Froth flotation is a widely used and versatile mineral processing method for concentrating metal ores. A finely ground ore feed is processed through a flotation circuit consisting of interlinked cells where mineral particles are separated from waste material by using their differences in surface properties. The layout of the circuit has been found to greatly affect the overall flotation performance. While industrial flotation circuit design has in the past relied on experience, due to the complex nature of the process, only small scale circuit optimisations using simple flotation models were reported in literature.

This work proposes a new system capable of automatic generation of optimal circuit designs for any given feed. The system combines a circuit simulator containing detailed froth-phase flotation models with a robust genetic algorithm to search through possible layouts and to produce the global optimal result.

An empirical model to predict the pulp phase flotation rate constant was developed, which was used together with physics-based models describing the froth recovery and entrainment factor to simulate the flotation process. Three feed models with different complexities were also developed. Comparable flotation performance was observed between a modelled 10-cell rougher circuit and the experimental results from Northparkes copper concentration plant.

Through the genetic algorithm, optimal layouts were obtained for circuits consisted of 3 to 10 cells. Layouts containing only rougher cells were able to recover a maximum amount of mineral and were found optimal for smaller circuits, whilst inclusion of cleaner cells in the optimal layouts was found more beneficial for larger circuits. This circuit modelling and optimisation system was used to study the sensitivity of flotation performance and optimal layouts to variations in feed particle size and grade. The results showed strong correlations between the variables and froth phase behaviours which determined the concentrate recovery and grade, the optimal layouts were, however, robust and relatively resilient to the changes in feed conditions.