Transitional behaviour prediction in iron tailings via artificial intelligence

• Artificial intelligence was used to predict transitional behaviour in iron tailings • Iron tailings are poorly-graded with high degree of transitional behaviour • No influence of sample preparation on the models developed for the prediction • ANN model is more reliable than ANFIS in predicting transitional behaviour • ANN has the highest prediction accuracy and has the best performance The significant contributions of mining of minerals to the development of any nation make the generation of tailings inevitable and therefore, understanding their characteristics is vital. The contribution of engineering granulometric signatures to the aspect of behaviour called transitional mode (non-convergent) is also crucial. This work presents the artificial intelligence based study for the prediction of transitional behaviour in iron tailings considering engineering granulometric indices. This was achieved by conducting laboratory tests on dry compacted DC, wet compacted WC and slurry SL iron tailings and re-analysis of data from previous studies to determine transitional behaviour as well as predicting their behaviour using artificial neural network and adaptive neuro-fuzzy inference system. The iron tailings are poorly graded with strong degree of transitional behaviour with m values ranging from 0.32 to 0.81. The ANN models for DC, WC, SL and combined samples CS have relative similar correlation values and ditto for the ANFIS models. This signifies that the influence of sample preparations is not significant. The ANN model is reliable and could be used to predict the transitional mode of behaviour in iron tailings. However, the ANFIS model is less suitable for the prediction of transitional behaviour in iron tailings. The ANN model has the best performance based on low model errors and highest accuracy in prediction.