The behavior of batch grinding of a dry binary mix using dolomite and quartz has been studied. Through several tests with different proportions of these minerals and different grinding cycles, the granular system evolution was observed in terms of sharpness coefficient and mean diameter d50. These parameters were fit into the Hill’s sigmoidal function of probability distribution to adjust the results. The results were used for the validation of a neural network system (multilayer perceptron-like algorithm), which will be presented in another article.
Selective comminution; grindability; artificial neural networks
