Size reduction processes, particularly fine grinding systems, in mineral processing and cement production plants constitute a great portion of energy consumption and operating costs. Therefore, the grinding systems should be designed properly and operated under optimum conditions to achieve productive and cost effective operations. The use of simulation based on kinetic mathematical models of grinding has proven useful in this respect. The kinetic models contain two essential parameters, namely, breakage rate and breakage distribution functions, that are to be determined experimentally, and preferably in laboratory, or by back-calculation from the mill product size distribution for a given feed size distribution. Experimental determination of the breakage parameters has been mostly carried out in laboratory batch mills using one-size-fraction material. The breakage rate parameter is obtained from the disappearance rate of this one-size-fraction material, while the breakage distribution parameters are estimated from the short-time grinding of the same material. Such laboratory methods using one-size fraction material, however, are not truly representative of industrial continuous mill operations where the mill contents have a distribution of particle sizes. There is evidence in the literature that the size distribution of the mill contents affects the breakage parameters. This thesis study was undertaken with the main purpose of investigating the effect of the size distribution of the mill hold-up on the brekage parameters of quartz and calcite minerals in lockedcycle dry grinding experiments. The locked-cycle and one-size-fraction experiments were performed in the Bond ball mill instrumented with a torque-measuring device. Different closing screen sizes were used in the locked-cycle work to produce different size distributions of the mill hold-up, and the operating conditions were changed in the one-size-fraction experiments to obtain different power draws. Particle breakage parameters were assessed for these changing conditions. Prior to the experiments related to the main purpose of the study, preliminary experiments were conducted for two reasons: (i) to find the power draw of the Bond mill in relation to the operating conditions with the intention of eliminating the use of costly torque-measuring devices by others / and (ii) to find the most accurate estimation method of breakage distribution functions among the three existing methods, namely, the &ldquo / zero-order production of fines&rdquo / method, the BII method, and the G-H method. The G-H method was found to be more appropriate for the current study. The locked-cycle grinding experiments revealed that the breakage rate function of coarse fractions increased with increasing proportion of fines in the mill hold-up. Breakage distribution functions were found to be environment-dependent and non-normalizable by size in one-size-fraction and locked cycle grinding experiments. It was concluded that the cumulative basis breakage rate function could sufficiently represent the breakage characteristics of the two studied materials in a wide range of operating conditions. Therefore, it would be more appropriate to evaluate the breakage characteristics of materials ground in ball mills by linearized form of the size-discretized batch grinding equation using single parameter instead of dealing with two parameters which may not be independent of each other.

The objectives of this project were to survey the flotation circuit of the Boliden concentrator, mass balance collected data and evaluate a scale-up model for laboratory flotation results. The model assumes that half of the recovery to cleaner middlings in a standard laboratory test would report to the final concentrate if it were done in closed circuit, as is the case in a full-scale plant. It has been used by Boliden Mineral AB since 1982 but its accuracy had not been studied since 1986. The model can be categorised as of open circuit type with scale-up factors. The project was based on a complex Ag-Au-Cu-Pb-Zn sulphide ore from the Kristineberg mine. Laboratory tests were done to produce concentrates of CuPb, Cu, Pb and Zn with pulp samples from the concentrator as feed material. The software HSC 9.3 was used to mass balance data from the plant survey. It was decided that the model would be deemed usable if it was able to predict the plant results with the same accuracy as in the survey of 1986. A simulated locked cycle test with split factors (Agar & Kipkie, 1978) was identified as an alternative scale-up model. The results showed that the model was able to predict the plant results with the same accuracy as in 1986. It was especially good at predicting grade and recovery of the main element in a concentrate. For example, it predicted an 18 % higher grade and 11 % lower recovery of Cu to the CuPb concentrate, while a 3 % lower grade and 11 % lower recovery of Zn was predicted to the Zn concentrate. The locked cycle model gave much worse predictions on grades, but more accurate recoveries. It was also better at predicting the behaviour of minor impurity elements such as As and Bi. A recommendation is to combine the two alternatives in a type of "mixed cycle" model. In this study, it would have predicted an 18 % higher grade and 7 % lower recovery of Cu to the CuPb concentrate, as well as a 3 % lower grade and 1 % higher recovery of Zn to the Zn concentrate compared with plant results. Such a model seems to give better figures, but should be put to the test on more samples and ores to confirm this belief. It could at the very least be used to check the reliability of results predicted by the current scale-up model. / Syftet med det här examensarbetet var att utföra en detaljprovtagning av flotationskretsen i Bolidens anrikningsverk, massbalansera data och sedan utvärdera en modell för uppskalning av resultat från laboratorieflotationer. Modellen antar att hälften av utbytet till returgodset i ett satsvis laboratorieförsök skulle rapportera till det slutliga koncentratet om det återcirkulerades, såsom i ett anrikningsverk. Den har använts av Boliden Mineral AB sedan 1982 men utvärderades senast 1986. Kategoriskt kan den ses som en uppskalningsmodell av typen öppen krets med skalfaktorer. Projektet baserades på en komplex Ag-Au-Cu-Pb-Zn sulfidmalm från gruvan i Kristineberg. Laboratorieförsök utfördes för att ta fram koncentrat av CuPb, Cu, Pb och Zn, med pulpprover från driften som utgångsmaterial. Programmet HSC 9.3 användes för att massbalansera datan från provtagningen. Det bestämdes att modellen skulle anses som godtagbar ifall den kunde förutspå driftresultatet med samma noggrannhet som 1986. Ett simulerat försök av typen sluten krets (Agar & Kipkie, 1978) identifierades som den mest intressanta alternativmodellen och även den utvärderades. Resultaten visade att modellen än idag ger godtagbara förutsägelser med samma noggrannhet som 1986. Modellen var särskilt bra på att förutspå halt och utbyte av den huvudsakliga metallen till dess eget koncentrat. Den förutspådde exempelvis en 18 % högre halt och 11 % lägre utbyte av Cu till CuPb-koncentratet, samt 3 % lägre halt och 11 % lägre utbyte av Zn till Zn-koncentratet. Den alternativa modellen gav sämre förutsägelser med avseende på halter, men bättre med avseende på utbyten. Den var bättre på att förutspå beteendet hos låghaltiga föroreningar såsom As och Bi. Rekommendationen är att kombinera de två modellerna till en "blandkretsmodell". I den här undersökningen hade ett sådant alternativ förutspått en 18 % högre halt och 7 % lägre utbyte av Cu till CuPb-koncentratet, samt 3 % lägre halt och 1 % högre utbyte av Zn till Zn-koncentratet jämfört med driftresultatet. En sådan modell tycks ge bättre förutsägelser, men bör testas på fler prover och malmtyper. Den borde åtminstone kunna användas för att kontrollera trovärdigheten hos resultaten förutspådda av den nuvarande modellen.