Representation of particle multimineral distributions

When considering particle behavior in a mineral processing plant, mineral processing engineers are concerned with many issues. Two important issues are how well grinding liberates minerals, and how these liberated minerals are separated from the barren particles. Research in mineral processing to date has focused largely on modeling particle behavior considering the distribution of minerals in terms of binary particles. That focus is placed on one mineral with all other minerals treated as the associated gangue. Such an approach is clearly limited, particularly when there are numerous valuable minerals. As the sophistication of mathematical models increases, it is then possible to model particle behavior by considering particles as consisting of numerous minerals. Thus particles can be considered as "multimineral" instead of just "binary". The new emphasis on multimineral particles has lead to the recognition of a number of problems. A paramount problem is how to represent multimineral particle composition distributions. This problem has a number of immediate applications. These include: (i) a simple method of extending binary stereological correction to multimineral particles, (ii) a way to carry out mass balancing for multimineral particle distributions in an efficient manner, (iii) a way to reduce data to represent multimineral particle distributions. This paper presents a methodology for reconstructing the particle multimineral composition distribution from broad statistical mineralogical information. The method is validated using ore particles from a hard-rock titanium deposit. The method for reconstructing the multimineral distribution is based on the principle of maximum entropy. Furthermore, to actually simulate particles the Markov Chain Monte Carlo (Hastings- Metropolis algorithm) is used. The agreement between the model and the actual distribution is found excellent