challenges in the production of mineral products— qemscan ... · dam. hmc from hillendale mine is...

CHALLENGES IN THE PRODUCTION OF MINERAL PRODUCTS 147

IntroductionThe Hillendale mine is located close to Richards Bay nextto the town of Esikhaweni. The orebody is varied in termsof mineral assemblage and thus also processing behaviour.As the mine moves into the final stages of its life, theability to blend out problematic assemblages through mineplanning becomes limited.

Heavy minerals are mined at Hillendale mine bymonitoring the high slimes ore which flows down trenchesto pumping stations. The ore is beneficiated in a gravityseparation circuit to produce heavy mineral concentrate(HMC) with sand tails being returned to reconstruct themined dunes with the slimes fraction disposed in a slimesdam. HMC from Hillendale mine is further beneficiated atthe minerals separation plant using a wet high intensitymagnetic separation circuit (WHIMS) to produce crudeilmenite. The crude ilmenite is dried and separatedmagnetically in a circuit using rare earth drum magneticseparators to produce a smelter feed quality ilmenite. Whenprocessing the crude ilmenite in the dry magnetic separationplant the main objective is to remove chromite. This is doneby rejecting a high magnetic susceptibility as well as a lowmagnetic susceptibility fraction to waste. Apart from thechromite, some other contaminant minerals are also rejectedwith the low susceptibility fraction in order to achieve thetarget chemical quality of the smelter feed.

The nonmagnetic fraction from the WHIMS circuit issubjected to gravity separation, electrostatic separation, hotacid leach and acid attritioning as well as high force

magnetic separation to produce final zircon and rutileproducts.

This paper reviews mineral assemblages found to giveseparation problems in the ilmenite processing plant as wellas in the dry mill’s rutile production circuits. The use ofQEMSCAN® analysis to identify the mineralogical driversleading to reduced plant performance is shown.

Ilmenite processing impact observedFor the purpose of this paper two periods of reducedilmenite processing performance will be discussed.Performance of the ilmenite processing plant can be definedin terms of the ilmenite mineral recovery as well as the finalsmelter grade produced. These two concepts are related,with the production of a higher grade ilmenite leading to alower recovery. In the case of smelter feed ilmenite there is,however, more than one critical contaminant dictatingpossible recovery. In the two distinct periods with lowrecovery it was first the Cr2O3 specification that constrainedrecovery and for a second period the constraint shifted toSiO2. Figure 1 shows the Cr2O3 and SiO2 trends for theilmenite smelter feed and illustrates how the plant controlshifted from Cr2O3 to SiO2 quality constraints.

From Figure 1 it can be seen how the Cr2O3 levels dropduring high periods. This is due to attempts by operators toimprove the SiO2 quality by rejecting more material to thewaste, in the process rejecting more Cr2O3. Metallurgicalinvestigations done on material from the first low recoveryperiod did not yield conclusive explanations for the reducedrecovery.

BEUKES, J.A., JANSE VAN RENSBURG, H., and RAMSAYWOK, P. Challenges in the production of mineral products—QEMSCAN® analysis as an aideto metallurgical troubleshooting. The 7th International Heavy Minerals Conference ‘What next’, The Southern African Institute of Mining and Metallurgy,2009.

Challenges in the production of mineral products—QEMSCAN® analysis as an aide to metallurgical

troubleshooting

J.A. BEUKES*, H. JANSE VAN RENSBURG*, and P. RAMSAYWOK†

*Exxaro KZN Sands†Exxaro Research and Development

For a process plant operator the question of ‘what next’ in terms of feed characteristics willalways be at the top of the list. Changes in feed material to a process plant happens as a rule and isgenerally worse towards the end of a mine’s life, when miners are looking for every possibletonne of ore. Quantitative evaluation of material by scanning electron microscopy (QEMSCAN®)has been a valuable aide in metallurgical troubleshooting. At Exxaro KZN Sands ilmenite is fed tosmelters for producing high titania slag and low manganese pig iron. The quality of the slag andmetal produced is largely dependent on the quality of the raw materials used. It was found thatilmenite quality as well as recoveries deteriorated as mining progressed into new areas. Qualityand recovery of rutile at the dry mill was affected even more negatively. An investigation wasconducted to find the root cause of the quality and recovery issues as well as to evaluate possibleoptions for improving the process performance. The first step was to go back to basics and ensureevery piece of equipment was operating normally. Secondly, laboratory as well as pilot-scale testwork supported by QEMSCAN® analysis was conducted to understand the mineralogical drivers.Following the identification of the mineralogical issues associated with the new mining area,processing options to improve quality as well as recovery, were evaluated.

HEAVY MINERALS 2009148

QEMSCAN® analysis used to identifymineralogical issues

QEMSCAN® is a SEM-based technology in which anelectron beam generates an energy dispersive X-rayspectrum which is detected and compared against areference library of mineral compositions and mixtures.Following the establishment of the QEMSCAN® capabilityin EXXARO, several samples were analysed to aid in theinvestigation into reduced ilmenite recovery. To illustratethe investigations, aided by QEMSCAN® analysis, the twomaterial types found to be problematic will be discussedseparately.

First material type with a quality constraintLaboratory tests were conducted on the material to find thecause of the performance issues on the plant. Samples ofthe problematic crude ilmenite as well as normal crudeilmenite were fractionated using a lift type Carpco magneticseparator and the fractions produced analysed on theQEMSCAN®. Figure 2 shows the Carpco fractionationresult for the two samples (normal and type 1 respectively).

From Figure 2 it can be seen that ‘type 1’ material hadprogressively lower mass pull to the magnetic fractionwhen compared to‘normal’ ilmenite. This would imply thatthe type 1 material has a slightly lower magneticsusceptibility. Figure 3 shows the TiO2 content per particlefrequency distributions for the two samples, normalilmenite and type 1, based on the QEMSCAN® analysis.

From the plots in Figure 3 the normal crude ilmenite hasa chemical distribution lower in TiO2 content compared tothat of type 1 material. The increase in type 1 material’sTiO2 content can explain the lower magnetic susceptibilitycompared to the normal crude ilmenite as seen in Figure 2.The presence of more ilmenite particles with high TiO2levels would be indicative of chemical weathering.

Figure 4 shows the distribution of Cr2O3 containingminerals in the different low susceptibility magneticfractions produced on the Carpco magnetic separator. Thelow susceptibility fraction is shown as it is normally theprocess stream where the most ilmenite is lost in theprocess plant. Figure 5 shows the ilmenite mineraldistribution for the same magnetic fractions as shown inFigure 4.

From Figure 4 it can be seen that a higher totalpercentage of the chromite resides in the low susceptibilityfraction for the type 1 material, compared to normal crude

ilmenite. From Figure 5 it can be seen that more of theilmenite in type 1 material reports to the low susceptibilityfractions when compared with the normal ilmenite. Type 1material thus yields lower ilmenite product recovery due toa greater extent of chromite and ilmenite mineral overlap interms of magnetic susceptibility compared to the normalcrude ilmenite processed.

The QEMSCAN® added value to the investigation byenabling the identification of ilmenite particles of elevatedTiO2 content. The QEMSCAN® also allowed quantification

Figure 1. Product quality trend showing %SiO2 and %Cr2O3 inthe ilmenite smelter feed product

Figure 2. Cumulative mass distribution as a function of theampere setting

Figure 3. TiO2 per particle analysed by the QEMSCAN frequencydistributions

Figure 4—Distribution of chromite to each of the Carpcomagnetic fractions that normally represent the low susceptibility

material


of different types of chromite minerals present in lowquantities in the material. The QEMSCAN® informationcombined with metallurgical test work assisted inunderstanding of the mineralogical reasons behind changesexperienced in the processing plants.

Second material type with a quality constraintAs shown in Figure 1, a second period was entered wherethe plant production constraint moved from Cr2O3 to therequired SiO2 specification. To investigate the high SiO2 inthe ilmenite, plant samples of the crude ilmenite as well asthe final ilmenite produced were analysed using theQEMSCAN®. Separation tests were also done to evaluateseparation characteristics of the material (type 2) and toallow comparison with normal crude ilmenite. Theseseparation tests were conducted using rare earth drum typemagnets similar to the equipment used in the actualmagnetic separation processing plant. For the test a multiplesplitter box was used in place of the conventional splitterarrangement. The multiple splitter box allowed severalfractions to be obtained from the mineral fan produced asmaterial passed over the drum magnet.

Figure 6 represents the QEMSCAN® data showing theelemental deportment of Si to the minerals present in themagnet feed and product. It can be seen that the majority ofthe Si in the feed is present as garnet minerals with coatingsand leucoxene (TiAlSi-intergrowth) also being significantcontributing minerals.

When the Si deportment in the product is reviewed, it canbe seen that the garnet mineral contribution to Si in thesample is minor with the siliceous FeAl coatings andleucoxene (TiAlSi-intergrowth) being the majorcontributors. It can thus be concluded that the garnetminerals are efficiently rejected in the magnetic separationprocess while the coatings and leucoxene intergrowths arenot rejected to the same extent. Table I gives the averageparticle sizes for selected SiO2 minerals in the crudeilmenite as well as the ilmenite product.

From Table I it is clear that the garnet in the feed is largein relation to the ilmenite particles which represent the bulkof the particles in the sample. The difference in average sizeof the particles explains the effective rejection on the largediameter rare earth drum magnets where centrifugal forceswould preferentially reject larger particles to thenonmagnetic fraction. Only fine grained garnet minerals

end up in the final ilmenite product. Other Si containingminerals are very small and if liberated would haveprobably been removed in the desliming stages of primaryprocessing. Figure 7 shows a particle view generated basedon the analyses from the QEMSCAN®. The particles wereordered as to focus on ilmenite particles containingsiliceous FeAl coatings and leucoxene. To assist visualinterpretation, the ilmenite phase was shaded out with thesiliceous FeAl coatings and leucoxene phase represented bya darker colour. From Figure 6 these two minerals wereidentified as the main contributors to the high SiO2 in thefinal ilmenite sample.

Figure 5. Distribution of ilmenite to each of the Carpco magneticfractions normally representing the low susceptibility material

Figure 6—Si deportment to minerals in the feed as well asilmenite product streams

Mineral Feed average particle Product average particlesize micron size micron

Ilmenite 80.0 86.0Garnets 95.9 18.0Siliceous AlFe coatings 12.3 11.7Leucoxene 9.0 8.3

Figure 7—QEMSCAN® particle map of the type 2 ilmeniteproduct sample showing ilmenite particles with surface coatings

and intergrowths of leucoxene and siliceous FeAl

Table IAverage mineral phase particle size as determined by

QEMSCAN® analysis


From Figure 7 the textural association between theSiliceous FeAl coatings, leucoxene and the ilmenite phasecan be seen. It is clear that the contaminants are attached tothe ilmenite grains and normal beneficiation separationtechniques will be inefficient. Figure 8 shows thecumulative SiO2 in the magnetic fraction as a function ofthe mass % to the magnetic fraction for normal as well astype 2 crude ilmenite based on the pilot rare earth drum testwork.

Figure 8 shows that the removal of SiO2 is efficient fortype 2 materials at low susceptibility fractions where themajor quantities of garnet and other SiO2 containingdiscrete mineral particles are rejected. This is evident fromthe slope of the curve at mass recoveries to the magneticfraction of above 75%. To clean the ilmenite further afterthe majority of liberated minerals are rejected is inefficientas the slope of the curve for values below 75% indicates. Ifthe SiO2 content needs to be reduced to levels of around0.9%, which is required for good quality smelter feed,significant loss of product mass is experienced. For thenormal crude ilmenite these is not a constraint with the feedsample’s SiO2 levels already within acceptable limits.

Processing options availableFor type 2 material the physical removal of coatings byattritioning was tested on a laboratory scale. Following theattritioning the material was processed using a similar testset-up on the rare earth drum magnets as done in previoustest work. Figure 9 compares the separation performance ofthe attritioned and unattritioned material to illustrate theimprovement achieved.

From Figure 9 it can be seen that by attritioning animprovement was achieved, increasing the mass yield fromaround 41% to 56%. Implementing this solution on theplant would, however, require significant capital andoperating cost. From geological information available itwas found that the section of the orebody prone to heavycoatings was not extensive enough to justify capitalexpenditure. Blending can also solve the SiO2 quality issueand allow higher yields by moving the quality constraintfrom SiO2 to Cr2O3. Even though there may not be scope inthe mining operation to blend different ore types, stockpiledcrude ilmenite available can be used for this purpose.Figure 10 illustrates the impact of blending on the massrecovered to the magnetic product.

From Figure 10 it can be seen that by blending the twomaterial types as feed to the processing circuit massrecovery of up to 73% can be obtained, compared to theaverage mass recoveries if the two material types wouldhave been separately processed of 58%. Due to theefficiency of the blending solution and the low cost, it wasimplemented as the solution to resolving the reducedrecoveries due to high SiO2.

Processing impact on rutile

Description of the dry mill of the mineral separationplantAs described earlier, the heavy minerals concentrate firstenters the feed preparation circuit where oversize material(larger than 0.85 mm) and magnetite are removed. Ilmeniteis produced by magnetic separation in the WHIMS circuitas described. The WHIMS nonmagnetic fraction is cleanedto reject quartz, some leucoxene and most of the aluminiumsilicate minerals using gravity separation. Figure 11 gives asummarized material flow for the dry mill plant.

The cleaned nonmagnetic fraction is dried and fed to theprimary electrostatic separation circuit for a primary split ofconducting minerals and non-conducting minerals.

The zircon rich, nonconducting stream is leached withsulphuric acid in a kiln reactor to remove iron staining fromthe surface of the zircon particle. After the acid is washedoff the material it is further cleaned by rejecting any

Figure 8—Ilmenite product grade as a function of cumulativemass recovery to the magnetic fraction. The dashed line

represents the allowable SiO2 level in the ilmenite product

Figure 9—Ilmenite product grade as a function of cumulativemass recovery to the magnetic fraction for attritioned as well as

unattritioned material

Figure 10—Ilmenite product grade as a function of cumulativemass recovery to the magnetic fraction as determined from a

pilot rare earth drum test


remaining low density contaminants such as quartz,leucoxene and aluminium silicate minerals. The upgradedzircon rich stream enters the final stage of electrostaticseparation where the final zircon product is produced. Theremainder is returned to the primary electrostatic circuit asa circulating load.

The rutile rich stream (from the primary electrostaticcircuit) is subjected to an acid attrition circuit. The purposeis to remove staining on particles in order to prepare themfor the final stage of electrostatic separation where finalrutile product is produced. The remainder is returned backto the primary electrostatic circuit as a circulating load.

The impact experienced

Two distinct periods were also experienced. The first periodwas characterized by heavy iron staining on the zirconparticles. This dropped the efficiency of separation in theprimary electrostatic circuit and led to lower quality feedgoing to the rutile and zircon circuits. Since no hot acidleaching is done on the rutile rich stream (only cold acidattritioning), the rutile circuit had great difficulty inproducing rutile to specification.

During the second period, however, iron staining was nota big problem. The rutile circuit had difficulty in producingrutile within the SiO2 specification, even when the zirconcontent was dropped to as low as 0.5%. The effect on rutileproduction is shown in Figure 12 where typical productionwas given a baseline value of 100% and the reduction inproduction shown relative to this baseline.

A secondary effect of low separating efficiencies is thatcirculating loads in the plant increase. As the rutile circuitwas affected most, circulation of especially rutile andleucoxene increased, resulting in higher feed rates to therutile circuit. Higher feed to the rutile circuit placedadditional load on the rutile attritioner, leading to reducedretention times and thus reduced attritioning efficiency. Thefirst reaction to this sudden drop in production was to dothe basic circuit checks and to do thorough plantinspections. Machines were surveyed and all machine andprocess parameters were optimized to ensure optimal

separation. Different circuit control philosophies weretested but no real improvements could be made.

Feed characterizationThe reason for the significant drop in production was notfully understood. Apart from some stained zircon, theoptical microscope could not provide the necessaryexplanation. Samples analysed by QEMSCAN®, however,gave some useful explanations. From the QEMSCAN®

analysis high levels of mineral coating and attachmentscould be identified. Since the electrostatic separationprocesses utilizes mineral particle surface conductivitydifferences as basis for separation coatings may severelyaffect the separation efficiency. This is shown on Figure 13on a few selected rutile particles.

Possible solutionsTo maintain production levels when treating theproblematic feed material, the processing plant efficiencyhad to be improved to compensate for the deterioratedmineral properties. The following two considerationsdirected the strategy followed:

Figure 11—General flow circuit of the dry mill in the mineralseparation plant

Figure 12—Effect of the deteriorated feed quality on daily rutileproduction tonnages

Figure 13—QEMSCAN® images of selected rutile particlesillustrating reasons for separation difficulties


• QEMSCAN® images revealed the existence of rutileparticles coated and intergrown by mainly leucoxeneand AlFe-silicates. Since the residence time in the rutileattritioner is not optimal, installing a bigger attritionerwas one obvious improvement to consider.

• Higher efficiency high tension roll (HTR) machinesexisted but their potential impact was unknown.

To test the effect of attritioning time on subsequentelectrostatic separation, a sample of the rutile rich streamfrom the primary electrostatic circuit was split into fiveequal fractions. Each fraction was attritioned in a laboratoryscale attritioner at different durations. The attritionedsamples were compared by subjecting them to the sameelectrostatic and magnetic separation circuit on laboratory-scale equipment. The circuit consisted of a single pass ofelectrostatic separation, followed by a single pass on a rare-earth roll separator and a final single pass of electrostaticseparation. Figure 14 below summarizes the resultsobtained from the test work. A plant attritioned sample wasalso taken and subjected to the laboratory electrostatic testcircuit. This was done to compare the plant attritioner to thelaboratory-scale attritioner and the data point for thissample is indicated on Figure 14 by a square.

The performance of an alternative HTR separator wascompared to existing HTR separators. A laboratory-scaleCoronaStat unit was used. It was tested against a full-scaleCoronaStat and the difference in separation efficiency waswithin 2%. The performance of the laboratory-scaleCoronaStat was then compared to that of a full-scale unit ofthe older HTR separator. The CoronaStat efficiency wasfound to be about 15–20% (absolute) higher.

A secondary benefit that could be obtained by replacingthe older HTR separators in the circuit with CoronaStatseparators was that the flow of the returns to the primaryelectrostatic circuit could be reduced. Reduced loading onthe primary separation circuit would result in improvedseparation efficiency. A third benefit was that the reducedcirculating loads throughout the plant would result in alower feed rate to the rutile attritioning circuit, thusincreasing residence time. Test work showed that byinstalling CoronStat separators the rutile yield wouldincrease by 14.5%. Since the installation of CoronaStatHTR units promised higher efficiencies, reduced circulatingloads and longer attritioning times, it was decided to onlyinstall the improved separators before any upgrade to theattritioner was considered. The impact of the lowerrecycling loads on the average retention time in theattritioner unit is indicated on Figure 14.

Post implementation results

Figure 15 illustrate the improvements in rutile productionachieved following the implementation of the plantimprovements.

During the onset of separation problems, rutileproduction dropped to an average of 38% below baselineproduction as seen on Figure 15. Following thecommissioning of the CoronaStat units, rutile productionwent almost back to baseline production. The returnsstream from the rutile circuit to the primary electrostaticcircuit reduced by about two-thirds. This significantlyreduced the loads in that circuit, presenting the opportunityto make a few process flow changes. After these changes,

rutile production made another step change to 19% abovebaseline production. This step change was due to improvedseparation efficiencies in the primary circuit, resulting inbetter feed quality to the attritioning circuit. Following theimprovements in separation, the feed rate to the attritioningcircuit reduced by 25% which caused the residence time inthe attritioner to increase.

Conclusions

This paper showed how QEMSCAN® technology can beused as an input to process troubleshooting. By studying themineralogy and getting quantitative data on mineralogicalproperties of samples, root causes for observed processingbehaviour can be determined. The QEMSCAN® can giveinformation about mineral associations, elementaldeportment to different minerals in a sample, mineral sizinginformation, and mineral assemblage of a sample. Theseanalyses are over and above the visual mapping of particlesin a sample. This paper presented some case studies toillustrate the value of the technique to augment traditionalmetallurgical test work in the problem solving process.

It is concluded that the QEMSCAN® as an analytictechnique can add value to minerals processing operationsas was illustrated by applications on ilmenite as well asrutile processing investigations.

Figure 14—Effect of attritioning time on electrostatic separationshowing recoveries at varying attritioning times for laboratory

test work

Figure 15—Improvement in rutile production on the problematicfeed material after firstly installing the CoronaStat machines and

secondly optimizing the circuit


Hendrik Janse van RensburgSenior Plant Metallurgist, Exxaro KZN Sands

1994–1996 Graduate in training at Durnacol coal beneficiation plant.1996–1998 Plant metallurgist at Durnacol coal beneficiation plant.1998–2000 SAP R/3 QM programming2000–2003 Plant metallurgist at Iscor, Newcastle bar- & rod mill2003–2004 Plant metallurgist at Exxaro KZN Sands, Hillendale mine.2004–2006 Senior plant metallurgist at Exxaro KZN Sands, Hillendale mine.2006–2009 Senior plant metallurgist at Exxaro KZN Sands, mineral separation plant.

challenges in the production of mineral products— qemscan ... · dam. hmc from hillendale mine is...

Documents