Dynamic measurements performed around an industrial gold ore leaching plant with two leaching tanks were used to perform a multivariate statistical analysis to extraction information from this data set. Performance indices related to leaching rate constants were generated in the data set pre-processing. Then, Principal Component Analysis (PCA) and Cluster Analysis (CA) were applied to find correlations between the process variables. It is shown that in the studied case, the behaviour of the two cyanidation tanks can be modelled separately. Despite correlated inputs, noisy measurements, and a few missing variables, it was possible, through multivariate regression based on the Projection to Latent Structures (PLS) technique, to propose an empirical model of the first leaching tank.
Keywords: gold ores, cyanidation, gold ore leaching, reactors analysis, PCA, principal component analysis, partial least squares, cluster analysis, performance monitoring, statistical analysis, modelling