Previous models of the drum scrubbing process have incorporated the size distribution of the feed ore, the scrubber's input rate. a combined breakage/discharge rate function and an appearance matrix which described the size distribution of broken particles for all unique ore types fed to the scrubber as variables in order to describe the size distribution of the products of scrubbing.
The first objective of this study was to rigorously construct a model of drum scrubbing for Argyle Diamond Mines where a previous study had resulted in the building of a model with a number of limitations. To implement this a statistically based sampling theory which allowed the estimation of sampling errors and suggested methods whereby these could be minimised was utilised. Accepting 10% relative error the sample masses which would give this accuracy in the critical size fraction, which was always shown to be the coarsest fraction present in a √2 sieve series were estimated. These proved to be too large for practical purposes so the sample masses were based on obtaining 10% relative error in all fractions except the coarsest one.
The drum scrubber circuit contained no sampling devices. Also, as large scale sample analysis facilities did not exist both of these facilities these were designed and installed following the theoretical guidelines as closely as possible. A test of the theory and equipment was built into the experiment which showed that the error incurred was less than expected due to the ore being coarser than was originally estimated. This also showed that a bias of the sample mass had been introduced into the analysis stage by the riffle used to reduce the quantity of 1.9 mm material for treatment. Never the less, it was concluded that the design of the experimental work had been successful.
Breakage with in the scrubber vas determined to be solely caused by attrition and the appearance functions were constructed to reflect this. A systematic method of construction for the five appearance functions each of which corresponded to one of the five levels of prior ore weathering was proposed.
The model incorporated the plant throughput rate, the speed of rotation of the scrubber and the size distribution of the feed stream into the regression equations which determined the values of the breakage/discharge rate functions. The output of the final model responded in the expected manner to all these variables, but the calculated size distribution of the scrubber's product stream consistently contained too little material in the finer size fractions.