Gold Liberation Modelling of Diagnostic Leaching Data Using Neural Network Analysis

This study presents a neural network approach to modelling the liberation of gold bearing ores. A complete mineralogical analysis of unmilled and milled ores, including gold deportment and gangue content are used as inputs to a self-organising neural net which generates order preserving topological maps. The arrangement and shapes of these clusters are coupled to unmilled free gold data to predict gold liberation in milled ores (absolute error: 4.2%). Moreover, the self-organising maps were diagnostic of the quality of data used, indicating that the relationship between particle size and gangue material content requires further investigation.