关键词: 卸荷速率, 非稳定阶段识别, 随机森林模型, 声发射监测

Abstract: Accurately identifying precursor information of rockbursts is a key issue in rockburst monitoring and early warning. Engineering practice has shown that rockburst disasters progress from static incubation to dynamic failure, with the unstable stage being critical for rock mass instability and mutation leading to rockburst occurrence. Studying this stage is crucial for understanding the essence of precursor information of rockbursts. The studies on the acoustic emission characteristics of diorite at different unloading rates and the identification of unstable stages are presented. The characteristics of amplitude, ringing count, frequency, etc. are analyzed, and the random forest (RF) model is used to identify the unstable stages of parameters such as amplitude, ringing count, and energy. The research results indicate that the stress-amplitude-ringing count-time relationship shows a quiet period, during which large, medium, and small values appear simultaneously. The initial frequency, reverberation frequency, cumulative initial frequency, and cumulative reverberation frequency increase sharply after the quiet period . These values are widely distributed and can serve as precursor information for rockbursts. This entire stage is unstable. The RF model can effectively identify the unstable stage of acoustic emission with a minimum accuracy of 81.2%. Conducting rock mechanics experiments to calibrate acoustic emission parameters and identifying unstable stages in a specific gold mine using the RF model is crucial for monitoring and early warning of rockburst disasters.

Key words: unloading rate, unstable stage identification, random forest model, acoustic emission monitoring