ebook img

Machine Learning for Subsurface Characterization PDF

442 Pages·2019·86.449 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Machine Learning for Subsurface Characterization

Machine Learning for Subsurface Characterization This page intentionally left blank Machine Learning for Subsurface Characterization Siddharth Misra Hao Li Jiabo He GulfProfessionalPublishingisanimprintofElsevier 50HampshireStreet,5thFloor,Cambridge,MA02139,UnitedStates TheBoulevard,LangfordLane,Kidlington,Oxford,OX51GB,UnitedKingdom ©2020ElsevierInc.Allrightsreserved. Nopartofthispublicationmaybereproducedortransmittedinanyformorbyanymeans, electronicormechanical,includingphotocopying,recording,oranyinformationstorageand retrievalsystem,withoutpermissioninwritingfromthepublisher.Detailsonhowtoseek permission,furtherinformationaboutthePublisher’spermissionspoliciesandourarrangements withorganizationssuchastheCopyrightClearanceCenterandtheCopyrightLicensingAgency,can befoundatourwebsite:www.elsevier.com/permissions. Thisbookandtheindividualcontributionscontainedinitareprotectedundercopyrightbythe Publisher(otherthanasmaybenotedherein). Notices Knowledgeandbestpracticeinthisfieldareconstantlychanging.Asnewresearchandexperience broadenourunderstanding,changesinresearchmethods,professionalpractices,ormedical treatmentmaybecomenecessary. Practitionersandresearchersmustalwaysrelyontheirownexperienceandknowledgeinevaluating andusinganyinformation,methods,compounds,orexperimentsdescribedherein.Inusingsuch informationormethodstheyshouldbemindfuloftheirownsafetyandthesafetyofothers, includingpartiesforwhomtheyhaveaprofessionalresponsibility. Tothefullestextentofthelaw,neitherthePublishernortheauthors,contributors,oreditors, assumeanyliabilityforanyinjuryand/ordamagetopersonsorpropertyasamatterofproducts liability,negligenceorotherwise,orfromanyuseoroperationofanymethods,products, instructions,orideascontainedinthematerialherein. LibraryofCongressCataloging-in-PublicationData AcatalogrecordforthisbookisavailablefromtheLibraryofCongress BritishLibraryCataloguing-in-PublicationData AcataloguerecordforthisbookisavailablefromtheBritishLibrary ISBN:978-0-12-817736-5 ForinformationonallGulfProfessionalpublications visitourwebsiteathttps://www.elsevier.com/books-and-journals Publisher:BrianRomer AcquisitionEditor:KatieHammon EditorialProjectManager:AleksandraPackowska ProductionProjectManager:SelvarajRaviraj CoverDesigner:MarkRogers TypesetbySPiGlobal,India Dedication This book is dedicated to our parents: Rabindra, Surekha, Tao, Jianying, Weidong, and Beihong. We are grateful to our parents for their uncon- ditional love, sacrifices, and encouragement. They were alwaystheretoholdour handsandsupportus through the most difficult times. We thank our parents for making our lives so beautiful! This page intentionally left blank Contents Contributors xvii Preface xix Acknowledgment xxvii 1. Unsupervised outlier detection techniques for well logs and geophysical data 1 SiddharthMisra,OghenekaroOsogbaandMarkPowers 1. Introduction 2 1.1. Basicterminologiesinmachinelearninganddata-driven models 3 1.2. Typesofmachinelearningtechniques 3 1.3. Typesofoutliers 4 2. Outlierdetectiontechniques 5 3. Unsupervisedoutlierdetectiontechniques 7 3.1. Isolationforest 8 3.2. One-classSVM 8 3.3. DBSCAN 10 3.4. Localoutlierfactor 11 3.5. Influenceofhyperparametersontheunsupervised ODTs 12 4. Comparativestudyofunsupervisedoutlierdetection methodsonwelllogs 14 4.1. Descriptionofthedatasetusedforthecomparativestudy ofunsupervisedODTs 15 4.2. Datapreprocessing 15 4.3. Validationdataset 17 4.4. Metrics/scoresfortheassessmentoftheperformances ofunsupervisedODTsontheconventionallogs 21 5. PerformanceofunsupervisedODTsonthefourvalidation datasets 26 5.1. PerformanceonDataset#1containingnoisy measurements 26 5.2. PerformanceonDataset#2containingmeasurements affectedbybadholes 28 5.3. PerformanceonDataset#3containingshalylayers andbadholeswithnoisymeasurements 29 vii viii Contents 5.4. PerformanceonDataset#4containingmanuallylabeled outliers 30 6. Conclusions 32 AppendixA. Popularmethodsforoutlierdetection 33 AppendixB. Confusionmatrixtoquantifytheinlier andoutlierdetectionsbytheunsupervised ODTs 34 AppendixC. Valuesofimportanthyperparametersofthe unsupervisedODTmodels 34 AppendixD. Receiveroperatingcharacteristics(ROC)and precision-recall(PR)curvesforvarious unsupervisedODTsontheDataset#1 34 Acknowledgments 36 References 36 2. Unsupervised clustering methods for noninvasive characterization of fracture-induced geomechanical alterations 39 SiddharthMisra,AdityaChakravarty,PriteshBhoumickand ChandraS.Rai 1. Introduction 40 2. Objectiveofthisstudy 41 3. Laboratorysetupandmeasurements 41 4. Clusteringmethodsfortheproposednoninvasive visualizationofgeomechanicalalterations 44 4.1. K-meansclustering 44 4.2. Agglomerativeclustering 45 4.3. DBSCAN 47 5. Features/attributesfortheproposednoninvasive visualizationofgeomechanicalalteration 48 5.1. Featureengineering 49 5.2. Dimensionalityreduction 52 6. Resultsanddiscussions 53 6.1. Effectoffeatureengineering 53 6.2. Effectofclusteringmethod 55 6.3. Effectofdimensionalityreduction 57 6.4. Effectofusingfeaturesderivedfrombothprefracture andpostfracturewaveforms 58 7. Physicalbasisofthefracture-inducedgeomechanical alterationindex 59 8. Conclusions 61 Acknowledgments 62 References 62 Contents ix 3. Shallow neural networks and classification methods for approximating the subsurface in situ fluid-filled pore size distribution 65 SiddharthMisraandJiaboHe 1. Introduction 66 2. Methodology 67 2.1. Hydrocarbon-bearingshalesystem 67 2.2. Petrophysicalbasisfortheproposeddata-drivenlog synthesis 68 2.3. Datapreparationandstatisticalinformation 69 2.4. Categorizationofdepthsusingflags 72 2.5. FittingtheT distributionwithabimodalGaussian 2 distribution 74 2.6. Min-maxscalingofthedataset(featuresandtarget) 77 2.7. TrainingandtestingmethodologyfortheANNmodels 78 3. ANNmodeltraining,testing,anddeployment 81 3.1. ANNmodels 81 3.2. TrainingthefirstANNmodel 81 3.3. TestingthefirstANNmodel 84 3.4. TrainingthesecondANNmodel 84 3.5. TestingthesecondANNmodel 86 3.6. PetrophysicalvalidationofthefirstANNmodel 86 3.7. ANN-basedpredictionsofNMRT distributionforvarious 2 depthintervals 87 4. Conclusions 89 AppendixA. Statisticalpropertiesofconventionallogsand inversion-derivedlogsforvariousdepth intervals 89 AppendixB. Categorizationofdepthsusingflags(categorical features) 96 AppendixC. Importanceofthe12conventionallogsand 10inversion-derivedlogs 97 AppendixD. Estimationsofspecificreservoirparametersfrom NMRT distributions 98 2 References 100 4. Stacked neural network architecture to model the multifrequencyconductivity/permittivity responses of subsurface shale formations 103 SiddharthMisraandJiaboHe 1. Introduction 103 2. Method 106 2.1. Datapreparation 106 2.2. Methodologyforthedielectricdispersion(DD)log synthesis 106

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.