Abstract: Rockbursts represent a formidable geological disaster in deep underground engineering endeavors, and their precise prediction is of paramount importance for safeguarding the security of deep mining activities. To overcome the constraints inherent in conventional acoustic emission (AE) monitoring, which assumes that a solitary AE signal corresponds to an individual fracture incident—a premise that often results in inaccuracies in parameter calculation, misidentification of fracture mode, and concealment of pivotal precursor characteristics of rock fracture—this study introduces an innovative methodology for discerning multiple fracture events grounded in AE spectral analysis. Rockburst simulation experiments were conducted on granite specimens to amass AE signal data. A spectral analysis model was subsequently formulated, employing short-time Fourier transform (STFT) alongside valley ray demarcation (VRD) techniques to identify multiple fracture events and execute parameter computations. The findings reveal that this approach effectively distinguishes multiple fracture events within a singular AE signal and precisely captures the precursor features of rockbursts. During the transitional periods of the rockburst precursor phase, low-frequency, high-energy events with relatively elevated absolute energy magnitudes were discerned; concurrently, critical features marked by sporadic mid-frequency signals and low-frequency, low-energy signals were observed preceding rock failure, thereby offering novel perspectives for rockburst prediction.

Key words: rockburst, acoustic emission (AE), spectral analysis, multiple fracture sub-events, precursor features