| Earthquake is one of the most serious geological hazards,which often causes huge economic losses.Its conception,development and outbreak are often accompanied by structural damage of the geological body,mainly in the form of rock rupture and frictional sliding of faults,etc.At the same time,it generates precursor signals such as topographic changes,magnetic field anomalies,seismic sound,groundwater anomalies,atmospheric anomalies and animal activity anomalies.Seismic sound waves have been of great research value as earthquake precursor signals.Based on this,this paper picks up the spectral characteristics of the acoustic information generated by the rock damage process through indoor tests,and uses data mining means to study the acoustic information and characterize the damage state of the rock structure,aiming to provide reference for the study of earthquake precursor information.In this paper,we design an acoustic test system for rock shear damage based on the doctrine of elastic jump back,conduct indoor simulation tests,apply deep learning algorithms for acoustic information pattern recognition,and study the acoustic information characteristics of rock shear damage.Firstly,an indoor rock shear damage acoustic test device is designed to simulate the rock damage process of elastic jump back.Secondly,sandstone was used as the test material,and 18 sandstone specimens of different sizes were made according to the test protocol.In the experiment,the acoustic information of the rock damage process was collected using acoustic emission device and microphone to analyze the acoustic characteristics of the specimen fracture-shear damage test.Then,using audio processing means,the spectral characteristics of the acoustic signals generated by rock damage are analyzed,and a prediction model is constructed based on deep learning algorithm to establish an intelligent model of the acoustic signals of rock folding and shearing damage.Finally,combining with the measured ground acoustic monitoring data,an automatic recognition model of ground acoustic signal based on convolutional neural network(CNN)is established to improve the intelligent recognition of ground acoustic signal.The results show that(1)the elastic wave propagation velocity of sandstone specimens with shear strength in the range of4.870 MPa-13.003 MPa is 5000-5573 m/s,and the number of acoustic emission events is the largest when the loading rate is 0.05 MPa/s.(2)During the folding shear damage test,the acoustic emission events gradually increased with the increase of force,the fracture breeding,expansion and penetration inside the rock structure,and the events were mainly generated in the fracture surface and the bottom tensile zone.(3)The acoustic signal of rock folding shear damage can be divided into noise signal,micro-rupture signal and damage friction,and their audio signals have obvious differences,specifically: the waveform fluctuation of noise signal is small;micro-rupture signal will appear concentrated fluctuation area,the duration of about0.01s;the damage friction section fluctuates sharply with a duration of about 0.3s.(4)In the folding shear damage test,the acoustic signal spectrum center of mass of the sandstone specimen is between 1024Hz-2048Hz;the energy center of mass frequency can be as high as4096 Hz when the specimen is micro-rupture and damage,and the highest value of the highest spectrum center of mass is the same for both.(5)According to the spectral characteristics of the acoustic information of rock folding and shearing damage,the deep learning algorithm is used to construct a model to identify the measured acoustic information intelligently,and its accuracy rate is relatively satisfactory.(6)Analyzing the measured geoacoustic monitoring signals,the convolutional neural network algorithm model is constructed to identify the abnormal values of geoacoustic monitoring signals effectively.The research results of this paper are of practical significance for acoustic information mining of sandstone damage process,which can further use acoustic signals to study the damage stages of rock masses and provide reference for predicting earthquakes. |
