Modeling the effects of ore properties on water recovery in the thickening process

International Journal of Minerals, Metallurgy and MaterialsVolume 21, Number 9, September 2014, Page 851DOI: 10.1007/s12613-01 4-0981-yModeling the effects of ore properties on water recovery in the thickeningprocessMajid Unesi'), Mohammad Noaparast-), Seiyd Ziaedin Shafaei , and Esmaeil Jorjani'1) Department of Mining Engineering. Tehran Science and Research branch of Islamic Azad University, Tehran 1477893855, Iran2) School of Mining Engincering, University of Tehran, Tehran 155556365 Iran(Received: 30 January 2014; revised: 21 March 2014; accepted: 12 April 2014)Abstract: A better understanding of solid-liquid separation would assist in improving the thickening performance and perhaps water recov-ery as well. The present work aimed to develop an empirical model to study the effects of ore properties on the thickening process based onpilot tests using a column. A hydro-cyclone was used to prepare the required samples for the experiments. The model significantly predictedthe experimental underflow solid content using a regression equation at a given solid flux and bed level for dfferent samples, indicating thatore properties are the efective parameters in the thickening process. This work confirmed that the water recovery would be increased about5% by separating the feed into two parts, overflow and underflow, and introducing two different thickeners into them separately. This is duoto the fact that thickeners are limited by permeability and compressibility in operating conditions.Keywords: mineral processing; ores; properties; tailings; dewatering; modeling; thickenersunderstanding of the thickening process and predict the rela-1. Introductiontionships between process parameters [15]. Thickeningtheories have been proposed in many literatures [15-25].Mineral processing is usually performed in environmentsFurthermore, chemical engineers have developed severalthat contain a significant amount of water. In many cases,models to simulate the dewatering behavior [26- -28]. Athe majority of the water is removed with tailings. Thus, it isone-dimensional model of dewatering was developed bynecessary to use dewatering equipment such as thickeners inBuscall et al; it quantified the solid volume fraction, theorder to avoid environmental contamination and facilitatecompresive yield stress, and the hindered settling functionwater recycling. Paste thickeners can minimize the volume[15]. Landman and White used a phenomenon model, aof water ulimately consumed and simultaneously reduce thecombination of empirical and theoretical models, to describevolume of solid wastes requiring disposal by dewateringthe solid flux function and effctive solid stress. They alsotailings. Reduction in the volume and cost of tailings dis-determined the subjected forces on particles in the sedimen-posal will increase the ease with which the site can be reha-tation and consolidation process [29- -30]. They found thatbilitated [1-9]. Paste is a term for the suspension of solidsonce solid particles came to rest, the particles began con-which are relatively nonsettling and nonsegregating. As thesolidating and produced a cohesive sediment bed [31].name implies, paste has a self-supporting structure, andThe Buscall and White (BW) theory, developed by Green,when deposited on the ground, will form a slope of varyingmodeled the comprehensive yield stress and hindered theinclination depending on the solid concentration. Pastesettling function in a suspension bed, but not sedimentationproperties are produced by the relatively high concentrationsabove the bed [32]. The Kynch theory was developed byof suspended solids [10-14]. Researchers have presentedGarrido et al, who modeled the sedimentation-consolidationnumerous models and experimental results to facilitate anprocess [33]. The Green theory was developed by De Kret-Corresponding author: Majd Unesit E-mail: majd.unesi@ gmail.cm◎University of Science and Technology Bejjing and Springer-Verlag Berlin Heidelberg 2014中国煤化工包SpringerMYHCNM HG852Int. J. Miner. Metall. Mater, Vol. 21, No.9, Sep. 2014ser et al, who presented a relationship between the bedpresent study aimed to develop an empirical model to studydepth and concentration [34]. Garrido et al. [35] developedthe effect of ore properties such as solid density, particle size,the software for the design and simulation of batches andand mineralogy on the thickening process in order to bettercontinuous thickening. Usher and Scales [36] developed aunderstand the dewatering behavior with thickeners. A pilotsoftware algorithm that account for the underflow concen-column was employed for continuous operation, and its un-tration and sediment bed based on fundamental suspensionderflow solid was measured in cases of different ore proper-properties. This algorithm could not predict the suspensionties to assess the thickening performance and water recov-properties in the shear process, so Gladman [37] improved itery.by shearing the suspension, but did not consider aggregateTo appropriately present relevant experimental datadensification. Then, Nasser and James [38] developed thegathered in this study, the conditions and validation of pilotLandman theory and investigated the effect of pH valuesexperiments were discussed. The pilot experiments providedand electrolytes on the thickener performance. Harbottle et al.greater control over experimental conditions and were, ac-[39- -40] studied the transportation and turbulence propertiescordingly，expected to support the assumptions 0of colloidal suspensions under different surface chemistriessteady-state strategies. Thus, after running the sedimentation(pH values, electrolyte type, and concentration), throughand thickening tests the empirical model based on the resultswhich the minimum transportation velocity of colloidalof those tests was developed.suspensions was well modeled and measured. The suspen-sion dewatering equations were developed by Usher et al.2. Materials and methods[41], based on aggregate densification, whereby the aggre-2.1. Sample preparation and characterizationgate compacted and became smaller when subjected tforces in the thickening process. They presented the liquidTailing samples from Sarcheshmeh Copper Flotationflow velocity around and through aggregates. The validationPlant, located in Iran, were used in all experiments. As aof Usher et al.'s algorithm was undertaken by Gladman et al,hydro-cyclone could separate particles based on their den-who found that it was most accurate at the shortest residencesity and size, the tailing samples were initially and continu-time and the lowest bed height, and the accuracy was re-ously introduced to the hydro-cyclone via a centrally locatedduced at longer residence times and higher beds. This waspipe to separate the tailings into two products: overflow (HO)due to changes in dewatering properties of flocculated ag-and underflow (HU) samples. Particle size analysis andgregates over time, which were not adequately considered inXRD test was then conducted using these samples. Labora-Usher et al.'s algorithm [42]. Lester et al. [43] developed atory screens and a cyclosizer were used for particle sizetwo-dimensional model of the BW theory, using the conti-analysis, and the results are presented in Table 1. The solidnuity, separation, and transport equations in their modeling.density (SG) was determined using a pycnometer (100 cm),Since it was difficult to solve this model by computationaland the results are presented in Table 2 by XRD analysis.methods, they used computational fluid dynamics for mod-eling. Van Deventer' s Kynch theory was based on the ag-Table 1. Particle size analysisgregate densification behavior, so the aggregate densifica-Particle Hydro cyclone Hydro-cycloneHydro-cyclonetion theory was used to predict the final equilibrium bed_sizefeed (HF)_overlow (HO)__ Underlow (HU)_height by densification rate and bed compression. Moreover,dso/ um202213the relationship between aggregate size and thickening timedhis/ um05110was obtained [44]. Birger et al. [45- -50] developed ad5o/ um .1047mathematical and numerical model that is the basis for real-_d2s/ umistic thickener simulators.Numerical, phenomenologicTable 2. XRD analysis and solid density of the samplesdescribed above were derived from routine laboratory batchMineralHOHUsettling and filtration experiments. All of these models wereAluminum silicate (clinochlore,based on fundamental physical laws, and all of them follow75.680.570.4ilite, chamosite, anorthite)/ wt%a series of assumptions. Another limitation of these studiesQuartz (SiO2)/ wt%I726.5is that the effect of ore properties has not been fully consid-Pyrite (FeSx) /w+th3.1ered; most of the research has focused on the effects of sus-Solid density中国煤化工，.8pension properties on the dewatering process. Thus, theYHCNMHG-M. Unesi et al, Modeling the effects of ore properties on water recovery in the thickening process853In order to study the effects of ore properties on the approximately 80% of the target height, at which the under-thickening process, the researchers attempted to run variousflow pump was turned on. This was done to compensate forexperiments in the plant conditions for all samples. Sincethe slow dynamical response in the column. In practice, thethe samples were taken as a suspension from the hy-speed of the underflow pump needed to be periodically ad-dro-cyclone, and were then treated by tailing wastewater,justed to keep the bed height constant due to the variation inthe thickening behavior can only be attributed to ore proper-underflow solid content. Therefore, the underflow rate var-ties. The samples were prepared to the required density (theied with the variation in speed of the underflow pump tofeed concentration in all experiments was 10% solid in sus-control the bed height. Over the course of a run, the bedpension) based on experimental conditions. The pH value ofheight was maintained within 10% of the nominal height. Inthe sample was 11, equal to the industrial pH value, in allthe first time of each run, the average underflow rate in-experiments. The anionic polyacrylamide with a high mo-creased, which corresponded to a decrease in solid contentlecular mass (NF43U from SNF) was used to flocculate theof the underflow as the system moved to the steady state.suspension and was prepared at 0.25 g/L一the industrialFurthermore, the underflow solid content was initially un-dosage. Since the amount of flocculent consumption in thesteady. After a while, this fluctuation died out and the un-Sarcheshmeh paste thickeners is usually between 20 and 30derflow solid content remained fairly constant.g/, the pilot tests were performed at 25 g/t.For each run, the column was allowed to operate for suf-ficient time to achieve the steady state prior to sampling com-2.2. Pilot columnmencing. On this basis, the steady-state conditions wouldA plexiglass column, which was 4 m in height and 0.2 mtake anywhere from 2 to 5 h, depending on the bed height.in intermal diameter, was used for pilot experiments. TheAfter this time had elapsed, column underflow samples werefeed and flocculent were introduced to the column at acollected at 15 min intervals. No solids were recorded in theheight of approximately 3 m via a feed-well by peristalticoverflow for each condition used in the experiments, and thepump. The dilution water was also introduced at this columnsupernatant was always observed to be clear.height. Overflow was then collected by a peripheral launder.The interface between the feed and clarified water could be3. Results and discussionobserved. Suspensions were then transferred to a hy-dro-cyclone equipped with overflow and underflow boxes.A series of experiments were performed in pilot scales toBased on the experimental condition, either HO, HU, or HFstudy and model the thickening behavior. Their results arewas transferred to the pilot column. Shortly after the columnpresented below.was flld, three zones were distinguishable. The top of the3.1. Discontinuous testscolumn was characterized by a clarified zone. In the middlezone, individual aggregates were observed to be settling.Fig. 1 presents the bed formation curves, where the col-The lower zone was an opaque region in which the solids umn was discontinuously operated at different solid fluxessettled at a slower rate. There was a marked interface be-(y4). Bed formation in the HF samples tends to the HO curvetween the middle and lower zones that is referred to hereinat low ψ (10 t:m-.d-) and tends to the HU curve at high ψas the“bed height" or“bed depth". It should be noted that(28.5 t:m-2.d). The long time at a lower ψ creates an op-this interface was not necessarily the point at which the sol-portunity for fine particles to settle easily. Therefore, highids became networked.solid density and particle size did not play a decisive role inIn addition, discontinuous tests were conducted in thebed formation, but their effects appear at higher y. In addi-start-up step. After column start-up, the bed of solids begantion to solid flux, bed formation depends on compressibilityto form, and the relevant interface could be readily detectedcaused by smooth slope changes in some parts of the curves.(by eye). The bed height gradually increased as solids acBed compression usually occurred at the 2- -2.5 m columncumulated, and when the solid bed reached the desiredheight. This was not observed in the HO and HU samples,height (0.5 to 2.5 m), the time was recorded using a chro-except at a higher ψ of 28.5 tm 2.d-' (with the HO samples).nometer. In continuous tests, several column experimentsThis effect was repeated in the HF samples, which could bewere performed. The results for solid flux (solid flow rate indue to coarser and finer particle interactions or the interac-feed or underflow per unit area), bed height, and ore proper-ion of clay, quartz, and metallic minerals. Therefore, it isties (particle size and solid density) for each run are pre-postulated that中国煤化工and the domi-sented in Table 3. The bed height was allowed to increase tonant minerals inMHCNMHG854Int. J. Miner. Metall. Mater, Vol. 21, No.9, Sep. 2014Table 3. Summary of pilot column experimentsActual level of variablesRun Bed depth,Solid flux, ψ1Solid density/ Run Bed depth, Solid flux, ψ1Solid density 1dso/ umdsao/ umh/m(tm2.d)(t.m7 3(tm 3.5122.8).52.7.028.5120.83214.350.72.0200382.540l08.530242130314445l61046781.59215(4.35C2324252729be noted that when the Qu values of the HU samples were 453.2. Continuous testsand 13 Lh, respectively, the difference was 32 L/h, becauseFig. 2 shows a series of curves of volumetric underflowthe compressibility of the HU samples was more than that ofrate (Qw) versus underflow solid content for the various pilotthe HO samples.column experiments outlined in Table 3. The Qu of the HOIn addition to compressibility, permeability is an effec-samples was more than that of the HU and HF samples attive parameter of paste thickener performance. This pa-different solid fluxes because of rapid bed formation in therameter was much more evident at a higher ψ. The relation-HO samples. Furthermore, the difference in Qu of the HOship between higher underflow solid content and bed heightsamples at different solid fluxes was more than that of theat a lower ψ is postulated to be due to the aggregate rear-two others. For instance, at a column height of 2.5 m, the Q .rangement over time, a phenomenon that is not well under-values of the HO samples at 28.5 and 10 t:m 2.dI were 94toodut is ko中国煤化residence timeand 27 Lh, respectively, the difference is 67 L/h. It shouldHowever, bed hlerflow contentTYHCNMHGM. Unesi et al, Modeling the effects of ore properties on water recovery in the thickening process855at high ψ, because there was not enough time for compres-Qu and settling rate. Thus, the thickener performance couldsion in the thickener, and underflow content was dependantdecrease, and changes in ore properties could significantlyon permeability. In this case, the thickener worked based onaffect the thickener performance.3.02.2.01.◆HF sample81.0■HO sample0.▲HU sample0.50.00.(04(5010) 60Time / minTime 1 min3.(|(c2.5(d,▲▲▲▲E 2.(E 2.蓉1.3营1.5家1.00.500.050020150200Time 1 mirFig. 1. Bed formation curves for different fluxes: (a) ψ= 28.5 tm-2.d'; (b) y= 20 tm-2.d'; (c) ψ= 14.3 tm-2:d'; (d) ψ= 10tm-:d".160- For 0.5 m height◆HTF (28.5t-m~2.d+)14■HO (28.5t-m2.d+)▲HU (28.5 tm-d)一For 2.5 m heightX HTF (20tm-2.d")0t* HO (20tm2.dI) .●HU (20 t:m2.d+)+ HTF (14.3 t:m2.d~)c 60--HO(143 tm-.dH)-HU(14.3 tm2.d1)0F◆HTF (10tm2.d)▲★▲二■HO(10tm2d-) .▲HU(10tm-2.d-)2025303540455()5:56065Underflow solid content 1 wt%Fig. 2. Volumetric underflow rate Q) as a function of underflow solid content (UF) at different ψ and h for three samples (HF,HO, and HU).The packing of particles on top of each other during set-more in the HO samples, typically stack into a honeycombtling influences the rheology and solid content of the thick-structure. Honeycomb packing retains large amounts of wa-ened underflow. The most familiar packing type is referredter, as water flls voids formed by the honeycomb structure.to as “edge-face”, which is also known as the “house ofGreat effort and highly effective dewatering systems are re-cards",“face-face", and“band structure”packing relation-quired to remov中国煤化工1 the thickenership [51]. Clays and finer particles, of which there wereperformance isHCNMHG856Int. J. Miner. Metall. Mater, Vol. 21, No. 9, Sep. 20143.3. Modeling(< 0.0001) and a very high determination coffcient (0.985).This indicates the precision and reliability in the experi-3.3.1. Regression modelAfter collcting the full data set, linear regression wasments. The model is restricted to steady-state conditionsemployed to model the relationships among the underflowwith an almost constant value of SG, and cases in whichsolid content, ore properties, and process parameters. In ad-there is no overloading (i.e, all fed solids are conveyeddition, the experimental results were analyzed statisticallythrough the underflow of the column).Results obtained from the experiments are summarized inusing the analysis of variance (ANOVA) to assess the ade-quacy of the model. This model was created using SPSS 16Table 4. The linear response function representing the un-and Minitab 14 software.derflow solid content can be expressed as a function of twoUsing SPSS, the effects of all model terms were calcu-processes and two ore parameters:lated, and a regression model was selected. The model wasUF = 529.776 + 6.007h - 0.305 ψ+ 0.629dso- 198.7SGhighly significant, with a very low probability value(1)Table 4. Observed and predicted values of the underflow solid contentVolumetric flow rate /(LhTUF/%Volumetric flow rate/(Lh-'UF1%Run Feed, Qr/Underflow, Qu/Actual Predicted| Run Feed, Qp/(Lh-)Lh-)(Lhl)3508636.9|31815223.923.442.039.9815327.026.444.442.9| 334730.029.45048.345.9| 344531.032.451.28.9|354133.735.4245635.99.5| 3(12725.024.742.8| 373327.745.245.5| 383131930.7272932.803451.951.5| 4034.636.775|443.223014.3| 4246.246.8| 4.49.2142649.050.50.552.255.2| 4:1553.916222342.6| 4(4245.83748.882047.548.6| 4851.8| 4956.554.8122555.854.6| 5(2858.357.86320.719.0|546.036313122.922.117553.225.54.553.62410228.028.1| 542157.156.62594| 5.1861.359.62549221.6| 561948.95426.024.627.6| 5854.929.00.7| 5<57.931.9|_ 6(中国煤化工;60.9YHCNMHGM. Unesi et al, Modeling the effects of ore properties on water recovery in the thickening process857The observed and predicted values using the model equa-were normally distributed with constant variance, it can betion (Eq. (1)) are given in Table 4. The predicted valuesconcluded that Eq. (1) fitted adequately to experimental data.match the observed values reasonably well. Table 4 indi-Further analysis of the results using the correlation plotcates the highest and lowest UF were 61.3% and 20.7% ob-showed that the ore properties have a more significant in-tained in Runs 60 and 21, respectively. The bed depth, solidfluence than the operating parameters (Fig. 5).flux, and particle size were 2.5 m, 10 t:m 2.dr", and 130 um70for Run 60 and 0.5 m, 28.5 tm 2.dr', and 50 μum for Run 21,respectively. It could be concluded that the underflow solidi 5(content was more significantly affected by ore properties;4than process parameters.! 3(The regression equation was then used to optimize thewater recovery within the studied experimental range. Th10optimum process parameters were h= 2.5 m and ψ= 14.3t:m 2.d', with a maximum predicted UF of 53.3%, whereas10203040506(the maximum UF was 55.2% in the experiments conductedUnderflow solid content observed / wt%as shown in Table 4 (ie, a 1.9% difference in UF could beFig. 3. Relationship between the experimental and predictedunderflow solid contents.obtained using the regression model1). .3.3.2. Model adequacy99.9The ANOVA results are summarized in Table 5. The fit-ted model was checked to determine whether it adequately二95.0estimated the true response surface. This was accomplishedusing the cofficient of determination (R). The R2 valueprovides a measure of how much variability in the observedresponse values can be explained by experimental variablesand their interactions. The closer the R value to 1, the better0-the model is believed to predict the response. The result im-.1 L-4-3-2-1012345plies that the regression is significant. Further, the overallResiduallack of fit test was significant at P = 0.01, using the MinitabFig. 4. Normal plot of residuals using Minitab.software.Table 5. Analysis of variance (ANOVA) for the significance ofSolid densitythe model点Particle sizeSumof Degree of MeanSignifi-Model .Fishersquaresfreedom squarecanceSolid fluxRegression 7615.6601903.915 875.770 0.000Bed depthResidual 119.569 .55Total59-0.4 -0.00.20.4).6.8.0Correlation efectsfig. 5. Correlation effects of operating parameters on thehe relationship between experimental and predictedunderflow solid content.underflow solid contents is presented in Fig. 3. The pre-dicted values are reasonably comparable to the experimental3.3.3. Significant factorvalues, with a linear correlation coefficient (R) of 0.992.3.3.3.1. Ore propertiesStatistically, this means that the model explains 99.2% ofThere is a significant correlation among particle size,the variability of the response data around its mean.solid SG and ore mineralogy. This suggests that the SG wasFig. 4 is a normal plot of residual values. As ilustrated indecreased by the reduction in particle size, which was due toFig. 4, all residuals lie on a straight line with a linear corre-the transferring of clay minerals to finer particles (in the HOlation coefficient of 99%, which indicates that the residualsamples). It is u中国煤化rifference in thevalues were normally distributed. Since the residual valuesformation of floYHCNMHGto the interac-858Int. J. Miner. Metall. Mater, Vol. 21, No. 9, Sep. 2014tion of particles. It was also observed that the thickener per-were about 55.5wt% (at the 0.5, 1, 1.5, and 2 m columnformance decreased with the reduction in particle size andheights, the underflow solid contents were 42wt%, 44.4wt%,the effects of ore properties were more significant than the46.8wt%, and 49wt%, respectively). Similar results wereprocess parameters described below.observed at all bed depths in the HF samples and at bed3.3.3.2. Solid flux (4)depths over 1 m for the HU samples, but were not observedSince the permeability of clay was very low, the thick-in the HO samples. It is postulated that this behavior wouldener performance was expected to decrease with the in-be observed at column heights more than 2.5 m or solidcrease of clay content in the ore. In this case, the lower ψfluxes less than 10 tm' 2.d 1 in the HO samples. These sam-was preferred, because the retention time increased andples need further time to be compressed, which will be ob-caused the thickener performance to increase. The bed depthtained at higher depths or lower solid fluxes. Furthermore,had significant effect on the underflow content, increasing itfurther pressure would apply at higher depths.at a lower y. In the HO samples, decreasing the solid fluxTherefore, it is understood that there was a difference infrom 28.5 to 10 t:.m °.d~ caused an increase in underflowthe formation of flock networks which was due to the inter-content from 30wt% to 34.6wt% at the 2.5 m column heightaction of particles. This interaction suggests a relationship(Tables 3 and 4). This phenomenon could also be observedbetween the solid bed depth and underflow solid contentin the HU samples, which had a less clay content. In thesethat was not observed in previous studies. Thus, it is sug-samples, decreasing the solid flux from 28.5 to 10 t.m 2.d-'gested that this phenomenon depends on ore properties, suchincreased the underflow content from 53.9wt% to 61.3wt%as particle size, solid density, and mineralogy (clay content,at the 2.5 m column height. The increase in solid content inquartz, and metallic minerals) that were considered in thethe HO samples was 4.6wt%; it was 7.4wt% in the HUpresent study. Based on the above, the optimum ψ and hsamples. This means that the compressibility and permeabil-could be obtained for desired underflow solid content.ity of the HU samples were much better than those of the3.4. Fractional water recovery (R)HO samples.According to measurements and calculations, the weightsFractional water recovery is defined as the water over-of solids in the HO and HU samples were 10% and 90% offlow flow rate divided by the total flow rate of water fed tothe HF samples, respectively. The underflow solid contentsthe column, assuming that all solids leave via the underflow.of the HF and HU samples at a yof 14.3 t:m-.drl wereTable 6 demonstrates that the higher solid bed depth and55.2wt% and 61.3wt%, respectively, at the 2.5 m columnlower ψ increased water recovery, and the fractional waterheight. The significant increase in underflow solid contentrecovery in the HU samples (RLHu) was more than that inwas also observed at other solid fluxes. Therefore, ore prop-the HO samples (RuHo). For instance, the water recovery forerties played a decisive role in the paste thickener perform-HF (Rur), RuHO, and Ruu at the 20 tmi 2.d 1 solid flux andance. It was also observed in batch tests that the thickening2.5 m column height were 63.80%, 63.36%, and 69.23%,behavior of the HF samples tends to the HO curve at low ψrespectively. Fractional water recovery would be an alter-(10 t:m?.d-) and tends to the HU curve at high ψ (28.5nate method to determine the thickener performance.tm 2.d-). However, in continuous tests, the increase in un-As mentioned in previous sections, by using a hy-derflow solid content behavior of the HF samples was thedro-cyclone, the weights of solids in the HO and HU sam-same as that of the HU samples at high ψ (28.5 t:m 2.d )ples were 10% and 90% of HF, respectively. Therefore theand the same as that of both the HO and HU samples at low amount of water recovery improvement (ARL) achieved byψ(10 t:m-2.d ). In other words, the impact of aluminosili-using the hydro-cyclone is given bycates on underflow solid content was increased by decreas-AR = 0.1RuHO +0.9RuHu- RuHr .(2)ing yto 14.3 t:m 2.d 1. Futher decreases in ψ did not haveThus, there were two options: (1) the feed was introducedsignificant effect on the underflow solid content.to the thickener directly; (2) the feed was initially introduced3.3.3.3. Bed depth (h)to a hydro-cyclone, and then the HO and HU were fed toAs described above, decreasing 4 impacted the under-two different thickeners.flow solid content and bed depth, but this impact was lim-The water recovery was increased, using a hydro-cycloneited, and further reduction in ψ eliminated the effect. Thisin the feed line (option 2) instead of the tailings being di-issue was considered in the HF samples at 10 and 14.3rectly introduced to the thickener (option 1). The improve-t:m 2-d-. As shown in Tables 3 and 4, the underflow solidment is shown中国煤化工yclone, the Rcontents for both solid fluxes at the 2.5 m column heightwas increased aHCNM HG'column heightM. Unesi et al, Modeling the efcts of ore properties on water recovery in the thickening process859Table 6. Fractional water recovery at different bed heights (h) and solid fluxes (4)_y/(t.m2.d)_ h/mHF samples HO samples HU samples| ψ/(t-m-.d-)_ h/m HF samples HO samples HU samples0.524.4133.4142.5146.19 .45.0249.451.46.1941.5846.1551.2753.4662.1328.51.248.1153.8514.31.555.6459.8764.102.058.2355.7857.69 .59.4461.6867.642.62.7763.132.568.3566.3672.780.30.381.047.9750.9152.7958.871063.3665.0357.9166.8064.9163.8069.2369.1667.716-that the particle size distribution and the dominant mineralsin the samples would cause the compressibility effect in bedformation.(2) The results of continuous experiments show that thevolumetric underflow rate (Qw) of the HO samples was morethan that of the HU and HF samples at different solid fluxesdue to rapid bed formation and less compressibility in theHO samples. In addition to compressibility, permeabilityI52025~3ψ/(tm-2.d-1was an effective parameter that was much more evident at aFig. 6. Amount of water recovery improvement as a functionhigher y, because the thickener worked based on Qu andof solid fluxes at 2.5 m column height.(3) Predicted values using the regression model equation(ie,△R = 3%- -5%) and will be the same as RuH at thewere in pretty good agreement with the observed values (R2 =solid flux of 28.5 t:m~2.d 1 (i.e.,△RL = 0). As seen in Table 6,0.985). The optimum process parameters were h = 2.5 mthe maximum Rur was obtained at the 2.5 m column heightand ψ= 14.3 tm- 2.d~, with a predicted maximum UF ofat the solid flux of 14.3 and 10 t:m-2.d-', at which R was53.3%, whereas the maximum UF was 55.2% in the ex-increased about 3% to 4% by this new methodology. Thisperiments conducted (i.e, a 1.9% difference in underflowbehavior can be attributed to the effect of ore properties onsolid content could be obtained using the regression model).the compressibility and permeability of the samples in theAnalysis of the data showed that interaction among factorsthickening process, as clarified in the model. It is predictedhad less influence on the thickener performance than orethat the thickener performance could be improved more thanproperties, and further analysis using the correlation plo5% by lowering the ψ or increasing the bed depth of the HOshowed that ore properties have a more significant influencesamples.than operating parameters.There are 12 thickeners in the Sarcheshmeh Copper Mine,(4) By using a hydro-cyclone, the water recovery in-and the tailings are introduced to these thickeners by distri-creased about 3% to 5% at the 2.5 m column height, and thebution tanks. Based on the results obtained, the recommen-recovery was the same as the feed samples at the solid fluxdation could be made to use and adjust a hydro-cyclone inof 28.5 tm 2.dr'. It is predicted that the thickener perform-the feed line and lead the HO to one or two thickeners andance could be improved more than 5% by increasing the bedthe HU to another.depth or lowering the ψ in the HO samples.4. ConclusionsReferences(1) The results of discontinuous experiments show that[1] C. Maurice, N.H. Fuerstenau, and N. Kenneth, Principles oflong residence time at a lower ψ created an opportunity forMineral Pro中国煤化工Mining, Metal-fine particles to settle easily. In addition, it was postulatedlurgy and E，p.340.YHCNM H G'860Int. J. Miner. Metall. Mater, Vol. 21, No.9, Sep.20142] A.B. Fourie,lian Centre for Geomechanics, University of Westem Austra-[23] D.A. Dahlstrom and E.B. Fitch, Mineral Processing Hand-lia, Perth, 2006, p. 5.book, N.L. Wiess, ed., Society of Mining Engineers of the3] J.S. Slottee and J. Johnson, Paste Thickener design and op-American Institute of Mining, Metallurgical, and Petroleumeration selected to achieve downstream requirements, [in12th International Seminar on Paste and Thickened Tailings,[24] T. Yalcin, Sedimentation characteristics of Cu-Ni mill tail-Chile, 2009.p. 6.ings and thickener size estimation, CIM Bul, 18( 1988), No.4] N. September and R. Kirkwood, Clermont coal mine project910,p. 69.selection of tailings paste thickener, [in] AusIMM - Technical25] E.G. Kelly and D.G. Spottiswood, Introducion to MineralMeeting, Rio Tinto, 2010. p. 10.Processing, Mineral Engineering Services, Denver, 1982.5] A.L. Mular, D.N. Halbe, and D.J. Barratt, Mineral Process-26] R. Burger, M.C. Bustos, and F. Concha, Stling velocities ofing Plant Design, Practice and Control Proceedings Hand-particulate systems: 9. Phenomenological theory of sedimen-book, Lttelton Colorado, 2002, p. 1295tation processes: numerical simulation of the transient be-6] htp://www.metsominerals.com, Brochure, 2011.haviour of flocculated suspensions in an ideal batch or con-7] J.H. Perry, Chemical Engineering Handbook, MeGraw-Hilltinuous thickener, Int. J. Miner. Process. 551999), No. 4, p.Companies, New York, 1950. p.250.267.8] htp://www .outotec .com, Paste Brochure, February, 201 1.[27] F. Concha, M.C. Bustos, and A. Barrientos, Sedimentation of9] F. Sofra and D.V. Boger, Slope Prediction for thickened tail-Small Particles in a Viscous Fluid, Computational Mechanicsings and paste, tailings and mine waste, [in] Proceedings ofPublications, Southampton, 1996, p. 51.the 8th International Conference on Tailings and Mine Waste,[28] M.C. Bustos, F. Concha, R. Birger, and E.M. Tory, Sedi-Rotterdam, 2001, p.75.mentation and Thickening, Kluwer Academic Publishers,[10] P. Newman and D. Landriault, The use of paste technology inDordrecht, 1999, p. 285.the surface disposal of mineral waste, [in] Waste Mininiza-[29] K.A. Landman, L.R. White, and R. Buscall, The continu-tion and Recycle Conference, Birmingham, 1997. p. 55.ous. flow gravity thickener: steady state behavior, AIChE J.,[11] T. Meggyes and A. Debreczeni, Paste technology for tailing34(1988), No. 2, p. 239.management, Land Contam. Reclamat, 14(2006), p. 41.[30] K.A. Landman and L.R. White, Solidiquid separation ofI 12] htp://www .dorrolivereimco.com, Brochure, 2004.flocculated suspensions, Adv. Colloid Interface Sci, 51(1994),[13] I. Arbuthnot, B. Garraway, R. Triglavcanin, T. Edwards, D.p.175Colwell, and K. Roberts K., Designing for paste thickening,[31] M.S. Nasser, FA. Twaiq, and S.A. Onaizi, An experimental[in] Proceedings of the Canadian Mineral Processors, Perth, .study of the relationship between eroded flocs and cohesive2005, p.597.beds in flocculated suspensions, Miner. Eng., 30(2012), p. 67.[14] W.A. Cinila, D.A. Landriault, and R. Verburg, Application32] M.D. Green, Characterisation of Suspensions in Settling andof paste technology to surface disposal of mineral wastes,Compression [Dissertation], University of Melbourne, Aus-tailings and mine waste, [in] Proceedings of the 4th Interma-tralia, 1997,p. 51.tional Conference on Tailings and Mine Waste, Rotterdam,33] P. Garrido, F. Concha, and R. Burger, Setting velocities ofparticulate systems: 14. Unified model of sedimentation, cen-[15] R. Buscall, P.D.A. Mill, R.F. Stewart, D. Sutton, L.R. White,trifugation and filtration of flocculated suspensions, Int. J.and G.E. Yates, The theology of strongly-flocculated suspen-Miner. Process, 72(2003), No.1-4, p. 57.sions, J. Non-Newton. Fluid Mech, 24(1987), No. 2, p.34] R.G. De Kretser, D.V. Boger, and PJ. Scales, Compressiverheology: an overview, Rheol. Rev, (2003), p.125[16] H.S. Coe and G.H. Clevenger, Methods for determining the35] P. Garrido, R. Burgos, F. Concha, and R. Biurger, Softwarecapacities of slime thickening tanks, Trans. AIME, 55(1916),for the design and simulation of gravity thickeners, Miner.p.356.Eng, 16(2003), No.2, p.85.[17] GJ. Kynch, A theory of sedimentation, Trans. Faraday Soc,[36] S.P. Usher and PJ. Scales, Steady state thickener modelling48(1952), p.166.from the compressive yield stess and hindered settling func-[18] W.P. Talmage and E.B. Fitch, Determining thickener unit ar-tion, Chem. Eng. J,, 111(2005), No.2-3, p. 253.eas, Ind. Eng. Chem, 47(1955), No. 1, p.38.[37] B.R. Gladman, The Effect of Shear on Dewatering of Floc-[19] N. Yoshioka, Y. Hotta, s. Tanaka, S. Naito, and s. Tsugami,culated Suspensions [Dissertation], The University of Mel-Continuous thickening of homogeneous flocculated slurries,bourme, Melbourne, 2005, p.55.Kagaku Kogaku, 21(1957), No.2,p 66.[38] M.S. Nasser and A.E. James, Numerical simulation of the[20] B. Fitch, Current theory and thickener design, Ind. Eng.continuous thickening of flocculated kaolinite suspensions,Chem., 58(1966), No.10, p.18.In. J. Miner. Process., 84(2007), No.1-4, p.144.[21] L. Svarovsky, Solid- Liquid Separation, Butterworth Heine-[39] D. Harbottle, M. Fairweather, and S. Biggs, The minimummann Press, London, Boston, 2000, p. 75.transport vel中国煤化工Tions, Chem. Eng.22] JH. Wilhelm and Y. Naide, Sizing and operating continuousSci, 66(201 IfYHCNMH G .M. Unesi et al, Modeling the effects of ore properties on water recovery in the thickening process6I[40] D. Harbtle, s. Biggs, and M. Fairweather, Fine particleequation for the flux Jacobian, SIAM J. Appl. Math, 70(2010),turbulence modulation, AIChE J.. 57(2011), No. 7, p.1693.No. 7, p. 2186.[41] S.P. Usher, R. Spehar, and PJ. Scales, Theoretical analysis of47] R. Birger, K.H. Kalsen, H. Torres, and J.D. Towers, Sec-aggregate densification: impact on thickener performance,ond-order schemes for conservation laws with discontinuousChem. Eng. J, 151(2009), No. 1-3, p.202.flux modelling clarifier- -thickener units, Numer. Math,[42] B.R. Gladman, M. Rudman, and PJ. Scales, Experimental116(2010), No. 4, p. 579.validation of a 1-D continuous thickening model using a pilot48] R. Birger, S. Diehl, and I. Nopens, A consistent modellingcolumn, Chem. Eng. Sci, 65(2010), No.13, p. 3937.methodology for secondary stting tanks in wastewater[43] D.R. Lester, M. Rudman, and PJ. Scales, Macroscopic dy-treatment, Water. Res, 45(2011), No. 6, p. 2247.namics of flocculated colloidal suspensions, Chem. Eng. Sci,[49] R. Birger, S. Diehl, S. Faras, and I. Nopens, On reliable and65(2010), No.24, p. 6362.unreliable numerical methods for the simulation of secondary[44] B.B.G. van Deventer, S.P. Usher, A. Kumar, M. Rudman,setting tanks in wastewater treatment, Comput. Chem. Eng.,and PJ. Scales, Aggregate densification and batch settling,41(2012),p. 93.Chem. Eng. J, 171(2011), No.1,p.141.[50] R. Birger, S. Diehl, S. Faras. I. Nopens, and E. Torfs, A con-[45] R. Birger, A. Coronel, and M. Sepulveda, Numerical solutionsistent modelling methodology for secondary settling tanks: aof an inverse problem for a scalar conservation law modelingreliable numerical method, Water Sci. Technol,. 68(2013), No.sedimentation, [in] E. Tadmor, J.G. Liu, A.E. Tzavaras, eds,1,p. 192.Proceedings of Symposia in Applied Mathematics, Provi-[51] A. Mcfarlane, K. Bremmell, and J. Addai-Mensah, Micro-dence, 2009, p. 445.structure, rheology and dewatering behaviour of smectite[46] R. Burger, R. Donat, P. Mulet, and CA. Vega, Hyperbolicitydispersions during orthokinetic flocculation, Miner. Eng.,analysis of polydisperse sedimentation models via a secular18(2005), No.12, p.1173.中国煤化工MYHCNMH G