Abstract:PreviousNext No Access2nd SEG Rock Physics Workshop: Challenges in Deep and Unconventional Oil/Gas Exploration, Qingdao, China, 25–27 October 2019Deep learning–assisted elastic parameter prediction from the digital rock images with geophysical constraintsAuthors: Rongang CuiaDanping CaoZhaolin ZhuQiang LiuYan JiaRongang CuiaChina University of PetroleumQingdao National Laboratory for Marine Science and TechnologySearch for more papers by this author, Danping CaoChina University of PetroleumQingdao National Laboratory for Marine Science and TechnologySearch for more papers by this author, Zhaolin ZhuChina University of PetroleumQingdao National Laboratory for Marine Science and TechnologySearch for more papers by this author, Qiang LiuChina University of PetroleumQingdao National Laboratory for Marine Science and TechnologySearch for more papers by this author, and Yan JiaChina University of PetroleumQingdao National Laboratory for Marine Science and TechnologySearch for more papers by this authorhttps://doi.org/10.1190/rpwk2019-035.1 SectionsAboutPDF/ePub ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InRedditEmail Abstract Digital rock images by modern X-ray computer tomography are widely used to calculate the elastic parameters of rocks, such as P-velocity and S-velocity. The numerical simulation method to calculate elastic parameters requires extremely accurate segmentation of the digital rock images and takes vast time. In the paper, the multi-scale heterogeneous convolutional neural network (MHNN) is proposed to predict the elastic parameters directly from the digital rock images with a limited training dataset. The MHNN is targeted for predicting elastic parameters and the rock matrix and pore types segmented from the gray-scale images are treated as a constraint. The cost function is designed into a composite form to accelerate the convergence speed, which contains the error of segmentation and the elastic parameter predicted from gray-scale images. The test result shows that the prediction accuracy of MHNN is significantly improved compared with the traditional convolutional neural network. Meanwhile, the MHNN is much more efficient than the numerical simulation method. By adding geophysical constraints to the CNN, an implicit representation from the grayscale image to the elastic parameters can be gained, which could improve the accuracy and efficiency of parameter prediction. Keywords: neural network, high-resolution, elastic, numerical, algorithmPermalink: https://doi.org/10.1190/rpwk2019-035.1FiguresReferencesRelatedDetails 2nd SEG Rock Physics Workshop: Challenges in Deep and Unconventional Oil/Gas Exploration, Qingdao, China, 25–27 October 2019ISSN (online):2159-6832Copyright: 2020 Pages: 58 publication data© 2019 Published in electronic format with permission by the Society of Exploration Geophysicists, China University of Petroleum (East China), Hohai UniversityPublisher:Society of Exploration Geophysicists HistoryPublished Online: 14 Apr 2020 CITATION INFORMATION Rongang Cuia, Danping Cao, Zhaolin Zhu, Qiang Liu, and Yan Jia, (2020), "Deep learning–assisted elastic parameter prediction from the digital rock images with geophysical constraints," SEG Global Meeting Abstracts : 46-49. https://doi.org/10.1190/rpwk2019-035.1 Plain-Language Summary Keywordsneural networkhigh-resolutionelasticnumericalalgorithmPDF DownloadLoading ...

Lithofacies Prediction from Core Images Using Bayesian Neural Networks

Reservoir Prediction Using Graph-Regularized Deep Learning

Application Of Artificial Neural Networks In Lithofacies Interpretation Used For 3d Geological Modelling

Lithology prediction in the subsurface by artificial neural networks on well and 3D seismic data in clastic sediments: a stochastic approach to a deterministic method

The Application of Deep Learning Method in the Prediction of Shale Content

Lithology identification from well-log curves via neural networks with additional geologic constraint

Data-driven lithofacies prediction in complex tight sandstone reservoirs: a supervised workflow integrating clustering and classification models

Automatic Lithofacies Classification with t-SNE and K-Nearest Neighbors Algorithm

Research and Application of Reservoir Classification Method Based on Deep Learning

Intelligent Sedimentary Lithofacies Identification with Integrated Well Logging Features

Deep Learning–assisted Elastic Parameter Prediction from the Digital Rock Images with Geophysical Constraints

Centimeter-Scale Lithology and Facies Prediction in Cored Wells Using Machine Learning

Reservoir Property Prediction Using the Dynamic Radial Basis Function Network

Direct estimation of lithofacies and geofluid parameters incorporating Gaussian mixture priori and prestack EVA inversion with bounding constraint

Joint Use of Multiseismic Information for Lithofacies Prediction Via Supervised Convolutional Neural Networks

A lithofacies prediction method based on stochastic simulation

Application of Machine Learning in Prediction of Lithofacies from Well Logs

Integrating deep learning and logging data analytics for lithofacies classification and 3D modeling of tight sandstone reservoirs

Inversion of reservoir parameters based on the rock physics model and neural ODEs

Bayesian Neural Networks for Geothermal Resource Assessment: Prediction with Uncertainty

Rock Type Prediction and 3D Modeling of Clastic Paleokarst Fillings in Deeply-Buried Carbonates Using the Democratic Neural Networks Association Technique