Artificial neural network prediction of Al2O3 leaching recovery in the Bayer process—Jajarm alumina plant (Iran)

The relationship between alumina leaching recovery in the Bayer process to the chemical modules of bauxite fed to the process has been studied using regression and artificial neural network (ANN) methods. The database for this study consisted of 332 sample analyses for bauxite and its subsequent red mud product, together with analyses of the leaching recovery in the Bayer process. The levels of Al2O3/SiO2, Al2O3/Fe2O3 and Al2O3/TiO2 were determined by SPSS software to be the appropriate predictors of Al2O3 leaching recovery in the stepwise variable selection procedure. The results of multivariable regression studies were not significant. The generalized regression neural network (GRNN) improved the correlation coefficient to an acceptable level of 0.86, with differences between − 0.98% and + 0.85% from the actual determined recovery. The proposed ANN method could be applied as a new method for the prediction of leaching recovery in the Bayer process, when the bauxite from different sources and with different chemical compositions is fed to the plant.