| The safe and intelligent operation of coal mining face is an important guarantee for efficient coal mining and personnel safety.However,there is still a certain gap between the current coal mining technology and the intelligent unmanned face.The coal-rock identification is the key technology to realize intelligent unmanned working face,and it is also the main difficult problem that restricts intelligent coal mining equipment.To solve the problem of coal-rock identification,domestic and foreign researchers have proposed many solutions,but there is still a lack of research on characteristics and universality of coal and rock.Therefore,an efficient,safe and highly applicable online identification method is urgently needed.Combined with the status of coal and rock identification research at home and abroad,we carried out the research on coal and rock identification method based on the difference of coal and rock mixing characteristics.The main research work is as follows:(1)Coal powder and rock powder are mixed in 39 different ratios and pressed into tablets,and THz signals of different samples were extracted by terahertz time-domain spectroscopy(THz-TDS).The characteristics of coal and rock mixture were studied by combining the fast Fourier transform(FFT),optical parameter extraction method and principal component analysis(PCA).Through characteristic analysis,it is found that with the change of coal content,the different characteristics of samples in the effective terahertz frequency band,such as amplitude,dielectric characteristics,absorption coefficient,etc.,have obvious change rule,which lays a theoretical foundation for the subsequent quantitative study of coal and rock.(2)Two kinds of optical properties of coal and rock mixed samples in the terahertz frequency band were used as input parameters of the quantitative model.Combined with back propagation neural network(BPNN),least squares support vector machine(LSSVM)and partial least squares(PLS),different coal and rock quantitative models were established respectively.The model prediction results show that the prediction effect of LSSVM quantitative model is the best among the three models,and when the input parameter is the refractive index,the prediction effect of any model is better than that based on the absorption coefficient as the input parameter.Therefore,the LSSVM quantitative detection model based on the refractive index as the input parameter has the best prediction effect,with correlation coefficient of 95.7% and root mean square error of 0.302.In addition,a theoretical model of coal-rock interface estimation is established for the shearer's mining face.Combined with the coal-rock quantitative detection model,the height of the roller embedded in the rock layer can be estimated by the theoretical model,which provides a reference for the application of coal-rock interface identification method.(3)Taking advantage of the near-zero terahertz absorption of high-density polyethylene(HDPE),HDPE is equivalent to air in the actual downhole sampling area,and the mixture of HDPE and coal/rock is used to simulate the suspension state of dust around the mining face.By comparing the simulation data with the experimental data,it is shown that there is light scattering phenomenon in the equivalent coal-HDPE tablet sample,but the scattering has little effect on THz signal.This conclusion confirms the feasibility of online identification of coal and rock.What's more,we prepared CoalHDPE and Rock-HDPE samples with different sparsity(the volume ratio of the medium to the sample),and uses the support vector machine(SVM)to classify the samples with different sparsity.The results show that when the sample sparsity is higher than 20%,the sample sparsity has almost no effect on the terahertz signal,that is,the sparsity in this range is more suitable for online coal and rock online identification.There are 42 figures,14 tables and 82 references in the thesis. |
PDF Full Text Request