UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Process analysis and energy efficiency improvement on Portland limestone cement grinding circuit Aguero, Sixto Humberto 2015

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2015_may_aguero_sixto.pdf [ 3.23MB ]
Metadata
JSON: 24-1.0166246.json
JSON-LD: 24-1.0166246-ld.json
RDF/XML (Pretty): 24-1.0166246-rdf.xml
RDF/JSON: 24-1.0166246-rdf.json
Turtle: 24-1.0166246-turtle.txt
N-Triples: 24-1.0166246-rdf-ntriples.txt
Original Record: 24-1.0166246-source.json
Full Text
24-1.0166246-fulltext.txt
Citation
24-1.0166246.ris

Full Text

PROCESS ANALYSIS AND ENERGY EFFICIENCY IMPROVEMENT ONPORTLAND LIMESTONE CEMENT GRINDING CIRCUITbySixto Humberto AgueroB.S. (Mechanical Engineering), Universidad Nacional Autonoma de Honduras, 1992MASc (Energy Management), New York Institute of Technology, 2008A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FORTHE DEGREE OFMASTER OF APPLIED SCIENCEinTHE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES(Mining Engineering)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)April 2015© Sixto Humberto Aguero, 2015iiAbstractWorldwide cement production is a high energy consuming industry; 90% is thermal and 10% iselectrical energy. This is the third most anthropogenic related carbon dioxide emitting industry inthe world. With a rising price of energy and a growing emphasis on environmental issues thecement industry is facing significant challenges to both remain a competitive and sustainable.Composite cement manufacturing is one alternative that is used reduce energy use and greenhousegas emissions. The dry grinding process used for finished product represents 40-50% of electricalenergy consumption. It is a very inefficient process generally ranging around 1% efficient.This research evaluated the process of a typical Portland cement grinding circuit in order to identifyinefficiencies in the process and how the operating parameters may be changed in order to improvethe system’s performance. Tests were conducted using samples from a B.C. cement producer andresults analyzed in order to characterize and build a high accuracy model that can be used as abench marking tool. Representative sampling and mass balance were performed on the circuitusing real steady state operative conditions data provided by process plant managers.Major research findings are: Air separator efficiency is rated 46.06% efficiency at fractions below 35 microns. High dust load feed and agglomeration are the main reasons for this low separator efficiency. Agglomeration effect is related to overgrinding, high energy impacts and the use of limestone.iii Whiten model is an adequate tool to fit and correct experimental data on cement air separatorsand to provide quantification of operating factors to evaluate the separation process. Low grinding kinetics at ball mill compartment 01, suggests improper size grinding mediaselection and high wear rate for the case studied (for media and liners).ivTable of contentsAbstract.......................................................................................................................................... iiTable of contents .......................................................................................................................... ivList of tables.................................................................................................................................. xiList of figures.............................................................................................................................. xiiiList of abbreviations .................................................................................................................. xviAcknowledgements ................................................................................................................... xviiDedication ................................................................................................................................. xviiiChapter 1: Introduction ................................................................................................................11.1 Circuit 03 cement production.......................................................................................... 21.2 Background ..................................................................................................................... 31.3 Thesis objective .............................................................................................................. 71.4 Thesis outline .................................................................................................................. 8Chapter 2: Literature review........................................................................................................92.1 Introduction to cement industry ...................................................................................... 9v2.2 Composite cement manufacturing ................................................................................ 102.2.1 History of the use of limestone additions ................................................................. 112.2.1.1 Europe ............................................................................................................... 112.2.1.2 North America .................................................................................................. 132.3 Finished grinding and quality of cement ...................................................................... 142.3.1 Agglomeration .......................................................................................................... 192.4 Grinding technologies in cement industry .................................................................... 192.4.1 Ball mill .................................................................................................................... 202.4.2 High pressure grinding roll ....................................................................................... 222.4.3 Vertical roller mill (VRM)........................................................................................ 232.4.4 Horizontal roller mill ................................................................................................ 252.5 Improving grinding efficiency in closed circuit cement ball mill................................. 262.5.1 Grinding aids............................................................................................................. 272.5.2 Optimum media ball size and mixing ratio............................................................... 282.5.3 Fill factor of grinding media ..................................................................................... 29vi2.5.3.1 Load power ....................................................................................................... 302.5.3.2 Blank height measurement................................................................................ 302.5.3.3 Run-time ........................................................................................................... 312.5.3.4 Ground tonnage................................................................................................. 312.5.3.5 Grinding efficiency ........................................................................................... 312.5.4 Pre-grinding of raw addition material....................................................................... 322.5.5 Air classifiers ............................................................................................................ 322.5.5.1 Performance of separators................................................................................. 362.5.5.2 High efficiency on separators ........................................................................... 372.5.6 Circulating load......................................................................................................... 402.6 Impact on carbon dioxide emissions............................................................................. 412.7 Modeling and simulation of Portland cement circuits .................................................. 452.7.1 Model of performance curves in separators.............................................................. 462.7.2 Model of two compartment ball mills....................................................................... 50Chapter 3: Experimental procedures.........................................................................................54vii3.1 Sampling and data gathering......................................................................................... 543.2 Bond standard ball mill grindability test....................................................................... 553.3 Breakage distribution function test and estimation using BFDS® software ................ 553.3.1 Clinker breakage function estimation test procedure................................................ 563.4 Particle size distribution analysis.................................................................................. 573.4.1 Rosin-Rammler distribution...................................................................................... 573.4.2 Whitens model for high efficiency separators .......................................................... 583.5 Selection function and its estimation using numerical grinding optimization tools inlanguage C (NGOTC®) software ............................................................................................. 593.5.1 Back calculation of selection function from continuous mill data ........................... 59Chapter 4: Results and discussion..............................................................................................614.1 Introduction................................................................................................................... 614.1.1 Particle size distribution and production................................................................... 634.1.1.1 Fresh feed size distribution ............................................................................... 634.1.1.2 Air separator size distribution........................................................................... 67viii4.1.1.2.1 Separator partition curve ............................................................................. 694.2 Ball mill, Bond work index, breakage and selection function ...................................... 774.2.1 Work index................................................................................................................ 784.2.2 Breakage function ..................................................................................................... 804.2.3 Selection function ..................................................................................................... 824.2.3.1 Grinding kinetics at compartments 01 and 02 .................................................. 824.2.4 Savings estimations................................................................................................... 85Chapter 5: Major research findings and conclusions...............................................................875.1 Major research findings ................................................................................................ 875.2 Conclusions................................................................................................................... 895.3 Recommendations for future work ............................................................................... 91References.....................................................................................................................................93Appendices....................................................................................................................................96Appendix A : Standard Bond work index calculation .............................................................. 96A.1 Procedure ....................................................................................................................... 97ixA.2 Sample preparation ........................................................................................................ 97A.3 Particle size analysis of the feed .................................................................................... 97A.4 Feed bulk density ........................................................................................................... 98A.5 Performing the grinding test .......................................................................................... 99A.6 Conditions for closure.................................................................................................. 101A.7 Particle size analysis of the product ............................................................................. 101A.8 Bond test grindability calculations............................................................................... 102Appendix B : Work indices..................................................................................................... 104B.1 Clinker 100%................................................................................................................ 104B.2 Limestone ..................................................................................................................... 105B.3 Clinker 95%/Limestone 5% ......................................................................................... 106B.4 Clinker 88%/Limestone 12% ....................................................................................... 107B.5 Clinker 60%/Limestone 40% ....................................................................................... 108Appendix C : Particle size distribution ................................................................................... 109C.1 Fresh feed to ball mill................................................................................................... 109xC.2 Air separator streams (data provided by plant) ............................................................ 110C.3 Fresh feed and circulating load feed to ball mill .......................................................... 111Appendix D : Bond equation for modeling throughput and savings ...................................... 112Appendix E : Specific rate of breakage .................................................................................. 113xiList of tablesTable 1-1: Cement types and production at plant site 2014............................................................ 1Table 1-2: Production circuit data in the studied cement plant in B.C ........................................... 2Table 2-1: Type designation for Canadian Portland Cement and ASTM equivalent ................... 10Table 2-2: European standard related composite cement ............................................................. 12Table 2-3: Electricity consumption during production of ordinary Portland cement ................... 26Table 2-4: Effect of Sika™ Polycarboxylate ether polymers grinding aid on energy use............. 27Table 2-5: Distinctive features of separators of different designs ................................................ 39Table 2-6: Influence of circulating load and type of separator on mill efficiency ....................... 40Table 2-7: Fuel savings and CO2 emissions reduction with PLC ................................................. 45Table 4-1: Circuit 03 equipment design specifications and operating condition on type 10 cement....................................................................................................................................................... 62Table 4-2: Separator efficiency parameters .................................................................................. 70Table 4-3: Ball mill grinding media charge details for C1 and C2............................................... 78Table 4-4: Work indices for research samples.............................................................................. 79xiiTable 4-5: Clinker average breakage function.............................................................................. 81Table 4-6: Savings estimate on electricity for fresh feed size reduction at ball mill 03............... 86Table 4-7: Profit estimate on increase in throughput at air classifier circuit 03 ........................... 86Table 5-1: Summary of current and modeled separator parameters ............................................. 91xiiiList of figuresFigure 1-1: Circuit 03 in the studied B.C plant............................................................................... 3Figure 1-2: Global production of cement........................................................................................ 4Figure 2-1: Clinker (left) and finished cement (right) .................................................................... 9Figure 2-2: Data on cement types produced in Europe................................................................. 13Figure 2-3: Relationship between compressive strength and uniformity factor according to Gates–Gaudin–Schuman plotting ............................................................................................................ 16Figure 2-4: Grindability of clinker and limestone ........................................................................ 17Figure 2-5: Grindability of limestone - clinker cement mixtures ................................................. 17Figure 2-6: Uniformity factor of inter-ground cement mixtures .................................................. 18Figure 2-7: Two-compartment tube ball mill. A- Compartment 01, B- Compartment 01/02 andseparating diaphragm .................................................................................................................... 21Figure 2-8: High pressure grinding roll ........................................................................................ 23Figure 2-9: Vertical roller mill...................................................................................................... 24Figure 2-10: HoroMill schematic diagram ................................................................................... 25Figure 2-11: Optimum ball mill void filling ................................................................................. 30xivFigure 2-12: First generation air separator.................................................................................... 33Figure 2-13: Forces balance in an air separator ............................................................................ 34Figure 2-14: Typical tromp curve ................................................................................................. 37Figure 2-15: Total global industry direct greenhouse gases emission .......................................... 41Figure 2-16: The IEA/CSI blue map for CO2 emissions reduction .............................................. 42Figure 2-17: Average kg of CO2 released per ton of cement produced........................................ 43Figure 2-18: Material balance - no raw limestone addition.......................................................... 44Figure 2-19: Material balance - PLC with 5% limestone addition ............................................... 44Figure 2-20: Efficiency to overflow vs size.................................................................................. 47Figure 2-21: Variation of efficiency curve related to α ................................................................ 48Figure 2-22: Variation of efficiency curve with β ........................................................................ 49Figure 2-23: Relation between (a) feed/bypass (b) and dust loading/bypass ............................... 49Figure 2-24: (a) Two compartment ball mill (b) model circuit array ........................................... 50Figure 2-25: Comminution on a closed circuit ............................................................................. 52Figure 3-1: Circuit 03 sampling points ......................................................................................... 54xvFigure 3-2: Representation of the distribution of particle breakage ............................................. 56Figure 4-1: Particle size distribution of circuit 03 fresh feed and clinker at CKP........................ 64Figure 4-2: Current d80 feed and product size at circuit 03........................................................... 65Figure 4-3: Particle size distribution on air separator ................................................................... 67Figure 4-4: Mass balance at air separator ..................................................................................... 68Figure 4-5: Separator efficiency reports to product fit using Whiten model ................................ 69Figure 4-6: Correlation between C, β parameters and relation to plant data ................................ 74Figure 4-7: Effect of separator dust loading on bypass and relation to plant data........................ 76Figure 4-8: Relation between sharpness and dust loading............................................................ 77Figure 4-9: Clinker breakage function at normalized size............................................................ 80Figure 4-10: Selection function for compartments 01 & 02 on 100% clinker.............................. 83xviList of abbreviationsAbbreviation DescriptionASTM American Society of Testing MaterialBM Ball MillCKP Chichibu Kawasaki Pre-grinderCSA Canadian Standard AssociationCSI Cement Sustainable InitiativeEIA U.S Energy Information AdministrationHPGR High Pressure Grinding RollIEA International Energy AgencyPC Portland CementPCA Portland Cement AssociationPFC Portland Flyash CementPLC Portland Limestone CementPPC Portland Pozzolan CementPSD Particle Size DistributionRPM Revolutions Per MinuteSSA Specific Surface AreaVRM Vertical Roller MillxviiAcknowledgementsI would like to express my deep appreciation to Dr. John Meech who trusted and gave me theopportunity of coming back to college. I would like to acknowledge Dr. Marcello Veiga, foradopting me as my co-supervisor and Dr. Akbar Farzanegan, for his right technical guidance andsupport throughout this research project.My gratitude for the support of all the sponsors involved in providing the logistics and fundsnecessary for this study.The assistance of Lo Pius, Aaron Hope, Maria Liu, Leslie Nicholson, Amit Kumar, NawoongYoon and Mike McClintock are deeply acknowledged.Finally, I would like to thank my loving wife and two children for their support during my studentduties.This research has been funded by NSERC (Natural Sciences and Engineering Research Councilof Canada) as an Engage Grants program whit a main purpose of the engage Grants programs theinterest of industry as partners on research activities related to production improvement, newtechnology/products development and knowledge transfer within universities.xviiiDedicationIn memory of Professor John A. Meech, who devoted his professional life to the development ofsustainable mining and a passion for teaching.1CHAPTER 1: INTRODUCTIONThe following research has been performed at a Portland cement plant located in the province ofBritish Columbia, Canada. This plant produces three types of cements: Type 10 (GU) is a generaluse hydraulic cement with 95% clinker and 5% gypsum, Portland Limestone Cement (PLC) witha composition up to 12% limestone addition, 5% gypsum and 83% clinker and Type III cementthat is a high early strength hydraulic cement with average productions. Gypsum is usually addedup to 5% in each cement type in order to control the rate of setting of the cement (Bhatty, 2011).A detail on production per type of cement is shown on Table 1-1.Table 1-1: Cement types and production at plant site 2014Cement type Annual Production tonnes/yearType 10 853,989PLC 181,594Type III 43,425The production of finished cement is performed at three grinding circuits identified as circuits 01,02 and 03. Circuits 01 and 02 are operated individually in a closed circuit grinding dry clinker andadditives each at a two compartment ball mills with a high efficiency air separator that classifiesthe finished product and rejects. Circuit 03 has a vertical pre-grinder that condition the clinker feedby reducing its size feeding to the ball mill. Limestone and gypsum are fed directly to a ball millthat operates in a closed circuit with a high efficiency air separator. For the purpose of this studythe research was focused on Circuit 03, since this circuit has the highest production capacity and2the potential for energy improvements like the current operation of Chichibu Kawasaki Pre-grinder(a vertical grinder) that process clinker and can be used also to reduce limestone.  A summary atthis B.C cement plant showing specific energy use, production rate and limestone addition may befound in Table 1-2. Blaine surface area is a quality parameter on finished cement that is relatedwith early strength development and water/clinker ratio should be brought in all these three circuitsto 3,700 cm 2 /g to assure an adequate strength on finished product. The low specific energyconsumption (energy in kWh used to produce a ton of finished cement) of Circuit 03 is basicallydue to the current operation of the pre-grinder before the ball mill.Table 1-2: Production circuit data in the studied cement plant in B.CCircuit Production rate, t/h Specific energy, kWh/t Limestone addition , %01 106 38.35 9.502 106 42.48 12.503 101-130 30.0-36.901 4.0-13.0Average BM systems 32-37 *(Seebach, 1996)* At Blaine 3,000-3,200 cm2/ g1 without including specific energy from CKP pre-grinder1.1 Circuit 03 cement productionCircuit 03 is composed of a vertical roller mill pre-grinder, feeding a two compartment ball milloperated as a closed circuit with a third generation high efficiency Osepa separator that rejects theoversize particles to the ball mill and separates the finished product. This circuit is used to producemainly type 10 cement (4% limestone) and is shown on Figure 1-1.3Figure 1-1: Circuit 03 in the studied B.C plant1.2 BackgroundCement is a key material used for the construction in housing and infrastructure. According to theInternational Energy Agency report (IEA, 2009) approximately 3.6 billion tonnes of cement wasproduced worldwide. The global production of cement actual and projected is shown in Figure1-2.4Figure 1-2: Global production of cement(EIA, 2009)The cement industry worldwide is facing challenges to conserve material and energy resources,and a demand to reduce CO2 emissions. According to Sustainable Cement Initiative CSI(Schneider et al., 2011) the main alternatives for cement producers are the increase in energyefficiency, clinker substitution and the use of alternative fuels.Cement production is an energy intensive process requiring an energy input of 850 – 1100 kWh/tof cement produced (Harder, 2003) and is the third most intensive anthropogenic industry relatedin terms of carbon dioxide (Abdel-Aziz et al., 2014). The thermal energy in cement productionrepresents approximately 90% of the total specific energy consumption with major fuel sourcesranging from coal, fuel oil to alternative residual fuels such as biomass, animal wastes anddiscarded tires. Electrical energy accounts for the remaining 10% of the total specific energyconsumption. The selection of the fuel source is primarily based on the cost. The electrical energy5consumed in a conventional cement production process is typically 95 – 110 kWh/t. The processof comminution, crushing and grinding of cement raw materials and finished cement, accounts for70% of the total electrical energy. The grinding stage for clinker and other additives accounts forapproximately 40 to 50% of total electrical energy consumption (Harder, 2003).Despite a high specific energy demand, two-compartment tube ball mills with an air classifier inclosed circuit have been used for the finish grinding of cement for over 100 years due to their goodreliability and favorable physical and chemical properties of the cement product such as a narrowerparticle size distribution (Aguero & Meech, 2014). Unfortunately, ball mills are one of the lowestenergy efficiencies of all the grinding mills. Ball mills suffer from considerable energy loss(approximately 98%) in the form of heat due to friction and collision in the tumbling mass of ballswhich transfers input energy to an unconfined bed of particles (Duda, 1976). Numerous impactsare required to produce effective breakage. Due to the high energy demand and the inherently lowenergy efficiency of conventional ball mill grinding, the cement industry is continually searchingfor new ways to reduce the energy use by improvements in mill design and circuit configuration.In recent years, the use of alternative fuels has already increased significantly, however thepotential for further improvements still exists. In cement, the reduction of the clinker duringfinished cement grinding by substitution with some specific materials having properties similar toclinker (such as limestone, pozzolan and blast furnace slag) remains a key priority. Remarkableprogress has already been made in this area. Nevertheless, appropriate materials are limited bytheir regional availability. New materials might be able to play a significant role as cementconstituents in the future, such as the use of synthetic pozzolan (waste material recovered from6combustion residues having an SiO2 /CaO ratio greater than 1,e.g. waste incineration/power plantstails, and having alkali oxides in amounts exceeding 1.5% by weight). Currently, the safeproportion of replacing clinker by alternative materials, such as raw limestone is up to 10%(Ramezanianpour et al., 2009), but the maximum extent of the substitution of pozzolanic andlimestone additives still needs to be evaluated.Initiatives to reduce the carbon footprint, reduce electrical use and the use of different additives tothe clinker are trending in cement industry worldwide. Portland Limestone Cement (PLC) isprogressively becoming a common product in the industry. PLC is produced by inter-grindingcement clinker, raw limestone and gypsum. The replacement of clinker with raw limestone inPortland cement production has resulted in a proportional reduction in the amount of fuel usageand CO2 emissions associated with cement production (Nisbet, 1996)The following study has been performed at a Portland Cement Plant in the province of BritishColumbia, Canada with an annual production of 85,989 t of type 10 or General Use cement (GU),181,594 t of Portland Limestone Cement and 43,425 t of Type III cement. Finished grinding ofcement is performed using three production circuits: 01, 02 and 03. Each of these circuits has aball mill which operates with high efficiency separators in closed circuit.This research is focused on circuit 03 because of a greater potential to improve circuit efficiency.Circuit 03 has the highest production capacity and is designed to operate in series with a verticalroller miller pre-crusher and individual storage capacity.7In order to characterize the grinding behavior of different parts of the circuit, particle sizedistribution (PSD) tests, breakage function, selection function and Bond grindability work index(Wi) tests have been carried out (Drosdiak, 2013).1.3 Thesis objectiveThe primary objective of this research is to evaluate circuit 03 of a B.C Portland cement plant inorder to identify energy inefficiencies and propose findings to improve operating parameters. Toprovide recommendations on improving the efficiency of the circuit, in order to accomplish thistarget the following secondary objectives have been defined, Analyze representative samples taken from the production circuit, their posterior analysisat the lab, interpretation of process and production data provided by the plant. Evaluate the breakage kinetics mechanisms on the two compartments of the ball millidentifying the breakage and selection function. Determine the classification function and performance of air efficiency separator usingprecise models, identify its probable causes and compare against normalizablebibliographic data. Define recommendations to improve the system and suggest possible areas of furtherresearch. Create a baseline for future energy improvements by the use of computer simulationsoftware in order to become a benchmarking reference for further research for the8improvement of Portland cement grinding circuits for ordinary and PLC cementmanufacturing.1.4 Thesis outlineChapter 2 provides a literature review on a brief overview of the worldwide cement industry,current technologies used in the grinding process and key factors involved with regards to thequality of cement. This chapter also discusses the importance of using PLC cements and its impactson greenhouse gas emissions. The chapter includes the modeling and simulation principles for airseparators and ball mills and examines researches on relating operating parameters on airclassifiers.Chapter 3 discusses the experimental procedures used for testing the different samples and alsothe description of software used to calculate parameters such as: uniformity factor, breakagefunction and selection function.Chapter 4 provides a description of the clinker used based on the work index, and other lab analysesand their relation within the existing circuit. The chapter also discusses operating parameters andeffects on the process.This thesis is concluded in Chapter 5 with a list of the major findings and recommendations forfuture work.9CHAPTER 2: LITERATURE REVIEW2.1 Introduction to cement industryCement is produced from a process of calcining a mixture of limestone and clay minerals withsome minor components as: iron ore, bauxite and sand. Raw materials are brought to a rotatingkiln with a temperature ranging from 1,3000C to 1,5500C (Bhatty, 2011). This sinteringtemperature generates a new product is defined as clinker shown at Figure 2-1.Figure 2-1: Clinker (left) and finished cement (right)(www.nachi.org)The final cement product is usually produced when the clinker is ground with usually 5% ofgypsum. The grinding process target is to produce a fine powder with 80% particle passing size(d80) of 30-40 microns. According to ASTM C-150 specifications (ASTM, 2011) there are eightdifferent types of Portland cement: Type I, IA, II, IIA, III, IIIA, IV and V and is based on: particle10fineness, chemical content, reactivity, early strength and final use. In Canada the types are relatedto the Canada Standard Association (CSA) A3000-03 standards and specified as type GU (GeneralUse and suitable for all operations), MS (moderate sulfate), MH (moderate heat), HE (high earlystrength), LH (low heat cement) and HS (high sulfate) detailed description and equivalents areshown in Table 2-1.Table 2-1: Type designation for Canadian Portland Cement and ASTM equivalent(CSA, 2013)2.2 Composite cement manufacturingComposite cement refers to the addition of raw mineral additives in the final clinker grindingprocess that reduces the use of clinker in the finished product. Usually cement manufacturers allowsubstitution of 5% to 35% of the clinker. The substitutes can be limestone, blast furnace slag, coalfly-ash, synthetic or natural pozzolan also identified as PLC (Portland Limestone Cement), PFC(Portland fly-ash cement) and PPC (Portland Pozzolan Cement). There are economic,environmental, and technical advantages related to the manufacture of composite cements. M InCanada manufacturers are slowly accepting the manufacture of composite cement after several11well-publicized trials and rigorous testing and standards development such as the EuropeanStandard EN 197-1 (EN, 2011) , ASTM C595 (ASTM C595, 2014), and Canadian CSAA3000/3001 (CSA, 2013) . Economic benefits relate to the fact that less fuel per unit of cementproduct is required in calcining since the clinker is replaced with a raw composite material thusreducing the thermal energy required to manufacture clinker. Environmental benefits are relatedto the emissions of greenhouse gases from combustion sources.2.2.1 History of the use of limestone additions2.2.1.1 EuropeThere have been several early documented experiences on the use of limestone addition in cementmanufacturing. In Europe, a number of countries allowed different percentages of limestone priorto the adoption of the European Standard EN 197-1 (EN, 2011) . For example, in Germany cementswith 20% limestone were produced by Heidelberg Cement as early as 1965 for specialtyapplications in industry (Nokken et al., 2007).In the 1987 draft of EN 197-1, a cement designated as PKZ (Portland Kalkstein Zement) wascomposed of 85+/-5% clinker and 15+/-5% limestone (Nokken et al., 2007). By 1990, 15+/-5%limestone blended cements were reported to be commonly used in Germany. In the UnitedKingdom, BS 7583 (BS, 1996) allowed up to 20% limestone cement in 1992. European Standardnumber EN 197-1 now allows all of the 27 common types of cement to contain 5% MinorAdditional Components (MAC) or mineral additives, which most typically are either limestone, orpozzolan as shown in Table 2-2 (EN, 2011).12Table 2-2: European standard related composite cement(EN, 2011)According to Table 2-2, six different types of cement allow higher amounts of limestone in twoclinker replacement levels, CEM II/A-L and CEM II/A-LL (6 - 20% limestone). The use of CEM13II limestone cements has grown from 15% in 1999 to 31.4% in 2004 and is now the single largesttype of cement produced in Europe as shown in Figure 2-2.Figure 2-2: Data on cement types produced in Europe(Nokken et al., 2007)2.2.1.2 North AmericaIn Canada, CSA (Nokken et al., 2007) has allowed up to 5% limestone addition in the clinker tomake composite cements and defined as Type GU under CSA 3,001 designation since 1983 (CSA,2013). This was related to the presentation of data from the Portland Cement Association in Canadato CSA that 5% limestone had no detrimental effect on concrete properties based on several studies(Sohoni et al., 1991). There have been attempts to allow a maximum of 12% of limestone addition14in the clinker grinding circuit, which is related to market driven forces and new regulatorygovernment standards in Canada.2.3 Finished grinding and quality of cementGrinding of clinker with additives is the final part of the cement production. It has a great impacton finished product quality. Specific surface area (SSA), particle size distribution (PSD) anduniformity factor (n) are important physical parameters affecting cement service properties. Theseparameters define the proportion of fine and coarse particles in the cement. Grinding technologies(Ball mill, VRM, HPGR or Horomill) have different effects on the particle (Celik, 2009). It is noteffective to excessive grinding in order to obtain a large surface area. The ground product mustfollow certain criteria relative to its particle size distribution in order to ensure the hardeningprocess. According to Duda (1976) , the technology of grinding clinker is based on the followingaspects: The particle size fraction from 3 to 30 microns is conductive to the most strength developmentof the cement. The particle size fraction below 3 microns contributes to the initial strength only. This particlefraction hydrates faster and after one day results in the highest compressive and flexuralstrengths. The fraction above 60 microns hydrates slowly and does not have significant contribution tothe strength of the cement.15The particle size distribution controls some cement quality parameters such as water demand,setting and hydration reaction (Schiller & Ellerbrock, 1992), heat release, capillary porositypercolation, diffusivity, shrinkage and microstructure (Bentz, 1999).The Uniformity factor, “n”, is defined by the slope of the graph representing the size distributionusing Rosin-Rammler mathematical function (Gupta & Yan, 2006) is shown on Figure 2-6. Itdefines the size distribution as “narrow” (sharp cut with a high slope or high uniformity value) or“wide” (prolonged slope with a low uniformity value). It is already established that narrow andwide particle size distributions under Rosin-Rammler plot have different influences upon cementproperties. Wider particle size distribution increases packing density and decreases water demand,while a narrower particle size distribution gives higher hydration rates for equal specific surfacearea (Celik, 2009). A narrow particle size distribution, produced by closed circuit grinding withhigh efficiency separators, influences both cement paste and concrete properties (Sumner, 1989).For samples having a constant position parameter, the 28 days strength remained unchanged evenfor increased slope. This is because the position parameter of a cement sample lies in the range of15–32 µm which is the determinant particle size range for strength development (Ellerbrock,1985).The relationship between the compressive strength and uniformity factor according to Gaudin-Schuman is shown in Figure 2-3. Higher compressive strength is obtained when the value ofuniformity factor “n” is higher than 2 (narrow slope). Lower than 2 (wide slope) the compressivestrength is almost a constant (Celik, 2009).16Figure 2-3: Relationship between compressive strength and uniformity factor according to Gates–Gaudin–Schuman plotting(Celik , 2009)Composite cements produced by grinding clinkers at a Bond work index of 13 kWh/t withlimestone have several benefits. Limestone is softer than clinkers having a bond grindability Bondwork index average, of 4.6-12.61 kWh/t (Bhatty, 2011) and therefore requires less energy to grindto the same fineness. Figure 2-4 shows the energy required to grind each of the two materials toobtain various specific surface areas. It can be observed that energy required to grind limestone ismuch less compared to clinker to obtain the similar specific surface area.17Figure 2-4: Grindability of clinker and limestone(Opoczky, 1996)As the content of limestone increases, the energy required to produce the same fine productdecreases under controlled lab conditions and shown in Figure 2-5 . It can be deducted fromreference (Opoczky, 1996) and Figure 2-5 that replacement of clinker with a material of lowerwork index as limestone decreases the mixed grindability and thus the energy usage.Figure 2-5: Grindability of limestone - clinker cement mixtures(Opoczky, 1996)18The particle size distribution of any mixture of harder and softer constituents inter ground togetheris affected by each of its respective grindabilities (Schiller & Ellerbrock, 1992). On Figure 2-6,Portland limestone cement gave a wider particle size distribution (lower slope) than other cementsthat were inter-ground with fly ash or natural pozzolan due to the softness of limestone comparedto fly-ash and pozzolan. PC represents Portland Cement 95% clinker and 5% gypsum.Figure 2-6: Uniformity factor of inter-ground cement mixtures(Voglis, 2005)When producing Portland limestone cements in order to provide equivalent compressive limestonestrengths (compared to just Portland cement) the grinding time required is increased and having agreater surface area to obtain the targeted compressive strength, this by having to increase theenergy use. It can be deducted from Figure 2-5 that the grain size was considerably reduced mainlydue to the effect of over grinding limestone which was caused by the presence of a harder clinkergrinding media that abrades easily the soft limestone (Opoczky, 1996).19Tsivilis et al. (1999) evaluated the production of fines when grinding clinker and limestone forvarious times. In general, the clinker is concentrated in the coarser fraction because it is moredifficult to grind it than limestone. Limestone reports to the finer fraction at an early grinding stageas the harder clinker abrades the limestone and returns to the mill in the circulating load as harderconcentrated circulating load.2.3.1 AgglomerationOvergrinding the cement in a ball mill can have a negative impact on production as energyincreases due to the agglomeration effect and the drag of fine particles. This is a reason why sizeclassification is important at the circuit. Agglomeration is the bonding of small particles one toanother under the Vander Wall forces principle. It is a consequence of breakage energy oversupplyby the effect of grinding impacts of high energy level generally from higher grinding media sizeimpacts (larger balls), temperature and  the crystals structure of material ground. Limestone isconsidered a highly agglomerative material (Tamás, 1983). Agglomeration can be reduced byimproving separator efficiency and reducing the oversized grinding media and the use of grindingaids.2.4 Grinding technologies in cement industryGrinding systems in cement industry play an important role in the particle size distribution andparticle shape. This affects the reactivity of the clinker and the temperature dependence ofdehydrating gypsum that is ground together with the clinker. These factors affect the mortar20properties of the cement product such as water demand, initial and final setting times and strengthdevelopment (Celik, 2009).Ball mills have been used as the main grinding equipment for finished cement production for over100 years. Although simple to operate and cost competitive relative to other technologies, the lowefficiency of ball milling is one of the main reasons for the development of more efficient grindingprocesses in recent years. Vertical Roller Mills (VRM), High Pressure Grinding Rolls (HPGR),Vertical Shaft Impact crushers (VSI) and more recently, the Horizontal Roller Mill (HOROMILL)(in which energy consumption is substantially reduced) has resulted in an improvement between45-70% in specific energy related a typical ball mill (Seebach, 1996).2.4.1 Ball millBall mills or tubular mills are built with diameters up to 6.0 m and lengths up to 20 m; the driveratings today are as high as 10,000 kW with stable operation and maintenance of a ball mill isrelatively simple. The maintenance cost and the capital cost are relatively low compared to othertechnologies. Due to the high levels of operational reliability and availability (~95%) ball millsremain the most frequently applied finishing grinding unit in cement plants. Compared with newermilling devices as VRM, HPGR, ball mills have the highest specific power consumption and thelowest power utilization (about 32–35 kWh/ton depending on the material hardness and to afineness between 3,000-3,200 cm2/gr) (Seebach, 1996) Most of the energy is lost as heat from thecollision of the steel balls among themselves and against the mill walls (Duda, 1976).21Portland cement production is usually finished using a two compartment ball mill as shown inFigure 2-7. First compartment or chamber 01 is known as the coarse chamber and in the secondcompartment material is finely ground. Between the two compartments there is a classificationdiaphragm that screens the fine form the coarse material.Figure 2-7: Two-compartment tube ball mill. A- Compartment 01, B- Compartment 01/02 andseparating diaphragmGenerally on cement mills, the product is ground dry in a ball mill has a relatively wider particlesize distribution; hence it is required to operate the ball mill in closed circuit with a size classifierwith an efficient or sharp cut of size separator. This happens especially when high levels of finesare generated, when mixtures have low Bond work index or grinding materials that have a tendencyto agglomerate due to overgrinding effect. The circulating loads range from 100% up to 600% thatare established based on the grindability of the new feed, the cut size, and the required productfineness in relation to reaching the adequate cement strength (Duda, 1976).22The energy efficiency of dry ball-mill grinding of cement depends on factors such as: ball chargefill-ratio, mill length/diameter ratio, size distribution of the ball charge, operating conditions of theair separators, air flow through the mill, production rate, use of grinding aids and the hardness andfineness of the feed and product (generally referred to as the Work Index (kWh/t) and the F80 andP80 sizes respectively) (Gupta & Yan, 2006).2.4.2 High pressure grinding rollIn High Pressure Grinding Roll (HPGR), the material is reduced by a highly compressive stresscreated by two counter-rotating rolls (one fixed and another floating). This creates a criticalfracture process that presses the material into a compact flow area. This flow area is shown inFigure 2-8. The grinding pressure between the rolls is 50 to 350 MPa, and the circumferentialspeed of the rolls varies between 1 and 2 m/s (Rosemann & Ellerbrock, 1998) on the grindabilitycharacteristics of the feed and the pressure applied to the roll, the compacted cake (consisting ofover 70% solids by volume) has a fine fraction below 90 µm. Up to 40% of these fines must berecovered by de-agglomeration of the compacted cake using another de-agglomerating device. Thespecific power utilization is between 14.6-19.8 kWh/t at a Blaine area 3,000-3,200 cm2/g(Seebach, 1996). HPGR are reported to be 45-60% more efficient than ball mills (Seebach, 1996).23Figure 2-8: High pressure grinding roll(KHD Humboldt Wedag, 2011)Trouble-free operation of an HPGR depends to a great extent on ensuring proper moisture below3% and the maximum particle size of the material should not exceed 1.5 to 2 times the gap width.Feed is distributed evenly along the rolls; and foreign material (scats) is not allowed to pass intothe rolls and is captured using a magnetic separator system. HPGR is covenient to comminutematerials that are not overly fine and have low moisture content. Material above 3% moisture mustbe pre-dried before feeding to the rolls. HPGR can be integrated into various circuitsconfigurations in new and existing grinding plants to increase the output of plants that have  onlyball mills with precrushing before a ball mill (Seebach, 1996).2.4.3 Vertical roller mill (VRM)VRMs with integrated classifiers have been used successfully for many years in cement plants togrind and simultaneously dry raw materials with moisture contents up to 20% by weight (Seebach,1996). Their production can be as high as 400 tph and have a drive power of 11.5 MW (S.L, 2014).24The feed is comminuted by pressure and friction between a horizontal rotating table and 2 to 4grinding rollers hydraulically pressed against the table as shown in Figure 2-9. Nowadays, thegrinding rollers have diameters as large as 2.5 m. The material being ground is carried bypneumatic and mechanical transport to the classifier located in the same housing directly abovethe grinding chamber. The classifier tailings (over-size rejects) are recycled back into the grindingchamber together with the fresh material. The grinding elements and mill settings are modified togrind harder materials such as clinker and granulated blast furnace slag. Power use is between 26-29 kWh/t when grinding to a Blaine 3,300 cm2/g using a VRM (Seebach, 1996).Figure 2-9: Vertical roller millVertical roller mills integrate the grinding, drying and separation processes into one unit. Thisintegration makes the VRM competitive in terms of specific electrical power consumption25compared against other technologies. According Seebach VRM are 50% more efficient than ballmills when comparing kWh/t used to grind same product under similar service properties(Seebach, 1996).2.4.4 Horizontal roller millThe horizontal tube (or horizontal roller) mill has a length/diameter ratio around 1.0 and issupported and driven on axial bearings. A solid single armored grinding roller is pressedhydraulically against the rotating inner drum surface within a cylindrical grinding zone as shownin Figure 2-10. The pressure is much lower than HPGR and is comparable to VRM. No compactedcake is produced that requires further deflaking. The grinding roller is supported on bearingsoutside the grinding tube. Internal fittings are subjected to heavy wear, however wear of thegrinding elements is still lower with VRM. Power consumption on horizontal roller mill whencompared against a ball mill is reduced by 10 to 25 kWh/t of cement depending on clinkergrindability and Blaine specific surface area (Aguero & Meech, 2014).Figure 2-10: HoroMill schematic diagram262.5 Improving grinding efficiency in closed circuit cement ball millFinal grinding of the cement is the most energy demanding part of the manufacture processconsuming almost 50% of the electrical energy (Bhatty, 2011). On a plant averaging consumptionof 110 kWh/t (electrical energy), use can be broken down according each main consumptionprocess as shown in Table 2-3. Cement grinding circuits operate more efficient in closed circuitconfigurations. Now with the implementation of high efficiency size separators, a more preciseparticle size cut product can be obtained, improving the quality of the cement. The separatorconfiguration can be arranged in different ways but it is usually related to the conserve heat bysending back or recirculating clean hot air from the grinding process (FLSmidth, 2014).Table 2-3: Electricity consumption during production of ordinary Portland cement(Nisbet, 1996)Process % kWh/t of cementQuarry 5 5.5Raw mix preparation 17 18.7Pyro-processing 29 31.9Finish Grinding 49 53.9Total 100 110.0Low-cost energy improvements can often be achieved with existing equipment with minor changessuch as the use of grinding aids and optimization of grinding media, improving the size separatorefficiency and the use of classifier liners (FLSmidth, 2014). These improvements can readily be27determined from circuit production surveys (particle size distributions, work indices) to defineequipment baseline energy levels for process stages and specific energy (kWh/t) of individualstages and materials to identify potential improvements in production strategies (Aguero & Meech,2014).2.5.1 Grinding aidsGrinding aids are chemical additives used to improve the production efficiency of cement plantsand energy consumption by reducing the boundary surface forces (Sohoni et al., 1991). Table 2-4shows the benefits of addition of grinding aid to the energy usage.Table 2-4: Effect of Sika™ Polycarboxylate ether polymers grinding aid on energy use(Sica, 2015)Scenario / feed ratio % Production(tph)Grinding energy use(kWh/t)Reduction inEnergy *%No grinding aid 80 50.2 0.00.018% grinding aid 85 47.3 4.30.035% grinding aid 90 44.7 8.6* includes the energy associated with manufacturing the polymerThese products enhance particle size distributions as well as powder "flowability" of the finishedcement (Opoczky, 1986). Two important mechanisms have been put forward to explain the actionof various grinding aids. Sohoni et al. (1991) explained a mechanism known as the "RehbinderEffect" which is based on the assumption that the action of grinding aids depend on the reduction28of the specific surface free energy of freshly ground material through the adsorption of a surface-active chemical. By reducing the surface free energy, the grinding aid helps propagate micro-cracks of fractured particles from impact that prevents the particles from binding together. Thismechanism also helps to explain the lack of coating on balls and mill liners with a fine particle bedthat absorbs impact energy. Beke (1983) has added to this mechanism with the idea that adsorptionof a grinding aid causes induced mobility of near-surface dislocations causing a lower hardness.2.5.2 Optimum media ball size and mixing ratioVarious formulas have been proposed by different researchers for optimum ball size. Based onthese formulas and a number of empirical rules, mixing phenomenon in the mill has been studied.In recent years, the ball size has decreased due to the adoption of improved classification linerconfigurations and to the use of a pre-grinder (Asia Pacific Partnership, 2011). According to Asia-Pacific Partnership on Clean Development & Climate Cement Task Force (Asia PacificPartnership, 2011), the percentage of tube mills using the smallest ball size of 17 mm  has increasedfrom 10%  to 80% from 1979 to 1991.A critical speed must be maintained in order to avoid centrifugation of grinding media duringcomminution. It is calculated according to Equation 1 and the mill speed should be around 60 to80% of that critical speed.Nc =42.3/ (D) 0.5Equation 1: Critical speed (Bhatty, 2011)29Where:Nc = Mill critical speed, RPMD=   Ball mill diameter, meters2.5.3 Fill factor of grinding mediaFill factor is the percent volume of a ball mill occupied by the grinding media. The fill factor ofgrinding media greatly affects the grinding capacity and power consumption of a mill. For cementgrinding, the optimum value is around 26-30% (FLSmidth, 2014) . To keep the fill factorappropriate for high grinding efficiency, continuous replenishment of grinding media is necessaryto compensate for the abrasion of the media. There are five methods to determine the mediareplenishment time. The graph effect of fill ratio on mill efficiency can be seen on Figure 2-11.30Figure 2-11: Optimum ball mill void filling(Bhatty, 2011)2.5.3.1 Load powerThe baseline power draw of the mill is an indication of the ball fill charge level. This can be usedas a set-point to maintain the fill factor on target. This method gives less variability in the fill factorover time, but it must be periodically checked using one or more of the remaining techniques.2.5.3.2 Blank height measurementThis method relies on a visual estimate of the height of the ball charge in the mill when it isshutdown. It is typically done once per shift or per day. The required addition of balls is calculatedfrom a fill formula based on the geometry of the charge (Gupta & Yan, 2006).312.5.3.3 Run-timeBased on the time the mill is turning under load, balls are added periodically based on a formulato predict ball wear rates. This method is not particularly accurate.2.5.3.4 Ground tonnageBased on the tonnage rate of ore being ground over a period of time, balls are added periodicallybased on a formula to predict ball wear rates. This formula is generally a little more accurate thanthat based on run-time.2.5.3.5 Grinding efficiencyBased on grinding efficiency measured periodically calculated using average power draw, totalore processed, and the average particle sizes of feed and product over the time period in question,a formula is applied that predicts ball wear rates, this formula is shown in Equation 2. Based onthe calculations grinding balls are added periodically (shift or daily basis) (Moly-Cop, 2012) .Wt = d (mb)/d (t) = - km AbEquation 2: Ball mill media charge wear rateWhere:Wt = mass wear rate, kg/hrmb =  ball weight, kg; after t hours of being charged into the mill.32Ab = exposed ball area, m2km =  mass wear rate constant, kg/hr/m2.2.5.4 Pre-grinding of raw addition materialMost ball mills operates in dry condition to grind cement have two chambers, one for coarse andone fine grinding. Ball size and distribution of sizes are designed and adjusted to take into accountraw material feed conditions and mill dimensions. However, energy efficiency in the coarsechamber is generally lower than in the second chamber. Furthermore, some plants feed cementadditives at particles sizes well above the maximum size (i.e. 12 mm) that a ball mill can process(FLSmidth, 2014). This can significantly limit the ability to improve both coarse and fine grindingperformances. Recently, a new system has been applied in which a pre-grinder (VRM or HPGR)is installed to perform coarse grinding ahead of an existing tube mill, which then is exclusivelyused for fine grinding. This system greatly reduces total specific power consumption and canimprove production as well (Seebach, 1996).2.5.5 Air classifiersAir separator is a key component for close circuit efficiency. The performance of a grinding plantdepends on the type of grinding technology. Modern grinding equipment incorporates airseparators to comminution devices in the upper part of equipment as the Vertical Roller Mills.  Thedispersion separator (static or dynamic), is the most used separator in the cement industry. A33distribution plate at the feed inlet is used to disperse evenly the feed into the separator (FLSmidth,2014).The operation principle is based on: the action of an air current of certain velocity upon a massparticle is proportional to the projected surface presented by this particle to the air current, so thesquare of the average diameter size of the particle. The action of the force of gravity upon a mass-particle is proportional to the volume, in other words to the cube of the mean dimensions of theparticle. Therefore the effect of the gravity increases faster than that of an air current of constantvelocity. If these two forces are concurrent, the gravity will prevail over the effect of the air currentas particle dimensions increase. On the other hand a properly adjusted air current will oppose theforce of gravity and lift up the smallest mass particles (Duda, 1976). A diagram of an air separatoris shown in Figure 2-12.Figure 2-12: First generation air separator(FLSmidth, 2014)34The separator inlet feed passes through the feed spout and drops by gravity to the distribution plate.The drive operates a rotating fan that promotes a continuous circulating internal air current (definedas circulating air separator) and the distribution plate disperses the feed evenly into the separator.Materials leaving the distribution plate are acted upon three forces as shown in Figure 2-13:1. - The centripetal force, Fc2. - The force of drag, Fd.3. - The force of gravity, Fg.Figure 2-13: Forces balance in an air separator(FLSmidth, 2014)35Air velocity, volume of air, density of material, particle size feed and speed of rotation areimportant factors in the separating fines to coarse particles. The distribution plate must exert to theparticles a centripetal force of adequate magnitude to send the particles to the classification zonefaster than the new feed is received at the air material entry. As more density material and largerparticles are sent to the outer body of separator and the particles centripetal force is decreased.They settle because of gravity. If the particles hits the body of cyclone wall the effect will forcethe particles into the rejects (Duda, 1976). Some small particles are entrained between largerparticles causing a “bypass” effect to the rejects. Another cause of bypass is agglomeration due tovan der Waal effects (Tamás, 1983).Undersize particles (finished product) are dragged up to the cut cyclone size and lifted by theascending air current and passing between the blades of rotating fan. Underneath the separatingzone the return air vanes are located to improve separation. The separation of the fines from thedescending air current in the outer separator cone is performed by decreasing the air velocity aswell as by the change air current. Because of the low rate of descent of the smallest particles, thesefractions are always suspended in the air stream and therefore a portion of finished product iscontinuously circulating resulting in fraction of the fines comes into separator rejects.Main fan, auxiliary fan and dispersion plate are mounted in a common shaft. The auxiliary fan actsagainst the intake air current caused by the main fan. This counteraction can be controlled by thenumber of blades of the auxiliary fan. A large number of blades cause a stronger counteraction.An adjustment in the number of blades is necessary when switching to other types of cement(Duda, 1976).36Another possibility for fine adjustment of dynamic separators is by the use of horizontal controlvalves which make it possible to change the cross section of the ascending air current. By adjustingthe control air valve is possible to strangle the air stream and shift the classification boundarycloser to the fines.2.5.5.1 Performance of separatorsThe performance of any type of separators is determined using the tromp curve (also identified as“partition curve”, “selectivity curve” and “probability curve”). Tromp curve is a graphic thatcombines the fractions in a sample and the fractions of particles of different sizes in the feed goingto coarse (rejects) or fines (product). Some important parameters are depicted on the Tromp curve: Cut Size (x50). Defined as the size of particles with equal distribution in the fine and as thecoarse fraction. This value is adjusted by setting the speed according the product size required. Sharpness factor (k). Is a measure for the steepness of the tromp curve. It is calculated as thesize of particles of which 25% pass into the rejects divided by the size of which 75% passk=D25%/D75%. A good separator has values between 0.52-0.58 Delta or bypass (δ). The lowest point on the curve, indicating the amount of bypass of goodproduct reporting to the rejects. This value should be between 10-15%. Bypass is affected by:agglomeration resulting from over ground small particles, poor dispersion at distribution plateand by the dragging effect of circulating larger particles with air into the rejects.A typical tromp curve for the rejects parts with plots for bypass, cut size and sharpness is shownin Figure 2-14.37Figure 2-14: Typical tromp curve(www.thecementgrindoffice.com)Agglomeration is the electrical attachment of very fine particles as a consequence of van der Waalsforces (on the order of 40-400 kJ/mol) in which charges of the crystal lattice suffer structuralchanges. Overgrinding and high impact energy events are some of the effects of agglomeration.This condition can be reduced by the use of a grinding aid (Tamás, 1983).2.5.5.2 High efficiency on separatorsHigh efficiency size separators improve the process by the application of air vortex allowing thecentrifugal and drag forces to interact effectively and perform a good classifying function. The38control of a particle motion is essential to the improvement of the separator performance. Limitingthe random motion of particles allows more fines to be removed from the mill. Also they operatewith cooler air, allowing a reduction in temperature of the finished product and rejects and themill.High efficiency separators reduce the energy consumption by: first removing the fines from thesystem sending the fines to the finished product preventing the fines from returning to mill andcause overgrinding, second by controlling the fines that cause a cushioning effect on breakage inthe mill. On average high efficiency separators reduces the specific power consumption in finishgrinding by 20-30% (Brugan, 1988).Modern separator manufacturers suggest that the fraction of fines in the distribution loading feedshould be 2.0- 2.5 kg of feed per actual cubic meter of air. For the finished product the effectivetransport concentration value should be between 0.75-0.85 kg of product per actual cubic meter ofair (FLSmidth, 2014).The separator finished product or efficiency is related to the amount of rejects which pass into thefines. The criterion for separator capacity is the amount of fines present into the rejects.The types of separators and characteristics of air classifiers according the generation type used inthe cement industry is shown in Table 2-5.39Table 2-5: Distinctive features of separators of different designs(Bhatty, 2011)Evolution First generation Second generation Third generationNomenclature ConventionalseparatorCyclone-air separators High EfficiencySeparatorsEfficiency 50-60% 60-75% 80-90%Bypass 30% 10% 2%Reclassification oftailingsNot efficient due tolack of fresh airBetter as recirculation aircontains fresh air alsoMore effective as freshair is usedFines collection In the outer cone ofseparatorIn the external cyclonesattached to separatorsIn cyclones or bag filterattached to mill systemCommercialexamplesSturtevantTurbopol (Polysius)Cyclopol (Polysius)ZUB (KHD-Wedag)O-sepa (Fuller)Sepax (FLS)Sepol (Polysius)Sepmaster (KHD)The finer the particle size of the finished product, the lower the separator’s production capacity.The efficiency of an air separator depends upon the type of the mill (BM, HPGR or VRM) workingwith the separator. It is possible an increase  production by 10-30% by replacing a poorperformance low efficient separator (first or second generation) by installing a high efficient (thirdgeneration) (Cleemann, 1986).402.5.6 Circulating loadThe circulating load has a big impact on mill efficiency. There is a relationship between the lowerenergy use to related  specific surface area and circulating load as reported by (Bhatty, 2011) andshown on Table 2-6.Table 2-6: Influence of circulating load and type of separator on mill efficiency(Bhatty, 2011)Mill Circuit Open Closed Closed ClosedSeparator type none 1st gen. 1st gen 3rd genProduct SSA, m2/kg 370 370 370 370Rejects SSA, m2/kg n/a 220 220 90Circulating load,% 100 300 500 300Mill exit SSA, m2/kg 370 270 250 183cm2/joule 22.9 26.2 26.8 27.5kWh/t 44.9 39.2 38.3 37.4Mill output,% 100 114 117 120For high efficiency separators there are smaller amount of fines in the rejects and as result SSA islower (90 m2/kg). As result the exit from the mill (183 m2/kg) is lower giving a similar comparativeproduct of 370 m2/kg, but using less energy at mill (37.4 kWh/t) and increasing production of120%.412.6 Impact on carbon dioxide emissionsCement production has a large CO2 footprint due to the tremendous use of cement around theworld, it is estimated that 3.6 billion of tons were produced on 2012 (Kline & Kline, 2014)representing the third largest CO2 emitting industry by anthropogenic sources (Abdel-Aziz et al.,2014) as shown on Figure 2-15. The vast quantities of cement (3.6 billion t/a) used around theworld today make cement production one of the leading sources of CO2 emissions and representsthe second most used commodity in the world (Kline & Kline, 2014).Figure 2-15: Total global industry direct greenhouse gases emission(Abdel-Aziz et al., 2014)There are several alternatives to reduce CO2 emissions on cement production such as, Carbon Capture/ Sequestration (CCS)42 Clinker substitution by similar properties minerals (limestone, pozzolan) The use of alternative fuels (biomass, used tires, industrial wastes) Energy efficiency (lighting, high efficiency motor, compressed air optimization, and highefficiency separators).The Cement Sustainability Initiative (CSI), in junction with the International Energy Agency (EIA)have developed a road map for CO2 reduction in the cement industry. The IEA/CSI “blue map”has targeted for a reduction of approximately 50% in the specific CO2 emissions per ton of cementby the year 2050 (IEA, 2009) this target is shown in Figure 2-16. 56% of the targeted reductionwill be from carbon capture and sequestration.Figure 2-16: The IEA/CSI blue map for CO2 emissions reduction(Kline & Kline, 2014)The CO2 emission from cement manufacturing is caused by calcining of limestone and from thecombustion of fuels at the kiln. The amount of limestone calcined in the cement manufacture is43relatively consistent across most cement plants. It can be decreased when alternative sources ofcalcium oxide are utilized, such as slag, fly-ash and/or bottom ash (Kline & Kline, 2014).Evaluation of the CO2 emissions in cement production excluding the emissions from electricityand found it be approximately 680 kg CO2/t of cement produced (Kline & Kline, 2014) as shownin Figure 2-17.Figure 2-17: Average kg of CO2 released per ton of cement produced(Kline, 2014)The CO2 from combustion depends on the system efficiency and the fuel which in turn dependson the technology and type of fuel used. Modern plants often use a 5 stage pre-calciner kiln systemwith an inline raw mill for maximum thermal efficiency. The amount of clinker in the cement hasa direct impact on the specific CO2 emissions per tonne of cement produced.44When calculating the impact on mass flows it is assumed that raw limestone added to the cementcomes from the same quarry as that used in the raw mix. If 5% raw limestone is added to thecement replacing clinker, the amount of clinker used in the final product (95%) is decreased by5.26% (M. Nisbet, 1996).The material balance for the finished Ordinary Portland Cement (OPC) product without anylimestone is shown in Figure 2-18. Produce PLC with 5% limestone is shown in Figure 2-19.Figure 2-18: Material balance - no raw limestone addition(Aguero & Meech, 2014)Figure 2-19: Material balance - PLC with 5% limestone addition(Aguero & Meech, 2014)45However, the total limestone needed per tonne of final cement product actually decreases slightlyfrom 1.216 to 1.202 tonne (1.152 for the raw mix plus 0.05 added to the finishing step) (M. Nisbet,1996).The fuel savings and reduction in CO2 emission by adding 5% and 20% limestone is shown inTable 2-7. It is assumed that the heat combustion of coal is 22.7 MMBtu/t of coal and CO2 emissionis 707 kg CO2 /t of cement produced (Carbon Dioxide Emission Factors for Coal, 2015).Table 2-7: Fuel savings and CO2 emissions reduction with PLC(Aguero & Meech, 2014)Coal as fuel source Units Limestone replacement5% 20%Fuel saved tfuel/tcement 0.014 0.043Reduced CO2 tCO2/tcement 0.0315 0.12582.7 Modeling and simulation of Portland cement circuitsThe modeling of a process is the use of mathematical equations to characterize an operationaccurately and to be able to simulate their impacts on production/efficiency from modifying theirdifferent variables. The model process has been used by several authors (Benzer et al., 2001) inthe characterization of cement production models in different processes. This technique has beenused to model grinding and separation technologies, since it is a low cost and high confidence way46to evaluate the improvements of any circuit based on powerful computers and the research overthe time from various authors (Napier-Munn et al, 1996).2.7.1 Model of performance curves in separatorsThere are several equations published to model the performance of separators. The most adequatedepends on the amount of variables affecting the process. Whiten’s models has been categorizedas the least sum of squares of deviations in particular at fines sizes as compared to others (Altun& Benzer, 2014). The mathematical model defined by Whitten’s is related to the overflowefficiency (finished product) and is defined by Equation 3 and concepts shown graphically onFigure 2-20.= ∗ 1 + ∗ ∗ ∗ 50 ∗ (exp ) − 1(exp ∗ ∗ ∗ 50 + exp − 2)Equation 3: Whiten efficiency of separator reporting to overflowWhere,Eoa : Actual Efficiency of fines to overflow, %C     : Fraction subject to real classification, %β      : Parameter that controls the initial rise of the curve in fine sizes (Also called fish hook)β* : Parameter that preserves the definition of d=d50 When E= (1/2) C47d     : size, mmd50c : Corrected cut size, mmα : Sharpness of separationFigure 2-20: Efficiency to overflow vs size(Napier-Munn et al., 1996)Alpha (α) is a variable from equation that is related to the slope the graph and represents thesharpness or steepness of the efficiency curve. It can varies from values between 0.25 to 10 andhigh values of “α” means a better and sharper (steeper cut) this can be seen on Figure 2-21.48Figure 2-21: Variation of efficiency curve related to α(Napier-Munn et al., 1996)Beta (β) has been identified also a “fish hook” because of its prolonged shape especially on finefractions (below 45 microns). It has been associated with agglomeration in cement classificationoperations. Napier-Munn (1996) high values represents high agglomeration and/or high feed rateto separator and shown on Figure 2-22.49Figure 2-22: Variation of efficiency curve with β(Napier-Munn et al., 1996)Air separator studies have shown there is a relationship which is directly proportional betweenbypass and separator feed load as shown on Figure 2-23. Also, there is evidence of cut size isproportional to airflow and inversely proportional to rotor speed (Altun & Benzer, 2014).Figure 2-23: Relation between (a) feed/bypass (b) and dust loading/bypass(Altun & Benzer, 2014)502.7.2 Model of two compartment ball millsTwo compartment dry ball mill used on cement finishing process can be modeled as multiple millsin series. The classifier diaphragm located between two chambers can be modeled as a screen thisis shown on Figure 2-24. These findings were analyzed using industrial scale experiments (Benzeret al., 2001).Figure 2-24: (a) Two compartment ball mill (b) model circuit array(Farzanegan et al., 2014)Perfect mixing models are based on the principle that the contents of the mill are fully mixed. Theyare  represent by either one perfectly mixed segment or a number of perfectly mixed segments inseries (Gupta & Yan, 2006).51The mathematical equation shown in Equation 4 describes the process of comminution of a particlefor a ball mill model. This relation uses feed and product matrices calculated for the breakage andselection function (Gupta & Yan, 2006):= ∗ ∗ + ( − ) ∗Equation 4: Basic equation model for open circuit (Gupta & Yan, 2006)Where:P     : Product vector for size distribution (mass)B : Breakage functionS    : Selection functionF     : Feed rate matrixI      : Unit diagonal matrixMany circuits operates in closed circuit where a classification device separates particles that needmore grinding (coarse) and sends material to the finished product (fines). This is shown on Figure2-25. Feed and Products are denoted by  F and P respectively, q the size distribution of classifier,B, S and C the breakage, selection and classifier functions respectively. All terms are consideredas vectors.52Figure 2-25: Comminution on a closed circuit(Gupta & Yan, 2006)Equation 5 shows the classification effect on operating in a closed circuit modifies to a newequation that is shown below (Gupta & Yan, 2006).= ( − ) ∗ ( ∗ + − ) ∗ [ − ∗ ( ∗ + − )]¯ ∗Equation 5: Equation model for a closed circuitWhere:P     : Product vector for size distribution (mass)B : Breakage functionS    :  Selection functionC    : Classification functionF1 : New feed53F2 : Mixed feed with classifier coarse and new feedq     : Size distribution of classifierI      : Unit diagonal matrix54CHAPTER 3: EXPERIMENTAL PROCEDURESThis chapter brings details about the experimental tests, sampling and software programs used tocharacterize the cement samples used in this study. These tests include the Bond standard ball milltest, breakage function test and the use of software for the determination of the selection function,Whiten model and Rosin-Rammler particle size distribution plot.3.1 Sampling and data gatheringThe circuit evaluated is identified as Circuit 03 which consists of a vertical pre-grinder, a ball mill,a high efficiency separator and several feeders of limestone, clinker and gypsum. A series of 22samples each with 50 kg from this circuit were taken and provided by the plant. The descriptionof sampling points and sample details are listed below on Figure 3-1.Figure 3-1: Circuit 03 sampling points553.2 Bond standard ball mill grindability testThe purpose of the standard Bond ball mill grindability test is to determine the Bond ball mill workindex (BWI), which can be compared with the work indices of known materials to evaluategrinding efficiency or mill design.The Bond work index is a measurement of the power required to reduce feed with a given 80percent passing size (d80) to product with a specified 80 percent passing size (d80). A completeprocedure on the Standard Bond work index test is included in the Appendix A.3.3 Breakage distribution function test and estimation using BFDS® softwareThe breakage function is a material specific property and denotes the relative distribution offragments after breakage of a monosize sample. It is almost found to be independent of initial sizeand usually is expressed in a matrix array in order to perform a further modelling using populationbalance methods. The breakage distribution mechanism is represented in Figure 3-2.56Figure 3-2: Representation of the distribution of particle breakage(Gupta & Yan, 2006)3.3.1 Clinker breakage function estimation test procedureFor the estimation of the breakage function during this research, the following procedures wereimplemented: Preparation of 300 g of monosize clinker sample. The monosized fractions for this testwere 1.400, 1.000, 500, 355 and 150 microns. A standard Bond ball mill was used for breaking the sample. Grinding at intervals was selected of 0, 5, 10, 25 and 30 seconds. Total content of sample removed after each time taking care to not lose any material andthe screened using at √2 factor screens for size distributions. After screening is done, mill sample is returned for further grinding on the next time period.57 Test is finished after obtaining 50% passing of the reduction of initial size fraction on thetop size screen. The use of a computer program is required to perform and apply any correction factors toaccount for any re-breakage that has occurred during the test. The software used is aBFDS® (Breakage function determination Software). This software is able to calculate thenormalized breakage function using Berube’s, Herbst/Fuerstenua and modifiedHerbst/Fuerstenau methods (Farzanegan, 2015).3.4 Particle size distribution analysisA size distribution is a quantitative representation of the proportion of particles in a sample. Resultsfrom particle size distribution are presented using algebraic forms to find the best fit for theexperimental parameters found for several feed, rejects and products. Results have been reportedin a log size axis to avoid congestion in representing the values and increasing the resolution ofcertain small particles areas in the plot.3.4.1 Rosin-Rammler distributionThe use of presenting data in cumulative percent retained exposes features of the data which areoften suppressed or entirely hidden in the cumulative passing form.This algebraic distribution describes the mass or volume distribution function in an exponentialform. This is suitable especially for very fine ground materials. Resolving the exponential by the58use of logarithms helps to expand the fine and coarse ends of the size range and compress the mid-range. The distribution is shown in Equation 6.= 100 exp − ᵇEquation 6: Rosin-Rammler distributionWhere,R= cumulative mass retained on size x, %x1 = size parameter, mmb = uniformity factorRosin-Rammler will be used specially on the finished product representation in order to evaluatethe uniformity factor as a quality control parameter.3.4.2 Whitens model for high efficiency separatorsThe most precise method to fit the curve of a classification device is the Whiten’s approach,because of its precision using the least sum of squares of deviations especially for the fines sizeswhen compared against others like Tromp curve.The equation is the same described in Chapter 2 on modeling and simulation and listed here again.59= ∗ 1 + ∗ ∗ ∗ 50 ∗ (exp ) − 1(exp ∗ ∗ ∗ 50 + exp − 2)3.5 Selection function and its estimation using numerical grinding optimization tools inlanguage C (NGOTC®) softwareThe selection function or specific rate of breakage is the probability of the breakage of certainparticulate on a breakage process and is a measure of the grinding kinetics. This is a machinespecific property. For modeling it is represented by a matrix and can be estimated by analyzingthe data for a plant process.NGOTC® is a software tool dedicated to calculate some variables including the selection function.The software was used to back calculate the selection function based on real breakage plant dataat ball mill 03 compartments 01 and 02.3.5.1 Back calculation of selection function from continuous mill dataThe estimation of selection function using NGOTC program consist of an algorithm to backcalculate a set or a vector of selection function elements based on a set of input data. The selectionfunction elements are back calculated by trial and error using a bisection search procedure. Theselection function elements are back calculated sequentially i.e. first the selection function elementfor the top size class is determined, and then using the estimated value, the selection function ofthe second size class is estimated. The core of the algorithm is in fact a single ball mill simulatorwhich produces product size distributions. The criterion to stop the iteration process is the60difference between the measured and predicted mass of current size class, which must be within atolerance interval set by the user. The estimated selection function elements, then, can be used asinput to other modules.61CHAPTER 4: RESULTS AND DISCUSSION4.1 IntroductionResults obtained from testing and the calculation of different variables from samples are presentedand discussed in this chapter. Special emphasis is made on evaluating the current cementproduction circuit 3. Modeling to air separator will be briefly presented and discussed. All samplesand operating condition parameters, reported values presented in this chapter and attached in theappendix were provided by the cement company surveyed. Data and samples gathered were takenbefore and during the plant crash stop on circuit 03 on May 2014. The ball mill and air separatordesign and operating condition details of circuit 03 equipment on producing type 10 cement (3%limestone) can be seen on Table 4-1.62Table 4-1: Circuit 03 equipment design specifications and operating condition on type 10 cementBall MillDiameter (m) 4.42Operating power (kW) 3,800Length first chamber (m) 4.17Length second chamber (m) 9.14Ball load  first chamber % 22.78Ball load  second chamber % 27.15Top ball size first chamber (mm) 63.5Top ball size second chamber (mm) 32Air ClassifierManufacturer FLSmidth OSEPA 3000Volumetric air flow (m3/h) 130,000Rotor speed (rpm) 205Installed power (kW) 223Feed tonnage (tph) 390Dust Load (kg/m3) > 2.6634.1.1 Particle size distribution and production4.1.1.1 Fresh feed size distributionFresh feed for the purposes of this study is defined as the feed (clinker, gypsum and limestone) tothe ball mill excluding the circulating load (rejects from the air classifier). Joint feed is the feed tothe ball mill considering the addition of the circulating load to the fresh feed.The average production on Type 10 cement on circuit 03 is 125.0 tph with a circulating load of211%. The size particle of clinker is reduced initially at a vertical pre-grinder (CKP) with a d80size from 15.0 mm to 1.98 mm before feeding the ball mill. Limestone and gypsum are fed directlyto the ball mill at d80 size of 13.70 and 61.14 mm respectively. The combined fresh feed has a jointd80 3.29 mm. Particle size distribution individually for limestone, gypsum and clinker before andafter CKP are shown in Figure 4-1.64Figure 4-1: Particle size distribution of circuit 03 fresh feed and clinker at CKPThe weight proportions of fresh feed entering the ball mill: clinker 91%, limestone 3% and gypsum6%. The ball mill product, with a flow rate of 390 tph, is fed to a high efficiency separator by abucket elevator where oversize rejects (circulating load) are sent back to ball mill for furthergrinding and undersize products are sent to a bag house filter to be recovered, transported and thefinished product stored as detailed on Figure 4-2.Specific energy consumption of the ball mill varies between 28.9 and 30.7 kWh/t. CKP is usedjust for pre-grinding clinker before feeding the ball mill. It has a specific energy consumption thatvaries between 5.00 to 5.42 kWh/t.Finished product is recovered after being classified in a high efficiency separator at a rate of 125.2tph and generates a d80 size of 0.0268 mm as seen on Figure 4-2.0.020.040.060.080.0100.00.1 1.0 10.0 100.0CUM passing, %Size, mmlimestone clinker before CKP Clinker after CKP Gypsum65Figure 4-2: Current d80 feed and product size at circuit 03Based on the assessment of the samples the following results can be obtained: The combined 125 tph of fresh feed to the ball mill (excluding circulating load) has a d80 of3.29 mm. This is the result of mixing 91.37% clinker at d80 of 1.98 mm, 2.95 % raw limestonewith d80 of 13.70 mm and 5.68 % gypsum with a d80 of 61.15 mm. The average circulating load (rejects from air separator) is 264.8 tph with a d80 of 0.063 mm,and represents a circulating load of 211%.66 Vertical pre-grinder CKP has a size reduction ratio of 7.6 .The feed of clinker to the CKPhas a d80 of 15.05 mm, and the product a d80 of 1.98 mm. Specific average energy of ball mill and vertical roller pre-grinder CKP is 30.77 and 5.42kWh/t respectively, adding up to a total specific energy of 36.19 kWh/t. There is a potential opportunity to optimize the fresh feed by reducing the limestone sizefrom a d80 of 13.70 to 1.98 mm at the CKP (currently just clinker is brought to CKP). Thelimestone has similar feed size that can be brought without making major modifications tocurrent CKP. The calculated fresh feed size d80 by pre-grinding limestone through CKPand replacing the feed by 3% limestone has been calculated as d80 of 2.98 mm. The estimated fresh feed size d80 by pre-grinding limestone through CKP and replacing thefeed by 12% limestone has been calculated as d80 of 2.38 mm. (this is when producing 12%PLC). Composite cement manufacturing cautions. When planning to increase the use of rawlimestone on composite cement manufacturing (PLC) extreme care should be taken on theoperating parameters of air separator, because of the change of densities and uniformityfactor of the mixture of the feed (limestone and clinker) that will behave differently. Theuse of extra grinding aid could be required in order to reduce agglomeration due toincreased limestone use.674.1.1.2 Air separator size distributionThe material milled at the ball mill fed the air separator at a rate of 390 tph with a d80 of 0.050mm. The oversize rejected particles with a d80 of 0.064 mm, are cool down and returned to the ballmill inlet as a circulating load. The undersize fines or finished product with a d80 of 0.027 mm arecooled down and sent to a bag house filter to be recovered and stored as finished product. Theparticle size distribution can be seen in Figure 4-3.Figure 4-3: Particle size distribution on air separatorThe calculated efficiencies of air separator at size fractions of 0.020, 0.035 and 0.045 mm are52.24, 64.74 and 45.43 % respectively. This calculation is based on the mass flow of finishedproduct divided by the mass flow of the feed for each fraction, and shown in Equation 7.0204060801000 20 40 60 80 100 120 140 160 180 200Cum Pass, %Size, micronsSeparator Product Separator Feed Separator Rejects68= 100( × )( × )Equation 7: Efficiency to overflowWhere Woi, Wfi are the proportions by weight of material of size “ith” in the overflow and feedsolids respectively, and Mo, Mf are the total solids mass flowrates of the overflow and feed streamsrespectively.The air separator rejects 109.89 tph of material in fraction less than 0.035 mm (most importantfraction for strength development), these fractions returns to the ball mill for overgrinding insteadof being recovered as finished product. This is an indication of a high degree of inefficiency at theseparator representing 46.06% efficiency, mainly caused by high dust load feed and agglomerationof fine particles to the feed of separator and shown on Figure 4-4.Figure 4-4: Mass balance at air separator694.1.1.2.1 Separator partition curveIn order to evaluate the total performance of the separator with regard to particle sizes the use ofthe partition curve for the selectivity process or probability of rejection for the whole system wasused. The measured and predicted efficiency curves are shown in Figure 4-5.Data from the feed, rejects and product to the separator were provided by the plant. This data wereused to calculate the actual efficiency curve (also known as Tromp curve) data points. Then,Whiten model was used to fit the Tromp curve obtained based on real plant data. The purpose wasto obtain a smooth curve that represents the most accurate description of the process by adjustingthe experimental errors and process instabilities.Figure 4-5: Separator efficiency reports to product fit using Whiten model0.0010.0020.0030.0040.0050.0060.0070.0080.0090.00100.000 20 40 60 80Reports to Product, %size, µmTromp Whitenα = 1.04105β = 1.61607d50C = 33.67 µmd50 = 30.36 µmBypass = 15.0 %70The area of the graph circled in red represents the “fish hook” region (Altun & Benzer, 2014),(Napier-Munn et al., 1996), this area is an indicative of loss of efficiency during the separationprocess and is related to agglomeration especially at finer fractions.Table 4-2: Separator efficiency parametersParameter Plant Data value Optimum values Referenced50c 0.030 mmD-limit 0.009 mmBy-pass(Delta), % 15.0 10 - 15(FLSmidth, 2014)Imperfection 0.42 0.2 < I < 0.3Sharpness, % 42.51 52-58Alpha value, α 1.04105 4.0(Napier-Munn et al., 1996)Beta value, β 1.61607 0.0According to Altun & Benzer (2014), the most effective model fitting is the Whiten model as ithas the least sum of squares deviation. The following findings are related to the fit of Whitenparameters: The Tromp curve reported by the plant has a strong fit with respect the whiten model at allsizes. Below 0.010 mm the model provided by plant shows a drop which decreases as size isreduced. This difference in both Tromp and Whiten curves is a result of experimental errorsduring sampling and analysis which is expressed in the Tromp graphic but corrected by thesquares differences at the Whiten curve.71 The region circled in red on Figure 4-5 represents the “agglomeration effect” or fish hook.This causes high tonnage of material from the feed going to the rejects (100.43 tph of fractionbelow 0.035 mm). Agglomeration is the result of overgrinding on the ball mill and high energyimpacts  with larger grinding media impacts (Tamás, 1983). These over ground particlesbecome to each other by weak electrical forces (van der Waals forces) creating a larger particlethat behaves, and also selected, as a coarse particle in the separator (Tamás, 1983).Some parameters calculated by the Whiten model are: Alpha “α” related to the sharpness of the airseparator, Beta “β” related to the effect of agglomeration and listed in, Table 4-2 with their comparative recommended values (Napier-Munn et al., 1996). A good Alpha “α” factor value should be around 4 (Napier-Munn et al., 1996). The separatorreported sharpness factor is 1.04105, denoting a low sharpness value. This low sharpnessfactor can be related to some operating conditions of the feed distribution at air separator i.e.air to feed ratio (has an impact on throughput), speed of rotor (has an impact of size ofclassification) (Benzer et al., 2001). The low Alpha “α” factor is double confirmed when comparing against the sharpness valueof 42.51. FLSmidth suggest a good value to be between 52-58%. The mass flow of material through separator with particles sizes between 0.003 to 0.035 mm(fractions related to strength development) are: 181.41, 87.58 and 93.81 tph for feed, rejectand finished product respectively. This balance represents an average of 51.71%(93.81/181.41) efficiency related to the final product at fractions below 0.035 mm. According72FLSmidth who is the manufacturer of air separator this efficiency should be at a maximum of85% for this fraction. The uniformity factor “n” for the finished cement product has a value of 1.30 (this is the slopeof Rosin-Rammler graph). This value represents a high quality product (narrow) with a highsurface area and is the result of the low cut off point of the air separator. But because ofagglomeration and low sharpness the cost of producing this quality product is high, resultingin lower production throughput, increase of recirculating load and higher energy productioncost. There is a close relationship between the effect of β (parameter that controls the initial rise ofthe curve in fine sizes and related to agglomeration effect) and C (fraction subject to realclassification). This relation was recently published by Altun and Benzer (2014) whencomparing several high efficiency air separators operating with cement and they found thatthe cause is related to operating factors. Low values of β represents high values of realclassification (β is related proportional to the bypass of particles). Data from the plantevaluated have been related to these findings and shown in Figure 4-6. The finding during thisresearch have similar trend found by Altun and Benzer (2014). The mass of particles between 0.001 to 0.010 mm and circled red on Figure 4-5 (these fractionsare related to early setting of cement) that goes through the separator are: 81.6, 29.25 and52.35 tph for feed, reject and finished product respectively. The 29.25 tph of fraction below0.010 mm that are rejected from the air separator and sent back to the ball mill forcomminution are of especial importance, because these particles have high probability of73being overground (mainly at compartment 02) and causing agglomeration and coating effecton grinding media. Results for Beta “β” and “C” (fraction to real classification) values obtained at the plantshowed a good relationship and similar trend when compared with results from Altun andBenzer (2014) and shown on Figure 4-6. A high value of Beta “β” are related to agglomeration,high dust load, and has a decrease and the real fraction of particles due to classification “C”.The β result of 1.61607 obtained from the Whiten model and shown in Figure 4-6, is anindication of the high agglomeration effect. This agglomeration effect can be related toovergrinding and to the physicochemical properties of limestone (Altun & Benzer, 2014;Tamás, 1983).74Figure 4-6: Correlation between C, β parameters and relation to plant data(Altun & Benzer, 2014) Altun & Benzer (2014) found a relation between the bypass of particles and dust load thisrelation is shown in Figure 4-7. Bypass is calculated from the Whiten model by subtracting100 from the obtained “C” value (Altun & Benzer, 2014). High values of bypass are the resultof high dust loading or/and low air flow throw separator feed; this can be deducted from Figure4-7. The bypass value obtained from Whiten model is 11.7%. By using Altun and Benzer (2014)relationship shown on Figure 4-7, it is feasible to relate the dust loading to the separator givinga value higher than 2.6 kg/m3. Dust load should be reduced from actual 2.6 to optimally 2.075kg/actual cubic meter to the feed of separator. Three main results from reducing the dust loadon the separator feed will happen. First, improvement of separator efficiency by sending morefinished particles to product, therefore increasing the production rate. Second, the rejects ratewill be reduced on the bypass and simultaneously increasing the D80 feed size of the circulatingload to the ball mill.  Third, fresh feed will be increased and in junction with upgradedcirculating load will increase the grinding kinetics on compartment 01 and reduce the highenergy impacts and agglomeration effect on overgrinding small particles rejects. According to the air separator’s manufacturer, the equipment is designed to handle 2.0-2.5kg/m3 efficient (this is the feed density of material to the air separator), but the actual load offines and agglomeration effect on the feed is high for this specific operation.(FLSmidth, 2014).76Figure 4-7: Effect of separator dust loading on bypass and relation to plant data(Altun & Benzer, 2014) A relation between sharpness and dust load is shown in Figure 4-8, it can be seen the actuallevels (red) and manufacturer suggested levels (green), it can be deducted that at lower dustload there is a higher sharpness factor.77Figure 4-8: Relation between sharpness and dust loading(Altun & Benzer, 2014)4.2 Ball mill, Bond work index, breakage and selection functionA series of standard bond ball mill tests, breakage and selection functions were performed on thesamples in order to categorize the different properties of the cement samples. The details of theparameters of the charge of the ball mill for compartment 01 and 02 are show in Table 4-3.78Table 4-3: Ball mill grinding media charge details for C1 and C2Compartment 01 has a 22.7% grinding media load and an equivalent ball size of 47.49 mm whereasCompartment 02 has a 26.18% grinding media load and an equivalent ball size of 18.54 mm.4.2.1 Work indexThe results of the Bond work index analysis conducted with the samples of limestone, clinker anddifferent proportions of limestone, clinker and gypsum are shown on Table 4-4. The followingresults obtained are discussed below:79Table 4-4: Work indices for research samplesSampleWork Index calculated forthis research (kWh/t)Reference dataWork Index (kWh/t)(Bhatty, 2011)100 % Limestone 5.29 4.60 - 12.61100 % Clinker 11.85 9.15 - 16.1995/5   %   C/L 11.12 ---88/12 %  C/L 10.86 ---60/40 %  C/L 9.46 --- Limestone used at the plant site has a Bond work index of 5.29 kWh/t. This value categorizesthe sample close to the lower ranges of another limestone found on bibliographic references(Bhatty, 2011). This can be one reason which limestone can be responsible for high rates ofagglomeration in the air separator and also to a fast selective breakage rate on the ball mill.According Beke (1983) Limestone has “free crystal movements side by side and a great scatterof sizes” that makes limestone a highly grindable material. Clinker with 13.03 kWh/t can be characterized as material with a medium work index incomparison with other clinkers shown in Table 4-4.  By adding more proportions of softerlimestone to clinker, there is a visible reduction on Bond work index. For 5%, 12% and 40%limestone addition Bond work index is 11.12, 10.86 and 9.46 respectively. This should betaking into consideration when planning the production of PLC with higher limestone contents80than 3%. This can increase the agglomeration effect on the separator due to softer limestoneand selective grinding in the ball mill at higher limestone additions. The addition of grindingaids after improving air efficiency classifier should be required in order to improve efficiency.4.2.2 Breakage functionBreakage function is a material related property, and by definition is the distribution of sizes from a singleparticle breakage event. The breakage function values for clinker were calculated from five different sizefractions from 1.4 to 0.325 mm. Results for the size normalized breakage function are shown at Figure 4-9and the average breakage function values are also in Table 4-5 (Farzanegan, 2015).Figure 4-9: Clinker breakage function at normalized size00.10.20.30.40.50.60.70.80.910 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1Cumulative Breakage functionNormalized size1400 500 150 1000 355 average81Table 4-5: Clinker average breakage functionSize,mmNormalizedSizeAverage cumulativebreakage functioncalculated for this researchBreakagefunction from bibliography(Farzanegan, 2015)1.400 1.00 1.00 1.01.000 0.71 0.47 0.540.710 0.51 0.29 0.190.500 0.36 0.20 0.120.355 0.25 0.14 0.090.250 0.18 0.10 0.070.150 0.11 0.07 0.060.106 0.08 0.07 0.050.075 0.05 0.06 0.050.063 0.05 0.05 0.040.045 0.03 0.03 0.03Results shows that the breakage function related to the clinker evaluated is normalizable andsimilar to other clinker references found.824.2.3 Selection functionSelection function is a machine related property, represents the breakage kinetics of the system.For this research selection function has been back calculated using real plant data. Selectionfunction has been estimated for each of the two ball mill compartment using NGOTC software.Selection function or specific rate of breakage provides information on how breakage kinetics areevolving inside each compartment and how to improve them (Farzanegan, 2015).4.2.3.1 Grinding kinetics at compartments 01 and 02Compartment 01 is the coarse grinding media compartment, and it receives a mixture of fresh ballmill feed and rejects (circulating load) from separator. This mass balance represents 125 and 264.8tph respectively accounting for a total ball mill feed mass of 389.8 tph with a feed size d50 of 0.05mm. The following discussion is related to these two compartments: For a feed size to compartment 01 d80 of 0.207 mm, the top calculated grinding media sizediameter should be 10.29 mm according Allis Chalmers’s formula (FLSmidth, 2014). The actual compartment 01 equivalent grinding media diameter is 47.49 mm and as shown inTable 4-3. This larger grinding media diameter selection of 47.49 mm instead of 10.29 mm,makes a high energy impact environment, creating an early agglomeration conditionaccording Beke (1983). The specific rate of breakage (on 100% clinker) for compartment 01, and represented in solidblack line and shown on Figure 4-10 denotes a drop in grinding kinetics for sizes larger than1.0 mm. and is related to an undersized grinding media.83Figure 4-10: Selection function for compartments 01 & 02 on 100% clinker The particles smaller than 1.0 mm represent 87.8 % (342.4 tph) of the feed. The grinding mediaselected to grind these size range is not efficient for these fractions, its grinding kinetics isreduced creating an increased wear to the media and high energy impacts (this condition wasconfirmed by the processing plant personnel). In order to improve the grinding kinetics on compartment 01, there should be before the ballmill a further and efficient size reduction of the limestone and gypsum top size feed. Afteroptimizing the feed the proper grinding media selection should be calculated.84 The reason for the tendency on using high top grinding media is for reducing the high freshfeed size of fresh limestone and gypsum (on composite feed to ball mill). A mass flow of 225.76 tph corresponding to 57.8% of the comminuted mass on compartment01 is below 0.063 mm, representing a high portion of the mass with a low grinding kinetics.These fractions are transported rapidly with low reduction ratio to internal diaphragm and tocompartment 02. This effect is mainly caused by a larger top size grinding media and thesaturation of very fine particles rejected from air separator. Compartment 02 has a feed size d80 of 0.084 mm. In addition 96.11% of the feed comprisesparticles below 0.15 mm showing high percentage of the amount of small particles. Selection function for compartment 02 is shown in Figure 4-10 (represented by a solid blueline) and denotes good reduction of particles between 0.001 to 0.1 mm. (less than 1% of theparticles on sizes 0.1-1.4 mm are comminuted on compartment 02). This condition creates aperfect comminution environment for particles rejected from the separator which reduces itssize to an even smaller size efficiently. In compartment 02, particles below 0.01 mm have a high tendency for agglomeration andshown in Figure 4-5, because of overgrinding and further rejection in the grinding circuit. It was estimated that 62.92 % of the feed to compartment 02 consists of particles below 45microns, showing a huge amount of particles that should have been separated efficiently at theair separator instead of being rejected. The overgrinding in compartment 02 of particles smaller than 0.003 mm is most likely  relatedto the use of raw limestone (Ludmilla Opoczky, 1996). Particles are rejected from the separator85and is the main reason of the agglomeration effect (Tamás, 1983). An efficient air separationis required followed by the use of grinding aid. By designing the right grinding media size according to joint feed. It is expected that thespecific breakage rate of compartments 01 and 02 be increased and that the difference betweenboth rates be reduced. Another benefits from selecting the right grinding media size is thereduction of wear rate and damaging internal parts of ball mill for direct impacts betweengrinding media and ball mill surfaces.4.2.4 Savings estimationsBased on Bond formula there is possible to calculate the production increase and the use in specificenergy for the ball mill by reducing the fresh feed size using the CKP. The effect of classifier isnot modeled, just the particle size reduction. As shown in Table 4-6, three scenarios under twodifferent conditions are calculated for a 3% and 12% limestone substitution under three reductionscenario: first is (just clinker through CKP), second is (clinker + limestone through CKP) and thirdis (clinker + limestone throw CKP and gypsum trough vertical shaft impactor).86Table 4-6: Savings estimate on electricity for fresh feed size reduction at ball mill 03Savings in CAD/yr.Scenario 01 Scenario 02 Scenario 033% limestone 0 13,641.56 12,651.6912% limestone 0 31,030.74 38,805.25* Assumptions: production of 800,000 t/yr. and cost of electricity at $0.03/kWh.The calculated profit on recovering fines in air separator at different percentages from the rejectsis shown on Table 4-7.Table 4-7: Profit estimate on increase in throughput at air classifier circuit 03Recovery of fines1% 5% 10%CAD/yr. 130,305 490,560 1,177,344* Assumptions: profit of $14/t and plant availability factor 0f 85%87CHAPTER 5: MAJOR RESEARCH FINDINGS AND CONCLUSIONS5.1 Major research findingsThe main objective of this research was to evaluate the process at grinding circuit 03 of a cementproducing plant by quantifying the most related causes of inefficiency. The objective was to obtaina relationship of the operational parameters, efficient models, comparing the results withbibliographical references. This assessment provides a series of recommendations to improveefficiency of the system. The following major findings from this research are as follows: Air separator operating with low efficiency. Separator is performing inefficiently with valuesrated at 64.74, 52.24 and 45.43 % on fractions of 0.020, 0.035 and 0.045 mm respectively, thisrange of fractions are fundamental for the strength development of cement. This operationalcondition makes that a high amount of fine particles, about 100.4 tph (on fraction below 0.035mm) are being rejected and sent back to ball mill for overgrinding. This bypass from the feedto rejects condition can be confirmed by the use of at least two comparative factors that definesoptimum operating values:- Alpha sharpness value using Whiten model. According to references (Altun & Benzer,2014), this model is the most precise because provides the least square differencerelated to other methods. The value of alpha obtained by fitting the Whiten model toTromp curve for the current separator has a factor of 1.0411. An efficient separatorshould have an alpha value around 4.0 according to references available (Napier-Munnet al., 1996).88- Sharpness by Tromp curve slope. Tromp or selectivity curve is defined as theprobability of defined partitions or portions of feed that reports to the products orrejects. The Tromp sharpness obtained from the evaluated separator is 42.51%. Basedon the literature (FLSmidth, 2014; Bhatty, 2011), it is recommended values between52-58%. Factors affecting separation efficiency. Two major causes for the low separation efficiency hasbeen found: low sharpness and high agglomeration levels on the separator caused mainly by,- Overloading separator feed. It is estimated that dust load density greater than 2.6kg/actual cubic meter of air is currently being fed into the separator. According toequipment manufacturer this feed rate should be between 2.0 and 2.5 kg/actual cubicmeter of air.- The reduction of the dust load on the feed will improve on the separator’s sharpness.Altun & Benzer (2014) found that the model that represents sharpness “α” and dustload “DL” is: α = 4.2044( . ) and shown Figure 4-8, it is deducting that byreducing the dust load the sharpness is increased. Agglomeration: This high agglomeration phenomenon is the effect related to mainly twooperating conditions: the first is by the use of oversized/larger grinding media on compartment01 creating high energy impact environment inside the mill allowing agglomeration of fineparticles. This group of agglomerated particles will attach one to another behaving like onebigger particle and once inside the air classifier will be rejected as a big particle. The secondcondition is related to overgrinding the particles under high breakage rates (especially on89compartment 02). This agglomeration effect is aggravated by the use of limestone as a mineraladditive on composite cements manufacturing. Limestone with a low Bond work index of 5.29kWh/t is comminuted at higher grinding kinetics and reduced faster than clinker producing finefractions of limestone that creates a narrower uniformity factor “n” slope during the grindingstage. This condition creates over expenditure of energy and the throughput is reduced by thehigh recirculation ratio and the agglomeration effect discussed in the first condition previously. Oversized grinding media. The grinding media size design diameter used in compartment 01is larger than the optimum for the current feed. This is based on Allis Chalmers formula thatgives a 10.29 mm diameter instead of the current 47.49 mm. The effects of this oversizing arethe expenditure of energy especially on top grinding media, probably increasing the wear rateof media and explaining the fast load degrading reported from last two mill surveys (Februaryand May 2014). This oversizing is mainly defined by the fresh feed top size of raw limestoneand gypsum which are d80 of 13.7 and 61.4 mm respectively.5.2 ConclusionsThe following research is unique because of the following results obtained: The use for the very first time at this plant of different tools like: Whiten model in analyzingmore precisely the process of air separation, the use of breakage/selection function to evaluatethe grinding kinetics at the ball mill and the application of updated researches on theoptimization of finished grinding in the cement industry.90 Based on Altun and Benzer (2014) who investigated the relationship of some operatingparameters on different high efficiency classifiers and FLSmidth (air separator manufacturer)the following upgrades can be estimated using Altun & Benzer models (Altun & Benzer,2014):- Sharpness “α” can be modeled by the use of the following relation α = 4.2044 ∗ .DL is dust load (density of the feed to separator) that according manufacturer should bebetween 2.0 and 2.5 kg/m3. Modeling to get the most efficient dust loading of 2.0 kg/m3 itis obtained a sharpness factor α=1.7948- Bypass “100-C” can be modeled by the use of the relation 100 − C = 10.467 ∗ .DL is dust load (density of the feed to separator) that according manufacturer should bebetween 2.0 and 2.5 kg/m3. Modeling to get the most efficient dust loading of 2.0 kg/m3 itis obtained a sharpness C=72.04% and Bypass=27.96%- The parameters “β” and “β*” can be modeled by the use of the relationβ = −0.0422( ) + 4.0907andβ∗ = 0.9878(β) + 0.8516obtaining the following values β=1.05 and β*=1.88Obtained values for all the modelled parameters are summarized in Table 5-1.91Table 5-1: Summary of current and modeled separator parametersParameter Current plant valueModeled best valueat dust load  DL=2.0α 1.04105 1.7948β 1.61607 1.05β* 0.24271 1.88C 63.67 72.04Bypass 36.33 27.96DL >2.6 2.0CL 211 100There is an increase in efficiency on operative air separator parameters modeled byadjusting the dust load to manufacturer design parameters. In order to calculate an accurate improvement on the grinding circuit, it will be required theuse of a calibrated simulation program that integrates all the different components of the circuitand its interactions and relate it to its impact on quality and production.5.3 Recommendations for future work Circuit simulation. The use of a calibrated simulation program is highly recommended as anoptimizing tool for this operation. A simulator will integrate all the key process componentsof the circuit and will provide a low cost modeling of changing several variables on searchingfor the right configuration.92 Composite cement manufacturing cautions. When planning to increase the use of rawlimestone on composite cement manufacturing (PLC) extreme care should be taken on theoperating parameters of air separator, because the change of densities and uniformity factor ofthe mixture of the feed (limestone and clinker) may kame the separator to behave differently.The use of extra grinding aid reagent could be required in order to reduce agglomeration dueto increased addition of limestone. Further research on the agglomeration effect during grinding and its implications should bedeveloped, especially if limestone content is expected to be increased in composite cementmanufacturing.93REFERENCESAbdel-Aziz, A., Acquaye, A., Allwood, J., Ceron, J.-P., Geng, Y., Kheshgi, H., … Tanaka, K.(2014). Industry. Climate Change 2014: Mitigation of Climate Change. Contribution ofWorking Group III to the Fifth Assessment Report of the Intergovernmental Panel onClimate Change, 114.Aguero, S., & Meech, J. (2014). Decreasing energy consumption in cement production. InSchetmann Conference, Cancun Mexico (p. 25).Altun, O., & Benzer, H. (2014). Selection and mathematical modelling of high efficiency airclassifiers. Powder Technology, 264, 1–8.Asia Pacific Partnership, I. (2011). Energy Efficiency and resource savings technologies incement industry.ASTM. (2011). Standard Specification for Portland Cement.ASTM C595. (2014). Standard Specification for Blended Hydraulic Cements.Bentz, D. P. (1999). Effects of cement particle size distribution on performance properties ofPortland cement-based materials. Cement and Concrete Research, 29(10), 1663–1671.Benzer, H., Ergun, L., Lynch, A. J., Oner, M., Gunlu, A., Celik, I. B., & Aydogan, N. (2001).Modelling cement grinding circuits. Minerals Engineering, 14(11), 1469–1482.Bhatty, J. (2011). Innovations in Portland Cement Manufacturing. Illinious, USA: PortlandCement Association, 773-790.Brugan, J. (1988). High Efficiency Separators. Zkg International, 41(07), 350–355.BS. (1996). Specification for Portland limestone cement.Carbon Dioxide Emission Factors for Coal. (2015). Retrieved on December 14, 2014 fromhttp://www.eia.gov/coal/production/quarterly/co2_article/co2.htmlCelik, I. B. (2009). The effects of particle size distribution and surface area upon cement strengthdevelopment. Powder Technology, 188(3), 272–276.Cleemann, J. (1986). Evaluation of the new high efficiency separators. ZKG International, 295–304.94CSA. (2013). CAN/CSA-A3000-13 - Cementitious materials compendium.Drosdiak, J. . (2013). MINE 331: Introduction to comminution and size classification. 2013.Duda, W.H., (1976). Cement Data Book. London, UK: Macdonald & Evans London, 130-258.Ellerbrock, H. . (1985). Particle size distribution and properties of cement, Part I: strength ofportland cements. Zkg International, (6), 136–145.EN. (2011). European Standards.Eng-Fr Sheets.qxd - CSA3000E.pdf. (n.d.). Retrieved January 19, 2015, fromhttp://cement.org/tech/pdfs/CSA3000E.pdfFarzanegan, A. (2015). BFDS.Farzanegan, A., Ghasemi, E., Valian, A., & Hasanzadeh, V. (2014). Simulation of clinkergrinding circuits of cement plant based on process models calibrated using GA searchmethod.FLSmidth. (2014). FLSmidth Operators manual.Gupta, A., & Yan, D. (2006). Introduction to Mineral Processing and Operation. Amsterdasm,Netherlands: Elsevier, 250-347.Harder, J. (2003). Advanced Grinding in the Cement Industry. Zkg International.IEA. (2009). Cement Technology Roadmap 2009.Kline, J., & Kline, C. (2014). Cement and CO 2 : What ’ s Happening, 1–9.Moly-Cop. (2012). Moly-Cop tools version 3.0.Napier-Munn, T. J., Morrell, S., Morrison, R., & Kojovic, T. (1996). Mineral comminutioncircuits their operation and optimization JKMRC.Nisbet, M. (1996). The reduction of resource input and emissions achieved by the addition oflimestone to Portland cement.Nisbet, M. A. (1996). Information The Reduction of Resource Input and Emissions Achieved byAddition of Limestone to Portland Cement. Portland Cement Association, 9781(847), 0–10.95Nokken et al. (2007). Portland-Limestone Cement : Stat e -of-the-Art Report and Gap AnalysisFor CSA A 300 0, 0–59.Opoczky, L. (1986). Grinding technology for producing high-strength cement of high slagcontent. Powder Technology, 48(1), 91–98.Opoczky, L. (1996). Grinding technical questions of producing composite cement. InternationalJournal of Mineral Processing, 44-45, 395–404.Ramezanianpour, A. a., Ghiasvand, E., Nickseresht, I., Mahdikhani, M., & Moodi, F. (2009).Influence of various amounts of limestone powder on performance of Portland limestonecement concretes. Cement and Concrete Composites, 31(10), 715–720.Rosemann, H., & Ellerbrock, H. . (1998). tecnica de molienda de cemento. Zkg International, 51.S.L, O. consulting. (2014). Market trends in vertical mills for the cement industry. ZkgInternational, 42–52.Schiller, B., & Ellerbrock, H.-G. (1992). Grinding and properties of cements with severalprincipal constituents. Zkg International, Edition B.Schneider, M., Romer, M., Tschudin, M., & Bolio, H. (2011). Sustainable cement production—present and future. Special Issue: 13th International Congress on the Chemistry of Cement,41(7), 642–650.Seebach. (1996). State of the Art of Energy Efficient Grinding Systems.Sohoni, S., Sridhar, R., & Mandal, G. (1991). The effect of grinding aids on the fine grinding oflimestone, quartz and Portland cement clinker. Powder Technology, 67, 277–286.Sumner, M. S. (1989). The influence of a narrow particle size distribution on cement paste andconcrete water demand.Tamás, F. D. (1983, March). The process of fine grinding. Cement and Concrete Research.Tsivilis, S., Voglis, N., & Photou, J. (1999). Study on the intergrinding of clinker and limestone.Minerals Engineering, 12(7), 837–840.96APPENDICESAppendix A: Standard Bond work index calculationThe purpose of the standard Bond ball mill grindability test is to determine the Bond Ball millWork Index (BWI), which can be compared with the work indices of known materials to evaluategrinding efficiency or mill design.The Bond work index is a measurement of the power required to reduce feed with a given 80percent passing size (d80) to product with a specified 80 percent passing size (d80).This procedure assumes a standard 100 mesh (150 µm) closing screen size. The Bond ball mill iscomposed of a steel shell with internal dimensions of 12-inch (30 cm) diameter x 11-inch (28 cm)length.  The shell has rounded corners, a smooth liner and no lifters. The feed hatch consists of aremovable cover plate on the curved surface of the mill.  The mill is set to operate at 70 rpm.  Theball charge consists of 20.125 kg of steel balls ranging from approximately 37 cm to 15 cm indiameter.Tolerance Ball Size Number Weight, ginch mm of Balls Total Avg. BallMax 1.50 38.1Avg 1.45 36.8 43 8,809 204.9Min 1.25 31.8Max 1.25 31.8Avg 1.17 29.7 67 7,215 107.7Min 1.06 26.5Max 1.06 26.5Avg 1.00 25.4 10 670 67.0Min 0.88 22.4Max 0.88 22.4Avg 0.75 19.1 71 2,003 28.2Min 0.63 16.0Max 0.63 16.0Avg 0.61 15.5 94 1,428 15.2Min 0.53 13.2Total 285 20,12597A.1 ProcedureThe procedure for a Bond ball mill grindability test depends on the following variables. Thevariables specified in the work order instructions for each test are as follows:• Closing screen size:  default is 100 mesh (150 µm)• Cycle 1 revolutions:  selected based on known sample hardness; default is 100 cycles butit typically corresponds to the closing screen size (e.g. 100 mesh corresponds to 100 cycles).A.2 Sample preparationPrepare 10 kg of sample to have 100% passing a 6 mesh screen. Split into twelve different feedsand store each split.A.3 Particle size analysis of the feedWeigh one “Feed” charge and record the weight on the manual worksheet. Screens to be usedaccording based on the closing screen size as specified in the test work order. Using the selectedscreen stack, add the sample and shake using the dry Ro-Tap® machine for 15 minutes. Weigheach size fraction on the manual worksheet. Recombine all fractions, bag, label (Feed PSA Reject)and set aside. Rejects may be used as supplementary feed for the grind test if needed.98The following conditions must be accomplished on the first stage before continuing to thefollowing stage:i. The feed d80 should be between 2,200 µm and 2,500 µm.ii. The feed size analysis should show less than 20% passing the closing screen size.A.4 Feed bulk densityTake two of the Feed charges and transfer the material to a 1000 mL graduated cylinder. Vibratethe sample for 10 minutes on the Vibro-Pad in order to minimize air pockets as shown on. Recordthe final volume level and the actual sample weight on the manual worksheet. Transfer and enterdata values into the computer spreadsheet. Riffle out the weight calculated by the computerspreadsheet which should be equivalent to 700 mL.  This is the feed for Cycle 1.  Bag the remainingsample, label (Feed Reject) and set aside (Gupta & Yan, 2006).99A.5 Performing the grinding testEnsure that the inside of the mill is clean (i.e. no foreign material) and follow the instructions:1. Place the ball charge in the mill according2. Place the equivalent weight of 700 mL of the “feed charges” samples (stored on bags) in themill according the density calculation.3. Secure the full cover plate with the two wing nut clamps.4. Grind the ore for 100 revolutions.5. At the end of the grinding during the 100 revolutions, replace the full cover.6. Making sure that the collection pan is in place, discharge the ball mill as shown7. Screen the product at the required closing screen size (as specified in the work order; default100 mesh).8. Ro-Tap® using the closing screen for 15 minutes.9. Collect the undersize material from the screens and set aside in a labelled ‘PRODUCT’ inplastic bag.10. Collect the oversize material from the three screens, combine and weigh in a metal pan.Record the weight on the manual worksheet as “Oversize #1”.11. Re-screen the oversize material for another 15 minutes using multiple screen-pan sets ifnecessary.10012. Collect the oversize material from the second stage of screening, weigh and record as“Oversize #2” on the manual worksheet. Enter this value on computer spreadsheet.13. The spreadsheet will forecast the number of revolutions and the mass of new feed to add tothe oversize material reserved in Step 12 for the next cycle.  Record these two set points onthe manual worksheet.14. Add the mass of new feed to the reserved oversize material.  Verify that the actual combinedmass is equal to the original mass equivalent to 700 mL.15. Place the combined new/oversize material in the mill and run for the determined number ofrevolutions.16. Repeat Steps 2 through 15 until all conditions for closure have been met, as described in thefollowing section 3.2.6.101A.6 Conditions for closureAll of the following conditions must be met for closure.• Minimum of seven cycles• A reversal in the Net Product per Revolution should be reported (calculated by thespreadsheet; grams of undersize product per mill revolution). A reversal is a trend in the NetProduct per Rev of either up/down/up (eg. 1.91 – 1.88 – 1.92) or down/up/down (eg. 1.81 – 1.85– 1.82).• Less than 3% difference between the highest and lowest values of Net Product perRevolution in the last three cycles.  This assures a 250% circulating load.  The following formulais used to calculate the difference:• Circulating load between 245 to 255% (250 +/- 5%).A.7 Particle size analysis of the productCombine the undersize products from the last three cycles. Blend the combined undersize productand then riffle-split a sub-sample of approximately 200 g.  Record this as the total product weighton the manual worksheet. Select the screens to be used according to based on the closing screensize as specified in the test work order. Wet screen at 400 mesh and dry the +400 mesh product in a clean pan in the over.  Discardthe -400 mesh material. Using the selected screen stack, add the dried +400 mesh product and shake using the Ro-Tap® machine for 20 minutes. Weigh and record the weight of each size fraction on the manual worksheet.Lowest ValueHighest ValueDifference = 1 -  x 100102 Transfer the feed and fraction weights from the manual worksheet to the spreadsheet created.A.8 Bond test grindability calculationsThe following calculations are automatically conducted by the spreadsheet.Ore Feed Density:weightfeedActualvolumeActualmLdensityfeedOre  700Target Recirculation Load WeightThe target recirculation load weight, also known as the Ideal Potential Product (IPP), correspondsto the target product weight to achieve a circulating load of 250%.5.3700 mLFeedWeightofproductpotentialIdeal Where 3.5 factor corresponds to 1 part (100%) target product and 2.5 parts (250%) circulatingload.Bond Ball Mill Work IndexThe BWI calculation is derived from F. C. Bond’s Third Theory of Comminution.103.110105.44808082.023.01 FPGprPBWiWhere:BWI = Bond Ball mill Work Index number in kWh/tP1 = Aperture of the closing screen size in micronsGpr =Average grams of undersize product per revolution from the last three cycles103P80 = Size at which 80 percent of the undersize product passes, in micronsF80 = Size at which 80 percent of the feed passes, in microns104Appendix B: Work indicesB.1 Clinker 100%105B.2 Limestone106B.3 Clinker 95%/Limestone 5%107B.4 Clinker 88%/Limestone 12%108B.5 Clinker 60%/Limestone 40%109Appendix C: Particle size distributionC.1 Fresh feed to ball mill110C.2 Air separator streams (data provided by plant)FINES (F) REJECT (R) FEED (A)Equivalent Corrected Corrected CorrectedSize µm % Passing % Passing % Passing150 99.70 97.08 97.87125 99.70 96.91 97.11100 99.70 94.31 95.4195 99.70 93.19 94.8390 99.70 91.82 94.1285 99.70 90.10 93.2480 99.70 87.98 92.1575 99.69 85.39 90.8070 99.63 82.22 89.1365 99.49 78.36 87.0560 98.98 73.63 84.4955 98.02 67.88 81.3450 96.62 61.10 77.5245 94.62 53.73 72.9440 91.75 46.00 67.5035 87.78 37.93 61.2730 82.38 29.98 54.5625 75.28 22.78 47.3920 66.24 17.06 39.9115 55.29 13.24 32.3910 41.81 11.05 24.779 38.66 10.69 23.138 35.30 10.29 21.377 31.67 9.63 19.456 27.68 8.74 17.325 23.24 7.67 14.904 18.30 6.37 12.133 12.83 4.85 9.022 7.15 3.23 5.711 2.11 1.64 2.63111C.3 Fresh feed and circulating load feed to ball mill112Appendix D: Bond equation for modeling throughput and savings113Appendix E: Specific rate of breakage

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0166246/manifest

Comment

Related Items