Authors: Contador Villegas, N; Huenchulao Catalán, H; Oliva Miranda, JM; Dubournais Donoso, F; Gallardo Arriagada, M

Open access courtesy of:


Cite As:
Contador Villegas, N, Huenchulao Catalán, H, Oliva Miranda, JM, Dubournais Donoso, F & Gallardo Arriagada, M 2021, 'Artificial intelligence applied to the detection and early warning of geotechnical instabilities in mining slopes', in PM Dight (ed.), SSIM 2021: Second International Slope Stability in Mining, Australian Centre for Geomechanics, Perth, pp. 227-234,

Download citation as:   ris   bibtex   endnote   text   Zotero

It is possible to detect and anticipate geotechnical instabilities of the slopes of an open pit from the systematic monitoring of surface displacements induced by mining excavation. In the last few years, the detection rate of large-scale instabilities has improved through the application of the radar interferometry technique. However, small-scale instabilities are the most frequent and they present a low detection rate and early warning, since they have a short activation period, and are eventually, eclipsed by environmental noise and/or the mining operation itself. For this particular case, how to improve the detection and anticipation rate was discussed based on the hypothesis that geotechnical instabilities show a ‘recognisable behaviour’, well-defined and mathematically expressible. This feature makes the application of artificial intelligence (AI) tools feasible, specifically neural networks, to generate models trained in the early detection of ‘recognisable behaviour’, under a supervised learning approach. To develop a predictive application based on AI tools, a collaborative work dynamic was proposed between Minera Los Pelambres (MLP), belonging to the mining company Antofagasta Minerals S.A. (AMSA), and E Mining Technology S.A., a mining technical services company. The results of the training process of neural network models showed they are capable of identifying the ‘recognisable behaviour’ of a geotechnical instability at an early stage. Moreover, these models, when complemented with a layer of geotechnical-mathematical criteria, were allowed to build an algorithm capable of improving the overall performance of the surface displacement monitoring system. The performance of the algorithm was evaluated over a period of eight months at the MLP open pit. In this period, it was possible to increase the detection and early warning rate from an effectiveness of 43–82% in smallscale instabilities and as a result, a decrease in the risk during the construction of slopes. In addition, a significant reduction of false positives was reached by minimising the effect in the environmental noise by 80% with respect to the performance of the current monitoring systems. In the future, it seems reasonable to predict that with an expansion of the dataset and auscultation of new AI models and/or architectures, it will be possible to further improve the efficiency of monitoring systems.

Keywords: slope monitoring, artificial intelligence, advanced analytics, geotechnical instabilities, early warning

Pohlmann, F, Waldron, A & Jain, SG 2018, Advanced Deep Learning with Keras, Packt Publishing Ltd, Birmingham, pp. 6–21.
Raschka, S & Mirjalili, V 2017, Python Machine Learning Second Edition, Packt Publishing Ltd, Birmingham, pp. 206–209.

© Copyright 2022, Australian Centre for Geomechanics (ACG), The University of Western Australia. All rights reserved.
Please direct any queries or error reports to