Microseismicity-based method for the dynamic estimation of the potential rockburst scale during tunnel excavation

The severity and harmfulness of a rockburst event are significantly correlated with the scale of rock mass ejection, especially when the rock mass are not supported. This paper presents a microseismicity-based method for the early estimation of rockburst occurrence and its potential scale, which is graded according to the volume of the rockburst pit (Rv). The establishment of the estimation method involves a rockburst database, a grading scheme of the rockburst scale, selection and clustering analysis of rockburst samples, training of an artificial neural network (ANN) model, and dynamic updating. Firstly, a rockburst database is established from cases that were collected from the tunnels at depths of 1900-2525 m in the Jinping II hydropower station, located in southwest China. A grading scheme regarding the rockburst scale is preliminarily proposed on the basis of statistical analysis. Next, seventy-four rockburst cases, collected in tunnels with microseismic (MS) monitoring from October 2010 to March 2011, are selected as typical rockburst samples by using cluster analysis, and the relationships between the microseismicity and rockburst scale are deeply revealed. Then, three MS parameters, namely, the cumulative number of events, the cumulative energy, and the cumulative apparent volume, are determined and used together as input indicators for the identification of the rockburst scale. The estimation model is trained and cross-validated by the ANN optimized through genetic algorithm (GA). Finally, the performance of this microseismicity-based method has been validated by thirty-one cases that occurred in the tunnels with a cumulative length of 1.85 km, excavated from April 2011 to November 2011. The result indicates that approximately 83.9% of the rockburst cases could be reliably estimated. This study provides a new and feasible method for rockburst scale estimation, which can be used separately or applied as a complementary approach to current prediction methods for risk assessment and management of rockbursts in drill-and-blast tunneling.

Automated summary

This study proposed a method for early estimation of rockburst occurrence and scale based on microseismicity, using an established rockburst database, selection of typical rockburst samples, training of an artificial neural network (ANN) model, and dynamic updating. The method demonstrated reliable estimation of rockburst cases in drill-and-blast tunneling, providing a new approach for risk assessment and management.