Research On Monitoring Method Of Dynamic Coal Gangue Composition Based On Near Infrared Spectroscopy

Coal is a non-renewable energy with dual attributes of energy and industrial raw materials,and it plays an indispensable role in the energy field.In coal exploration,mining,washing and utilization,the quality of coal needs to be monitored.Near-infrared spectroscopy is a technology that combines near-infrared spectroscopy and chemometrics to achieve material composition analysis.Spectroscopy technology can realize the monitoring of coal gangue composition.At present,the analysis of coal and gangue based on near-infrared spectroscopy is mostly based on certain mining area or through analysis to select a certain particle size and speed.In actual use,most of them are multiple samples and multiple particle sizes.Coal gangue samples from multiple places exist same time and same coal gangue samples from different producing areas will have differences due to environmental reasons.These differences will be reflected in the near-infrared spectra of the samples,making their spectral characteristics different,and then affecting The stability and accuracy of the built model,in addition,the speed of different coal gangue conveyors are also different,there will be situations where the built model is not universal,and the model needs to be replaced.In order to study the prediction of coal gangue ash content based on near-infrared spectroscopy,a spectrum integrated probe independently developed by the laboratory was used to build an online near-infrared spectroscopy collection system.Three types of coal(gas coal,lignite,anthracite)and three types of gangue(fine sandstone,Carbonaceous shale,gangue)is used as the research object,the PLS multivariate analysis algorithm is used to establish the coal gangue ash prediction model.In order to make the final model more general,the same sample is collected from Shandong,Henan,Shaanxi and other places,analyzed and combined The influence of particle size and speed difference on the spectrum is corrected,and a general model for predicting coal gangue ash content under dynamic conditions is constructed.(1)Analyze the absorption position,curve slope,absorption valley depth and slope of the coal/gangue diffuse reflectance spectrum curve under different particle sizes,and obtain the effect of particle size on the diffuse reflectance spectrum of the sample;for he effect of particle size on the spectrum curve,convolution smoothing and multiple scattering are used Correction,baseline correction and other preprocessing methods are used for correction,and the performance parameters of the model,such as the coefficient of determination,root mean square error,etc.,are determined by comparing and calculating the various methods used to determine the preprocessing algorithm;the basis for determining the preprocessing algorithm In the above,use continuous projection,competitive adaptive re-weighted sampling and other wavelength screening algorithms,use partial least squares regression analysis algorithm for modeling,and determine the characteristic wavelength screening algorithm based on the processing results;finally,establish a multi-granular coal sample ash prediction model,which R~2_cand R~2_pare 94%and 95%,respectively,RMSEC and RMSEP are both 0.06;a multi-granular gangue lime content prediction model is established,and the R~2_cand R~2_pare respectively 91%and 87%,and RMSEC and RMSEP are 0.02 and 0.04 respectively.(2)Analyze the absorption position,curve slope,absorption valley depth and slope of the coal/gangue diffuse reflectance spectrum curve at different speeds,and obtain the influence of speed on the sample diffuse reflectance spectrum;for the influence of speed on the spectrum curve,convolution smoothing and multiple scattering are used Correction and other preprocessing methods are used for correction,and the performance parameters of the model,such as the coefficient of determination,root mean square error,etc.,are determined by comparing the calculation and analysis of the various methods used.The preprocessing algorithm is determined on the basis of determining the preprocessing algorithm.Wavelength screening algorithms such as continuous projection and competitive adaptive re-weighted sampling are modeled by partial least squares regression analysis algorithm,and the characteristic wavelength screening algorithm is determined according to the processing results;finally,a multi-speed coal sample ash prediction model is established,with R~2_cand R_p~2Respectively 90%,91%,RMSEC,RMSEP were 0.06,0.07;Establish a multi-velocity gangue lime content prediction model,the R_c~2,R~2_pwere 99%,92%,RMSEC,RMSEP were 0.006,0.02(3)Based on the above influence of particle size and speed on the diffuse reflectance spectrum of coal/gangue,the determined preprocessing and characteristic wavelength screening algorithms for correcting the particle size and speed respectively are adopted,and the PLSR algorithm is used for modeling,and the prediction effect of the calculation model is compared to establish The general model of coal gangue ash prediction has RMSEP and RMSEC of 0.1,and R_c~2 and R_p~2 of 95%and 87%respectively.Using external data for model verification,it is concluded that the verified RMSEP and R_p~2 are 0.11 and 87%respectively.