Prediction of Metallurgical Performances as a Function of Fineness of Grind

"Grinding simulators are now widely available to assist in the optimization of industrial circuits. They provide circuit operating conditions or flowsheet designs that yield a given product size distribution. However, they do not provide information of the impact of a change of product size distribution on the subsequent flotation or leaching stages. This paper presents a method to predict grade and recovery indices for a flotation plant as a function of the flotation feed size distribution. The application of the method to various size distributions allows the determination of the optimal fineness of grind. A case study for a CuPb rougher circuit of a New Brunswick complex sulphide processing plant is presented to illustrate the method.IntroductionIs the grinding circuit product fineness optimum for the subsequent separation steps? This is a question that is certainly very difficult, but critical, to answer. Laboratory and pilot plant testworks can provide clues as to whether grind should be finer or coarser. For a full scale operation there is no other way than trying to change the setup of the grinding circuit and verify the effect on flotation. True effects can, however, be masked by the numerous and unexpected disturbances inherent to grinding and flotation circuit operations. A good preliminary investigation requires a noise free environment to elaborate possible solutions for the question.Plant simulators provide this noise free environment. However, despite grinding and flotation simulators having evolved significantly for the last 20 years, very little is available on the integration of both simulators to analyze the effect of fineness of grind on metallurgical performances. For instance, a grinding circuit simulator may help to find the tuning of a grinding circuit that will yield the desired fineness of grind, but does not inform on the effect the new product size distribution will have on the separation stages. Linking grinding and flotation would require a flotation plant simulator with the rate and infinite recovery parameters for each mineral in each size interval considered. This approach is seldom used because of the tedious work required to estimate the model parameters. Recently Leroux(1992) proposed a method to predict mineral liberation from grinding ore of known mineralogical texture. Although the method is still at its infancy it may provide, when combined with a reagent conditioning model (e.g Niemi et al., 1982; Bazin et al., 1988), a way to simulate flotation without the need of entering the numerous parameters of a flotation model."