text

Synopsis By making use of genetic programming, empirical models for metallurgical processes can be evolved that are more cost-effective than models determined by means of classical statistical techniques. These methods explore populations of candidate models assembled from sets of variables, parameters and simple mathematical operators, and do not require explicit specification of model structures. The application of the proposed strategies is illustrated by means of a case study pertaining to the leaching of nickeliferous chromite ores. Modelling of the nickel and cobalt extraction from the ores yielded models of similar accuracy compared to those obtained by non-linear regression and artificial neural networks. However, in the case of the iron extraction from the ores, the genetic programming model was significantly more accurate than both the regression and neutral network models