Study On Coal-gas Outburst Risk Evaluation Based On Association Rules And Intelligent Algorithms

Coal-gas outburst is a kind of complex disaster,which often occurs in the underground of coal mine and has a great threat to the safety production of coal mine in China.From occurrence,development,until the end of coal-gas outburst,there is usually some regularity,in its essence,these regularities are determined by the natural gas and geological conditions in and around the disaster site.From this perspective,the gas geological condition also affects the distribution of coal-gas outburst hazard and risk level.The information contained in the gas geological conditions is huge,how to find out the main information from the huge gas geological information data base,and optimize the process of these major information.It is very important to get the magnitude of coal and gas outburst hazard reflected by the geological information,this can also improve the safety level of the mine.Through the use of field experiments,field investigations,research experiments and theoretical analysis and other research methods,this dissertation describes the association rules and intelligent information of coal-gas outburst risk assessment means.In the beginning,a series of works have been carried out including the survey of mining area,the collection of geophysical data and the study of gas geology.The mine geological structure appears fold,fracture and so on,tectonic coal has many characteristics.Gas geological factors affecting coal and gas outburst in the study area are summarized,when the collected data are integrated and analyzed.At the same time,the characteristics of gas geological data of coal-gas outburst in the study area are extracted by mathematical geological method.On the basis of analyzing and comparing the classical Apriori algorithm and FP-growth algorithm,a new algorithm for mining association rules is proposed,which is calculation method for Class-FP-growth.Based on this study,the association rules of gas and geological factors in the coal mine area and the association rules of outburst intensity in the study area are worked out.In addition,the information processing of coal-gas outburst in the intelligent algorithm is studied,the Harmony Search(HS)is used to to optimize BP neural network intelligent algorithm,also the Genetic Algorithm(GA)is used to optimize ELM intelligent algorithm.Then the HS-BP intelligent algorithm and GA-ELM intelligent algorithm of coal and gas outburst risk assessment method are proposed.Finally,chosing a working face in the study area to test,extracting gas geology data of the corresponding working face,forming the HS-BP and GA-ELM evaluation evidence body of the Bayes information fusion,a more effective coal-gas outburst risk assessment method is formed according to the Bayes fusion for the working face,and it provides a new technology support for coal-gas outburst risk assessmentThis dissertation mainly studies from the following five aspects:(1)Analysis of the influencing factors of coal-gas outburst in the study areaBased on the study of the tectonic background and geological characteristics in the study area,the outburst intensity,geological conditions and gas conditions of twenty coal-gas outbursts are statistically analyzed of the Sijiazhuang mine in the study area.Take Sijaizhuang and Xinjing coal mine as example,gas parameters including coal seam gas pressure,gas concentration and gas adsorption characteristics are discussed in the studied area.The parameters such as pore volume,specific surface area,median pore diameter,specific surface area median pore diameter and porosity,are tested by mercury injection experiment of five different coal types of xinjing coal mine,and its influence on gas adsorption characteristics are analyzed.At the same time,the four groups of coal structure parameters and twenty six groups of drilling data in the No.1 coal mine,No.2 coal mine,No.5 coal mine and Xinjing coal mine are used to analyze the coal seam buried depth,thickness of coal seam and coal structure on the influence of coal-gas outburst risk level.Then the controlling effect of different geological structures on coal and gas outburst is analyzed according to the coal seam floor contour and coal thickness contour data of the Xinjing coal mine.On this basis,the gas data and geological data of twenty coal-gas outburst in Sijiazhuang coal mine are formed,there are six factors including gas pressure,gas concentration,coal seam depth,coal seam thickness,coal body failure type and geological structure,also the quantification of geological structure is determined by comprehensively considering the combination of structures,number of faults and fault drop.(2)Study on association rules mining theory and algorithm of coal-gas outburstUnder the condition of expounding the feasibility of outburst risk assessment,the outburst factors are excavated by association rules.This dissertation compares the advantages and disadvantages of Apriori and FP-growth with the classical mining association rule algorithms.When the frequent item set data volume is large,Apriori algorithm not only produces a large number of candidate item sets,but also repeatedly scans the database,resulting in low running efficiency.Fp-growth algorithm does not generate candidate item sets,which reduces the computational complexity.However,each category attribute is built in FP-tree as a node,which will increase the spatial overhead when the amount of data to be mined is large.Aiming at the shortcomings of the two algorithms,a new association rule mining algorithm is proposed,the Class-FP-growth association rule mining algorithm,this algorithm not only overcomes the lack of Apriori algorithm when dealing with a large number of frequent itemsets database,and reduce the redundancy branches caused by using FP-growth algorithm.And according to this,coal-gas outburst risk assessment model for mining association rules is established based on the Class-FP-growth algorithm.Choosing a total of three algorithm database which are Webdosc,AuchoDocs and Accidents to carry out simulation test,the performance of FP-growth algorithm and the Class-FP-growth algorithm is compared with the minimum support of one over ten thousand,one over one thousand,one percent,ten percent,twenty percent,thirty percent,forty percent,fifty percent,sixty percent,seventy percent and eighty percent.It is proved that Class-FP-growth algorithm has certain advantages than FP-growth algorithm when mining large databases or in strict constraints,and the model is applied to the dangerous grade data analysis in Sijiazhuang coal mine of the study area,which mine has twenty times of coal-gas outburst.The association rules of coal-gas outburst evaluation factors with thirty percents support,and the association rules of outburst intensity are obtained.(3)Research on the intelligent algorithm of HS-BP neural network for coal-gas outburst risk assessmentOn the basis of study on structure and basic principle of BP neural network algorithm for coal-gas outburst risk assessment,the problem of long training time,easy training paralysis and falling into local minimum value of BP neural network is analyzed.Compared with traditional optimization algorithms,HS algorithm has advantages of fewer parameter settings,easy to share information,etc.In order to overcome the shortage of BP neural network,according to its own characteristics,the HS algorithm is used to improve BP neural network,so as to make it better.The original slow convergence is improved,so that it can meet the target requirement on its own local minimum problem.And three standard benchmark test functions including unimodal Sphere function,Rosenbrock function and multi-peak Griewank function is selected to test the algorithm performance of numerical optimization,so that to verify the correctness and effectiveness of the HS-BP algorithm,after two hundred independent experiments,the average value and average standard deviation of the optimal fitness function value of HS-BP algorithm are better than that of PSO-BP neural network algorithm,GA-BP neural network algorithm and BP neual network algorithm,and the HS-BP model has better prediction accuracy and generalization ability.At the same time,the HS-BP neural network algorithm is used to compare its target value,and the result of 0.000125 is more accurate than that of 0.0086,0.0072 and 0.0364 obtained by HS-BP,PSO-BP,GA-BP and BP to some extent.The results showed that BP network improved by the harmony algorithm has scientific nature and superiority to some extent.(4)GA-ELM intelligent algorithm for coal-gas outburst risk assessmentThe basic principle of extreme learning machine(ELM)algorithm is expounded first,it is compared with SLFN algorithm and BP algorithm,ELM algorithm has activation function and can use discontinuous function,it only needs to set the number of hidden neurons,by reducing the training time of the calculation method,the extreme value problem that SLFN is prone to gradient descent and situation can be avoided,with strong generalization ability.However,in the process of ELM algorithm implementation,it is easy for the input value matrix and the randomness of hidden layer deviation to affect ELM,therefore,genetic algorithm(GA)is introduced to optimize it,then a new method for evaluating outburst risk of coal-gas is proposed.In addition,GA-ELM algorithm,ELM algorithm and BP algorithm are respectively used to analyze the risk evaluation model of coal-gas outburst sample data in Sijiazhuang coal mine.Compare the evaluation results with each other,the result shows that the evaluation accuracy of ELM algorithm is sixty percent,that of BP algorithm is sixty percen,and that of GA-ELM algorithm is eighty percen,GA-ELM algorithm has high accuracy.Therefore,the model combined with genetic algorithm(GA)and extreme learning machine(ELM)has high evaluation accuracy and strong accuracy in the evaluation effect of coal-gas outburst.(5)Field application of coal-gas outburst risk assessment method based on association rules and intelligent algorithmAccording to the main controlling geological factors and the actual mine in the research area,fifteen data information samples of coal-gas outburst risk assessment are constructed,The HS-BP evidence body and GA-ELM evidence body of Bayes information fusion evaluation are formed,and a more effective evaluation method of coal-gas outburst risk in experimental working face is established.Based on the research of association rules of coal-gas outburst data and intelligent algorithm optimization technology,the evaluation process of coal-gas outburst risk is proposed.In the risk assessment of coal-gas outburst,the predicted results of HS-BP neural network algorithm,GA-ELM algorithm,association rules and intelligent computing methods and the actual test are compared.The result shows that the accuracy of coal-gas outburst risk evaluation based on association rules and intelligent algorithm optimization is higher,up to 93.33%,which provides a new technolog support for coal-gas outburst risk assessment.