Research On Acoustic Emission Signal Characteristics And Prediction Model Of Rockburst Tendenc

With the demand for deeper mineral mining,tunnel and hydropower station construction,rock bursts occur more and more frequently in deep underground engineering.As a phenomenon of severe rock ejection dam age,rock bursts have significant hidden dangers for the safety of engineering construction.Rock burst prediction has become an important problem that needs to be solved in deep underground engineering construction.Therefore,this article conducts resear ch on rock burst prediction based on acoustic emission technology and machine learning algorithms.Acoustic emission technology can effectively reflect the internal damage and energy conversion laws of rocks.By utilizing the characteristic parameters of a coustic emission during rock fracture and instability,it can provide theoretical basis for early warning and precursor information of deep underground engineering rock bursts;Establish a self-organizing feature mapping neural network(SOFM)algorithm to predict the rock burst level of actual engineering cases.The combination of acoustic emission(AE)technology and SOFM algorithm provides a comprehensive theoretical basis for rock burst warning and prediction.Given this,the main research content and co nclusions are as follows:(1)The laboratory uniaxial compression and Brazilian splitting experiments of marble,red sandstone and dolomite were carried out.The strength brittleness index,the improved brittleness index BIM coefficient,the maximum storage elastic strain energy,and the elastic energy index are used to determine and comprehensively determine the rockburst tendency of rocks.The results show that marble has weak rockburst tendency,red sandstone has medium rockburst tendency,and dolomite has strong rockburst tendency.(2)Uniaxial compression and acoustic emission tests were conducted,and it was found that the AE time-domain characteristic parameters have a certain correlation with the rock burst tendency of different rocks.The AE event rate of marble with weak rockburst tendency rises gently to the maximum,and the AE phenomenon is most obvious in the fracture compaction stage;There are few AE events in the elastic stage of the red sandstone with medium rock burst;The AE events of strong rock burst dolomite during the crack compaction stage are weak,while during the full compression process,the internal crack propagation and penetration phenomenon of dolomite is obvious.The AE events of medium rock burst red sandstone and strong rock burst dolomite show alternating upward and downward phenomena before the rock burst.The variation trend of AE cumulative energy in rocks with different rockburst tendencies is different during the stable and growth stages.The cumulative energy change curve of AE during the loading process of rocks shows a trend of "slow pull-up,transition rise,and rapid growth".(3)Analyze AE parameters and study the characteristics of acoustic emission parameters under different rock burst tendencies.The study found that the cumulative energy of AE fitted with time conforms to the Boltzmann function and has a strong correlation.As the rock burst level increases,the A value(maximum AE cumulative energy)gradually increases,and the energy storage performance increases＿The T value(ratio of damage parameter to total uniaxial compression time)increases,the earlier the energy storage time is,the smaller the C value(time required for AE cumulative energy to suddenly increase to peak),and the shorter the rock burst time.There is a significa nt correlation between the raw data of RMS-ASL(effective value voltage average electrical signal)and the fitted curve,and it conforms to the Bradley curve model.The variation pattern of constants a and b decreases with the increase of rock burst level.(4)Conduct fractal feature research on AE feature parameters.The research shows that when the dimension of phase space m=4,it is the most suitable embedding dimension,and each rock type has obvious fractal characteristics in different uniaxial loading stress intervals.The higher the rockburst grade is,the smoother the curve between the Correlation dimension D value and different stress intervals will be.On the contrary,the lower the compactness is,the greater the rise and fall of the curve will be.(5)The macro state of rock after failure is sketched,and the cracks in the process of rock compression are classified by using the acoustic emission parameters RA and AF values.The results show that,with the increase of rockburst tendency,most of th e weak rockburst marble’s RA value density distribution is small,and the AF value is relatively large;The distribution of RA value density and AF value of red sandstone with medium rockburst tendency is relatively average,which shows the joint action of tensile failure and shear failure;The RA value distribution density of dolomite with strong rock burst tendency is relatively high,and the AF distribution density is relatively high,indicating that its failure mode is mainly RA-AF mixed failure.(6)Establish SOFM models under three different competitive layer topologies,and use variance and neural clustering effects to select the optimal SOFM model.The results indicate that the competitive layer is the number of neurons of 4×4 is the optimal model for predicting rock burst.Nine representative algorithms,including BP,Bayes,KNN,discriminant analysis classifier,Decision Tree,LDA,SVM,RBF,and Random forest,were introduced for comparison and validation.The results showed that the SOFM algorithm performs better and has the highest accuracy in hierarchical and overall prediction.To further verify the engineering applicability of the established SOFM model,domestic and foreign engineering cases were introduced for prediction analysis.The results showed that the algorithm proposed in this paper is superior to the Russell criterion,fuzzy comprehensive evaluation method,and grey whitening weight function clustering method.(7)Based on the rockburst grades of marble,red sandstone and dolomite comprehensively determined by the test in this paper,the prediction results are completely consistent with the rockburst grades comprehensively determined by the test using SOFM algorithm,which indicates that the accuracy of SOFM model in rockburst prediction is high.