No. 1998.100 Hydrocyclone Model Simulation: A Design Tool for Dewatering Oil Sands Plant Tailings A.I.A. Salama, CANMET, Devon, Alberta, Canada; and T. Kizior, Syncrude Canada, Fort McMurray, Alberta, Canada Abstract Design of classifying hyd rocyclone d ewatering ci rcuits for oil sands plant tailings requi res careful evaluation of d ewatering levels as a function of the cyclone cut size, variations in the cyclone me chanical dimensions, and th roughput. The availability of reliable models can simplify this p rocess and p rovide valuable information that enables the design engineer to understand the p rocess and evaluate di fferent options ( e.g., single- or two-stage cyclone ar rangement, cut size, per cent solids in underflow and feed, and number of cyclones). A hyd rocyclone model simulation app roach has been developed at the CANME T Western Resea rch Cent re (CWRC) for designing d ewatering and fine particle sepa ration circuits. The app roach is based on an existing empirical model. Utilizing a cyclone manufactu rer s published data, some modifications have been d eveloped and int egrated into the model. Plant tailings characteristics, cyclone me chanical dimensions, and ope rating conditions a re utilized in the CWRC modeling simulation. The computer results are presented in 3- D graphs and cor responding 2-D maps showing the cyclone mass recovery, per cent solids by mass in underflow and ove r- flow, and cut size as functions of the cyclone ap ex diameter and cyclone th roughput (i. e., inlet p ressure or pressure drop). The g raphs and maps a re useful in visualizing and illust rating the effects of ope rating conditions on cyclone performanc e. The p roposed computer simulation app roach has been demonst rated th rough the design of d ewatering ci rcuits for oil sands plants. The design includes evaluation of hyd rocyclone performance and d ewatering l evels. A summary of the results is presented. Introduction Conway, 1985 adopted an approach based on Plitt s model (Plitt, 1976), where the cyclone mechanical dimensions are adjusted in a prescribed manner until the desired cut size is achieved. Also the manufacturer s cyclone capacity data are used to determine the required number of cyclones. No attempt was made to check the cyclone underflow per cent solids (e.g., cyclone operating in rope mode or plugged). While the approach may be valid in some cases, it does not provide a thorough insight into cyclone performance as the operating conditions are changed. Again, Plitt s model is used in the computer simulation approach adopted in this paper. The cyclone mechanical dimensions are kept fixed except the apex and vortex finder diameters are changed according to selected manufacturer values. It is known that the apex diameter has significant effects on cyclone performance (per cent solids and mass recovery). For a particular cyclone size, inlet diameter, and overflow diameter, two sets of apex diameters and feed volumetric flow rates are used to study their effects on cyclone performance (cut size, underflow mass recovery, and underflow and overflow per cent solids by mass). In general, the hydrocyclone model simulation objective may be stated as: start with given oil sands plant tailings characteristics and try to predict hydrocyclone performance, in particular, cyclone underflow and overflow per cent solids by mass, underflow mass recovery, and cyclone cut size. These steps are summarized in the following table. Input Data Solids particle size distribution Feed solids and fluid mass rates Feed solids (mass) Solids specific gravity Carrier fluid specific gravity Carrier fluid viscosity Desired cut size, D 50 * Model and Intermediate Data Number of cyclones geometrical dimensions Cut size range Underflow/overflow mass split Sharpness of separation Pressure drop/capacity** Model tuning parameters Output Data Overflow solids (mass) Underflow solids (mass) Overflow solids and fluid mass rates Underflow solids and fluid mass rates Underflow mass recovery Product particle size distributions 1
* D50 is the cyclone cut size (i.e., particle size which has a 50 50 chance of reporting to either the underflow or the overflow streams). **Pressure drop is defined as the difference between inlet pressure and overflow pressure. CWRC Hydrocyclone Model Simulation Approach In Plitt s model, the cyclone model predictions can be determined by utilizing four fundamental parameters expressed in terms of the operating design variables (Plitt, 1976). These parameters are: Separation cut size D 50 Flow split between overflow and underflow Sharpness of separation Capacity/pressure drop By determining these parameters, a complete mass balance together with size distribution of the cyclone products can be achieved. Plitt s empirical model was developed based on a large amount of data collected over wide ranges of operating conditions. The CWRC computer simulation utilized Plitt s model with some changes. These changes were derived based on Krebs Engineers (KE) published data. The details of Plitt s model are given elsewhere (Plitt, 1971, 1976, and further work by Plitt and Kawatra, 1979; Flintoff et al, 1987, see Svarovsky, 1984); however, the modifications are presented next. Pressure Flow Rate Size Correlation Based on Krebs published data the following correlation has been derived 2.12 (1) Q = K cdc Dp 0.498 where K c is a coefficient and is dependent on the cyclone size and p (pressure drop across the cyclone) and Q (cyclone feed volumetric flow rate) are expressed in psi and USgpm, respectively. However, it is straightforward to adjust K c so that p and Q can be expressed in different units. A set of nominal values of K c are given in this table. D c 4" 6" 26" 30" 33" 44" 50" K c 0.481 0.603 0.417 0.374 0.468 0.527 0.619 0.616 0.475 0.539 The K c values could be changed around the nominal values to give similar relationships between p and Q for the same cyclone size. If, for a given cyclone size and known p and Q at a particular operating conditions, then by back substitution of these values in Equation 1 a new K c can be determined. Mular and July 1978 (and reported in Wills, 1992) utilized Krebs published data and derived a similar form for determining the cyclone maximum capacity as 2.12 0.498 Q = K c D c D p (2) Cut Size Correlation Based on KE published data and using nonlinear regression, a correlation for the corrected cut size D 50c can be determined as = F. Dp D 50c D50c 0. 301 0.683 c D m ( 1-1.9C) 1.48 Dr where a viscosity term is added and F D50c includes units conversion. KE engineers have derived a similar form with slight changes to the powers (Gottfried et al, 1982). The difference between the actual cut size D 50 and the corrected cut size D 50c is the actual cut size obtained using the actual classification curve (i.e., including water bypass to underflow). 0.5 0.5 (3) where K c = 9.4 x 10-3 and using SI units. 2
Case Study In late 1997, Syncrude Canada Ltd. decided that the hydrocyclone could be a viable process technology for dewatering plant tailings. The plant approached CWRC to carry out a computer modeling simulation and preliminary engineering design, and to produce in-depth data that would provide a basis for the engineering and installation phases of the development. The plant tailings sand has a typical direct particle sizemass distribution as shown in Table 1. This particle size distribution is used as part of the input files to the CWRC modeling simulation. Note that a top particle size of 250 mm is used in the cyclone computer models. The direct mass distribution is used to generate the cumulative mass distribution. Close examination of the cumulative mass distribution indicates that the desired cut size is in the range of 20 40 mm. Computer Input Data Based on the tailings characteristics, slurry flow rate, and particle size distribution, a set of cyclone sizes (,, ) and a set of feed per cent solids by mass (35, 40, 50, 60) were selected. The solids and liquid specific gravities are 2.65 and 1, respectively. Using a nominal pressure drop across the cyclone, a set of feed volumetric flow rates (using Equation 1) was selected. This allowed better presentation of the results and facilitated visualization of the effects of feed volumetric flow rate and apex diameter on the underflow and overflow per cent solids, cut size, and mass recovery. Using plant tailings total volumetric flow rate in USgpm and Krebs published data, the number of cyclones can be determined. Such values are determined by assuming nominal values of cyclone feed volumetric flow rate. There was no attempt to adjust the calculated number of cyclones to meet KE design. This adjustment can be made at the design stage. KE cyclone mechanical dimensions, in particular, the vortex finderbapex distance and cyclone inlet diameter for the 10-, 15-, and 20-inch Krebs cyclones, were kept constant during computer simulation. Two settings around Krebs cyclone overflow diameters were selected. A set of underflow diameters was selected rather than two settings around Krebs nominal values. This made it possible to evaluate the effect of underflow orifice diameter on cyclone performance. performance is very sensitive to apex diameter and to lesser extent to cyclone inlet pressure, as will be shown later. Computer Simulation Results The single-stage cyclone performance predictions at maximum underflow mass recoveries for different operating settings are summarized in Table 2, where * indicates that the apex is overcrowded (small apex) and sends coarse particles to the overflow stream, and recovery indicates mass recovery. The maximum mass recoveries were obtained at low overflow and apex diameters. The corrected D 50c is dependent on the actual cut size D 50 and the water split (bypass) to the underflow. A selected set of three-dimensional (3-D) graphs and corresponding two-dimensional (2-D) maps (for single-stage and cyclones and 40 and 50 feed solids by mass) are presented in Figures 1-6. These figures are typical of the computer results obtained and are presented to demonstrate the effects of apex diameter and cyclone pressure drop (feed volumetric flow rate, Equation 1) on cyclone performance. Figures 1 and 3 are 3-D graphs for the cyclone with 40 feed solids and overflow diameters of 5" and 6", respectively. Figure 2 is the 2-D map corresponding to the 3-D graph in Figure 1. Figures 4 and 5 are 3-D graphs for the cyclone with 50 feed solids and overflow diameters of 7" and 8", respectively. Figure 6 is the 2-D map corresponding to the 3-D graph in Figure 5. In general, the simulation results for the 10-inch cyclone showed flat surfaces for the underflow per cent solids (i.e., constant high values) over the selected ranges of feed volumetric flow rate and apex diameter. This is because the 10- inch cyclone tends to separate at a low cut size and, as a result, the apex becomes overcrowded resulting in higher underflow per cent solids. The underflow mass recovery and per cent solids over the selected ranges of apex diameters exhibited opposing trends (see Figures 1, 3, 4, and 5). The underflow mass recovery and cyclone actual cut size D 50a over the selected ranges of apex diameters exhibited opposing trends (see Figures 1, 3, 4, and 5). performance (in particular the underflow mass recovery, underflow per cent solids, and cut size D 50 ) is strongly affected by the apex diameter. In the selected range of 20 of the nominal feed volumetric flow rate, the model predictions show unexpectedly small effects of the feed volumetric flow rate (see Figures 1, 3, 4, and 5). Low settings of overflow diameters and high settings of apex diameters produce higher underflow mass recovery with lower per cent solids. This is due to forcing more solids and water to the underflow stream. On the other hand, low settings of overflow and apex diameters produce higher underflow mass recoveries and per cent solids. Furthermore, as the feed per cent solids increases the underflow mass recovery decreases. A two-stage cyclone circuit was simulated where the overflow of stage I is fed to stage II. Stages I and II underflows were combined to form the circuit underflow. Based on the results of single-stage cyclone simulation the suitable apex settings did not change very much in relation to the feed per cent solids. As a result, the same settings used in the singlestage simulation were repeated for the two-stage simulation. The two-stage cyclone circuit performance predictions at the selected settings are summarized in Table 3. The results indicated that the two-stage cyclone circuit mass recoveries are much higher than the single-stage cyclone recoveries; however, the underflow per cent solids is lower in the two-stage cyclone circuit than in the single-stage cyclone. The per cent solids differential decreases as the feed per cent solids 3
increases. Therefore, the two-stage cyclone circuit is recommended for feed with higher per cent solids (50 60). The computer simulation results are predictive in nature and supplementary experimental work is needed to support the predictions Conclusion A hydrocyclone model simulation approach for designing dewatering and fine particle separation circuits has been developed at the CANMET Western Research Centre (CWRC). The approach utilizes an existing empirical model and some modifications based on a cyclone manufacturer s published data are integrated into the model. The published data on cyclone mechanical dimensions and operating conditions are used in the development of the modeling simulation. The computer results are presented in 3-D graphs and corresponding 2-D maps showing the cyclone mass recovery, mass per cent solids in underflow and overflow, and cut size as functions of the cyclone apex diameter and cyclone throughput. The results obtained from an oil sands plant utilizing this computer simulation approach are briefly summarized and presented. The usefulness of the CWRC model simulation has been demonstrated by the utilization of the results in the course of full-scale implementation at several oil sand plants applications. References 1. Conway, T.M., 1985. A computer program for prediction of hydrocyclone performance, parameters, and productsize distributions, Mintek-Report No. M233, Randburg, South Africa. 2. Flintoff, B.C., Plitt, L.R., and Turak, A.A., 1987. modeling: a review of present technology, CIM Bulletin 80:905, 39 50. 3. Gottfried, B.S., Luckie, P.T., and Tierney, J.W., 1982. Computer simulation of coal preparation plants, U.S. Dept. of Energy, DOE/PC/30144-T7, DE 83004279, December. 4. Mular, A.L. and Jull, N.A., 1978. The selection of cyclone classifier, pumps and pump boxes for grinding circuits, Mineral Processing Plant Design, AIMME, New York. 5. Plitt, L.R., 1971. The analysis of solid-solid separation in classifiers, CIM Bulletin 64:70: 42 47. 6. Plitt, L.R., 1976. A mathematical model of the hydrocyclone classifier, CIM Bulletin 69:776: 114 123. 7. Plitt, L.R. and Kawatra, S.K., 1979. Estimating the cut size of classifiers without particle size measurement, Int J Min Proc 5: 369 378. 8. Svarovsky, L., 1984. Hydrocyclones, Holt, Reinhart, and Winston, New York. 9. Wills, B.A., 1992. Mineral Processing Technology, 5 th Edition, Pergamon Press, New York. Acknowledgment This work was supported in part by the Federal Panel on Energy Research and Development (PERD). The authors would like to thank Syncrude Canada Ltd. for permission to publish the present work. 4
Size (mm) Cumulative Mass () Size (mm) Direct Mass () - 10 9.02-10 9.02-20 12.54 10-20 3.52-40 20.15 20-40 7.61-80 26.54 40-80 6.39-90 32.46 80-90 5.92-100 39.31 90-100 6.85-150 78.69 100-150 39.38-200 95.77 150-200 17.08-250 100.00 + 200 4.23 Total 100 Table 1: Sand Cumulative Mass and Direct Mass Distributions With Particle Size Feed Solids Diameter Underflow Cut Size D50a (mm) Solids Recovery 75.3 82.8 38 35 63.3 87.2 37 64.7 79.1 52 40 77.2 67.3 69.4 74.1 82.3 76.3 65 * 42 60 50 78.0 75.1 75.8 57.6 75.7 67.6 117 * 62 87 60 79.1 76.3 73.7 46.9 61.2 50.8 158 * 109 150 * Overcrowded apex (small apex) Table 2: Single-Stage Simulation Results 5
Feed Solids Diameter Stage I Underflow Stage II Underflow Two- Stage Circuit Solids Recovery D 50a mm Solids Recovery D 50a mm Solids Recovery 35 75.3 63.3 64.7 82.8 87.2 79.1 38 37 52 15.9 11.4 18.1 45.2 53.3 50.7 26 23 36 57.0 47.6 49.6 90.6 94.0 90.0 40 77.2 67.3 69.4 74.1 82.3 76.3 65 * 42 60 31.8 18.5 24.3 57.7 55.5 52.9 29 25 39 62.3 52.6 54.9 89.0 92.1 88.8 50 78.0 75.1 75.8 57.6 75.7 67.6 117 * 62 87 61.0 33.8 42.1 66.5 61.1 57.5 41 30 49 71.5 62.5 64.6 85.8 90.6 86.2 60 79.1 76.3 73.7 46.9 61.2 50.8 158 * 110 150 75.3 58.3 63.3 60.6 64.2 55.3 70 53 97 77.5 70.0 69.7 79.0 86.1 78.0 Overcrowded apex (small apex) Table 3: Two-Stage Simulation Results Figure 1: 3-D Simulation Results for 40 Feed Solids,, 5" D o 6
Figure 2: 2-D Simulation Results for 40 Feed Solids,, 5" D o Figure 3: 3-D Simulation Results for 40 Feed Solids,, 6" D o 7
Figure 4: 3-D Simulation Results for 50 Feed Solids,, 7" D o Figure 5: 3-D Simulation Results for 50 Feed Solids,, 8" D o 8
Figure 6: 2-D Simulation Results for 50 Feed Solids,, 8" D o 9