In this research, we address the problem of evaluating physical stability (PS) to close tailings dams (TD) from medium-sized Chilean mining using artificial intelligence (AI) algorithms. The PS can be analyzed through the study of critical variables of the TD that allow estimating different potential failure mechanisms (PFM): seismic liquefaction, slope instability, static liquefaction, overtopping, and piping, which may occur in this type of tailings storage facilities in a seismically active country such as Chile. Thus, this article proposes the use of four machine learning algorithms, namely random forest (RF), support vector machine (SVM), artificial neural networks (ANN), and extreme gradient boosting (XGBoost), to estimate five possible PFM. In addition, due to the scarcity of data to train the algorithms, the use of generative adversarial networks (GAN) is proposed to create synthetic data and increase the database used. Therefore, the novelty of this article consists in estimating the PFM for TD and generating synthetic data through the GAN. The results show that, when using the GAN, the result obtained by the ML models increases the F1-score metric by 30 percentage points, obtaining results of 97.4%, 96.3%, 96.7%, and 97.3% for RF, SVM, ANN, and XGBoost, respectively.

Chile is one of the major producers of copper in the world, and as such is responsible for 1.7 million tons of tailings per day. While the most commonly used deposit to store this type of mining waste is historically tailings sand dams, the mining industry has over the last two decades been inclined toward thickened tailings dams (TTD) because of their advantages in water resource recovery, lower environmental impact, and better physical and chemical stability over conventional deposits. Within the geotechnical area, one key requirement of TDD, is the need to monitor moisture content (w%) during operation, which is today mostly performed in situ - via conventional geotechnical or simple visual means by TTD operators - or off site, via remote sensing. In this work, an intelligent system is proposed that allows estimation of different classes of in-situ states and w% in TTD using Machine learning algorithms based on Artificial Neural Networks (ANN), Support Vector Machine (SVM) and Random Forest (RF). The results show an accuracy of between 94% and 97% in the classification task of the Dry, Semisolid, Plastic and Saturated classes, and between 0.356 and 0.378 of the MAE metric in the regression task, which is sufficient to estimate the w% with ML methods.