Research And Application Of Wall Rock Borehole Crack Identification Technology Based On Deep Learning

The realization of intelligent mining is based on more efficient access to more effective mineral geological information,and the goal of achieving a transparent earth has become an important part of the mining discipline.The development characteristics of surrounding rock fissures are one of the important indexes for evaluating the stability of the project.At present,the acquisition of cracks in surrounding rocks of rock layers is mostly obtained by direct observation method,which has low identification efficiency and lacks a unified crack division standard.In this paper,how to realize the intelligent identification and quantitative treatment of rock drilling cracks is the research goal.Combining artificial intelligence and mining engineering knowledge,a new method for identifying rock cracks is proposed.In this paper,the development plan of the fracture plane of the borehole in the rock stratum is taken as the research object,the types of cracks in the plane development map of the borehole detection image are divided,and the cracks are quantitatively divided;The fissures in the plane development diagram are divided into five categories: longitudinal large fissures,longitudinal small fissures,lateral large fissures,lateral small fissures,and fracture zones.Through deep learning,the Deep CORAL framework of the convolutional neural network was used to deeply analyze the distribution characteristics of the surrounding rock drilling cracks,and the targeted network structure was designed.The convolutional neural network model was obtained.Intelligent identification system for cracks in surrounding rock drilling pictures.First,pixel-level annotation of the cracks in the surrounding rock picture,build a deep learning convolutional neural network,and then after 1900 iterative training according to the crack features in the surrounding rock borehole picture to obtain the model.After 10 verifications,the average accuracy of the network is 0.86,which is high.On this basis,using Microsoft Visual Studio 2015,Matlab 2016,Python and other development tools,combined with C #,Matlab,Python language design intelligent identification system of surrounding rock drilling cracks.The system includes five functional modules: login,human-computer interaction,crack identification,database storage,and connection.It truly realizes the fully automated process from picture input to output of crack identification results.The intelligent identification system of surrounding rock cracks is applied to a coal mine tunnel support to study the deformation and failure characteristics of the tunnel.A total of six borehole images from two stations were input into the system for identification,and the fracture characteristics of the borehole picture of the coal mine roadway were analyzed to obtain the surrounding rock fragmentation and fracture development.Combined with the destruction characteristics of the roadway,reasonable and reliable supporting countermeasures were proposed,which verified the engineering practicability of the research results in this paper.As an auxiliary method for the identification of surrounding rock cracks applied in this project,the intelligent identification system of surrounding rock bore cracks has the characteristics of intelligence and automation,which provides certain methods and ideas for the development of smart mines and unmanned mines in later mines.There are 44 figures,2 tables and 70 references in the paper.