measures the Jg near the pulp-froth interface (i.e., measurements obtained from bubbles that likely floated into the froth). The gas profile approach was implemented by allowing the option to control the N2 addition in cascade mode, i.e., the operator sets the total Qg for a given bank and the controller calculates the individual Qg for each cell to give a target profile. The calculation compensates for the discrepancies described above. Operating in cascade has become the standard practice since it was implemented. --------------------------------------- 5
CASE STUDY 2: BUBBLE SIZE OPTIMIZATION Background A study was conducted to test the effect on overall circuit metallurgy of controlling the size of the bubbles generated in the two flotation columns of the Cu/Mo separation circuit [4]. One of the columns TM (Column #2) was retrofitted with a Microcel sparging system. In the Microcel system, slurry from the bottom of the column is pumped through a static in-line mixer. The gas is injected into the in-line mixer where slurry and gas mix under high shear flow. The turbulence created when the gas-slurry mixture flows through the mixer results in the formation of small bubbles. In this study, the Microcel pump was equipped with a variable frequency drive (VFD). By controlling the flow rate of the Microcel pump, it was possible to vary the turbulence (i.e., gas-pulp shear rate) in the in-line mixer and, thus, to adjust the size of the generated bubbles. Experimental The tests were designed to compare the metallurgical performance of the rubber-sleeve sparger (Column #1) and the Microcel system (Column #2), and to evaluate the effect on metallurgy of bubble size in Column #2. A typical testing sequence began by measuring the bubble size (Db) in both columns. A metallurgical survey was then carried out by collecting pulp samples from the feed, concentrate and tailings streams of each column. Additionally, pulp samples around the rougher, scavenger and first cleaning stages were collected to verify that the circuit was operating under steady conditions. Timed lip samples were collected to measure the concentrate solids flow rate from the columns. At the end of the survey, the bubble size in Column #2 was changed by changing the motor frequency (VFD) of the Microcel pump. The operating conditions in Column #1 were kept constant. Sufficient time was allowed to permit the circuit to stabilize after the change was made prior to conducting the next bubble size measurement and metallurgical survey. Results Before conducting the metallurgical surveys, Column #2 was characterized in terms of bubble size by varying the Jg and Microcel pump speed. Figure 4 shows examples of bubble size distributions including typical measurements from Column #1 and, for reference, from a mechanical (scavenger) cell. A vertical dashed line is located at the 1-mm size class to help assess the proportion (volume fraction) of this critical size [4]. The data reveal considerable differences in bubble size generation. In the case of the mechanical cell, all the gas is dispersed into sub-1 mm bubbles, whereas the conventional rubber-sleeve sparger (Column #1) generates an insignificant quantity of sub-1 mm bubbles. The proportion of sub-1 mm bubbles generated in the Microcel column varies from 10% at high pump speed (70 Hz) and high Jg (1.5 cm/s) to almost 50% at high pump speed (70 Hz) and low Jg (0.5 cm/s). Table 1 shows a summary of the testing conditions for the metallurgical surveys. Surveys 1, 2 and 3 were conducted on the same day, whereas surveys 4 and 5 were carried out on a different day. In survey 2, the speed of the Microcel pump was increased from 30 to 67 Hz, while the froth depth (Hf) was kept constant. This increase in pump speed caused an increase of sub-1 mm bubbles (F1). In survey 3, the Microcel pump speed was set to 67 Hz and the Jg was reduced to half of that used in survey 2. This resulted in a significant increase of sub-1 mm bubbles. The froth depth in survey 3 was adjusted to keep the froth gas residence time (Äf) constant. The operating conditions in Column #1 were kept constant during the three surveys. Because the Mo grade in the rougher feed was low (< 0.9% Mo) during surveys 4 and 5, the concentrate flow rates from the rougher and first-cleaning stages were not sufficient to permit operating both columns simultaneously. Thus, the test work focused on evaluating the effect of bubble size following the same sequence applied in surveys 1 and 2.
-------------------------------------- 6
Figure 4 - Examples of size distributions from various bubble generation methods: a) Mechanical cell; b) Microcel at high pump speed and low Jg; c) Microcel at low pump speed and low Jg; d) Microcel at high pump speed and high Jg; e) rubber-sleeve sparger Table 1 - Column operating conditions during the metallurgical surveys Column #2 Operating Column #1 variable Survey 1-3 Survey 1 Survey 2 Survey 3 Survey 4 Survey 5 [1] Pump, Hz - 30 67 67 30 62 Jg, cm/s 0.8 1.0 1.0 0.5 0.9 0.9 [2] F1, % 1.8 19.1 25.1 40.8 17.2 22.6 Hf, cm 50.8 50.8 50.8 25.4 50.8 50.8 Äf, s 64 51 51 51 56 56 [1][2] Microcel pump speed; volume fraction of sub-1 mm bubbles The results from the metallurgical tests (grade vs. recovery) are shown in Figure 5. It is evident that Column #2 is capable of significantly improving both grade and recovery compared to Column #1, i.e., the grade increased from 49% to 51% Mo, and the recovery increased by a factor of 3-4. The metallurgical grade-recovery results from Column #1 agreed with historical survey data and, therefore, were considered to be characteristic of a rubber-sleeve sparger column. Since the operating conditions in Column #1 were constant, the observed variation in grade and recovery for that column was considered to be caused by experimental error. The Microcel results from surveys 4 and 5 were not as impressive as those obtained from surveys 1 to 3. However, as mentioned earlier, the Mo head grades were quite low. The results in Table 1 and Figure 5 suggest that bubble size was the key factor to improve column performance. In all cases, increasing the proportion of minus 1 mm bubbles resulted in better metallurgical results.
--------------------------------------- 7
Figure 5 - Metallurgical results (survey number indicated within the symbols) In addition to the improved grade and recovery, analysis of the concentrate production rates demonstrated that the Microcel column recovered material 4-5 times faster than the conventional column. As can be seen in Table 2, in surveys 1 to 3, Column #2 produced concentrate at a rate of 550-710 kg/h, while the rates in Column #1 were 120-150 kg/h. This increase in concentrate production rate reduced the circulating load in the circuit causing a significant increase in Mo recovery. An analysis of historical production data [4] showed a strong relationship between final concentrate production rate and overall Mo recovery. The analysis indicated a 4% increase of Mo recovery (from 90% to 94%) and an increase of 1-2% Mo (absolute) in concentrate grade. Table 2 - Concentrate solids flow rate (kg/h) Survey Column #1 Column #2 1 120 590 2 130 710 3 150 550 4 - 440 5 - 610
--------------------------------------- 8
CASE STUDY 3: ON-LINE MEASUREMENT OF PULP DENSITY Background It is well known that adequate control of pulp density is critical for successful operation of a flotation circuit. At HVC, controlling the pulp density in the bulk flotation circuit presents several challenges. Due to the low Cu and Mo head grades, the viability of the operation is largely dependent on maximizing throughput. To maximize the capacity of the primary mills, the main control strategy is to adjust the mill feed rate to maintain a power set point. The control logic also checks the hydrostatic pressure to prevent the mill from overloading. One limitation of this approach is that it can introduce large tonnage variations over short periods, which often cause significant pulp density fluctuations in the bulk flotation circuit via the secondary grinding cyclone overflow stream. Another challenge is that the scavenger concentrate and cleaning circuit tailings streams are recycled to the rougher feed. To adjust the flotation pulp density, the operator manipulates the water added in the cyclone overflow tub and also the scavenger concentrate pulp flow rate (controlled by changing the froth depth, air flow rate, launder water and frother dosage). Because of the lack of a method to measure directly the pulp density in a flotation cell, the operator decides how to manipulate these variables based on visual “clues” and an “estimation” of the residual reagent concentration (mainly frother) recirculating in the system. The author proposes to approach this problem by concentrating efforts to implement a method that can provide reliable on-line measurements of pulp density in a flotation cell. Approach The approach starts by noting that if the bulk density of the dispersion (Áb) and gas holdup (µg) are known, then the pulp density (Áp) can be calculated directly: Áp = Áb / (1 - µg) (1) Since the specific gravity (SG) of the solids is virtually constant in the bulk flotation circuit, equation (1) can be expressed in terms of percent solids: 1-ÁSG p % solids =˜×100 (2) Áp1-SG Results The McGill University gas holdup and bulk density sensors were used to test the approach above. The approach was tested on March 20, 2009, in the third rougher cell of A2 flotation bank (i.e., the bank that receives fresh feed from #2 ball mill in A-line). The results are shown in Figure 6. It can be seen that the pulp density (% solids) was increasing at the time the sensors started collecting data. At 12:37 pm, the pump that recirculates cleaning circuit tailings to the head of the bank (A-line cleaner tails pump) was shut down. This caused a sharp increase of % solids. Then, at 1:35 pm, all the scavenger launder water valves were closed which caused another sharp increase of % solids. At 2:30 pm, A-line cleaner tails pump was restarted, the scavenger launder water valves were reopened, and additional water was added to the bank via cyclone overflow. This resulted in a dramatic pulp density decrease. To validate the data, Marcy scale density measurements were performed on pulp samples collected from the flotation cell (close to the gas holdup sensor). The data were compared with measurements from the proposed approach. The results are shown in Figure 7. It is evident that the proposed method provides reliable data.
--------------------------------------- 9
Figure 6 - Direct measurement of pulp density (% solids) in a flotation cell using the gas holdup and bulk density sensors (the bold line indicates a moving average). A: Shut down the cleaner tails pump in A-line; B: Closed the scavenger launder water valves in A-line; C: started cleaner tails pump in A-line, resumed operation of scavenger launder water in A-line and opened cyclone overflow water valve (25%) Figure 7 - Validation of pulp density measurements
--------------------------------------- 10
CONCLUSIONS " The knowledge gained from using the McGill University gas dispersion sensors has helped develop and implement improved operating practices at HVC; " The Jg sensor was used to devise a gas profiling control strategy for the rougher-scavenger and first cleaner cells in the Cu/Mo separation circuit. The data showed a 40% reduction of down-the-bank Cu recovery at constant Mo recovery which resulted from operating the cells at the target profiles; " The bubble size analyzer was used to better understand the bubble generation mechanisms of the TM conventional rubble-sleeve sparger and the Microcel system. The case study showed that controlling bubble size in the Microcel column resulted in a significant increase in concentrate quality (1-2% Mo, absolute) and a dramatic (4-5 fold) increase of concentrate production rate over the typical operation. The estimated economic benefit resulting from the improved column flotation operation was an overall 4% increase in Mo circuit recovery; " Real-time measurements obtained from the gas holdup and bulk density sensors were used to infer the pulp density (% solids) in a rougher cell of the bulk flotation circuit. Validation tests demonstrated that this approach provides reliable pulp density measurements. Although promising, more work is required before implementing this approach at HVC. ACKNOWLEDGEMENTS The author wishes to acknowledge the management and plant personnel of HVC for permitting the publication of the data in this paper and for the support received during test work. I would also like to express my gratitude to Jeet Basi, Scott Reddick and Frank Laroche for the valuable contributions to the studies presented here, and to Prof. Jim Finch, Jan Nesset and Dr. Cesar Gomez for the many fascinating discussions we have had over the years. REFERENCES 1. R. Dahlke, J.A. Finch, C.O. Gomez, M. Cooper and D. Scott, “Impact of Air Distribution Profile on Banks in a Zn Cleaning Circuit”, Proceedings of the 36 Annual Meeting of the Canadian Mineral Processors, J. Abols, Ed., CIM, Ottawa, 2004, 525-539. 2. J.E. Nesset, J.R. Hernandez-Aguilar, C. Acuna, C.O. Gomez and J.A. Finch, “Some Gas Dispersion Characteristics of Mechanical Flotation Machines”, Minerals Engineering, Vol. 19, 2006, 807-815. 3. J. Pyecha, B. Lacouture, S. Sims, G. Hope and A. Stradling, “Evaluation of a Microcel Sparger in the Red Dog Column Flotation Cells”, Minerals Engineering, Vol. 19, 2006, 748-757. 4. J.R. Hernandez-Aguilar, R. Thorpe and C.J. Martin, “Experiences in Using Gas Dispersion Measurements to Understand and Modify Metallurgical Performance”, Proc. of the 38 Annual Meeting of the Canadian Mineral Processors, C. Hardie, Ed., CIM, Ottawa, 2006, 387-402. 5. J.R. Hernandez-Aguilar and S. Reddick, “Gas Dispersion Management in a Cu/Mo Separation Circuit”, Proceedings of the 6 International Copper-Cobre Conference, R. del Villar, J.E. Nesset, C.O. Gomez and A.W. Stradling, Eds., Vol. 2, CIM, Toronto, 2007, 173-184. 6. J.R. Hernandez-Aguilar and J. Basi, “Improving Column Flotation Cell Operation in a Cu/Mo Separation Circuit”, Proceedings of the 41 Annual Meeting of the Canadian Mineral Processors, R. Henderson, Ed., CIM, Ottawa, 2009, 39-61.
7. C.O. Gomez and J.A. Finch, “Gas Dispersion Measurements in Flotation Machines”, CIM Bulletin, Vol. 95, 2002, 73-78. 8. J.R. Hernandez-Aguilar, R.G. Coleman, C.O. Gomez and J.A. Finch, “A Comparison between Capillary and Imaging Techniques for Sizing Bubbles in Flotation Systems, Minerals Engineering, Vol. 17, 2004, 53-61. 9. J.R. Hernandez-Aguilar and J.A. Finch, “Validation of Bubble Sizes Obtained with Incoherent Imaging on a Sloped Viewing Window”, Chemical Engineering Science, Vol. 60, 2005, 3323- 3336. 10. J.R. Hernandez-Aguilar and J.A. Finch, “An Experiment to Validate Bubble Sizing Techniques Using Bi-Modal Populations of Known Proportions”, Proceedings of the Centenary of Flotation Symposium, G.J. Jameson and R.-H Yoon, Eds., AusIMM, Brisbane, 2005, 465-472.
