Rule-based characterization of industrial flotation processes with inductive techniques and genetic algorithms
By making use of machine learning techniques, the features of flotation froths and other plant variables can be used as a basis for the development of knowledge-based systems for plant monitoring and control. Probabilistic induction and genetic algorithms were used to classify different froth structures from industrial copper and platinum flotation plants, as well as recoveries from a phosphate flotation plant. Both algorithms were equally capable of classifying the different froths at least as well as a human expert. The genetic algorithm performed significantly better than the inductive algorithm but required more tuning before optimum results could be obtained. The classification rules produced by both algorithms can easily be incorporated into a supervisory expert system shell or decision support system for plant operators and could consequently make a significant impact on the way flotation plants are currently being controlled. * Author to whom all correspondence should be addressed.