Rockburst time warning method with blasting cycle as the unit based on microseismic information time series: a case study

摘要

Rockburst warning includes prediction of the position, intensity, and timing of potential rockburst. Rockburst time warningrefers to the prediction of the moment at or time period during which a rockburst may occur. Due to the complex rockburstmechanism and many influencing factors, several key difficult-to-solve scientific problems currently remain in rockbursttime warning research. In this article, microseismic (MS) monitoring is performed, and blasting is implemented as an iconicevent to study the warning method for rockburst with blasting cycle as the unit of time. Focusing on this research goal, a deeplearning method is applied to establish an MS information prediction model (MSIPM) and a rockburst time warning model(RBTWM) based on a long short-term memory network (LSTM). The MSIPM predicts the MS information for subsequentblasting cycles through the MS information time series of historical blasting cycles. The RBTWM predicts the potentialrockburst intensity and which blasting cycle a rockburst may occur through the MS information time series obtained byfusing MS information from historical and subsequent blasting cycles. The developed method is applied in a railway tunnelexcavated with the drilling and blasting method. The warning results of the test set demonstrate that the rockburst warningaccuracies for the first, second, and third subsequent blasting cycles are approximately 74.6%, 71.2%, and 63.1%, respectively.In addition, further application and verification are carried out in the construction of another new railway tunnel. Therockburst warning accuracy for the first subsequent blasting cycles is approximately 80.0%. The application results show thatthe MSIPM and RBTWM provide warnings regarding the immediate rockburst time in blasting cycle units. The combinationof MS monitoring and artificial intelligence represents a new idea for rockburst time warning.