Research On Multi Parameter Comprehensive Warning Of Coal Pillar Rockburst Based On Neural Network Analysis

The 3303 working face,3305 working face and 1310 working face of Yangcheng Coal Mine formed irregular pillar due to geological conditions and other reasons.In addition,the working face was buried deep and affected by the fault environment,which has a high impact risk.In this dissertation,the law of instability of coal pillars was studied through theoretical analysis,and the main control factors of coal pillar-type rock burst are studied by using the Artificial Neural Networks model.And using the research results of coal pillar instability law and impact factors of coal pillar impact rock pressure,design a reasonable impact risk monitoring program.A deep neural network model was established by collecting microseismic,drill cuttings,and stress on-line monitoring data during mining of the logging face.And the characteristics of these monitoring data are extracted.This method was used to establish a multi parameter comprehensive early warning evaluation system,which provides a more reliable basis for the coal pillar type shock and ground pressure protection.The main conclusions are as follows:(1)The geological factors and mining disturbance factors that affect the stability of coal pillars were analyzed.Through the establishment of the Artificial Neural Networks model,the dangerous samples of pillar-type rockburst of Yangcheng coal mine and nearby mines were collected to calculate the weight of each influencing factor.The main control factors for the impact instability of the coal pillar on the coal mine are the shape of the coal pillar,the depth of burial,the distance between the pillar and the geological structure.(2)The Deep Neural Networks model was established.Through the training of microseismic,stress online and drill cuttings monitoring data,the mean,variance,mean square error,slope,peak,origin moment,crest factor and skewness data characteristics of the shock hazard monitoring data were obtained.And according to its characteristics,the evaluation and classification are carried out.On the basis of this method,the multi parameter integrated early warning system of coal pillar type rock burst based on the Deep Neural Networks was studied.The application of deep neural network in the comprehensive early warning of impact ground pressure was realized.(3)On the basis of the multi parameter integrated early warning software for the coal pillar type rock burst of Deep Neural Networks,the evaluation and classification of the impact hazard monitoring data during the heading of 3305 working face are carried out.The results of the two-dimensional classification of microearthquakes and stresses and the results of microseismic,stress and pulverized coal classification are obtained.The reliability of the early warning method is verified.