Analysis Of High Sulfur Coal Property Based On Machine Vision

At present,the automation level of domestic coal preparation plants at home and abroad was basically in a state of single control,such as automatic launching and stopping machine,automatic dosing control,liquid automatic level control of medium density barrels.The real-time monitoring of raw coal and products property processes has not been realized,hence the whole coal preparation process cannot be controlled in time.In coal preparation plants,coal property information such as size composition,density composition and sulfur ash content is usually measured manually.With much experiment,complicated analysis process and huge consuming time,manual off-line measurement can not feedback the change of coal property information in time,it can not be directly used in the production process control and reduce the product qualification rate..The research group had done the coal property composition analysis of Taixi anthracitel based on machine vision,this paper applied this technolgy to the composition analysis of high-sulfur coal.Due to the huge appearance differences between high-sulfur bituminous coal and low-sulfur anthracite,this paper proposed new measures in terms of specific image analysis methods.Histogram equalization was used to enhance image contrast,smoothing filter was used to eliminate image noise,watershed algorithm based on tag control was used to segment coal particle areas,the minimum bounding rectangle method was used to intercept coal areas;The two-dimensional shape and size characteristic parameters of coal particles were extracted,the multiple regression method was used to construct the estimation model of coal particle thickness.The volume estimation model of the coal particle was obtained based on the area of the coal particle,and the relative error of the model was 9.7131%;Seven physical quantities related to particle size were compared,and their characterized correct rate was calculated,the best equivalent particle size named the minimum bounding rectangle width()was selected,its characterized correct rate was 87.08%;The estimation method of size composition suitable for Nantong high-sulfur coal was proposed by combining the best equivalent particle size with the volume estimation model.Forty-two kinds of superficial color and texture characteristic parameters of coal particles were extracted and processed by normalized treatment and outlier recognition,the boxplot was used to explore the variation tendency of characteristic parameters with coal particle size and density,and then selecting features initially.The genetic algorithm was used to screen and optimize the characteristic parameters further,and the BP neural network optimized by genetic algorithm was used to establish the estimation model of coal particle density level.Finally,the estimation method of density composition suitable for Nantong high-sulfur coal was proposed by combining the best equivalent particle size,volume and density level estimation models.The results of the study indicated that the sulfur content of coal samples for testing was mainly pyrite,but pyrite could not be discerned in the image of coal particles because of finer dissemination size.At the same time,it was found that there was positive correlation between sulfur content and density,so the characteristic parameters used for predicting density were selected as the characteristic parameters of sulfur content estimation.They were processed by normalized treatment and outlier recognition,the boxplot was used to explore the variation tendency of characteristic parameters with coal particle sulfur content,and then selecting features initially.The genetic algorithm was used to screen and optimize the characteristic parameters further,and the BP neural network optimized by genetic algorithm was used to establish the estimation model of coal particle sulfur content and particle group sulfur content.Finally,an estimation method of coal pile sulfur content was proposed by combining the best equivalent particle size,volume,density with particle group sulfur content estimation models.