Study On The Application Of Multi-source Information Fusion Technology In The Coal And Gas Outburst Prediction

As one of the serious geologic disasters threatening safety production in coal mine,coal and gas outburst restricts the sustainable development of mining industry.However,because the mechanism of coal and gas outburst has not revealed completely,there is no effective method to totally eradicate the coal and gas outburst.Therefore,deeply understanding the regulation and feature of gas disaster,and improving the accuracy of coal and gas outburst prediction is a prerequisite for coal disaster prevention and control,efficiently decreasing the occurrence of accident and reducing the casualties.Since collecting mine gas information is always disturbed by environment and affected by noise,it is necessary to fusion the multi-source information from mine to increase the accuracy of prediction.This thesis analyses the mechanism of coal and gas outburst and model the prediction of multi-source information fusion of coal and gas outburst based on Least Square Support Vector Machine and D-S Evidential Theory,to the question of coal and gas outburst,the uncertain and nonlinear geological dynamic phenomenon.Firstly,on the basis of analyzing the feature of coal and gas outburst in research area,this thesis collects historical outburst data in diggings,and researches factors influencing coal and gas outburst.After the acquiring of basic data,this thesis sorts the degree of association of indicators applying Grey Relational Analysis,and selects main controlling factors closely relating to coal and gas outburst.In the next process,in consideration of the particular advantage Support Vector Machine,a classification learning algorithm,in data processing in small sample,this thesis applies it to predict coal and gas outburst preliminary.Because the parameters of Support Vector Machine affects its classification performance,while Particle Swarm Optimization is an optimization algorithm based on swarm intelligence simulating birds preying and optimizes by iteration.Hence,this thesis presents an optimization method based on sub-dimension Particle Swarm Optimization for Least Square Support Vector Machine.Experimental result indicates optimized model has a high prediction accuracy and improves the multi-source information fusion.To solve the problem of inherent defect of Least Square Support Vector Machine,this thesis researches D-S Evidential Theory,having the ability of handling and fusing the uncertainty of multi-source information,to increase the reliability of prediction.After this,this thesis researches deeply D-S Evidential Theory and summarizes the main existing problems in traditional D-S Evidential Theory.For the conflict among evidences occurring in D-S combination rules,this thesis refers the conception of evidence distance,and presents a modifiedcombination rules,modifying the source of evidence according to the confidence level and support degree of evidences.To test the validity of modified combination rules in dealing with high conflicting evidences,this thesis performs a numerical experiment.At the end of this thesis,Least Square Support Vector Machine,single indicator method and aggregative indicator method are combined to input the prediction of D-S Evidential Theory.After that,5 coal and gas data are chosen to test the feasibility and effectiveness of the model of predicting coal and gas outburst.