MASTER'S THESIS. Replacement of Dichromate with Dextrin during Copper-Lead Separation

Transcription

1 MASTER'S THESIS 2010:130 CIV Replacement of Dichromate with Dextrin during Copper-Lead Separation Malin Hagemalm Luleå University of Technology MSc Programmes in Engineering Chemical Engineering Department of Chemical Engineering and Geosciences Division of Mineral Processing 2010:130 CIV - ISSN: ISRN: LTU-EX--10/130--SE

2 1 Abstract At the Garpenberg Concentrator the separation of copper and lead minerals is performed with flotation. The method flotation uses the minerals surface properties to separate different minerals. By forming slurry, stirring it and blowing air into the slurry a separation is possible. The copper-minerals who are more hydrophobic to their nature will follow the bubbles to the surface of the slurry and form froth. The froth is removed and with it the copper-minerals. To further the separation, different reagents like collectors or depressants can be added. Collectors render the surface of the mineral more hydrophobic and more inclined to attach to the bubbles. Dichromate is a depressant with the mission to depress the lead in the flotation. The dichromate attaches to the lead-minerals and render them more hydrophilic. The more hydrophilic the lead-minerals are the better the separation. Dichromate performs excellent in the separation but it is a highly toxic substance that is also carcinogenic. Needless to say a safer and less toxic depressant would be preferred. The object of this master thesis was to investigate the depressant dextrin effect on the separation and if it would be possible to replace dichromate with dextrin. The dextrin investigation started with designing of experiments with the help of the program MODDE. The designed experiment was executed in the proper order and the result from the experiment was analyzed with both MODDE and Excel. A reference experiment with dichromate was also performed in order to be able to compare the results from dextrin experiments. The results from the investigation compared with the reference experiment gave indications that dextrin can replace dichromate at least in the case of batch laboratory scale flotation. The parameters that affected the dextrin s performance the most were the ph and the amount of dextrin added. The most promising results came when the dosage of dextrin was kept low, around 40 g/ton of solid and the ph was kept around 10 and a weak collector was employed. 2

3 Preface This master thesis report was written for Boliden Mineral at their office in Boliden. I would like to thank my supervisors Johan Hansson at Boliden for giving me a long leash enabling me to think outside the box, and Bertil Pålsson at Luleå Tekniska Universitet for keeping me on a short leash and making sure that I don t stray too far from the box. I would also like to thank everyone at TMP for putting up with my endless questions and for always helping me out: special thanks go to Micke Mus with his expertise knowledge of Bettan, to Amang who analyzed all my samples and to Rolf who made the experiments possible, and to Lizzie for her mean right hook. My love and care goes out to my family who has supported me through all of this, my mother with lots of food and a good listening ear. My father who read my entire report twice! My sister who always helps me even though she usually doesn t want to ;). To Robert, who always seem to be able to cheer me up and make me feel like superwoman. This would not have been possible without you! My last thanks goes to my partner in crime: Hassna Aitahmed-Ali. Malin Hagemalm Boliden Figure 2. The TMP- crew: From the back, Anders Rundström, Mikael Widman, Roger Lundström, Mikael Eriksson. From the front, Jenny Hagemalm, Carina Andersson and Amang Saleh. Behind the camera and therefore missing in this picture is Rolf Danielsson. 3

4 Table of Content 1 Abstract Introduction Background Theory Comminution Flotation basics Literature review Objective Material and Equipment Depressants Di-chromate Dextrin Collectors Dithiophosphate Danafloat Xanthate KAX Regulators NaOH Frother Nasfroth Other Equipment Analysis Method Preparation of flotation Flotation Sample treatment COD and Dextrin COD Dextrin Results Determining the parameters The DOE method The Model COD and Dextrin analysis Discussion Conclusion Future work Reference Books Personal communications Reports Internet Appendix 1: Flotation Theory Appendix 2: Experiments Appendix 3: Analysis and Calculations Appendix 4: the Model

5 15.1 Creating a worksheet Objective in MODDE, Model and Design Method of Fit Goodness of Fit How good is the Model COD and Dextrin analysis Appendix 5: Chemsoft

6 2 Introduction 2.1 Background Boliden Mineral is a metal company, with zinc and copper as their main metals. Their operations focus on the initial stages of the processing chain: exploration, mining and milling, smelting, refining and recycling. Metal recycling is a field in which Boliden is a global leader and more development on this front is to be expected. Boliden employs at the moment 4,400 people. Below a list of the mines and smelters that constitutes Boliden Mineral: Mines: Tara (Ireland) Boliden Area Garpenberg Aitik Smelters: Kokkola (Finland) Harjavalta (Finland) Odda (Norway) Rönnskär Bergsöe Garpenberg is the oldest mine in Sweden still operating. At Garpenberg about 1.4 million tonne/year of complex ore is mined. The complex ore is a mixture of zinc, copper and lead sulphides, with additional gold and silver minerals (Table 1) ( The most common metal bearing minerals are sphalerite ( ZnS ), galena ( PbS ), chalcopyrite ( CuFeS 2 ) and pyrite ( FeS 2 ). It is called a complex ore since it contains metal sulphides that are difficult to separate. Therefore the complex ore are among the most difficult sulphide ores to treat ( Table 1. The assay of the copper-, lead - and zinc-concentrate during December 2007 at the Garpenberg Concentrator. Malm: G9 dec-07 Slig Halter Ton Au g/t Ag g/t Cu % Zn % Pb % Cu Pb Zn Ing flot

7 At the concentrator the ore is crushed and ground into smaller fractions, the ore is then treated with a series of flotation cells (Figure 3), the flotation method is explained under the headline flotation basics. From the first stage of flotation cells a copper-lead-concentrate and a middling product is extracted. The middling product is treated further in flotation cells to produce a zinc concentrate. The copper-leadconcentrate is also further treated with a series of flotation separation cells that separates the lead and the copper. Dichromate is added to the separation as a depressant for lead when extracting the copper. The lead and copper concentrates are sent to Rönnskär and the zinc concentrate is delivered to Kokkola Zink or Odda Boliden. Figure 3. Process scheme of Garpenberg Concentrator. 7

8 2.2 Theory Comminution The ore from the mines contains valuable minerals and gangue. Gangue is rock with no economic value. The minerals that exist in the ore are usually finely dispersed and well integrated with the gangue making the process of separating the minerals difficult. To be able to separate the gangue and the minerals comminution of the ore is performed. This means that the particle size of the ore is decreased until liberated particles of the valuable minerals can be removed. Blasting in the mine can be regarded as the first stage of comminution, and later in the mill plant the ore is further crushed and ground until the mineral particles are liberated Flotation basics The finely ground minerals and gangue are mixed with water to form slurry with about 40 % solids by weight. The slurry (pulp) is placed in a tank with a certain width and length. Air is bubbled through the tank, which is stirred with an agitator to increase the chances for bubble and mineral to meet. The minerals attach to the bubbles and float to the surface where the bubbles form a froth phase. The froth is scraped off and with it the valuable minerals. In some cases the valuable minerals are the minerals left in the tank and the froth contains the unwanted minerals. Different additives can be used to achieve a more effective flotation. Collectors are added to promote the minerals attachment to the bubbles, depressants are added to decrease a certain minerals attachment to the bubbles, frothers are added to increase the froth stability and ph regulators are added to control the pulp environment. Appendix 1 gives a more thorough explanation Literature review Separation of the copper-lead concentrate at Garpenberg concentrator plant is in the form of flotation and potassium dichromate is used as a depressant. The dichromate attaches to the galena particles and render them hydrophilic and therefore unwilling to attach to the air bubbles. The copper-particles are hydrophobic (with the help of collectors, also explained more in Appendix 1) and attaches to the bubbles and forms a stable froth at the top of the flotation cell. The froth is then removed and with it the copper minerals. The lead concentrate is what remains in the cell after the flotation. Potassium dichromate is a hexavalent compound and is harmful to health and must be handled and disposed of appropriately (Wikipedia). New directive from EU states that toxic and environmental unfriendly substances should be phased out and replaced with more suitable substances (Eurlex.europa.eu). There have been several attempts to replace potassium dichromate with a less toxic depressant. The more successful trials involved dextrin. Tests on ore from 8

9 Renström/Långdal in the Boliden Concentrator plant with dextrin as a substitute for potassium dichromate in the flotation separation of copper-lead-concentrate have been performed with good results (Bolin et al, 1989). The dextrin was combined with NaOH as a ph regulator and the results indicated that dextrin+naoh gave better depression of lead but gave lower yield of copper in the Copper concentrate. The addition of dextrin could be lowered to at least 40 g/ton and the ph 11.9 was more than sufficient according to the tests. The decrease in ph with increasing addition of dextrin indicates that a dextrin-lead surface complex is formed as a result of the interaction of dextrin with lead hydroxide Lead hydroxide-dextrin interaction is at optimum at ph 11; at this ph the solution conductance also decreases with increasing dextrin concentration in the lead-dextrin mixture (Laskowski and Liu, 1989). This conclusion is supported by Forsling et al (1997) and Forsling et al (1998) who concluded that adsorption of dextrin is strongly ph dependent with high adsorption densities appearing around the ph of metal hydroxylation and that the interaction between dextrin and metal hydroxyl was found to be chemical complexation. Dextrin has exhibited difficulties to interact selectively when the mineral surfaces have a similar variety of metal ions (E. Bogusz et al, 1997), (Jan Drzymala et al, 2002). In a complex ore the mineral surfaces are often contaminated with numerous different metals. 2.3 Objective The purpose of this project was to further investigate the possibility to exchange the hazardous potassium dichromate with the non-toxic dextrin in the separation flotation of copper-lead concentrate. To evaluate dextrin s ability to depress lead minerals laboratory scale flotations with dextrin will be performed as well as reference flotations with potassium dichromate as a depressant. The products from the flotation experiments will be analysed and compared with the products from the reference flotation with potassium dichromate. The flotation solution will also be analysed in order to determine dextrin s ability to attach to the lead minerals. 9

10 3 Material and Equipment The copper-lead concentrate from Garpenberg was taken out continuously from the feed stream to the copper-lead separation during a longer time period. Between the samplings the excess water was decanted leaving the buckets to contain a high concentrate of solids with some water on top. Five buckets of concentrate was then sent from Garpenberg concentrator to Boliden Pilot Plant. A sample of the concentrate was sent for analysis and the result is presented in Table 3 under the headline feed concentration (6.1). 3.1 Depressants Di-chromate ( 2Cr2O7 K, Na 2Cr2O7 ) Figure 4. Potassium dichromate. Chromium was first discovered in 1797 in the mineral Crocoite (Figure 1, Wikipedia). Potassium Dichromate (Figure 4) is an orange, crystalline substance with a specific gravity of The solubility in water is 11, 7 % at 20 C. Dichromate solutions have 2 an acidic reaction because Cr 2O 7 reacts with water in the following way: H 2 + 2O + Cr2O7 H CrO 2 4 Salts of chromic acid in acid medium are oxidants (Cr(VI) changes to Cr(III)) according to Bulatovic (2007).Potassium dichromate is an oxidant (oxidizing agent). The reduction half-equation is: Cr 2 O 7 2 (aq) + 14H + + 6e 2Cr 3+ (aq) + 7H 2 O (E = V) The compound is corrosive and can cause damage to the eye or even blindness. It consist of Cr (VI), which means it is also toxic and carcinogenic (Wikipedia). 10

11 3.1.2 Dextrin Figure 5. Dextrin. Dextrin (Figure 5) is a non-toxic organic polymer with a molecular weight that ranges from 800 to 79,000. Organic polymers are often used as modifiers in flotation of sulphide and non-sulphide minerals. The organic polymers can be divided into four major groups: Non-ionic, Cationic, Anionic and Amphoteric when looking at the character of their polar group. But because polymers often undergo modifications the characterisation is very loosely applied. The polymers role in flotation is highly dependent on the composition of the polymer. Some polymers are used as dispersants, flocculants or depressants (Bulatovic, 2007). The organic polymers are the most complex of all reagents used in flotation. Because of the vast number of different types of each polymer the chemical structure is often undefined and this is why some type of polymer can work perfectly in a flotation and another not work at all according to Bulatovic (2007). Dextrin belongs to the non-ionic polymers. It is derived from potato starch through heating in an acidic environment. Depending on the degree of conversion, dextrin may exhibit varying solubility in water in which it forms colloidal system. As a polyhydroxy compound, dextrin is able to participate in a number of chemical reactions characteristic of alcohols (etherification or esterification through substitution on hydroxyl groups, or chemical complex formation with the hydroxyl groups (Laskowski and Liu, 1989). 11

12 3.2 Collectors Dithiophosphate Danafloat 871 Alkyl and aryl dithiophosphoric acids and their alkali salts are widely used as sulphide collectors known as Aerofloat (Bulatovic, 1997) The sodium (or potassium) salts of dialkyl dithiophosphoric acids can be prepared by reacting phosphorous pentasulphide with the right alcohols and sodium (potassium) hydroxide (Wills et al, 2006). Dithiophosphates are usually considered as a more selective collector then xanthates (Adams et al, 1986). Danafloat 971 is a mixture of the sodium salts of disopropyl-dithiophosphate and disec-butyl-dithiophosphate with mercaptobezothiazole dissolved in water (Appendix 5) Xanthate KAX Figure 6. Image of KAX. Xanthates are derivatives of carbonic acid, H CO 2 3 in which two oxygen have been replaced with sulphur and one hydrogen group with an alkyl (or aryl) group. KAX (Figure 6) are in another word O-alkyl (or aryl) dithiocarbonates (Leja, 1982). In the presence of moisture, xanthate hydrolyses and forms unstable xanthic acids, which further decompose into carbon disulfide and the corresponding alcohol. In solution the decomposition of xanthates increases with lowering of ph. In an alkaline medium xanthates are relatively stable (Bulatovic, 2007). KAX, potassium n-amyl xanthate is often used for sulphide-ores especially for copper-, lead- and zinc-containing ores which are Boliden`s main minerals. KAX is commonly used in the flotation processes within Boliden Mineral. 12

13 3.3 Regulators NaOH Sodium hydroxide is considered the strongest alkaline ph regulator. It is a very active substance and highly corrosive. Most sodium hydroxide is manufactured by electrolysis of saturated brines (NaCl). The sodium hydroxides regulating capacity stretches from ph 7 to ph 14 (Bulatovic, 2007). 3.4 Frother Nasfroth 240 Nasfroth is a glycol ether that is used to create a stable froth. 3.5 Other The water used during washing, flotation and vacuum separation was regular Boliden tap water. The process water from the concentrator plant was not used. 3.6 Equipment A stainless steel rod mill (Figure 7) was used to refresh the surface of the particles. The rods were of different sizes and weighed together 8,015 kg. The standard load consisted of 500 ml of water and 1 kg of solid. Figure 7. Rod mill displayed with rods. 13

14 The flotation apparatus (Figure 8) was a Wemco model number with an agitator speed of 1200 rpm and a cell volume of 2.7 L. Figure 8. Wemco flotation device. Other equipment used was vacuum flasks, hose, funnel, filter paper, water, trays, jars, syringes and a special syringe with a particle filter. 4 Analyses Solid samples All dried froth samples were sent to Garpenberg for analysis. The analysis is performed with a Bruker axs S4 explorer. Solution samples COD (Chemical oxygen demand) is a measurement of the amount of oxygen that is used during total oxidation of organic substances in water. This means that it will also measure other compounds other then dextrin. If the COD-values present a similar response to the change in factors as the Dextrin-values then one can assume that the change in COD is due to the dextrin present in the solution. Dextrin will oxidize and COD measures the oxygen needed. COD is measured with a spectrometer, a Hach Lange DR The concentration of dextrin in the solution is measured at the wavelength 489 nm with the same spectrometer. The spectrometer uses the curve in Figure 9 to estimate the dextrin level in the analyzed samples. The values in Table 2 are contributed by Amang Saleh, development engineer at Boliden Mineral and used by the spectrometer to create the curve in Figure 9. 14

15 Table 2. Values that contribute to the curve in Figure 9 used for measuring the dextrin level. Dextrin mg/l Abs Figure 9. Curve of dextrin versus the wavelength used by the spectrometer to estimate the dextrin content. 15

16 5 Method 5.1 Preparation of flotation Buckets with the concentrate were first decanted to remove the excess water on top. The buckets were then subjected to a stirrer who turned the concentrate into a thick slurry. The buckets were stirred until all the solid had entered the slurry phase. The slurry s solid content was calculated and since 1 kg material is usually used for laboratory scale flotation (according to the laboratory handbook) the concentrate was divided into samples containing 1 kg of solids. A pump was used and the material was evenly distributed. The calculations below show how the amount of slurry needed from the bucket in order to achieve 1kg of solids was calculated. The volume and mass of the samples was different for each bucket (5 buckets all and all) but the calculations where the same. Only the first buckets calculations will be presented. First the density of the sample was calculated by taking a small sample of slurry. The volume and the weight of the sample were measured. V m sample sample = 252mL = = 757.5g = kg L = dm The densities for the water (aquatic) and the density for the solid (s) assumed to be: 3 ρ = 1kg dm aq ρ = 4.5 kg dm s 3 3 Equation 1 m V = ρ Equation 2 m = m + m tot aq s Equation 1 and 2 combined gives: m = m + m sample s aq m aq = m m = kg m sample s s V sample m m s = + ρ ρ s aq aq ms kg m = + ρ ρ s aq s = ms 4,5 kg dm kg m + 3 1kg dm s Thus the solid weight in the extracted sample can be calculated. 16

17 m s = kg From the weight the volume needed of the pulp in order to achieve 1 kg of solid can be calculated. 1 kg solid = 387,735 ml of pulp The needed volume was measured up with water (388 ml) and the water level was marked on the sample container and then used to control the amount of pulp needed in all the sample containers taken from the same bucket. This leads to a rather rough estimate of the solid in the sample but due to the many other uncontrollable factors the fault will probably be so small that it can be overlooked. During this experiment several buckets where divided in this way. Scrubbing and washing In order to remove the remaining reagents from previous flotation the concentrate was subjected to 1 min of grinding. The theory is that interaction between the particles and the rods will scrub the mineral surfaces clean from reagents and therefore making the particles more reactive to the reagents used in this experiment. The concentrate is poured into a rod mill and water is added until the combined water content reaches 550 ml. The mill is run for 30 seconds and the concentrate is then poured out into a container. The rods and the mill are thoroughly rinsed and the rinsing water is also poured into the concentrate container. This leads to a vast amount of water that needs to be removed. In order to avoid loss of concentrate the slurry was filtered under vacuum to remove the treated excessive water and the wet cake was then washed thoroughly with 100 ml tap water twice. The washed cake is then added to the flotation cell and the cell is moved to the flotation machine. 5.2 Flotation First the reagents were prepared and the amounts of reagents needed were calculated. The ph-electrode was calibrated and the vacuum pumps were prepared. The NaOH was weighed in order to estimate how much was needed to reach the desired ph. The agitator was lifted into place and water was used to fill the cell to the lip which gives about 60 weight percent solution when the solid weighs about 1 kg. The agitator was turned on and the solid and solution was mixed to slurry. The ph was measured and solid NaOH was added until the desired ph value was reached, giving the NaOH five minutes to react. The estimated amount of dextrin solution was added to the slurry and given a certain number of minutes to react. The collector was added and given 1 minute to attach itself to the copper minerals. The flotation segment (pull) starts when opening a valve that sucks air into the agitator and releases air into the slurry. The minerals attach to the bubbles and float to the surface where the bubbles form a froth phase. The first flotation lasted for 1, the second one lasted for 1.5 and 17

18 the last flotation lasted 2 minutes. The froth was carefully scraped of and assembled (Figure 10). The froth was then filtered to remove water (Figure 11). The solid part of the froth is placed on Tablets and placed for drying over night in an oven. Figure 10. Picture of flotation in progress. Figure 11. Vacuum treatment of the froth from the flotation. 18

19 5.3 Sample treatment The dried samples are weighed, treated so that all the particles are uniform and splitted into equal fractions. One of the fractions was sent to Garpenberg for analysis. It is important that the fraction has a small particle size and that all the particles are of roughly the same size. It is also important to have a fraction that represents the entire sample accurately. 5.4 COD and Dextrin The solution extracted with the vacuum pump from the lead concentrate was collected and a part of the solution was extracted with a syringe that was equipped with a special particle filter ensuring that no particles was present in the samples taken. The samples were then prepared differently depending on the analytical method. All samples were analyzed for dextrin and COD. The dextrin that failed to attach to the lead-minerals will be present in the lead concentrate solution and the analysis will determine the amount of dextrin present in the solution. The purpose is to have as little dextrin in the solution as possible to avoid overdosing. Overdosing leads to accumulation of dextrin in the process water which could lead to other complications. Also overdosing is not very economic COD 2 ml of the sample is added to the special test-tube containing the prepared reagents used. The tube is gently turned upside down to mix the solutions. The tube is then inserted into a heater and heated to 148º C for about 2 hours. A blank test-tube is always prepared, in other words a sample with 2 ml of water instead of lead concentrate solution. The blank sample is later used to adjust the spectrometer. The result from the analysis is presented in Table Dextrin Two ml of the sample is added together with 0.2 ml 40 % phenol and 5 ml concentrated sulphuric acid, The test tubes are place into a heater for about 10 min. The heated samples are taken out to cool and then analyzed with the spectrometer using a special program that detects dextrin. A zero sample is also used here to adjust the spectrometer. The dextrin levels was measured and documented in Table

20 6 Results To evaluate dextrin s ability to depress lead minerals laboratory scale flotations with dextrin was performed as well as reference flotations with potassium dichromate as a depressant. The investigation was concerned with the combination of dextrin and NaOH at ph: 11 with the lowest addition of dextrin at 40 g/ton. The assay and distribution of the copper and lead in the formed copper and lead concentrates were analyzed and compared. The assay is how much of a concentrate that consist of a specific metal expressed in weight per cent. Distribution is the amount of metal in the feed that will end up in any product given as percentage. A high quality lead concentrate contains a high level of lead (high assay) and has a high distribution of lead. But when the distribution is too high it usually means that more of other metals will also end up in the concentrate causing the assay of lead to decrease. For instance, if the copper-lead separation was to malfunction and not separate at all the distribution of lead in the lead concentrate would be a 100 % but then all the copper would still be present causing the assay of lead to be low. If the assay of lead is too high then the distribution of lead will be low because a large part of the lead will go to the copper concentrate. To have a high quality concentrate the assay and distribution should be balanced. The flotation solution was also analysed in order to determine dextrin s ability to attach to the lead minerals. Feed Concentration The concentrate used was analysed and the results is presented in Table 3. The assays for the concentrate are roughly the same as the Garpenberg Concentrator produces at present. Table 3. Table of Copper-lead-concentrate. 20

21 6.1 Determining the parameters Flotation is a method where many different parameters can have an influence. The parameters can be altered in many ways but to be able to do a proper investigation some parameters must be kept constant. The obvious parameters that were kept constant were: size of the flotation tank and the agitator speed. The flotation device used during these experiments has only one size of cell and one speed for the agitator. The frother was decided to be the same used at the Garpenberg plant. According to previous attempts at Boliden Mineral (Bolin et al, 1989) the most suited regulator would be sodium hydroxide. The parameters that were assumed to be of the most importance for the outcome of the flotation with dextrin were: Depressant, Dextrin (initially it was planned to use other organic depressants as well) Amount of depressant Mixing time for the depressant ph Type of collector Since the project was about exchanging potassium dichromate with dextrin, dextrin is the obvious choice for depressant. The literature suggested that the amount of dextrin added could be as low as 40 g/ton (Bolin et al, 1989). The span that was chosen was from 40 g/ton to 120 g/ton. The mixing time for a depressant is usually (Laboration handbook, 1998) 2 min but in order to fully investigate the impact the mixing time a span from 2-12 min was selected. Mixing time means the time the slurry has to interact with the reagent until a new reagent is added or the flotation begins. The ph should be around according to the previous attempts and based on the many reports concerning metal hydroxides. The collector used for the experiment was a strong collector, xanthate and a weak collector, dithiophospate. A strong collector will float less selective then a weak collector and since dextrin according to the literature might be less selectively a weak selector could be the better choice. Dithisphosphate is a weaker collector meaning that it is more selective and might suite dextrin better. 21

22 6.2 The DOE method To ensure a statistically correct laboratory testing the method Design of experiments, DOE was utilised. The method is commonly used in all kind of industries for development of a new product/process or optimising a current product/process. DOE insures that the selected experiments are maximally informative. First the factors are decided and inserted. The factors are in this case the ph during the flotation, the depressant used during flotation, the amount of depressant (addition g/ton) used and the collector used during the flotation, the reaction time for the depressant (mixing time, min). Using the factors given and with the help of the program MODDE a worksheet (Table 4) for the experiments was obtained. The worksheet is constructed in a way that all relevant factors are varied systematically and thus avoiding unnecessary experiments. The experiments were carried out in the run order suggested by the worksheet. Table 4. Worksheet of experiments. Exp No Run Order ph Collector Addition Mixing time Xanthate Xanthate Ditiophospates Ditiophospates Xanthate Xanthate Ditiophospates Ditiophospates Xanthate Xanthate Ditiophospates Ditiophospates Xanthate Xanthate Ditiophospates Ditiophospates Xanthate Xanthate Ditiophospates

23 Prior to the experiments with the Dextrin a reference experiment with dichromate as the depressant was executed (Table 5). At Garpenberg the ph is not altered during the flotation with any kind of ph-regulators, the collector used is xanthate and the mixing time is 2 min. The same frother reagens was used during all flotations and can be found in Table 6 below. The amount of dichromate used during laboratory batch scale flotation is 1000 g/ton concentrate. Table 5. Worksheet for reference experiment. Exp No ph Collector Addition Depressant Mixing time (g/tonne) (min) Xanthate 1000 Dichromate 2 Table 6. Type, amount of frother reagent used for all flotation experiments. Frothergreagens Nasfroth 240. Amount 1 drp Experiments All results from the flotations and the total analysis of the flotation products can be found in Appendix 2. The analysis result was also calculated further with the help of a template (Appendix 3). The result from the analysis and the calculations was used to investigate the importance of the different parameters. 23

24 Selectivity Selectivity Pb dist % Cu dist % Dextrin experiment, 18 Dextrin experiment, 19 Dichromate experiment, 0 Figure 12. Selectivity diagram of the reference experiment and two dextrin experiments for the copper concentrate. The purpose with this graph (Figure 12) is to demonstrate the difference in selectivity between the different experiments. All results can be found in Appendix 3. The distribution of lead is plotted against the distribution of copper in the copper concentrate. Distribution means how much of the incoming metals will end up in the specific stream in percentage. Each experiment had three pulls that contributed distribution values to the curve. For each pull the amount of copper will increase but the amount of lead will also increase. To have good selectivity, the amount of lead should be low and the amount of copper should be high so the more curved graph the higher selectivity. The most selective experiments with dextrin were experiment nr 18 and nr 19 which are presented in Figure 12. The other experiments can be found in a selectivity Figure in Appendix 3. The parameters for experiment nr 18 and nr 19 can be found in Table 7. The assay results are presented in Table 9 and Table 10. The assay result for the experiment nr 0, the dichromate experiment is in Table 8. The parameters for the dichromate experiment can be seen in Table 6. Both experiment nr 18 and nr 19 gave a higher assay of lead in the lead concentrate and copper in the copper concentrate compared to the dichromate experiment. Table 5. Experiment parameters for experiment nr 18 and nr

25 Table 6. Assay of the copper and lead concentrate for the dichromate experiment (exp 0). Table 7. Assay of the incoming copper-lead concentrate, the copper concentrate and the lead concentrate from experiment nr 18. Table 8. Assay of the incoming copper-lead concentrate and the copper concentrate, lead concentrate from experiment nr

26 6.3 The Model The results from Appendix 3 are added to the MODDE- program as responses (Table 11). The responses were the assays of silver (%Ag), copper (%Cu), zinc (%Zn) and lead (%Pb) in the copper concentrate and the assays of lead (%Pb) and copper (%Cu) in the lead concentrate. The difference between lead and copper distribution (DiffPb/Cu) in the lead concentrate was also added. All responses representing the values after the third pull. Table 9. Responses added to the model. Regression analysis The data given is analysed with regression analysis giving a model that relates the changes in the factors to the changes in the responses. The model will indicate which factors are important and their influence on the responses and also predict the best operating conditions. The full analysis can be found in Appendix 3. Due to the low amount of copper in the feed concentrate the main focus of the simulation was on optimising the lead concentrate. The model used to analyse the results was a screening with a screening interaction model, a more thorough explanation of the model is presented in Appendix 4. 26

27 Scaled and Centered Coefficient plots Figure 13. Coefficient plot for the lead concentrate. With a coefficient plot the factors that affect the responses the most can be distinguished. The size of the coefficient shows the change in the response when a factor varies from 0-1. High coefficient indicates that the factor contributes to the change in the response and if there are any interactions between the factors (the expanded factors). The 95 % confidence interval (the line) determines the uncertainty of these coefficients and the size determines the noise level. For the lead concentrate in Figure 13 the most important terms seems to be ph, add and Co2. In other words the ph-level, the amount of dextrin added and which collector used during flotation. Interaction plots Figure 14. Interaction plots showing the interaction between the ph and the addition of dextrin for the lead concentrate. 27

28 Interaction plots displays predicted values of the response, when a factor varies from its low to its high level, plotted for all the combinations of levels of the other factor. There are no interaction between the ph-level and the addition of dextrin for the lead concentrate according to Figure 14. Main effects plots Figure 15. Main effects plot of ph on lead concentrate. The main effects plot presents the effect a certain factor has on a specific response by varying the factor between high and low values. In Figure 15 the ph dependence for the different responses can be seen. To increase the amount of lead in the lead concentrate and decrease the amount of copper the ph should be kept around ph 9. The amount of lead in the copper concentrate will also increase but the increase can be seen as negligible. 28

29 Figure 16. Main effect plot for addition of dextrin for the lead concentrate. From Figure 16, a low addition of dextrin to the flotation contributes to a higher ratio between the lead and copper distribution in the lead concentrate. It also gives a higher amount of lead and lower amount of copper in the lead concentrate. Contour plots A contour plot is a graphic representation of the relationships between numeric variables in two dimensions. Two variables are for x- and y- axis and a third variable z is for contour levels. The contour levels are plotted as curves and the area between is colour coded for estimation of the value. In this case the highest values are marked with red and lowest values are marked with blue. The plots are hard to distinguish but an enlargement of the contour plot from Appendix 4 that gives the highest ratio between the lead and copper distribution in the lead concentrate can be seen in Figure 17. The collector used was dithiophosphate and the mixing time 2 min. 29

30 Figure 17. Enlargement of the contour plot of the difference between lead and copper distribution in the lead concentrate with mixing time 2 min and ditiophosphate as collector. The ratio between the distribution of copper and lead in Figure 17 is at its highest value when the collector dithiophosphate is used, the mixing time 2 min and the ph held around 10 and the addition of dextrin around 40 g/l. In order to achieve a high amount of lead in the lead concentrate the collector should be dithiophosphate and the mixing time 12 min according to the contour plots in Appendix 4. An enlargement of that plot can be found in Figure 18. Figure 18. Enlargement of the contour plot of the assay of lead in the lead concentrate with dithiophosphate as a collector and the mixing time 12 min. 30

31 To achieve a high value of lead according to Figure 18 the addition of dextrin should be kept low (around 40 g/l) and the ph around 10. Optimising Although the screening interaction model is not suitable for optimisation it was used as a tool to further investigate the most suitable setting of the parameters. The restrictions used during optimisation were the same as used for the worksheet. Table 10. Optimum values for the factor within the given restrictions. ph Collector Addition (g/ton) Depressants Mixing time (min) 10 Dithiophosphates 40 Dextrin 12 The optimum values in Table 12 are the values that according to the model and the given restrictions will most likely give the best outcome. The given restrictions are the limits provided by the span of the parameters. The responses that where optimized was the difference between the lead and the copper assay in the lead concentrate (Diff Pb/Cu), the assay of copper in the lead concentrate (%Cu_Pb) and the assay of lead in the lead concentrate (%Pb_Pb). The difference between the lead and copper assay was maximized and the assay of copper was minimized. Prediction plots The prediction plot works pretty much the same as the contour plots above. The black square indicates the area covered by the experimental values. A higher ratio in lead and copper distribution in the lead concentrate is achieved when the mixing time is kept around 7 min and the collector dithiophosphate according to the prediction contour plots in Appendix 4. To examine the ph-level and addition of dextrin an enlargement of the plot in question can be found in Figure 19. Figure 19. Enlargement of the prediction plots of the difference between the lead and the copper distribution in the lead concentrate with mixing time 7 min and dithiophosphate as collector. 31

32 The square representing the experiment in Figure 19 indicates that the experiment should perhaps have spanned into lower ph in order to reach a higher difference in distribution. The addition of dextrin seems to have the right span. Addition lower then 10 g/ton and a longer mixing time than 17 min seems to be most efficient in increasing the amount of lead in the lead concentrate according to the prediction contour plots in Appendix 4. Figure 20 is an enlargement of the graph. Figure 20. Enlargement of the assay of lead in the lead concentrate with dithiophosphate as a collector and the mixing time 17 min. Again here in Figure 20 it becomes apparent that a lower ph and a lower addition of dextrin could give a higher amount of lead in the lead concentrate. 32

33 7 COD and Dextrin analysis If the depressing was successful the dextrin should be on the surface of the leadparticles and not in the solution. Table 11. COD and dextrin values for the solution part of the lead concentrate. The COD and Dextrin values are missing in experiment nr 11 is due to a human error. Some of the Dextrin values are zero: this could be due to large amount of dextrin present in the solution. Very high concentrations disturb the measuring. The values from Table 13 were added as responses in the designed model (Table 14). Table 12. Responses for the model, with dextrin and COD added as responses. 33

34 Scaled and Centered Coefficients plots Figure 21. Coefficient plot over dextrin and COD. The significant factors for dextrin is according to Figure 21 the ph, add and add*ph. Thus meaning that the ph and the amount of added dextrin is important to the level of dextrin in the lead concentrate solution. The significant factor for COD are ph*add. Interaction plots Figure 22. Interaction plot displaying the interaction between the ph and the addition of dextrin for dextrin and COD.. In Figure 22 there is a clear interaction between the addition of dextrin and the ph for COD. 34

35 Main effects plot Figure 23. Main effects plot for ph of dextrin and COD. Figure 23 indicates that a ph around 10 will give a higher part of dextrin in the lead concentrates solution. COD will be lower at a lower ph. Figure 24. Main effect for addition of dextrin and COD. From Figure 24, a low addition of dextrin will give a lower quantity of dextrin and a lower COD in the lead concentrate solution. This seems reasonable. Contour plot The contour plot in Appendix 4 for the response dextrin provides the information that in order to achieve a low amount of dextrin in the lead concentrate solution the collector used during the flotation should be a dithiophosphate. The different mixing time plots show very little difference indicating that the mixing time for dextrin has a negligible effect but is still important during optimisations. An enlargement of the contour plots with dithiophosphate as a collector and the mixing time 12 min can be seen in Figure

36 Figure 25. Contour plot of the response dextrin with dithiophosphate as a collector and the mixing time 12 min for dextrin. According to Figure 25 the addition of dextrin should be kept low and the ph around 12 in order to keep the dextrin level in the lead concentrate solution low. However, the best flotation response is for lower ph levels according to what is shown in Figs The contour plots in Appendix 4 of the response COD indicates that in order to achieve a low COD level the mixing time plays a vital role. To keep a low COD-level in the lead concentrate solution the mixing time should be kept low. The collector used in the flotation seems to be less important to the outcome of the COD-levels. An enlargement of the contour plot with the mixing time 2 min and the collector dithiophosphate can be observed in Figure 26. Figure 26. Contour plot of the response COD with the mixing time 2 min and the collector dithiophosphate for COD. 36

37 From the enlargement in Figure 26 the ph and the addition of Dextrin effects on the COD-level in the lead concentrate solution is presented. A low ph and a low addition of Dextrin will give a low COD-level. Prediction plots Again this model is not suited for optimising but in order to further investigate the factors some prediction plots were added. The prediction plot in Appendix 4 of the response dextrin indicates that to achieve a low dextrin value in the lead concentrate solution the collector should be dithiophosphate and that the mixing time is not crucial. An enlargement of the prediction plot with dithiophosphate as a collector and the mixing time 2 min is presented in Figure 27. Figure 27. Prediction plot of the response dextrin with dithiophosphate and 2 min mixing time for Dextrin. In order to minimise the level of dextrin in the lead concentrate solution the addition of dextrin should have been lower and the ph kept higher then the implemented trials. 37

38 8 Discussion Dichromate can be replaced by dextrin providing that the factors that affect the outcome the most, the ph level and the amount of dextrin added is carefully monitored and adjusted. A ph around 12 gives a lower dextrin level in the lead concentrate solution but a ph around 10 gives more lead in the lead concentrate and a better separation result. High concentration of lead in the lead concentrate indicates better depression of the lead minerals, presumably that the dextrin has attached to the lead -minerals surfaces and formed complexes. But when the dextrin level in the solution is also higher it is contradicting. What also further complicates things is that the COD-levels are low. If dextrin was depressing the lead and causing the lead to form complexes the dextrin level in the pulp liquid of the lead concentrate should be low. The dextrin should be happily attached to the lead particles and not staying around in solution. According to previous studies dextrin is more likely to form hydroxide complexes with lead at ph 11. Perhaps the formation of hydroxide complexes is not the only way for dextrin to depress lead? Also dextrin may absorb to various degree on other sulphide and gangue minerals, especially talc. Most of the earlier studies were conducted during single mineral adsorption tests not flotation test which may explain the contradiction. Another possibility is the handling of the concentrate. The solution and the solid part of the concentrate is separated with vacuum, maybe the complexes formed are not very stable and the separation damages the complexes and the dextrin to report back to the solution. But a higher level of dextrin in the lead concentrate solution should also cause a higher COD-value. Perhaps this anomaly is because something else than what the COD measures became lower and thus neutralized the effect of the rise in dextrin level and caused lower COD-value. In other words, the measured COD-value is not a good indicator for the copper-lead separation result. A low level of added dextrin seems to give a better depression of lead than a high level, which further corroborates the report (Bolin and Norén, 1989) from the Renström/Långdal concentrator plant with dextrin as a substitute for potassium dichromate in the flotation separation of copper-lead concentrate. By adding more depressant the amount of lead being depressed should increase not decrease. This effect could be because more of the other minerals will be depressed when more dextrin is added and thus diluting the stream. More lead could be depressed but the assay of lead will not increase because the size of the stream will also increase. When looking at the distribution of lead and copper in the experiments in Appendix 3, the distribution of copper and lead will increase in the lead concentrate with a higher addition of dextrin. Also when comparing the experiments with all the other parameters constant almost all the experiments exhibit an increase in size in the lead stream with a higher addition of dextrin. This is a clear indication that when adding more dextrin more minerals in general were depressed. Low addition levels of dextrin also give lower dextrin and COD values in the lead concentrate solution, which is perfectly understandable. 38

39 Due to the contradicting results, no real correlation between the dextrin in the pulp liquid and the separation results could be determined. To use measurements of dextrin in the pulp liquid as a quality response for the separation of copper-lead concentrate is therefore not recommended. The model designed is not the best for prediction analysis but some predictions can be made. The prediction plots suggests that a ph around 9, an addition lower then 40 g dextrin/tonne concentrate and a mixing time spanning from 7 to 17 minutes should give a higher assay of lead in the lead concentrate. The collector that should be used is dithiophosphate. The plots also suggests that in order to keep the dextrin level lower in solution the addition should be lower then 40 g dextrin /tonne concentrate and the ph higher then 13. The ph difference makes the goal to achieve a high lead assay in the lead concentrate and at the same time have a low dextrin level in the solution difficult. However, this is only a problem if one believes in a direct coupling between separations results and the rest dextrin solution levels. The separation of copper and lead is currently performed in the ph span of 9-12 and to lower it or increase it beyond this span would have implications on the steps before and after the separation. This is not something to recommend. The copper concentrate from the best experiment, nr 19 gives a copper concentrate with 5.6 % copper and a lead concentrate with % according to Table 13. The Garpenberg concentrator produced a concentrate with 22 % Cu during December 2007 (Table 14). But considering that the concentrator plant uses a series of cleaner flotations stages (the plant scheme in Figure 2) instead of a laboratory batch cell like the one used for this experiment the results can not be compared. To be able to compare a trial run with dextrin as a depressant in the copper-lead -separation at Garpenberg Concentrator Plant should be performed. Table 13. Assay of the incoming copper-lead concentrate and the copper concentrate, lead concentrate from experiment nr 19. Table 14. The assay of the copper-, lead - and zinc-concentrate during December 2007 at the Garpenberg Concentrator. Malm: G9 dec-07 Slig Halter Ton Au g/t Ag g/t Cu % Zn % Pb % Cu Pb Zn Ing flot

40 The incoming copper-lead concentrate from Garpenberg has 1.80 % copper present (Table 14). This is considered to be a very low copper value and it raises questions. Is the separation of the copper-lead concentrate really necessary? Will separate lead and copper concentrates bring more profit than the copper-lead concentrate, considering the cost of separating the concentrates? Is the separation economically supportable? This question is not addressed in this report and will not be further discussed because it is not within the scope of this master thesis. 40

41 9 Conclusion Dextrin exhibited ability for depressing lead during copper and lead separation in a laboratory scale batch flotation equivalent to dichromate in both selectivity and assay. The factors that have the highest impact on the amount of lead in the lead concentrate were the ph of the solution and the amount of Dextrin added to the flotation. To achieve the highest assay of lead and a low copper assay in the lead concentrate within the given restrictions the collector used should be dithiophosphate (Danafloat 871), the ph should be kept around 10, the amount of dextrin used around 40 g/ton concentrate and the mixing time 2 minutes. Both the collector and the mixing time are weak variables but they still have an optimum. No real correlation could be determined between the content of dextrin in the pulp liquid and the quality of the copper-lead separation. The content of dextrin in the lead concentrate solution was most affected by the ph level and the addition-level of dextrin during the flotation. The CODlevel was generally affected by the cross factor ph and addition. If it is desirable to achieve a low dextrin level in the lead concentrate solution within the given restrictions, the collector used should be dithiophosphate and the ph should be around 12 and the addition of dextrin around 40 g dextrin/tonne concentrate. The mixing time have little to no effect on the dextrin-level in the solution. The collector used during the flotation has very little effect on the CODvalues. 41

42 10 Future work A new model The model designed is not equipped for prediction analysis so to further investigate a new model with less parameters should be created. The span for the ph and the addition of dextrin should be lower. Full scale trial A full-scale trial should be performed in order to investigate dextrin s depression ability in a concentrator plant. Only then can the question as to whether or not dextrin can replace dichromate as depressant for lead minerals in the copper and lead separation. 42

43 11 Reference 11.1 Books Adams, R.W., Burley, D.R., Cappuccitti, F., Carlson. A.W., Day, A.,Farwel, F.W., Foster, T., Hjartens, H., Holme, R.N., Mingione, P.A., (1986). Mining Chemicals Handbook. USA: American Cyanamid Company. Bulatovic, M.Srdjan., (2007). Handbook of Flotation Reagents. Amsterdam: Elsevier. Clarke, N.Ann., Wilson, J.David., (1983). Foam Flotation New York: Marcel Dekker, Inc. Eriksson, L., Johansson, E., Kettaneh-Wold, N., Wikström, C., Wold, S., (2000). Design of Experiments. Forssberg, K.S.Eric., (1985). Flotation of Sulphide Minerals. Amsterdam: Elsevier Leja, Jan., (1982). Surface Chemistry of Froth Flotation. New York: Plenum Press. Shaw, J.Duncan., (1992). Colloid and Surface Chemistry. Eastbourne: Anthony Rowe. Will, B.A., Napier-Munn, T.J., (2006). Mineral Processing Technology. Great Britain: Elsevier. Laboration handbook Mineralteknik Metodpärm, Försök lab och pilot, okt 1998, flotation på malm och pulp 11.2 Personal communications Bertil Pålsson, Lecturer at Luleå Tekniska Universitet. Johan Hansson, Development Engineer at Boliden Mineral. Rolf Danielsson, Process Engineer at Boliden Mineral. Jan-Eric Sundkvist, Development Engineer at Boliden Mineral. Nils-Johan Bolin, Development Engineer at Boliden Mineral. 43

44 Amang Saleh, Development Engineer at Boliden Mineral. Michael Eriksson, Laboratory Assistant at Boliden Mineral. Daniel Eklund, Geologist at Boliden Mineral Reports Bogusz, E., Brienne, S.R., Butler, I., Rao, S.R., Finch, J.A., (1996). Metal ions and Dextrin Adsorption on Pyrite. Minerals Engineering, 10, (4), Bolin, Nils.Johan., Norén, Peter., (1989). CuPb-koncentrat Renström/Långdal, Försöksanrikningskampanj juni 1989/CuPb-sep med NaOH+Dextrin. Boliden: Boliden Mineral, Division engineering. TM 37/89 Drzymala, Jan., Tomasik, Piotr., Sychowska, Beata., Sikora, Marek., (2002). Dextrins as Selective Flotation Depressants for Sulphide Minerals. Physiochemical Problems of Mineral Processing, 36, Forsling, Willis., Holmgren, Allan., Raju Baskar, G., (1997). Adsorption of Dextrin at Mineral/Water Interface. Journal of Colloid and Interface Science, 193, (CS975004), Forsling, Willis., Holmgren, Allan., Raju Baskar, G., (1998). Complexiation of Dextrin with Metal Hyrdoxides. Journal of Colloid and Interface Science, 200, (CS975332), Laskowski, S.J., Liu, Qi., (1989). The Interactions between Dextrin and Metal Hydroxides in Aqueous Solutions. Journal of Colloid and Interface Science, 130, (1), Internet LexUriServ.do?uri=OJ:L:2009:309:0071:0086:SV:PDF

45 12 Appendix 1: Flotation Theory Flotation is a method that uses the difference in surface properties for the minerals and the gangue to separate them. Flotation was at first used mostly for low grade and complex ore bodies which otherwise would have been regarded as to expensive to mine. The process is selective and can be used for separation of specific minerals from complex ore bodies such as copper-lead-zinc ore according to Wills and Napier-Munn (2006). Flotation basics The finely ground minerals and gangue are mixed with water to form slurry with about 40 weight procent solid. The slurry is placed in tank with a certain width and length. Air is bubbled through the tank, which is stirred with an agitator to increase the chances for the bubble and the mineral to interact. The minerals attach to the bubbles and float to the surface where the bubbles form a froth phase. The froth is scraped off and with it the valuable minerals. In some cases the valuable minerals are the minerals left in the tank and the froth contains the unwanted minerals. Flotation principle The flotation has selective attachment to air bubbles due to solid solution properties. The particles that attach to the bubbles must exhibit hydrophobic tendencies (aversion towards water), the mineral or at least the mineral surface should be non polar. A nonpolar particle is hydrophobic because water is a highly polar solution. The hydrophobic mineral will be more eager to attach it self to the non-polar air bubble (Shaw, 1992). The air bubble will carry the mineral up to the surface if the particle size of the mineral is not too large. The froth phase must be stable in order for the air bubble to continue to carry the mineral particles in the froth phase. By adding certain chemicals (frothers) the stability of the froth phase can be improved according to Wills and Napier-Munn (2006). To achieve selective flotation the different particle surface properties for the minerals are utilised. In other words selective flotation can only occur when there is a difference in surface properties between the mineral particles (Bulatovic, 2007). With chemical reagents the difference in properties can become more defined and a selective flotation more successful. Collectors are added to increase a certain minerals affinity for air and depressants are added to decrease a certain minerals affinity for air. The ph is often regulated with different reagents in order to achieve the desired environment. There are different forces that operate on the surface of the particle that affect the surface properties. 45

46 Forces at work during flotation Adsorption of a collector causes reduction of the surface energy and a change in the electrical double layer (charge and potential), adhesion of water. The flotation is dependent on many parameters and it is not possible to determine their exact mechanism of action. All parameters have only a positive effect within a certain span and that certain span is different for each case (Laskowsky and Liu, 1989) The flotation thermodynamics Young s equation for particulate material (Wills and Napier-Munn, 2006). γ = γ + γ cosθ s a s w w a γ s γ s a w = Surface energy between solid and air = Surface energy between solid and water γ w a = Surface energy between water and air Θ = The contact angle between the particle surface and the air bubble To break the particle bubble interface a force called work of adhesion W is required and that force is equal to the work needed to separate the solid air interface and produce separate air-water and solid-water interfaces. s a Ws a = γ w a + γ s w γ s a Thermodynamic criterion of flotation ( 1 Θ) W = γ cos s a w a From the equation it becomes apparent that the flotation of a solid on a solution depends on the contact angel between the air bubble and the solid particles according to Wills and Napier-Munn (2006). The greater contact angle the greater the work of adhesion between the particle and the bubble, meaning stronger attachment between the particle and the bubble. Collectors can be added to achieve a larger contact angle. The equation is not related to the rate of the flotation stated Laskowsky and Liu (1989). The attachment of minerals to the bubbles is the purpose of the flotation but entrainment in the water that passes the froth and physical entrapment of other particles to the bubbles also occurs. The degree of entrainment is dependent on the particle size and becomes more apparent for particles less then 50 µm according to (Smith and Warren 1989) These side effects affect the separation efficiency because both gangue and other minerals can be victims to entrainment and physical entrapment. The drainage of unwanted particles occurs in the froth phase. To achieve an adequate separation the flotation is often performed in several stages called circuits according to Will and Napier-Munn (2006). 46

47 13 Appendix 2: Experiments Table A2:1. Experiment 1 with dextrin. 47

48 Table A2:2. Experiment 2 with dextrin. 48

49 Table A2:3. Experiment 3 with dextrin. Table A2:4. Experiment 4 with dextrin. 49

50 Table A2:5. Experiment 5 with dextrin. 50

51 Table A2:6. Experiment 6 with dextrin. 51

52 Table A2:7. Experiment 7 with dextrin. 52

53 Table A2:8. Experiment 8 with dextrin. 53

54 Table A2:9. Experiment 9 with dextrin. 54

55 Table A2:10. Experiment 10 with dextrin. 55

56 Table A2:11. Experiment 11 with dextrin. 56

57 Table A2:12. Experiment 12 with dextrin. 57

58 Table A2:13. Experiment 13 with dextrin. 58

59 Table A2:14. Experiment 14 with dextrin. 59

60 Table A2:15. Experiment 15 with dextrin. 60

61 Table A2:16. Experiment 16 with dextrin. 61

62 Table A2:17. Experiment 17 with dextrin. 62

63 Table A2:18. Experiment 18 with dextrin. 63

64 Table A2:19. Experiment 19 with dextrin. 64

65 65

66 14 Appendix 3: Analysis and Calculations Table A3:1. Experiment 0 with sodium dichromate. Table A3:2. Experiment 1 with dextrin. Table A3:3. Experiment 2 with dextrin. 66

67 Table A3:4. Experiment 3 with dextrin. Table A3:5. Experiment 4 with dextrin. Table A3:6. Experiment 5 with dextrin. 67

68 Table A3:7. Experiment 6 with dextrin. Table A3:8. Experiment 7 with dextrin. Table A3:9. Experiment 8 with dextrin. 68

69 Table A3:10. Experiment 9 with dextrin. Table A3:11. Experiment 10 with dextrin. Table A3:12. Experiment 11 with dextrin. 69

70 Table A3:13. Experiment 12 with dextrin. Table A3:14. Experiment 13 with dextrin. Table A3:15. Experiment 14 with dextrin. 70

71 Table A3:16. Experiment 15 with dextrin. Table A3:17. Experiment 16 with dextrin. Table A3:18. Experiment 17 (31) with dextrin. 71

72 Table A3:19. Experiment 18 with dextrin. Table A3:20. Experiment 19 with dextrin. 72

73 Selectivity Pb dist % Cu dist % Figure A3:1. Selectivity plot with the distribution of Pb vs. the distribution of Cu for the Cuconcentrate for all experiments. 73

74 15 Appendix 4: the Model 15.1 Creating a worksheet To create a worksheet the factor are decided and inserted (Table A4:1). The factors are in this case the ph during the flotation, the depressant used during flotation, the amount of depressant (addition g/ton) used and the collector used during the flotation, the reaction time for the depressant (mixing time, min). The factors are specified and divided into different categories. A quantitative factor is a factor that may change according to a continuous scale (ph that varies from 10-12), a qualititative factor can only assume certain discreet values (the depressant used: Dextrin) Table A4:1. Factors added to the model. The program MODDE creates a worksheet (Table A4:2) using the factors. The worksheet is constructed in a way that all relevant factors are varied systematically and thus avoiding unnecessary experiments. The run order is the order in which the flotation should be performed 74

75 Worksheet Table A4:2. Worksheet of experiments. The results from attachment 3 (experiments) is added to the MODDE- program as responses to further investigate. The responses chosen for this simulation was mention before and are the amount of -silver (%Ag), -copper (%Cu), -zinc (%Zn) and the amount of lead (% Pb) in the copper concentrate and the amount of -copper (%Cu), - lead (% Pb) and the difference between the lead-distribution and the copperdistribution in the lead- concentrate (diff Pb/Cu) also called difference index. The responses can be found in Table A4:3. Table A4:3. Responses added to the model. 75

76 The data given is analysed with regression analysis giving a model that relates the changes in the factors to the changes in the responses. The model will indicate which factors are important and their influence on the responses and also predict the best operating condition. Due to the low amount of copper inbound the main focus of the simulation was on the optimising the lead concentrate Objective in MODDE, Model and Design The objective is what determines the choice of model and design. The objective used for this simulation was screening. Screening finds the important factors in the first phase of a project using interaction models. The design is a full factorial (2 levels), an orthogonal (balanced) design with all combinations of the factor levels. Main effects and all interactions are clear of each other (not confounded). In the present case, only the interaction ph*addition is significant. Table A4:4. Selected objective, model and design.!!" # $% &' ( ' )*+ $,!!" # &-'.-&-+--'./+ $, &' ( ' 76

77 15.3 Method of Fit The investigation was fitted with MLR, multiple linear regression. The coefficients of the model are calculated to minimize the sum of squares of the residuals, the sum of squared deviations between the observed and the fitted values of each response. The least squares regressions method gives little variance for the coefficients and small prediction errors. MLR fits one response at a time and assumes that the responses are independent Goodness of Fit Summery of fit plot Figure A4:1. Summery of fit plot over the responses. The plot above is a summery of fit-plot. The R2-value (first, green bar) is a measure of how well the data fits the model, the Q2-value (second, blue bar) is a measure of how well the model can predict new data. The model validity value (third, yellow bar) must be higher when 0.25 then there is no Lack of Fit of the model. A validity value 77

78 of 1 represents a perfect model. All yellow bars are well above 0.25 indicating that the model is valid. By adding factors like ph*add and expanding the model the predicition and the fit of the model can be improved. The danger with expanding the model is that you could be adjusting the model to noise. Noise is not important information that disturbs the model. Table A4:2. Model validity. 0%1 02%,201,,%210%$0$1$,01(((2 01( 01%220%%%$ 0%$, $0($( 011%%11 $, ($, 02%20(( 0$%% $0%222 $0($%$1 $,0%12 0,,,$!0%2(2$2022$ 0,1 20%(1% (0, $,0%2, 0,2%$% 01$2$,$0%2%, 011$ $0(,( 01($ $,0%%,1 0,2%2( "" 0%%$($ 02($ 012, $(0(%,0$% $,0$(,, 0$$2!#!01$(0%$($ 02%1,$ 0(($2%2 0$,( $,02,$$ 0122!!021,,,%0$2% 01%1 $01,11 $0(1( $,011( 02% )*$, &0 $0$$ 0*!*$( 3* 78

79 15.5 How good is the Model FigureA4:2. Normal probability plot of the lead concentrate. The residuals represents the experiments. If the residuals are random and normally distributed they will form a straight line between 4 and +4. In the assay of copper and lead in the lead concentrate some curvature in the residuals is noticed but not enough to cause alarm. 79

80 Figure A4:3. Normal probability plot of copper concentrate. The residuals in Figure A4:3 are shoving some pattern (the snakelike pattern most presented in the graph showing the assay of zinc in the copper concentrate. Since the model was based more on the lead concentrate the pattern may be ignored. 80

81 Coefficient plots Figure A4:4. Coefficient plot for the lead concentrate. The size of the coefficient shows the change in the response when a factor varies from 0-1. High coeffient indicates that the factor contributes to the change in the response and if there are any interactions between the factors (the expanded factors). The confidence interval (the line) determines the uncertainty of these coefficients and the size determines the noise level. For the copper concentrate in Figure A4:4 the most important terms seems to be ph, add and Co2. The ph-level, the amount of dextrin added and which collector that is used during the flotation. Figure A4:5. Coeffient plot for the copper concentrate. 81

82 The most important terms for the copper concentrate according to Figure A4:5 seems to be ph*add and ph. Effects plots Figure A4:6. Effects plot of the copper concentrate. With the effects plot the factors that affect the responses the most can be distinguished. The size of the bars shows the change in the response when a factor varies from 0-1 and all other factors are held constant at their intermediate. High bar indicates that the factor contributes to the change in the response and if there are any interactions between the factors (the expanded factors). The confidence interval (the line) determines the uncertainty of these bars and the size determines the noise level (not important information that disturbs the model). From Figure A4:6 the terms ph*add and ph seems to have the greatest effect on the copper concentrate. 82

83 Figure A4:7. Effects plot of the lead concentrate. For the lead concentrate the most important terms seems to be ph and add according to Figure A4:7. Interaction plots Figure A4:8. Interaction plots showing the interaction between the ph and the addition of dextrin for the lead concentrate. Interaction plots displays predicted values of the response, when a factor varies from its low to its high level, plotted for all the combinations of levels of the other factor. There are no interaction between the ph-level and the addition of dextrin for the lead concentrate according to Figure A4:8. 83

84 Figure A4:9. Interaction plots presenting the interaction between the ph and the addition of dextrin in the copper concentrate. All plots in Figure A4:9 indicate that for the copper concentrate there is a clear interaction between the ph-level and the addition of dextrin. Main effects plots Figure A4:10. Main effects plot of ph for the lead concentrate. 84

85 Figure A4:11. Main effects plot of ph for the copper concentrate. In Figure A4:10 and A4:11 the ph dependence for the different responses can be seen. To increase the assay of lead in the lead concentrate and decrease the assay of copper the ph should be kept around ph 10. The assay of lead in the copper concentrate will also increase but the increase can be seen as negligible. This will also provide higher assay of copper in the copper concentrate. Most of the zinc and the sulphur will end up in the copper concentrate. Unfortunately the assay of silver will decrease in the copper concentrate. It is more favourable if the silver reports to the copper concentrate instead of the lead concentrate. Figure A4:12. Main effects plot for addition of dextrin for the lead concentrate. 85

86 From Figure A4:12, a low addition of dextrin to the flotation contributes to a higher ratio between the lead and copper distribution in the lead concentrate. It also gives high assay of lead and low assay of copper in the lead concentrate. Figure A4:13. Main effects plot for addition of dextrin on copper concentrate. From Figure A4:13, a low addition of dextrin to the flotation effects the copper concentrate and provides with a lower assay of copper, higher assay of, higher assay of lead and a higher assay of sulphur.. Contour plots Figure A4:14. Contour plot for difference of lead and copper distribution in lead concentrate. A contour plot is a graphic representation of the relationships between numeric variables in two dimensions. Two variables are for x- and y-axis and a third variable z 86

87 is for contour levels. The contour levels are plotted as curves and the area between is colour coded for estimation of the value. In this case the highest values are marked with red and lowest values are marked with blue. The plots are hard to distinguish but an enlargement of the contour plot that gives the highest difference between the lead and copper distribution in the lead concentrate can be seen below. The collector used was dithiophosphate and the mixing time 2 min. Figure A4:15. Enlargement of the contour plot with mixing time 2 min and ditiophosphate as collector for the difference between lead and copper distribution in the lead concentrate. The difference between the distribution of copper and lead is at its highest value when the collector dithiophosphate is used, the mixing time 2 min and the ph held around 10 and the addition of dextrin around 40 g/l. 87

88 Figure A4:16. Contour plot of the assay of lead in the lead concentrate. Figure A4:16 gives the contour plots for the assay of lead in the lead concentrate. The collector dithiophosphate and the mixing time 12 min seems to give the highest value of lead. An enlargement of that plot can be seen below. Figure A4:17. Enlargement of the contour plot of the assay of lead in the lead concentrate with dithiophosphate as a collector and the mixing time 12 min. To achieve a high value of lead the addition of dextrin should be kept low (around 40 g/l) and the ph around 10 according to Figure A4:17. 88

89 Figure A4:18. Contour plot of the assay of copper in the lead concentrate. The copper should be kept low in the lead concentrate and the collector dithiophosphate and the mixing time 12 min seems to give the lowest values. An enlargement of Figure A4:18 can be seen below. Figure A4:19. Enlargement of the contour plot from Figure A4:18 with dithiophosphate as a collector and the mixing time 12 min. Furthermore the ph should be kept around 10 and the addition of dextrin around 40 g/l in order to keep the copper low in the lead concentrate. 89

90 Optimising Although the screening interaction model is not suitable for optimisation the optimality was used as a tool to further investigate the most suitable setting of the parameters. The responses are given the right criteria (Table A4:6). In order to optimise the lead concentrate, the responses for the copper concentrate were excluded. The diffpb/cu was maximized, the % Pb was maximized and the % Cu in the lead concentrate was minimized. Table A4:6. Optimiser Table of the factor and responses. The optimiser optimises an overall desirability function combining the individual desirability of each response. The results are represented in Table A4:7 (below). Table A4:7. Optimiser Table giving the condition most likely to give a high lead content in the lead concentrate within the given restrictions. The values for factors that according to the optimiser will give the best lead concentrate within the given restrictions are given in Table A4:7 and also in Table A4:8. 90

91 Table A4:8. Optimum values for the factor within the given restrictions. ph Collector Addition (g/ton) Depressants Mixing time (min) 10 Dithiophosphates 40 Dextrin 12 Predicting contour plots Figure A4:20. Predicting contour plot of the difference in lead and copper distribution in lead concentrate. The prediction plot works pretty much the same as the contour plots above. The black square indicates the area covered by the experimental values. Figure A4:20 indicates that a higher difference in lead and copper distribution in the lead concentrate is achieved when the mixing time is kept around 7 min and the collector dithiophosphate. To examine the ph-level and addition of dextrin an enlargement of the plot in question can be found below. 91

92 Figure A4:21. Enlargement from Figure A4:23 with mixing time 7 min and dithiophosphate as collector. The square representing the experiment in Figure A4:21 indicates that the experiment should perhaps have spanned into lower ph. The addition of dextrin seems to have the right span. Figure A4:22. Prediction contour plot of the assay of lead in the lead concentrate. The predicted optimised area is just below and to the right of the experiment area. Thus indicating that perhaps a lower addition of dextrin and a lower ph should be recommended. Addition lower then 10 g/ ton and a longer mixing time then 17 min seems to be most efficient in increasing the assay of lead in the lead concentrate according to the prediction contour plots. An enlargement of the graph to high right from Figure A4:22 can be seen in Figure A4:23. 92

93 Figure A4:23. Enlargement of Figure A4:22 with dithiophosphate as a collector and the mixing time 17 min. Again here in Figure A4:23 it becomes apparent that a lower ph and a lower addition of dextrin could have given a better outcome. Figure A4:27. Prediction contour plot of the assay of copper in lead concentrate. Figure A4:27 supports the earlier indications. Dithiophosphate as a collector and a lower ph (around 9) and a low addition of dextrin should give a lower assay of copper in the lead concentrate. The mixing time seems less important in this scenario. 93

94 15.6 COD and Dextrin analysis The solution in the lead concentrate was analyzed with a spectrometer. If the depressing of the lead minerals was successful the dextrin should be found on the surface of the lead particles and not in the solution. COD (Chemical oxygen demand) is a measurement of oxidizing particles. Table A4:9. COD and dextrin values for the solution part of the lead concentrate. The COD and Dextrin values are missing in experiment nr 11 because of a human error. Some of the Dextrin values are zero, this could be due to a large amount of dextrin present in the solution making the analyses faulty. The values from Table A4:9 were added as responses in the previous model. Table A4:10. Responses for the model, with dextrin and COD added. 94

95 Table A4:11. Worksheet with dextrin and COD values inserted. Summery of fit plot Figure A4:25. Summery of fit plot. From the Figure above (Figure A4:25) the response dextrin seems to fit the model but not the COD (the explanation to interpreting the summery of fit plot can be found under Figure 9). COD is very difficult to predict and so the COD results are perhaps not as liable as the dextrin results and therefore the result will mostly focus on dextrin. 95

96 Normal probability plots Figure A4:26. Normal probability plots residuals of dextrin and COD. Both dextrin and COD show snakelike patterns in their residuals but both are within the 4 and +4 limits according to Figure A4:26. Coefficient plots Figure A4:27. Coefficient plot over dextrin and COD. The significant factors for dextrin is according to Figure A4:27 the ph, add and add*ph. Thus meaning that the ph and the amount of added dextrin is important to the level of dextrin in the lead concentrate solution. The significant factor for COD are ph*add. 96

97 Interaction plots Figure A4:28. Interaction plots displaying the interaction between the ph and the addition of dextrin for dextrin and COD. According to Figure A4:28 the only interaction occurring is between the ph and addition of dextrin for the COD. Main effects plot Figure A4:29. Main effects plot for ph of dextrin and COD. Figure A4:29 indicates that a ph around 10 will give a higher part of dextrin in the lead concentrates solution. The COD will be lower at a lower ph, but the effect is less than the experimental error shown by the error bars. 97

98 Figure A4:30. Main effect for addition of dextrin and COD. From Figure A4:30, a low addition of dextrin will give a lower quantity of Dextrin and a lower COD in the lead concentrate solution. This seems reasonable. Contour plots Figure A4:31. Contour plots of dextrin values in the lead concentrate solution. The contour plot in Figure A4:31 provides the information that in order to achieve a low amount of dextrin in the lead concentrate solution the collector used during the flotation should be a dithiophosphate. The different mixing time plots show very little difference indicating that the mixing time for dextrin is of little importance in this scenario. An enlargement of the contour plots with dithiophosphate as a collector and the mixing time 12 min can be seen below in Figure A4:32. 98

99 Figure A4:32. Enlargement of contour plot from Figure A4:31 with dithiophosphate as a collector and the mixing time 12 min for dextrin. According to Figure A4:32 the addition of dextrin should be kept low and the around 12 in order to keep the dextrin level in the lead concentrate solution low. Figure A4:33. Contour plots of COD. Figure A4:33 indicates that in order to achieve a low COD level the mixing time plays a vital role. To keep a low COD-level in the lead concentrate solution the mixing time should be kept low. The collector used in the flotation seems to be less important to the outcome of the COD-levels. An enlargement of the contour plot with the mixing time 2 min and the collector dithiophosphate can be observed in Figure A4:34. 99

100 Figure A4:34. Enlargement of a contour plot from Figure A4:33 with the mixing time 2 min and the collector dithiphosphate for COD. From the enlargement, Figure A4:34 of the contour plots in Figure A4:33 the ph and the addition of dextrin affect on the COD-level in the lead concentrate solution is presented. A low ph and a low addition of dextrin will give a low COD-level according to Figure A4:34. Table A4:12. Optimising Table for dextrin and COD. 100

101 Table A4:13. Optimising results for dextrin and COD. Prediction contour plots Figure A4:35. Prediction plot of dextrin. The prediction plot in Figure A4:35 indicates that to achieve a low dextrin value in the lead concentrate solution the collector should be dithiophosphate and that the mixing time is not crucial. An enlargement of the prediction plot with dithiophosphate as a collector and the mixing time 2 min is presented in Figure A4:

102 Figure A4:36. Enlargement of the prediction plot in Figure A4:35 with dithiophosohate and 2 min mixing time for dextrin. In order to optimise the level of dextrin in the lead concentrate solution the addition of dextrin should have been lower and the ph kept higher then the implemented trials from Figure A4:

103 16 Appendix 5: Chemsoft 103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125