Anglo research (AR) experience with integrated comminution and flotation plant modelling

Transcription

1 SAIMM_Oct_37-46:Templte Jounl 11/5/07 8:11 AM Pge 641 Anglo esech (AR) expeience with integted comminution nd flottion plnt modelling by S. Nik* nd W. vn Dunick* J o u n l Synopsis To dte, Anglo Resech hs cied out sevel pojects, bsed on the techniques developed in the AMIRA P9 (AMIRA Intentionl Limited), to develop comminution nd flottion models fo sevel Anglo Bse Metls nd, to lesse degee, Anglo Pltinum flottion plnts. Plnt optimiztion is chieved though stedy-stte smpling, model-fitting nd simultion, using the JK SimMet nd JK SimFlot softwe pckges. The suveys cn vy depending on the needs of the plnt/client, nd cn theefoe include s little s single piece of equipment, n entie milling/flottion cicuit, o complete cicuit fom pimy comminution to finl flottion concentte (i.e. n integted suvey ). The pesence of the AR tem on site usully pecipittes fcevlue obsevtions tht cn led to initil impovements. This stems fom the knowledge-shing on best pctice tht tkes plce in pllel with the suveys. Numeous (esonble) chnges cn then be simulted. These indicte the optimiztion step most likely to esult in the best efficiency impovements. The simultions cn include chnges in the flow sheet, oe chcteistics, mill lods, ginding medi size, gind, flottion esidence time nd i hold-up. Plnt suveys cn esult in pocess chnges tht povide the plnt with highe plnt tonnge, impoved plnt gind, dditionl flottion ecovey nd/o impoved poduct gde. An dditionl benefit of pefoming n integted plnt suvey is tht the effects of n optimiztion chnge in the comminution cicuit cn be simulted using the flottion model. This ppe descibes numbe of cse studies of individul nd integted plnt suveys conducted by AR fo Anglo Bse Metls. Intoduction Anglo Resech (AR), on behlf of Anglo Ameicn plc ( sponso of the AMIRA P9 poject), hs become involved in the AMIRA P9 comminution nd flottion modelling pojects ove the lst few yes. Duing this peiod, AR hs cied out detiled smpling nd mesuement cmpigns in both the comminution nd flottion concenttos of vious Anglo bse metl opetions on n individul bsis, nd integted suveys s descibed in Tble I. The focus of this ppe is not to exploe in detil ny pticul spect of this pocess. Insted the uthos im to descibe (biefly) numbe of ctivities tht compise typicl modelling cmpign, nd to povide exmples nd evidence of how pocessing plnts hve benefited fom this integted suveying ppoch. The on-site wok descibed in this ppe ws cied out by AR metllugists with the ssistnce of the pesonnel woking on the site in question. JKTech stff wee pesent duing the fist cmpigns fo both comminution nd flottion, which lso functioned s tining execise. As esult, AR now cies out typicl P9 execises, lthough JKTech stff hve sometimes emined involved s consultnt. AR hs gined expeience duing the couse of this poject which hs helped these execises to become incesingly efficient. Desciption of modelling pocess Typiclly, pio to the ctul smpling cmpign, fist tip to the site is mde to infom ll concened bout the ims of the execise s well s to meet potentil ssistnts nd to gthe infomtion bout the cicuit. A geneic plnning document is issued t the site to fmiliize pesonnel with the suvey pocess, equipment, nlyses equied, etc. Theefte, futhe fou to six weeks e nomlly equied to llow stff t the plnt to pepe fo the execise. This usully involves ceting ccess fo smpling nd mesuements with the mills, cyclones nd the flottion cells. A typicl integted smpling execise tkes on vege one week, duing which the following ctivities e cied out: detiled suvey of evey vilble stem in the comminution cicuit dditionl smpling fo BWI, RWI nd JK dop weight tests mesuement of mill dimensions, lines specs, etc. * Anglo Resech. The Southen Aficn Institute of Mining nd Metllugy, SA ISSN X/ This ppe ws fist published t the SAIMM Confeence, Bse Metls 2007, July P p e The Jounl of The Southen Aficn Institute of Mining nd Metllugy VOLUME 107 NON-REFEREED PAPER OCTOBER

2 SAIMM_Oct_37-46:Templte Jounl 11/5/07 8:11 AM Pge 642 Anglo esech (AR) expeience with integted comminution nd flottion plnt modelling Tble I List of AR flottion modelling execises Dte Opetion Cicuits : Suvey Type 2002 Lisheen Pb, Zn : Flottion 2003 Hudson By Cu, Zn : Flottion 2004 El Solddo Cu : Flottion 2004 Ls Totols Cu : Flottion 2004 Mntos Blncos Cu : Flottion 2004 Los Bonces Cu : Comminution 2005 Lisheen Pb, Zn : Integted 2006 Ctlo Nb : Integted 2006 Blck Mountin Cu, Pb, Zn : Flottion 2007 Copebs P2O5 : Integted tht smpling be postponed if non-typicl oe is being un when the execise is conducted. Lbotoy btch flottion tests e cied out to collect dditionl kinetic dt fo modelling nd, in ddition, to confim cetin pmetes such s esidence times nd dditionl clening steps. Mill csh-stops nd gind-outs e conducted to detemine mill dimensions nd lod/bll volumes ccutely. In totl, ppoximtely smples e typiclly collected duing the week-long test pogmme. This equies ceful plnning in the smple peption lbotoy. Assy nd sceening esults e genelly eceived ppoximtely five to seven weeks fte the end of cmpign. Mss blncing, modelling nd nlysis typiclly tke futhe two to thee months. Collection of Scd dt to ssist in mss blncing, powe modelling, stedy stte nd pocess contol nlyses, etc. Detiled suvey of evey stem in the flottion cicuit (fo model building) Lbotoy btch flottion tests hot flots of citicl stems (to collect dditionl kinetic dt) Cell hydodynmic mesuements (to chcteize the cell behviou); - Supeficil gs velocity (Jg) - Ai hold-up - Bubble size Residence time studies (to confim mss blnce nd to detemine cell mixing chcteistics) Altente cicuit suvey (fo dditionl model vlidtion, if possible). Fo the smpling execise itself, AR hs found it essentil to do s much wok s possible on single dy on ech of the comminution nd flottion cicuits. To dte, expeience hs shown tht no plnt opetion is sufficiently stble to llow consistent detiled smpling fom dy to dy. Typiclly, few plnts cn be elied upon to hve less thn five pe cent eltive chnge in hed gde fom one dy to the next, nd whilst this my hve only limited effect on the ovell pefomnce of the plnt, the chnge in ssys fom smples of individul cells cn be mked. Hence AR involves s mny people s possible in the suvey. Cetin mesuements such s bubble size nd supeficil gs velocity, fo which no sluy smples e tken, e less sensitive to oe chnges. These mesuements cn be mde on othe dys, lthough idelly they e best tken on the sme dy. The sluy smples, which epesent snp-shot of the pocess, e used to model the equipment nd the pocess. The im of this execise is to genete n ccute detiled mss blnce of the plnt to fit pmetes fo modelling. This diffes to some extent fom the im of obtining n oveview of vege plnt pefomnce. Although this is conty to conventionl wisdom, expeience to dte hs shown tht single-cut snp-shot smpling cn be s good s, if not bette thn, multi-composite smpling. In ddition, mine mngement must be infomed tht the cuent models e oe-specific. So, in ode fo the model to be of mximum benefit, the mine must ttempt whee possible to un on typicl mteil. If necessy, it is dvised Modelling ppoch The comminution section is modelled in JK SimMet vi nge of vilble models. The choice of model depends on the ntue of the mchine nd the qulity of the dt. Fo bll mills, the pefectly mixed bll mill model is typiclly used. Seven diffeent models e vilble fo ech of the sceen nd cyclone modules. SAG/AG mills hve fou models vilble, of which the vible tes model is typiclly used. The genel spectum of comminution-type equipment cn be modelled in the flow sheet, fom cone nd jw cushes to HPGRs nd the ssocited clssifiction devices. The genel pinciple of the flottion modelling ppoch is the decoupled model of flottion s shown in Eqution [1] below. Cell pmetes such s bubble sufce e flux, esidence time, wte ecovey nd the ovell ecovey cn be mesued nd clculted fo ech cell fom the mss blnce infomtion collected duing smpling. To lesse extent, entinment nd foth ecovey cn eithe be pedicted o mesued. Pi the fundmentl flottion esponse is then fitted s distibution. Mny dt points thoughout the cicuit, nd the ccompnying on-site lbotoy btch flottion tests, e used to ensue high degee of confidence in the fit. The AMIRA P9 ecovey model fo flottion fo species i in pefectly mixed cell: whee: R i P i Sb τ R f,i R w ENT i = ecovey = flotbility = bubble sufce e flux = esidence time = foth ecovey = wte ecovey = entinment pmete Hving chcteized the oe, AR cn use the model to simulte cicuit chnges nd to ssess vious wht-if scenios. Howeve, mny of the mesuements nd intemedite esults e often s useful s the finl simulto which esults fom this pocess. These e discussed in tun. [1] 642 OCTOBER 2007 VOLUME 107 NON-REFEREED PAPER The Jounl of The Southen Aficn Institute of Mining nd Metllugy

3 SAIMM_Oct_37-46:Templte Jounl 11/5/07 8:11 AM Pge 643 Anglo esech (AR) expeience with integted comminution nd flottion plnt modelling Modelling esults The ims of plnt model cn be eithe specific o genel. Fo exmple, one my equie model to detemine the optiml cicuit configution to poduce tget gind, gde o ecovey. A genel im would be to nlyse the ovell pefomnce of the plnt in tems of the oe s well s the pefomnce of the individul mills/cells nd to investigte possible es of impovement. In this section, fo ech of the ctivities typiclly cied out on the plnt duing modelling execise, some exmples of esults e pesented nd discussed. Mill pefomnce nd efficiency pmetes A numbe of mesuements nd efficiency pmetes e mesued fo evey comminution cmpign. Thoughput nd powe dw e insufficient in detemining the efficiency of mill with espect to its design pmetes. The ovell pefomnce of mill is collectively bsed on numbe of fctos, tht include: line design, dischge gte design, pecentge of open e t the mill dischge, espective ginding medi (type, size nd loding), poduct PSD, etc. Figue 1 below shows typicl mill bekge nd dischge te cuve with espect to size. The dischge of sluy is most efficient below ppoximtely 400um, fte which the sluy dischges moe slowly. This is pticully useful fo lteing the mill dischge design fo impoved thoughput. In ddition to this, the bekge te cuve fo SAG mill displys dip between 20 nd 70mm. This indictes the citicl size mteil size nge, which is citicl to coectly sizing dischge tommel nd lso setting the ecycle cushe gp, fo instnce. empiicl coeltions to pedict the bubble sufce e flux (e.g. Goin, Fnzidis nd Mnlpig, 1999) nd hence, indiectly, bubble size in mechnicl cells, pediction of bubble size in columns hs been less well quntified theefoe mesuement is essentil. Figue 3 pesents typicl bubble sizes of between mm (d 32 ) in the conventionl oughe cells. This is in geement with litetue souces (Deglon, Egy-Mensh nd Fnzidis, 2000; Powe, Fnzidis nd Mnlpig, 2000), which indicte tht bubble sizes of less thn 2mm e suitble fo flottion. Inteestingly, in the secondy cicuit, the bubble size deceses down the oughe bnk: this my be the esult of the fothe ddition down the bnk. The d 32 of the pimy oughe 9 cell ws mesued s high t 2.5mm, indicting fothe equiement. J o u n l P p e Flottion pefomnce pmetes: gs dispesion mesuements Bubble size The elie testwok mde use of the McGill Bubble Size (see Henndez-Aguil et l., 2002 fo moe detils) while moe ecently AR hs mde use of the Anglo Pltinum Bubble Size pesented in Figue 2. Although thee e published Figue 2 Anglo pltinum bubble size 8 30 In (bekge te) Dischge function Size ln R Dischge Rte Figue 1 SAG mill bekge nd dischge functions The Jounl of The Southen Aficn Institute of Mining nd Metllugy VOLUME 107 NON-REFEREED PAPER OCTOBER

4 SAIMM_Oct_37-46:Templte Jounl 11/5/07 8:11 AM Pge 644 Anglo esech (AR) expeience with integted comminution nd flottion plnt modelling Bubble size - d Rgh 1 Rgh 3 Rgh 5 Rgh 7 Rgh 9 Pimy Roughes Secondy Roughes Figue 3 Bubble size distibution down the oughe bnks Bubble size mesuements impove the nlysts confidence in the simultions becuse of the moe ccute detemintion of the bubble sufce e flux (Sb). Bubble size esults lso demonstte whethe the cells nd the egent egimes e ceting the typicl bubble sizes needed fo flottion. To dte, this hs been the cse in ll the cells mesued. The Bubble Size nlyse my povide tool to evlute diffeent egent nd/o cell conditions. Howeve this hs not been done in ny AR studies s yet. Supeficil gs velocity (Jg) The P9 ecovey eqution contins pmete Sb, the bubble sufce e flux. This equies knowledge of the supeficil gs velocity Jg nd the Sute men bubble size (d 32 ). Mesuing Jg in cell lso llows the opetion of the cell to be comped with tht of othe cells in the sme plnt nd in othe plnts (eg Powe et l., 2000), nd it llows the use of the cell fo othe puposes to be pedicted in simultions. The Anglo Pltinum Bubble Size is lso used to mesue this pmete. Although moden cells e often fitted with i flow metes, the Jg pobe emins useful fo clibting the ccucy of the mesuement. Sometimes sevel cells use common i souce: the Jg pobe cn identify bis in the mount of i fed to ech cell. In one cse, the Jg mesuement quickly eveled tht the i flow to cetin cell ws constined. Additionlly, fo self-eted cells (e.g. Wemco), the Jg pobe is essentil to mesue nd/o compe the i flow tes. In Figue 4, compison of the Jg mesuements in plnt shows tht thee is oom fo impovement in the oughe nd clene scvenge cells by dding moe i. Typiclly, foth ecovey pemitting, the cells my pefom bette t deepe foth depths nd highe i flows. Ai Hold Up (AHU) The pinciple of the AHU mesuement device (shown in Figue 5) is tht seled volume of eted pulp is cptued fom cell, nd then elesed into mesuing cylinde. The diffeence in volume to tht of the smple coesponds with the i fction. This mesuement is useful in esidence time detemintion, becuse it povides coection fo the volume of i within cell. This impoves the ccucy of the models. The AHU smple seves n dditionl pupose s mteil fo foth ecovey clcultions. As dignostic tool, AHU hs hd only limited ppliction. Typiclly, on the plnts visited to dte, i holdup nd Jg hve been ppoximtely popotionlly elted: cells with low i flows hve lso hd low i hold-ups. High i hold-up mesuements (~30 pe cent) cn be n indiction of high pulp viscosities. Howeve, this hs not been encounteed to dte. Residence time studies Solution tce studies e conducted s pt of the plnt modelling execises to detemine the liquid esidence time in vious stges of the cicuit. While these e not the highest pioity in the execise, they e useful fo comping with the mss blnce, nd pticully in detemining the wte blnce, which is typiclly less ccute thn eithe the solids o minel blnces. Both LiCl nd NCl hve been used s tces. The clcultion of the equied concenttion is bsed on the minimum needed to mke distinguishble pek, even in the wost-cse scenio (when the slt dispeses pefectly). Jg (cm/s) TYPICAL OK TANK CELL ROUGHERS Figue 4 Compison of Jg vlues (cm/s) fo vious Outokumpu tnk cells in cicuit Figue 5 Ai Hold Up (AHU) device CLNR. SCAVS 1st CLEANERS 644 OCTOBER 2007 VOLUME 107 NON-REFEREED PAPER The Jounl of The Southen Aficn Institute of Mining nd Metllugy

5 SAIMM_Oct_37-46:Templte Jounl 11/5/07 8:11 AM Pge 645 Anglo esech (AR) expeience with integted comminution nd flottion plnt modelling Foth ecovey detemintion Foth ecovey eflects the loss of ecovey of ttched pticles in the foth phse. Foth ecovey (Rf) of 100% (s ssumed in the shllow foths in lbotoy btch flottion tests) implies no loss of ecovey due to the foth phse. A numbe of techniques e vilble to mesue foth ecovey. In the AR execises to dte, the mss blnce ppoch suggested by Alexnde, Fnzidis nd Mnlpig (2003) hs been dopted. Simultneous mss blnces of ttched nd entined mteil, using feed, pulp nd tils smples fom flottion cell, combined with smple fom the top of the foth, llow the foth ecovey of ttched minels to be clculted. To dte owing howeve, this ppoch hs yielded little success, pehps to the need fo extemely ccute smpling. Tiplicte smples my be needed to identify the subtle chnges in gde. This is often not pcticl fo complete cicuit of mny cells. At best, the ppoch my wok fo cetin cells, which e then used s indictos fo the emining cells in the cicuit. The impliction of this difficulty in mesuing foth ecovey consistently is tht Rf becomes nothe fitted vible in the flotbility detemintion fit. This mkes it ll the moe impotnt to collect sufficient dt to ensue n ccute mthemticl fit. Detiled suvey smpling Micosoft Excel Solve techniques e used to fit the mss blnce flows ound the cicuit fom the ssy dt of the vious elements. A common finding in ll of these plnt modelling execises is tht detiled cell-by-cell mss blnce is in itself citicl dignostic tool. Fo exmple, Figue 6 shows the mss distibution coss the oughe bnk, while Figue 7 pesents the distibution of coppe down the bnk while vey little led is collected. Additionl collecto (both SIBX nd Sscol 61) hs been dded t the mid-point of the oughe bnk. Note how the mss pull to the concentte inceses fte the ddition of the egents, indicting tht thee my be meit in stge ddition of the egents fo this oebody, in ode to impove the mss pull to the concentte. Figue 8 pesents the ovell ecoveies (by size) fo one of the bse metl concentto plnts. In this exmple the use of ssy-by-size mss blnces identifies the citicl fctions which ccount fo the losses. A mjoity of the losses e (ppoximtely 65%) in the pimy oughe til in the +53 μm fction. In ddition, most of the vluble mteil lost in the clene 7 tils is in the -10 μm fction (ppoximtely 86%). The use of minelogy futhe identifies the potentil esons fo the losses. Exmintion of the clene tils stem showed tht ove 40% of the vluble mteil in this fction is libeted, while the eminde is eithe locked o ttched to the gngue minels. Lbotoy btch flottion tests Lbotoy btch flottion tests e cied out on citicl smples, typiclly feed, concenttes nd tils, fom ech stge of the pocess, using exct duplictes of the smples sent fo sceening, chemicl nd minelogicl nlysis. The kinetic ecoveies mesued in these tests povide futhe infomtion, which is used to incese confidence in the fit of the oe s flotbility distibution. Typiclly lbotoy flottion tests e cied out pe cicuit, nd these genete n dditionl 112 to 160 dt points which cn be used in the model fitting. The esults of the lbotoy btch flottion tests on the oughe tils stem e n indiction of the content of non-floting mteil in the oe. Vlubles which e not ecoveed in these tests in the ten pe cent foth ecovey conditions of the lbotoy btch cell e consideed non-floting. This is useful numbe with which to evlute the ctul plnt ecovey, nd to quntify ny potentil scope fo impovement in the cicuit. Figue 9 pesents the possible ecoveies fom the feed nd concentte smples s well s the potentil to ecove some of the vlubles fom the exit stems. The lbotoy tests e lso citicl in evluting nodes t which thee is chnge in flotbility, fo exmple t points in the cicuit whee thee is egent ddition o egind mill. In genel, AR wok to dte suggests tht egent ddition within cicuits hs only limited effect on flotbility. The effect of eginding is demonstted in J o u n l P p e Mss pull (%) RC1 RC2 RC3 RC4 RC5 Cell # RC6 RC7 RC8 RC9 Mss Pull Figue 6 Mss ecoveies down the pimy oughe bnk The Jounl of The Southen Aficn Institute of Mining nd Metllugy VOLUME 107 NON-REFEREED PAPER OCTOBER

6 SAIMM_Oct_37-46:Templte Jounl 11/5/07 8:11 AM Pge 646 Anglo esech (AR) expeience with integted comminution nd flottion plnt modelling Distibution (%) Coppe Cell No. Led Led Coppe Figue 7 Bse metl distibution down the oughe bnk Unsized +53 μm +10 μm 10 μm Distibution (%) Figue 8 Distibution by size clss Pim CT Feed Pim Comb conc. RT9 Pim CT7 RC1 Re-Cln Conc Stem Figue 10, which shows kinetic ecovey cuves befoe nd fte egind mill. The fine gind leds to n impovement in both the kinetics nd ecovey. Model fitting Once the mss blnce of the comminution cicuit hs been completed, it is then possible to model-fit the cicuit. This is done by fitting pmetes to the equipment models, usully stting with n intelligent estimtion of the pmete. The feed dt fo the equipment e then enteed into the model, nd good fit is obtined when the model ccutely pedicts the smpled poduct of the equipment. Figue 11 shows the typicl nge of pmetes fitted to hydocyclone nd SAG mill. With espect to the flottion esults, once ll the detils of the mesuements nd the mss blnce nd lbotoy flottion test dt hve been collected, the next stge in the pocess is to fit the flotbility distibution of the feed oe. An exmple of flotbility distibution fo glen nd sphleite in led cicuit is shown in Tble II. With feed of this flotbility distibution, it will not be possible to ecove the slow-floting glen without lso ecoveing ppoximtely nine pe cent of the sphleite. This is n exmple of how this ppoch helps to quntify nd hence identify the cuses of poblems within cicuit. % Pb ecovey Time (min) Pb Food Pb Cil 1 Conc Pb Col Tils Pb Scv Tils Figue 9 Pb Recoveies in btch flottion cell A Sphleite ecovey (%) 100% 80% 60% 40% 20% 0% Time (min) Figue 10 Sphleite ecovey in the lbotoy btch test befoe nd fte e-ginding 646 OCTOBER 2007 VOLUME 107 NON-REFEREED PAPER The Jounl of The Southen Aficn Institute of Mining nd Metllugy

7 SAIMM_Oct_37-46:Templte Jounl 11/5/07 8:11 AM Pge 647 Anglo esech (AR) expeience with integted comminution nd flottion plnt modelling J o u n l P p e Figue 11 Typicl model pmetes fo hydocyclone cluste nd SAG mill % Cumultive % Pssing - Exp nd Cl Figue 12 Typicl SABC cicuit simultion flow sheet Cicuit simultions The finl stge in the pocess is the development of cicuit simulto using ll the infomtion descibed in this ppe. Once the models hve been fitted, simultions cn be un on host of chnges in eithe the pocess flow sheet, opeting conditions o equipment dimensions. It is lso possible to conduct simultions using combintions of ll of these chnges (within esonble limits). Figue 12 shows typicl SABC cicuit simultion flow sheet in JK SimMet. This pticul cicuit simulto ws used to detemine (mong othe things) the effect of totl plnt thoughput on the finl gind. Tble II Flotbility distibution of sphleite nd glen in led cicuit Glen Sphleite Pi Xi Pi Xi Fst Medium Slow Non Totl minel The Jounl of The Southen Aficn Institute of Mining nd Metllugy VOLUME 107 NON-REFEREED PAPER OCTOBER

8 SAIMM_Oct_37-46:Templte Jounl 11/5/07 8:11 AM Pge 648 Anglo esech (AR) expeience with integted comminution nd flottion plnt modelling Figue 13 shows the effect of incesed thoughput on the cyclone oveflow PSD. AR uses JK SimFlot (His et l., 2002) to un the flottion simultions. A JK SimFlot sceen is shown in Figue 14. The fist step, once the simulto is woking, is to un chnged cicuit in ode to vlidte the plnt model. At ech site, equest is mde to chnge the cicuit nd to suvey this chnged cicuit. The plnt must be peped to opete this cicuit until it stbilizes. In genel, AR hs obtined good model vlidtion when the chnged cicuit uns on simil oe to tht used in the oiginl suvey. Howeve, thee hve been occsions when the oe hs chnged, nd the pediction does not mtch the ctul cicuit pefomnce. This expeience emphsises the need to do s much smpling s possible on single dy. Note tht the eltive chnges in pefomnce pedicted by the simulto e believed to be f moe obust nd to pply even in cses whee the feed oe is diffeent. Howeve, when comping the outcome of the chnged cicuit with the esults of the oiginl detiled suvey, it is essentil to hve simil oe so s to evlute the quntittive effect of the chnge on plnt pefomnce. A typicl bsic simultion is to vy the esidence time. Although dditionl cells e often not n option on the plnt, this gives n indiction of ny bottlenecks in the cicuit. Othe simultions which e typiclly tested, nd e moe likely to be possible to implement in elity, e: chnged i flows wte ddition open cicuit/closed cicuit clening nd the tils etun point oughe-scvenge split in bnks, nd multiple stges of clening. An exmple of n AR simultion which ws implemented in pctice is shown in Figue 15. The simultion showed tht etuning the fist clene tils to the fouth oughe cell feed Mill cicuit poduct size with chnging feedte Cum % pssing Size (mm) 170 t/h 180 t/h 190 t/h 200 t/h 211 t/h 220 t/h 230 t/h Figue 13 Effect of thoughput on finl cicuit gind Figue 14 Typicl JK SimFlot sceen showing inputs nd the flow sheet in the bckgound 648 OCTOBER 2007 VOLUME 107 NON-REFEREED PAPER The Jounl of The Southen Aficn Institute of Mining nd Metllugy

9 SAIMM_Oct_37-46:Templte Jounl 11/5/07 8:11 AM Pge 649 Anglo esech (AR) expeience with integted comminution nd flottion plnt modelling would give the sme ecovey nd gde s tht chieved when it is etuned to the feed of the cicuit. Howeve, this chnge would getly benefit the plnt in tems of the impoved stbility of the fist two oughe cells, which e esponsible fo the mjoity of the cicuit ecovey. Hence the plnt hs implemented nd mintined this chnge. Unfotuntely, the benefits chieved though incesed stbility e difficult to quntify exctly. The second exmple illusttes how the simulto cn be used to identify not only the need of dditionl clening cpcity but lso whee to oute the concentte stem. The need fo dditionl etention time ws identified in the fist clene stge (in thee stge clening cicuit). Howeve, with the id of the simulto it becme cle tht in ode to mintin the finl concentte gde, the concentte fom the dditionl cell should be sent bck to the feed going to the fist clenes the thn the feed to the second clene stge (i.e. the cell is cting s clene scvenge). The cicuit configutions e illustted in Figue 16. Developments Although the JK SimMet nd JK SimFlot pckges e ble to poduce elible simultions, the development of the models is still on going. Much of the esech nd development of the models is done though the sponso-funded esech of Ami P9. The oiginl JK mill nd cyclone models hve been diclly impoved in ecent decdes, nd the cuent vesion of JK SimMet includes the best of these models. Howeve, owing to the diffeent spect tios of mills nd the vey lge dimete of the mills tht e cuently being built, the J o u n l P p e Roughes Scvenges Cicuit Feed Fist Clenes Finl Clenes 2nd Clenes Finl Conc Figue 15 Cicuit chnge s esult of simultion Additionl Clene Concentte To Clene 1 Feed Figue 16 Addition of dditionl clening cpcity The Jounl of The Southen Aficn Institute of Mining nd Metllugy VOLUME 107 NON-REFEREED PAPER OCTOBER

10 SAIMM_Oct_37-46:Templte Jounl 11/5/07 8:12 AM Pge 650 Anglo esech (AR) expeience with integted comminution nd flottion plnt modelling models e constntly being upgded. Powell is cuently developing the UCM (Unified Comminution Model), which ims to ccount fo bode nge of mill dimensions, including the effects of size on the bekge te of n oe component nd lso the effect of segegtion in low spect tio (pooly-mixed) mills. The mjoity of the dt pesented in this ppe wee souced fom execises in which the flotbility component ppoch ws used to epesent the diffeent flotbility clsses of ech minel. The most ecent AR execises hve included ssy by size nlysis to impove the ccucy of the models. This is pticully necessy in cicuits including egind milling stge, nd possibly lso column cells, which e believed to be moe sensitive to pticle size. In these cses, the potentil uses of the mss blnce esults will hve n dditionl level of complexity. The futhe inclusion of minelogicl nlysis, which hs considebly ided in the undestnding of the flottion kinetics, is clely n exmple of dditionl complexity. Howeve, it is dvised tht modelling should lwys poceed sequentilly in ode of incesing complexity (nd cost) of nlysis. Fistly, this will llow intemedite epoting of esults to the plnt, nd secondly will povide check tht only ppopite smples e sent fo nlysis, s cn be the cse in the event of chnging plnt hed gdes fo exmple. Integted suveys The following is desciption of ecent exmple whee the benefit of integted suveys ws evident. On pticul led-zinc mine, the flottion suvey esults indicted tht the plcement of the e-gind mill ws inppopite. The egind mill ws teting the column tils, which then fed bnk of scvenge cells. The concentte fom these cells ws then sent to the column feed. Minelogicl exmintion of the column feed showed tht mjoity of the minels wee in biny phses with pyite; hence it would be beneficil to e-gind the column feed the thn the column tils to libete the minels. This would enble the mine to chieve highe concentte gdes. The stff t the plnt wee eluctnt to ccept this chnge s they felt tht the efficiency of the e-gind mill would be hmpeed by the dditionl mss flow though the mill. Howeve, since the e-gind mill ws lso suveyed s pt of the integted comminutionflottion suvey, the new thoughputs wee simulted. This esults showed tht with chnges to the steel medi lod nd bll size, thee ws sufficient cpcity to tet the column feed nd obtin the desied gind. The chnges wee implemented by the plnt with success. In ddition to this exmple, thee hve lso been cses whee the flottion suvey hs ecognised cetin size clsses tht do not flot well (becuse they e eithe too cose o too fine). The milling optimiztion theefoe imed t educing the mount of mteil outside the desied size nge by ceting steepe poduct PSD. Thoughput is often the desied tget of mny ginding plnts, but the effect of educed flottion esidence time is difficult to estimte without flottion model. The integtion of the flottion suvey thus mkes it possible to detemine the effects of tonnge nd gind on ecovey nd gde. In ddition, the flottion simulto is ble to simulte which chnges of the cell conditions e ble to cte fo the nticipted chnges. Conclusions This ppe hs given n oveview of some the wok done nd the esults obtined in the AR comminution nd flottion modelling poject ove the lst few yes. The potentil benefit of ech of the phses of typicl plnt modelling cmpign hs been demonstted. Comminution cicuits e modelled to the extent tht the effects of cicuit chnges cn be confidently nticipted (whethe the output is n ccute pediction o eltive chnge in the pocess). The chnges tht cn be simulted include: flowsheet configution, flow tes, opeting conditions, equipment pmetes nd oe competency. The flottion cell chcteiztion equipment is useful in detemining the ctul pefomnce of the cells, in tems of i flow nd bubble size. Typiclly, these esults e comped with industy veges. Advice cn be given in cses whee non-typicl pefomnce is mesued. In AR s expeience, the detiled cell-by-cell mss blnce nd the lbotoy btch flottion tests e often the most useful stges of the modelling execise. Not only e they the most esily undestood nd tusted by plnt mngement, but they cn be poduced within few weeks of eceiving the dt nd hence e elevnt to cuent plnt pefomnce. Assy by size hs futhe enhnced the mss blncing execise, nd lso the undestnding of the vious units within concentto. The comminution nd flottion simultion stge povides mens to quntify the effect of chnges to complex cicuits in tems of potentil benefit. This ppoch, despite the need fo cution nd undestnding of the softwe, offes vible solution fo plnt impovements. Refeences ALEXANDER, D., FRANZIDIS, J., nd MANLAPIG, E. Foth ecovey mesuement in plnt scle flottion cells, Minel Engineeing, vol. 16, no DEGLON, D., EGYA-MENSAH, D., nd FRANZIDIS, J. Review of hydodynmics nd gs dispesion in flottion cells on South Aficn pltinum concenttos, Minels Engineeing, vol. 13, no. 3, pp GORAIN, B, Selection of cell opeting conditions to optimize pefomnce of flottion cicuits with lge cells, in Poceedings Seventh Mill Opetos Confeence, (The Austlsin Institute of Mining nd Metllugy: Melboune) pp GORAIN, B., FRANZIDIS, J., nd MANLAPIG, E. The empiicl pediction of bubble sufce e flux in mechnicl flottion cells fom cell design nd opeting dt, Minels Engineeing, vol. 12, no. 3, pp HARRIS, MC., RUNGE, K.C., WHITEN, W.J., nd MORRISON, R.D. JKSimFlot s pcticl tool fo flottion pocess design nd optimiztion, in Minel Pocessing Plnt Design, Pctice, nd Contol, (Society fo Mining, Metllugy nd Explotion: Littleton). vol. 1, pp HERNANDEZ-AGUILAR, J., GOMEZ, C., nd FINCH, J. A TECHNIQUE FOR THE DIRECT MEASUREMENT OF BUBBLE SIZE DISTRibutions in industil flottion cells, in Poceedings 34th Annul Meeting of Cndin Minel Pocessos. J. Knesset (ed.), pp LYNCH, A., JOHNSON, N., MANLAPIG, E., nd THORNE, C. Minel nd Col Flottion Cicuits: Thei Simultion nd Contol, (Elsevie: Amstedm), pp POWER, A., FRANZIDIS, J., nd MANLAPIG, E. The chcteistion of hydodynmic conditions in industil flottion cells, in Poceedings Seventh Mill Opetos Confeence, (The Austlsin Institute of Mining nd Metllugy: Melboune) pp Ve, M, Mthee, Z, Fnzidis, J., et l. 2001, The modelling of foth zone ecovey in btch nd continuously opeted lbotoy flottion cells., in Intentionl Jounl of Minels Pocessing (Elsevie), pp OCTOBER 2007 VOLUME 107 NON-REFEREED PAPER The Jounl of The Southen Aficn Institute of Mining nd Metllugy