Search

Due to the high environmental risks and negative impact of a failure, tailings storage facilities (TSFs) need constant monitoring. Advanced mathematical models have been developed in the past to predict the behavior of TSFs and raise alerts if needed. To be precise and reliable, such models need a spatial distribution of soil types within the dam as an input. Getting this data from laboratory measurements is time and cost-consuming. In this article, we propose an ANN-powered algorithm, which allows us to accurately estimate the soil distribution based on a cone penetration test (CPT)., {"references":["Bhattacharya, B., Solomatine, D.P., Machine learning in soil classification, in Neural networks, 19(2), 2006, pp. 186-195186–195","Blagus, R., Lusa, L. SMOTE for high-dimensional class-imbalanced data. BMC Bioinformatics 14, 106 (2013).","Chawla, N. V., Bowyer, K. W., Hall, L. O., Kegelmeyer, W. P., SMOTE: synthetic minority over-sampling technique, Journal of artificial intelligence research, 321-357, 2002.","Douglas, B.J. and Olsen, R.S. (1981) Soil Classification Using Electric Cone Penetrometer. Proceedings of Conference on Cone Penetration Testing and Experience, St. Louis, 26-30 October 1981, 209-227.","Hulse JV, Khoshgoftaar TM, Napolitano A: Experimental perspectives on learning from imbalanced data. Proceedings of the 24th international conference on Machine learning. 2007, Corvallis, Oregon: Oregon State University, 935-942.","IlluMINEation project website www.illumineation-h2020.eu","Koperska, W., Stachowiak, M. Duda-Mróz, N. et al. The Tailings Storage Facility (TSF) stability monitoring system using advanced big data analytics on the example of the Żelazny Most Facility, Archives of Civil Engineering, 68(2).","Kurup, P. U., and Griffin E. P., Prediction of soil composition from CPT data using general regression neural network in Journal of Computing in Civil Engineering, 20(4), 2006, pp. 281-289","Robertson, P. K., Soil classification using the cone penetration test in Canadian Geotechnical Journal, 27(1), 1990, pp. 151-158","SEC4TD - Securing tailings dam infrastructure with an innovative monitoring system – project website sec4td.fbk.eu","Stefanek, P, Engels, J, Wrzosek, K, Sobiesak, P & Zalewski, M 2017, Surface tailings disposal at the Żelazny Most TSF, today and into the future, in A Wu & R Jewell (eds), Paste 2017: Proceedings of the 20th International Seminar on Paste and Thickened Tailings, University of Science and Technology Beijing, Beijing, pp. 213-225","Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., & Salakhutdinov, R. (2014). Dropout: a simple way to prevent neural networks from overfitting. The journal of machine learning research, 15(1), 1929-1958."]}

Approximately 30 million tons of tailings are being stored each year at the KGHMs Zelazny Most Tailings Storage Facility (TSF). Covering an area of almost 1.6 thousand hectares, and being surrounded by dams of a total length of 14 km and height of over 70 m in some areas, makes it the largest reservoir of post-flotation tailings in Europe and the second-largest in the world. With approximately 2900 monitoring instruments and measuring points surrounding the facility, Zelazny Most is a subject of round-the-clock monitoring, which for safety and economic reasons is crucial not only for the immediate surroundings of the facility but for the entire region. The monitoring network can be divided into four main groups: (a) geotechnical, consisting mostly of inclinometers and VW pore pressure transducers, (b) hydrological with piezometers and water level gauges, (c) geodetic survey with laser and GPS measurements, as well as surface and in-depth benchmarks, (d) seismic network, consisting primarily of accelerometer stations. Separately a variety of different chemical analyses areconducted, in parallel with spigotting processes and relief wells monitorin. This leads to a large amount of data that is difficult to analyze with conventional methods. In this article, we discuss a machine learning-driven approach which should improve the quality of the monitoring and maintenance of such facilities. Overview of the main algorithms developed to determine the stability parameters or classification of tailings are presented. The concepts described in this article will be further developed in the IlluMINEation project (H2020).