Study On Rockburst Characteristics Of Acoustic Signal And Dynamic Warning Method

Rockburst is a common geological hazard during excavation of deep rock mass.At present,with the vigorous development of underground engineering construction in China,the occurrence of rockburst is increasing in intensity and frequency,which poses a great threat to the safety of deep underground engineering.Due to the high complexity and influence factors of rockburst,the current prediction and early warning level of rockburst is not enough to meet the requirements of engineering practice,and the early warning method of rock burst needs to be further developed.Before the rockburst,there were often phenomena such as particle ejection or rock plate splitting accompanied by obvious sound signals.This is a potential important information that could be beneficial to rock burst warnings,but related research is currently weak.The study of sound characteristics and dynamic early warning method of rock explosion process is of great practical significance and academic value for improving the early-warning level of rock burst and guaranteeing the safety of deep underground engineering.The research content and related conclusions of this paper are as follows:1.Acquisition and pretreatment of sound signals during rockburst.The rockburst process is reproduced indoors using the true triaxial rock burst test system.The rockburst sound signal is collected by sound monitoring equipment and the sound signal is preprocessed.In view of the diversity of acoustic noise in the field,the acoustic signal of common materials in rock burst is analyzed.The sound of the indoor rockburst process is combined with the common material failure noise at the rockburst site to simulate the complex sound of the rockburst.Using wavelet denoising to denoise the synthesized sound,the process of de-noising the rockburst's live sound is simulated,and the "clean"rockburst sound signal is obtained,which provids a theoretical basis for the sound signal analysis of rockburst typical damage phenomena.2.Study on sound signal characteristics of rockburst process.The sound signals of three kinds of damage phenomena,including particle ejection,splitting of rock plate,and rapid film ejection,are analyzed during the detonation process of rockburst.Through the analysis of waveforms,spectrums,voiceprints,melphatic cep strum coefficients(MFCC),short-term average zero-crossing rates,spectral centroids,and short-term energy characteristics of three types of destruction sound signals,three types of sounds of typical damage phenomena were discovered,which have significant differences in signal duration,waveform shape,frequency,energy,and amplitude.This difference is enough to quickly distinguish the three types of rockburst damage,which lays the foundation for using this difference to identify the rockburst damage phenomenon and establish the intelligent identification of the rockburst damage phenomenon based on sound signal.3.Intelligent identification of rockburst process failure phenomena based on sound signals.Using a Gaussian process machine learning method suitable for dealing with small samples and nonlinear classification problems,MFCC,spectral centroid,short-term average zero-crossing rate,etc.are used as feature vectors to establish a destructive phenomenon of intelligently identified Gaussian process machine learning model based on the destruction of rockbursts before the occurrence of rockbursts.The GPC model identification procedure for the rockburst failure phenomenon is given,and the quantification and automatic identification of typical rock brittle failure phenomena such as particle ejection before rockburst and rock plate splitting are realized.The results show that the model has high accuracy in identifying typical failure phenomena before rockburst.This model overcomes the problems of large error and low reliability in human experience recognition,which is easy to implement and not easily affected by noise.It provides a scientific basis for establishing a rockburst warning method based on intelligent recognition of damage phenomena.4.Study on dynamic early warning method of rockburst based on sound signal.Using the "intelligent recognition + trend prediction"dynamic identification strategy,based on the intelligent identification of rockburst process sound signal GPC model for intelligent destruction of rockburst process typical damage phenomenon,and taking the changes of sound characteristics such as the quiet period,harmonic mean value and chromatogram vector average on the eve of rock burst as rockburst precursor information,a multi-level progressive rockburst dynamic warning method is proposed.It gives a multi-level warning standard for rock bursts.This method not only overcomes the traditional prediction method of rockburst failure that not directly predict whether the rockburst eventually occurs.It also solved the problem that the single trend forecasting method was only applicable to the issue of early warning on the eve of the occurrence of rock bursts,but the time for evacuation of personnel and important equipment was insufficient.It solves the limitation that the current state of development in the process of rock blasting cannot be dynamically identified and is not conducive to the timely release of early warning information on the occurrence of rock bursts.Compared with the traditional method,the dynamic early warning method of rockburst can improve the accuracy and reliability of rockburst warning.5.Research on quantitative evaluation index of rockburst.Traditional rockburst intensity evaluation indicators based on sound signals are mainly identified by human ear and qualitatively described by experience,rockburst sound strength is weak,reliability is not high,and sound signal amplitude "peak clipping" caused by overvoltage protection of sound signal monitoring equipment itself.Through the analysis of short-term energy and kinetic energy of rock burst,a new index can be proposed as a quantitative evaluation of rock burst,That is,local acoustic total energy(TELS).It reflects the release characteristics of a piece of eclipse energy with a certain duration,and overcomes the fact that the traditional indicators such as the maximum amplitude of the waveform,short-term energy,and maximum decibel can only reflect the instantaneous strength of the rockburst sound in a very short time,and cannot reasonably reflect the cumulative energy of a sound process that has a certain duration as the rock burst occurs.It provides a new way for rationalizing the evaluation of rock burst intensity.