Prediction Method Of Coal-gas Outburst Based On Geological Data Mining And Information Fusion

Coal-gas outburst is a complicated gas geodynamic disaster caused by comprehensive effects,including ground stress,coal bed methane,and coal physical properties in coal mine,which gravely threatens the safety of coal production.Studies have shown that the distinct dangerous degree of coal-gas outburst in parts of and among diverse geological structure is controlled by geological condition.Therefore,predicting coal-gas outburst in different locations is particularly important.In this dissertation,coal-gas outburst prediction method based on geological data mining and information fusion is studied by means of field investigation,theoretical analysis,indoor experiment and field experiment.Firstly,the geological and geophysical prospecting data of mine area,coal mine and working face are collected.Then the characteristics of geological structure including folds and faults,coal seam thickness and tectonic coal development and distribution are investigated.The collected data are systematically collated,compared and analyzed synthetically,and the control effect of geological conditions on coal-gas outburst is summarized in the study area.At the same time,the geological data characteristics of coal-gas outburst in the study area are extracted by mathematical geology method.Based on the gray relational analysis and apriori correlation analysis model of data mining technology,the geological factors of coal-gas outburst in the study area are discussed,the geological data association rules of coal-gas outburst are mined,also the geological data and outburst strength and the geological data and outburst type association rules are analyzed.In addition,the algorithms on feature level and decision level of information fusion are studied.Then the algorithm on feature level is proposed by combining the third quantitative theory,genetic projection pursuit clustering and particle swarm optimization neural network.The algorithm on decision level is also proposed by using the confilct degree weight distribution rule.Then the algorithms of information fusion including genetic projection pursuit clustering,particle swarm optimization neural network and confilct degree weight distribution rule are simulated.Finally,according to the main controlling geological factors in the study area,the experimental surface is selected and the geological data of coal-gas outburst prediction are formed.The identification framework of coal-gas outburst prediction is developed,the evidence body of the third quantification theory,genetic projection pursuit clustering and particle swarm optimization neural network is also constructed.Therefore,the coal-gas outburst prediction risk level of the experimental working face is achieved by the decision fusion.Besides,the construction process,logical structure and function structure of the prediction system of coal-gas outburst are designed,the system software was developed,and the theoretical and technical support of prediction for coal-gas outburst is provided.The following four aspects are mainly studied in this dissertation:?1?Study on the control effect of geological condition on coal-gas outburstBased on the study of the tectonic setting and the characteristics of the geological structure of the coal mine,the influencing factors such as outburst strength,outburst type,geological structure,coal-gas outburst position and omen of one hundred and twenty three coal-gas outbursts in the study area are analyzed statistically.Then the structural characteristics of coal-gas outburst of Wu_9,10 and Ji₁₅ coal seam in the Pingdingshan No.8 coal mine are studied,such as fault structure,fold structure,coal seam thickness and change and coal seam inclination angle.At the same time,research is mainly focus on the faults of one hundred and nine faults in the Wu_9,10 coal seam and the faults of one hundred and five faults in the Ji₁₅ coal seams,including the fault characteristic,fault drop,the direction of the large and medium faults,the distribution of the trend and tendency of the small faults and the dip angle of the faults.The contour maps of coal seam thickness and coal seam inclination angle are drawn with the coal seam drilling data.Then the distribution of coal-gas outburst points in the study area was compared with the contour map.In addition,on the basis of observing the macroscopic characteristics of coal structure in the field,the structure of the coal body in the study area is classified.Through the laboratory measure of protodyakonov strength of coal?f?and initial speed of gas emission of the coal??P?of the fifty fresh coal samples collected in the study area,the distribution relationship between the parameters of coal-gas outburst?f and?P?and the coal failure type in the Wu_9,10 and Ji₁₅ coal seams.Then the two coal samples of Wu_9,10 and Ji₁₅ coal seams are collected for industrial analysis,low temperature liquid nitrogen adsorption experiment and gas pressure content calculation,also the effect of adsorption characteristics on coal-gas outburst is discussed.?2?Coal-gas outburst prediction geological data mining technologyBased on the study of geological conditions on the control of coal-gas outburst,the possibility of data mining technology applied to coal-gas outburst geological data mining is expounded,the structure and process of geological data mining is analyzed,and the preliminary geological data mining model is established.In order to combine the data mining technology and coal-gas outburst geological data in-depth,functional methods including association analysis,classification and prediction,cluster analysis,exception analysis and evolutionary analysis are analyzed.Geological factors of coal-gas outburst are preliminary determined,which includes geological structure and coal structure.The aspect of geological structure is coal seam depth,fault number,coal seam thickness variation coefficient,coal seam inclination angle change,soft coal seam stratification change and surrounding rock combination of a total of six factors.The aspect of coal structure is coal body failure type,coal seam thickness,soft coal seam thickness,coal seam inclination angle,crumpled coefficient,initial speed of gas emission of the coal and protodyakonov strength of coal.Then the geological structure data are characterized by gas geology and mathematical geology,which forms the forty six geological data samples of coal-gas outburst.In view of the complexity,irregularity and large volume of the geological data,the gray relational analysis is choosen to analyze the geological data of coal-gas outburst.The amount of outburst coal which reflects the intensity of coal-gas outburst is selected as the parent factor of the gray relational analysis,and the other as the sub factors.And the correlation coefficient and correlation degree of coal-gas outburst geological factors are obtained.In addition,on the basis of analyzing the effctitiveness measure index includes the support degree and confidence level,and the practicability measure index of the life degree of the geological data.The data mining process of the geological data of coal-gas outburst prediction is constructed.And the process of apriori associational rules of geological data is developed through the apriori associational rules.Since the data of samples should be nonnumerical,so thirteen attribute geological data are discretized.Then the associational rules of coal-gas outburst geological data is analyzed using the apriori model of SPSS Modeler software at forty and thirty percents of the support degree.At the same time the amount of outburst coal and the outburst type are separately selected as the succedent item,the associational rules of coal-gas outburst intensity are mining at twenty five percents of the support degree and seventy five percents of the confidence degree.Then the associational rules of pour out are mined at four percents of the support degree and ninty percents of the confidence degree,the associational rules of extrusion are mined at forty percents of the support degree and eighty percents of the confidence degree,and the associational rules of outburst are mined at ten percents of the support degree and sixy percents of the confidence degree.?3?Research on geological information fusion technology of coal-gas outburst predictionIn order to apply the information fusion technology to coal-gas outburst prediction more effectively,the basic principles of information fusion including data level,feature level and decision level are clarifies.Then the basic algorithm of probabilistic method,logical reasoning method and learning method are analyzed.Due to the qualitative and quantitative data of geological data,the third quantification theory algorithm which can deal with both qualitative and quantitative variables is introduced at the feature level of geological information fusion.At the same time the genetic algorithm is used to optimize the projection pursuit direction with the cluster method,the optimizing process is constructed,and the genetic projection pursuit cluster method is proposed on the feature level.Thus the control parameters of genetic algorithm are set up,and the algorithm simulation test before and after the optimizing is carried out using computer simulation method.Since the BP neural network is easy to fall into the local extremum,the BP neural network algorithm based on particle swarm optimization is proposed.Then the algorithm flow of particle swarm neural network is established.The parameters of particle swarm optimization and BP neural network are selected.Then the forty six groups of coal-gas outburst geological data and ten groups of non-outburst geological data are used as training samples.After the optimazion of BP neural network by the particle swarm optimization toobox of Matlab R2008a,the effectiveness of the particle swarm neural network is simulated with the test of training results,training times,maximum error,minimum error and average error.Besides,the Dempster-Shafer rule of evidence synthesis is prone to produce the result of conflict with the actual situation,so the conflict degree weighting evidence theory synthesis rule is proposed,and the rationality of synthesis rule is verified through the case analysis.?4?Coal-gas outburst prediction method field applicationIn order to verify the effectiveness of the prediction method of coal-gas outburst based on geological data mining and information fusion,the Wu_9,10-21030 tunnelling working face is selected as the experimental face,and the sixteen sets of geological data for the prediction of coal-gas outburst in the working face are developed according to the main geological factors in the study area and the geological conditions in the working face.In addition,based on the study of geological data mining and geological information fusion technology,a new process for the coal-gas outburst prediction including determination of identification frame,obtaination of evidence body and fusion on decision level is proposed.Thereinto the fusion results on feature level using the third quantification theory and genetic projection pursuit cluster method are disposed using membership degree and normalization,then the fusion results on feature level using particle swarm neural network are normalized.Therefore,the three evidence bodies are used as the input of decision level fusion,the results on decision level based on the conflict degree weighting evidence theory synthesis rule are obtained,and the comparison between the field prediction method,prediction methods,and the actual level is carried out.The result shows that the prediction method of coal-gas outburst based on geological data mining and information fusion is accurate and effective.In addition,the system requirement,system construction flow and system structure design by the method are discussed,the system construction process,system logical structure and system function structure are designed,the system software was developed.It provides technical support for the rapid and accurate prediction of coal-gas outburst.