Intelligent Prediction of Critical State Parameters for Non‑plastic Tailings and Soils Using Evolutionary Algorithms - Mining, Metallurgy & Exploration (2024)

Tailings from mines and industries impose danger to the environment as evidenced by the recent failures of tailing storage dams. The tailings are mostly non-plastic with silt-size particles, but their behavior is largely dependent on the extreme void ratios of the material. Hence, the critical state provides an appropriate perspective in explaining the behavior of these geomaterials. Based on a comprehensive dataset of non-plastic tailings and non-plastic soils, an attempt has been made here to correlate critical state parameters to extreme void ratios and stress parameters. A hybrid algorithm with multi-objective feature selection (MOFS) and extreme learning machine (ELM) was used to establish the correlation and to identify the most influential input parameters. Based on the information from the proposed hybrid algorithm, a comprehensive prediction model is proposed using multi-gene genetic programming (MGGP) algorithm to facilitate quick estimation of critical state parameters. The performance of both models has been assessed using several performance metrics. The results from the study indicate that hybrid MOFS algorithms perform very well in the prediction of critical state line parameters and the identification of influential features. The findings of the study have significant implications for future research on the safe management of tailing storage facilities.