A Novel Dynamic Recognition Method of Rock Burst Precursor Information Based on Adaptive Denoising and Object Detection

Abstract

Acoustic emission (AE) and electromagnetic radiation (EMR) can reflect the precursor information of rock burst and play important roles in rock burst monitoring, early warning, and prevention. However, the existing denoising methods of AE and EMR monitoring signals are poor, and the recognition of precursor information lacks comprehensiveness, accuracy, and real-time. This paper presents a novel method combining adaptive denoising and object detection to realize dynamic recognition of rock burst precursor information. Successive Variational Mode Decomposition (SVMD) adaptively decomposed the AE and EMR monitoring signals such as pulse and intensity into different mode components and Kalman Filter (KF) performed on each mode component to eliminate redundant noise. Furthermore, the YOLOX object detection algorithm recognizes the precursor information in the time–frequency domain after noise removal, including the time interval, frequency band, and energy. The case study illustrates that the precursor response of the AE and EMR monitoring signal in time–frequency domain is highlighted by denoising, and the average accuracy of different types of precursor recognition reaches 96%. Finally, the consistency of the identified precursor information and field records shows the feasibility and effectiveness of the method, which has practical guiding significance for improving the level of rock burst prevention.