Development of a procedure for predicting real-time seismic wave velocity in underground mines using discrete physical laboratory modelling and explainable artificial intelligence (XAI)
Samadi H. Suorineni F.
February 2026Elsevier Ltd
International Journal of Rock Mechanics and Mining Sciences
2026#198
Seismic event source locations in underground mines are crucial for safety, production efficiency, and mine profitability. Common microseismic monitoring systems rely on a constant input velocity model and periodic updating for seismic event source locations calculation. This approach often results in significant errors because of the constantly changing underground mine conditions from mining activities, discrete geological features, intersection of different rock lithologies, and the complex underground mine infrastructure, such as a changing network of excavations and the use of backfill. To address this problem, this study used discrete physical models mimicking snapshots in time to track the velocity changes of the constantly changing underground mining environment due to mining activities. The data generated in the laboratory is then analyzed using machine learning to develop a method for predicting input velocity in real-time that reflects the ground condition at any time, for use in the seismic event source determination algorithm, particularly SIMPLEX. The discrete physical models used to mimic the constantly changing underground mine conditions were concrete blocks (synthetic rocks) and granite in the form of cubes of diverse sizes and physical conditions. The SAEU3H Acoustic Emission system was used to generate and track the acoustic signals in various known locations around the blocks in their different physical conditions. One sensor served as the event source (pulse generator), while others served as receivers. Given the data generated from various mimicked ground conditions, smart computational techniques, including explainable artificial intelligence (XAI), were applied to predict wave velocity accurately for given ground conditions. The developed models showed a strong correlation between predicted and actual wave velocity values, particularly with the stacking-ensemble algorithm that gave the best performance (R2 = 0.97, NRMSE = 0.002, MAPE = 0.0001). The findings of the study have the potential to be implemented in seismic monitoring systems for real-time velocity prediction in underground mines to improve seismic event source location accuracy.
Acoustic emission , Discrete physical modelling , Explainable artificial intelligence , Microseismic source location accuracy , Real-time seismic wave velocity , Underground mining
Text of the article Перейти на текст статьи
School of Mining and Geosciences, Nazarbayev University, Astana, Kazakhstan
School of Mining and Geosciences
10 лет помогаем публиковать статьи Международный издатель
Книга Публикация научной статьи Волощук 2026 Book Publication of a scientific article 2026