Probabilistic and statistical approach to accident forecasting in underground mines

This article addresses the pressing challenge of forecasting geomechanical accidents in underground mines. A key issue is the integrated analysis of heterogeneous data (geodetic measurements, expert assessments, statistical data) obtained from existing monitoring systems. As a solution, a method is proposed for the data processing unification based on converting the data into time series followed by the calculation of a universal risk indicator– the probability of a critical value of a control parameter within a specified time interval. The novelty of the work lies in the development of an integrated probabilistic–statistical approach that enables the comparison and aggregation of risks from diverse processes within a single risk-oriented model. The article presents mathematical forecasting models for different data types, a methodology for transitioning to probabilistic assessment, and demonstration calculations using real-world examples. It is shown that this approach establishes a foundation for building comprehensive decision support systems in the mining industry. The advantages of the approach include: its universality—the approach is applicable with data of any physical nature recorded in time; compatibility—comparison of risks of different processes; proactivity—the approach ensures a qualitative basis for the proactive decision-making. Regarding the main limitations and introduction issues, the quality of a forecast drastically depends on the correctness of a selected model and on the data representativeness; the critical values are determined in specific geological conditions and can be subjective. The promising trends of the further research are: more complex models of time series; machine learning for automated detection of trends and anomalies; development of methods for aggregating partial probabilities Pfail into an integrated risk criterion of a whole object.