Research Of The Risk Assessment Of Water Inrush From Coal Seam Roof And Floor And Prediction Of The Dewatering Rate

China is the most coal producing country in the world,coal plays an important role in the national economy,it occupies a leading position in the energy structure of our country,and the exploitation and utilization of coal resources is the realistic demand of the national economic development of our country.Because of the relatively complex geological structures and hydrogeological conditions in China,the problem of water inrush is one of the major hazards that seriously affect the safety mining of coal resources in China.According to the statistics,water inrush is the second-biggest disaster in the heavy and especially big accidents in coal mines only after gas explosion,meanwhile,it is characterized by serious loss of casualties,difficulties in rescue and long period of recovery.The coal seam roof and floor water inrush are the most important form of coal mine water disaster,therefore,the comprehensive prediction and evaluation of the risk of water inrush from seam roof and floor provides theoretical guidance and technical support for the work of mine water control,it is of great practical significance for the effective control of mine water disaster and the mining safety in China.Water inrush from coal seam floor is a nonlinear dynamic phenomenon which is controlled by many factors and has complex formation mechanism.In order to reveal the complex relationships between the main controlling factors and the influences on floor water inrush,two kinds of evaluation models named variable weight unascertained evaluation model and stochastic forest evaluation model for floor water inrush are established by using modern information and mathematical technology,both of them are based on the multi-source information fusion technology of GIS.Compared with other mathematical evaluation methods,unascertained measure theory uses confidence recognition criteria,it is more reasonable than the maximal membership criterion which is frequently used in measure function and could improve the reliability and accuracy of the evaluation results.The theory realizes the"orderliness"of the segmentation of the evaluation space,satisfies the principle of"additivity"and"polarity",it makes the evaluation results more credible.In the process of determining the weight,the variable weight theory is introduced,which fully considers the control effect of a variety of main control factors on the seam floor water inrush under different combined state levels.The stochastic forest model can handle high dimensional data,it need not make feature selection,and has strong generalization ability and could balance the error,especially for the unbalanced classification sample,it can still keep the high accuracy and overcome the shortcomings of the slow convergence speed and local minimum which are universally exist in other machine learning algorithms.Finally,the two models are applied to the evaluation of seam floor water inrush of Datong Tashan Coal Mine and Panjiayao Coal Mine respectively.It can effectively improve the prediction accuracy of seam floor water inrush,lay a solid theoretical foundation for the evaluation of the risk of seam floor water inrush,and improve and perfect evaluation method of water inrush from seam floor.With the increase of mining scale and mining intensity,the threat of seam roof water disaster in the western part of North China type coalfield and Early and Middle Jurassic coal beds of northwest of China is becoming more and more serious,which seriously affects the normal production of the mine,and brings serious potential dangers.Based on the method of"three maps",water inrush risk of No.2 seam of Hongliu Mine which belongs to Ningxia Coal Mine Group is evaluated.On this basis,the water gushing amount of the aquifer of the No.2 seam roof is predicted.In the process of establishing the groundwater numerical simulation model which is used to predict water inflow,the artificial test and error method is commonly used as the parameter estimation method,however,due to the shortcomings of time consuming and the subjectivity,it is difficult to give a real quantitative standard,and could not answer whether there is a better combination of parameters.In order to solve this problem,the Bayesian method is presented to analyze and optimize the uncertainty of osmotic coefficient in the numerical simulation of the water flow in the roof aquifer.As a probability analysis method,the Bayesian method could make the best use of all data and known information,and consider the uncertainty of the model,parameter,model input and observation data,by using the likelihood function,the input parameters can be modified by the observation data,meanwhile,the final parameter distribution is consistent with the actual data,and the parameters value probability distribution can be defined in the domain.That is more likely to get the global optimal solution of parameters,and it could obtain more accurate evaluation for the uncertainty of numerical simulation.numerical model is established with the optimized parameters,and water gushing amount of the No.2 seam roof is predicted,a reasonable prediction is obtained.Specific research as follows:(1)Combining unascertained measure theory with variable weight model,an evaluation model of water inrush risk of seam floor based on variable weight and unascertained measure is proposed and is applied to the predict and evaluate risk of floor water inrush of No.8 seam in Tashan Coal Mine,Datong coal field.The eight main controlling factors of water inrush in seam floor are constructed according to the three aspects characteristics of study area,including water filled aquifer of seam floor,outburst prevention performance of water resisting rock section between seam floor and underlying aquifer roof,geological structure.They are water pressure of aquifers,water abundance of aquifer,effective thickness of water-resisting layer,brittle rock thickness,fault and fold distribution,fault and fold intersection point distribution,collapse column distribution and fault scale index.The variable weight model is established,and constant weights of main control factors are calculated by the analytic hierarchy process,a partition state variable weight vector is constructed,the variable weight threshold is determined,finally,the weight of variable weight for each main control factor is fixed.The evaluation space of each evaluation index is constructed,the risk grade of each evaluation index is divided,the single index unascertained measure function is established,the multi index measure of all factors are obtained by calculation of variable weights for each index.Finally,the confidence level recognition criterion is adopted to identify the risk grade of the evaluation target,and the risk assessment area map of the water inrush for the No.8 seam floor in the study area is obtained.The result shows that the risk of water inrush increased from east to west.In order to verify the validity of the proposed model,the achievement is compared with the result obtained by the traditional constant weight model,and the local differences of the two methods are analyzed in detail.The result shows that the variable weight unascertained measure method proposed in this paper is a more feasible and reasonable method,which has more practical guiding significance in engineering application.(2)The random forest theory is introduced into the evaluation of water inrush from seam floor,and an evaluation model based on random forest is put forward,and the model is applied to the risk evaluation of water inrush of No.8 seam floor in Panjiayao coal mine.The sample selection is mainly carried out by the following three methods:(1)Risk levels are quantified according to the water inflow by collecting 41 field samples;(2)Due to the limited number of measured samples,in order to increase the number of samples,synthetic minority oversampling technique is introduced,and 159samples are synthesized on the basis of field measured water inrush samples;(3)According to the actual geological conditions in the study area,the thresholds of the main control factors in each risk grade are determined by the K-means clustering method,and the classification system of risk grade is established.Accordingly,20groups of samples were randomly generated by using the uniform distribution within each risk level threshold of the main control factors,and 100 groups of samples were generated in five grades.A total of 300 measured and synthetic samples are used for model training and verification.Using the Gini index and error average reduction precision out of bag,the importance of each main factor is measured.It is found that water pressure and effective thickness of water resisting stratum are the two most important among the eight main controlling factors.The evaluation result shows that the overall risk of the study area was relatively high,and the whole risk trend increases gradually from the surrounding of research area to its center.In order to compare with the established random forest model,the Probabilistic Neural Network model is built with the same training and verification samples,and the performance of the two models is compared with the obfuscation matrix,and the local differences of the two results are analyzed in detail.The results show that the accuracy of random forest is 6.67%higher than that of PNN,and it has a better prediction ability in the risk assessment of seam floor water inrush.The proposed model provides a new method for the risk assessment of water inrush,and the evaluation results provide some references for risk management,prevention and treatment of water inrush from seam floor.(3)Aiming at the problem of seam roof water disaster of Hongliu coal mine which belongs to Ningxia Coal Mine Group,water inrush risk of No.2 seam is evaluated under the guidance of“three maps”theory.By analyzing the hydrogeological conditions and the water filling conditions of the Hongliu coal mine,the main seam roof aquifers which affects the mining safety of No.2 seam are determined to be the"Qilizhen"sandstone aquifers in the lower section of the Jurassic Zhiluo group.Main control factor system of hydrous quality in aquifers are constructed,the permeability coefficient,the thickness of sandstone in the lower section of the Zhiluo group,the consumption of the flushing fluid,the core adoption rate and thickness ratio of brittle ductile rock are considered in the system above.The thematic maps of the control factors are established correspondingly,where weights of the factors are determined by the analytic hierarchy process.Integrated powerful spatial analysis function of GIS with the hydrous quality index coupled by analytic hierarchy process,the hydrous quality zone map of No.2seam roof aquifer is obtained,the hydrous quality zone map is verified by using the data of unit water inflow calculated by pumping test.The crack height of the seam roof in the study area is calculated by the empirical formula,and the calculation formula is corrected by the field measured value.The zone map of the seam roof crack safety is obtained,the result shows that the whole research area is in the unsafe zone of the roof crack.The comprehensive water inrush risk evaluation zone map of the No.2 seam roof aquifer is generated by overlying the hydrous quality zone map of aquifer and the crack safety zone map of the seam roof.The evaluation results show that the hydrous quality of the No.2 seam roof aquifer increases gradually from the southeast to the northwest of the mining area,and the risk of water inrush gradually increases.(4)In view of the uncertainty of hydrogeological parameters in the groundwater numerical simulation of the No.2 seam roof aquifer,an Bayesian-based uncertainty-analysis method is proposed,it could optimize the hydrogeological parameters,this paper lays emphasis on the permeability coefficient.In order to calculate the Bayesian posteriori distribution,the Markov Chain-Monte Carlo sampling method is introduced to calculate the posterior distribution density of the permeability coefficient.In the framework of Monte Carlo simulation,the method updates the Markov chain continuously,and ultimately converges to the posterior probability distribution of the model parameters.The performance of MCMC method is mainly determined by the sampling effect of its sampling algorithm,adaptive Metropolis algorithm is used to sample in this paper.When determining the likelihood function in the MCMC method,we need to calculate the drawdown of the specified observation well corresponding to each Sampling permeability coefficient,so a lot of sample calculations are needed(usually recache thousands of times).The cost is very high,if we call the groundwater numerical simulation model directly.In order to solve this problem,a substitution model is introduced in this paper,whose idea is similar to the established numerical simulation model.In the process of MCMC sampling,the substitution model can be invoked directly,thus the calculated load and cost can be greatly reduced.Firstly,the permeability coefficient is sampled by Monte Carlo method,and the input value of the replacement model is generated.Then the alternative model is established by the random forest method,which greatly reduces the calculation during calling numerical model and improves the sampling efficiency.The optimum combination of the permeability coefficient of 11 parameter zoning in the study area is obtained with the method of MCMC.(5)A roof aquifer flow numerical model is established for No.2 coal seam of Hongliu coal mine,and the water drainage of seam roof is predicted.Combined with the measured data of the observation holes in the study area,the optimal penetration coefficient calculated by Bayesian method is used to verify the model,the fitting error is within the acceptable range,it shows the feasibility of using Bayesian method to analyze the uncertainty of groundwater simulation parameters.Using the identified groundwater numerical model,the water drainage of No.2 seam roof sandstone aquifer is predicted.Water head should be dewatered to the elevation of seam roof,therefore,the height of the water head that need to dewater could be obtained according to the elevation of seam roof and the height of water head,the dewatering holes are arranged according to the height of water head and its distribution,then dewatering holes position and dewatering amount could be adjusted in accordance with dewatering consequence,the final forecast of mine drainage is 1380m~3/h.The prediction results are reasonable and have certain reference value to the treatment design of mine drainage.