| Infrastructure construction of our country in recent years more and more flourishing,permeability parameters is an important parameter of rock and soil mass,the permeability coefficient can be used for quantitative evaluation of water resources and prediction of groundwater flow and contaminant migration,the permeability coefficient of inaccurate often make water forecast with the actual situation,thus influence the design of the drainage system,safety hazard.In the field of infrastructure construction,especially tunnel engineering,it is very important to study the permeability properties and permeability parameters of rock.Seepage will seriously affect the mechanical properties of rock,change the stress of rock mass,cause deformation and softening of rock mass,and thus lead to rock mass fracture,endangering the safety and stability of rock mass engineering.In the mining process,if it penetrates the water-isolation fault or encounters water storage caverns and hidden rivers suddenly,it will lead to the steady loss and influx of groundwater,which will lead to mine disasters such as mine water inrush and water gushing,resulting in casualties and catastrophic consequences.In addition,reservoir plays are increasingly becoming major players in today's energy market,and reliable values of key physical properties of rocks,including permeability,are required to estimate raw oil and gas reserves.Permeability determines the migration ability of oil and gas in a reservoir.As one of the parameters to measure permeability,permeability represents the migration ability of oil-gas reservoir and controls the direction movement and velocity of reservoir fluid in the formation.It is an important parameter to evaluate the reservoir productivity,especially the porosity of tight rock,which has always been a difficulty in research.Therefore,it is of great significance to study the permeability of reservoir rocks for the exploration and development of oil and gas reservoirs.This paper relying on cranes big high-speed Korea tunnel sandstone,using scanning electron microscopy(sem)test,image processing technology,image semantic segmentation,and the method of depth study of image features,the study of sandstone pore structure characteristic,based on the fractal theory and micro seepage theory through pore permeability and permeability coefficient is calculated.The paper also directly measured sandstone permeability data through indoor pulse attenuation test,and compared the permeability data obtained by these two different methods.The comparison results are quite consistent,so as to prove that the artificial intelligence method is reliable.The specific research process of the paper is as follows:(1)The scanning electron microscope images of sandstone samples were processed by image denoising,image sharpening and other image enhancement techniques,and then the number of images was increased by segmentation and rotation.All images after pretreatment were taken as data sets,and the data sets were divided into training set images and test set images in the ratio of 3:1.(2)Taking the training set as the input layer,the training model was trained based on caffe and FCN-8S.The trained model had the characteristics of efficient and accurate recognition of pore structure features of images.(3)After the pores in the SEM Image of sandstone were automatically identified by the model,the area and perimeter data of the pores were derived by Image Pro Plus and Excel,and then the fractal coefficient N and fractal dimension D were calculated by fractal theory.(4)Based on the micro permeability theory,the permeability k and permeability coefficient K of sandstone are calculated by substituting the fractal coefficient N and fractal dimension D into the formula,so as to discuss the relationship between the micro pore structure and macro permeability parameters of rock.(5)The permeability of sandstone is measured directly by pulse attenuation laboratory test,and compared with the permeability data obtained by artificial intelligence method,it is found that the two sets of data are in high agreement,thus demonstrating the reliability of the artificial intelligence method adopted in this paper.The innovation of this paper lies in the use of full convolutional neural network FCN8 s for pixel level semantic segmentation of sandstone SEM images,which overcomes the limitations of traditional image segmentation methods(such as threshold method)such as poor segmentation effect and low segmentation accuracy,so as to obtain more accurate pore structure characteristics.Based on deep learning algorithm and full convolutional neural network to train the model,the model can recognize the new input image after learning image features,and can efficiently process a large number of images with good recognition accuracy. |
PDF Full Text Request