Research On Dynamic Response Characteristics Of Coal Gangue Granular Impact And Identification Of Vibration Signals

Coal plays a dominant role in China's primary energy mining and consumption structure.The fully mechanized sublevel caving has become the first choice for thick seam mining in China for its high mining intensity and strong adaptability.At present,the degree of caving of the top coal is basically determined and controlled by manual visual inspection,which inevitably leads to the owing caving and over caving for a long time,how to control the optimal opening and closing time of caving mouth according to the degree of coal-drop is the key issue of realizing automation and intellectualization in fully mechanized caving mining.In fully mechanized caving mining,when coal or rock impacting the tail beam of hydraulic support,the tail beam shows different dynamic response from that under quasi-static load.In this thesis,the mechanism and difference of coal-gangue response are studied,the vibration response characteristics of coal-gangue are obtained.On this basis,the recognition of coal-gangue is realized,which provides a new idea for coal-gangue interface recognition based on tail beam vibration.This thesis is supported by the National Natural Science Foundation project "Research on the identification technology of top coal caving based on tail beam vibration signal".Firstly,the basic theoretical research on coal-gangue impacting metal plates is carried out,and establishing the contact dynamics model of elastic single coal-gangue particle impacting metal plates.Based on Hertz contact theory,a mass-spring-damping nonlinear contact oscillator model which considering gravity is established.the numerical solutions of contact force and dynamic response including displacement,velocity and acceleration are obtained by the fourt-fifth order Runge-Kutta with variable step size.Based on this,the response differences of elastic contact force and damping contact force under nonlinear and linear conditions are studied,and the effects of different contact stiffness and contact damping on vibration response are studied.The explicit dynamic analysis software Is-dyna is used to simulate the vibration conditions of coal and gangue impacting rigid metal plates vertically,the vibration response characteristics of rock are studied,and the simulation results are compared with the theoretical results,which verify the correctness of the simulation method.Secondly,the transient response simulation of single coal-gangue particle impacting elastic metal plate is simulated,and the influence of different impact parameters on vibration response are studied,including three working conditions:different impact energy with the same mass of rock particles,different contact area and different impact energy with different mass of rock particles.The results show that the dynamic response,contact force and energy interaction of coal gangue under different working conditions are quite different.The difference of impact parameters will also changes in contact stiffness and contact damping,resulting in differences in vibration response.Based on this,a vibration table simulation model is established.The rock material is based on the Johnson-Holmquist constitutive model,,and use piecewise linear plastic material model for metal plate of table.The response characteristics of the rock impact vibration table are studied and the vibration response of rock under different mass,impulse energy and shape are researched,including the dynamic response,energy transfer and dissipation response between rock and table vibration plate.The stress response of the selected element on the metal plate surface and the difference of damage and fracture of rock under high strain rate are also obtained.At the same time,the simulation of the multi coal-gangue particles impacting table are carried out to study the difference of rock breakage and the response characteristics of metal plate under different coal and gangue mixing ratios.The particle rock shock vibration test acquisition and analysis system is built.The acceleration and vibration signals generated by the particle coal gang impact table are monitored and monitored,including 1000 sets of coal and gangue signal samples,and then the signals are preprocessed to obtain the same signal length.The traditional time domain features of the signal are extracted,and the variance,peak-to-peak and kurtosis indicators with high sensitivity are selected as feature samples.The Hilbert-Huang Transform(HHT)is used to extract the time-frequency features of the vibration signal.The signal is decomposed by empirical mode decomposition(EMD)to obtain the intrinsic mode function(IMF),and then the Hibert transform is performed.The Hilbert spectrum and the Hilbert marginal spectrum are obtained.The results show that the coal-gangue signal also sho ws a large difference in the energy characteristics of the time spectrum,and the energy characteristics of the first six items in the IMF and the marginal spectrum in the IMF are selected as feature samples.In this thesis,the machine learning algorithm is used to realize the identification of coal gangue vibration signal.Two kinds of data samples are used as the database:one is the original acceleration signal sample with length 5000,and the other is composed of 10 time-frequency domain features extracted by signal processing.The algorithm of random forest,extreme gradient boosting(XGBoost)and support vector machine(SVM)is used to identify the coal gangue signal.Three models are integrated by stacking integrated learning algorithm,and the meta-classifier model is generated by logistic regression.Integrate learning to obtain the final coal shovel identification model,and use decision tree(DT),long short term memory neural network(LSTM),factoring machine(FM)as reference group,and find out that the rate of the identi fication of random forest,XGBoost,SVM and Stacking are higher than other recognition algorithms,and the stacking integration has the highest recognition accuracy,and the test set can reach 98.12%.Although the recognition effect using the original sample is higher than the domain feature sample,it will greatly increase the time consumption.The coal gangue identification of the domain feature sample can significantly improve the recognition efficiency by sacrificing a small amount of accuracy,and improve the recognition performance as a whole.In this thesis,theoretical analysis,numerical simulation,experimental acquisition,signal processing and machine learning recognition are used to study the dynamic response of coal-gangue particles during impact and the feasibility of using coal and rock vibration signals to realize the identification of coal-gangue interface.These studies can be used as the basic research of automatic and intelligent caving mining,and have important significance to promote the construction of digitized mine in China.