Cover
Machine Learning
for Subsurface
Characterization
Copyright
Dedication
Contributors
Preface
Machine learning-driven success stories
Challenges and precautions when using machine learning
Recommendations for harnessing the power of machine learning
Concluding remarks
References
Further reading
Acknowledgment
1
Unsupervised outlier detection techniques for well logs and geophysical data
Introduction
Basic terminologies in machine learning and data-driven models
Types of machine learning techniques
Types of outliers
Outlier detection techniques
Unsupervised outlier detection techniques
Isolation forest
One-class SVM
DBSCAN
Local outlier factor
Influence of hyperparameters on the unsupervised ODTs
Comparative study of unsupervised outlier detection methods on well logs
Description of the dataset used for the comparative study of unsupervised ODTs
Data preprocessing
Feature transformation: Convert R to log(R)
Feature scaling: Use of robust scaler
Validation dataset
Dataset #1: Containing noisy measurements
Dataset #2: Containing bad holes
Dataset #3: Containing shaly layers and bad holes with noisy measurements
Dataset #4: Containing manually labeled outliers
Metrics/scores for the assessment of the performances of unsupervised ODTs on the conventional logs
Recall
Specificity
Balanced accuracy score
Precision
F1 score
Receiver operating characteristics (ROC) curve and ROC-AUC score
Precision-recall (PR) curve and PR-AUC score
Performance of unsupervised ODTs on the four validation datasets
Performance on Dataset #1 containing noisy measurements
Performance on Dataset #2 containing measurements affected by bad holes
Performance on Dataset #3 containing shaly layers and bad holes with noisy measurements
Performance on Dataset #4 containing manually labeled outliers
Conclusions
Popular methods for outlier detection
Confusion matrix to quantify the inlier and outlier detections by the unsupervised ODTs
Values of important hyperparameters of the unsupervised ODT models
Receiver operating characteristics (ROC) and precision-recall (PR) curves for various unsupervised ODTs on the Dataset #1
Acknowledgments
References
2
Unsupervised clustering methods for noninvasive characterization of fracture-induced geomechanical alterations
Introduction
Objective of this study
Laboratory setup and measurements
Clustering methods for the proposed noninvasive visualization of geomechanical alterations
K-means clustering
Agglomerative clustering
DBSCAN
Features/attributes for the proposed noninvasive visualization of geomechanical alteration
Feature engineering
Dimensionality reduction
Results and discussions
Effect of feature engineering
Effect of clustering method
Effect of dimensionality reduction
Effect of using features derived from both prefracture and postfracture waveforms
Physical basis of the fracture-induced geomechanical alteration index
Conclusions
Acknowledgments
Declarations
References
3
Shallow neural networks and classification methods for approximating the subsurface in situ fluid-filled pore size distrib ...
Introduction
Methodology
Hydrocarbon-bearing shale system
Petrophysical basis for the proposed data-driven log synthesis
Data preparation and statistical information
Categorization of depths using flags
Fitting the T2 distribution with a bimodal Gaussian distribution
Min-max scaling of the dataset (features and target)
Training and testing methodology for the ANN models
ANN model training, testing, and deployment
ANN models
Training the first ANN model
Testing the first ANN model
Training the second ANN model
Testing the second ANN model
Petrophysical validation of the first ANN model
ANN-based predictions of NMR T2 distribution for various depth intervals
Conclusions
Statistical properties of conventional logs and inversion-derived logs for various depth intervals
Categorization of depths using flags (categorical features)
Importance of the 12 conventional logs and 10 inversion-derived logs
Estimations of specific reservoir parameters from NMR T2 distributions
Abbreviations
References
4
Stacked neural network architecture to model the multifrequency conductivity/permittivity responses of subsurf ...
Introduction
Method
Data preparation
Methodology for the dielectric dispersion (DD) log synthesis
Evaluation metric/measure for log-synthesis model
Data preprocessing
ANN models for dielectric dispersion log generation
Results
Sensitivity analysis
Generalization capability of the DD log synthesis using the SNN model
Petrophysical and statistical controls on the DD log synthesis using the SNN model
Conclusions
References
5
Robust geomechanical characterization by analyzing the performance of shallow-learning regression methods usin ...
Introduction
Methodology
Data preparation
Data preprocessing
Metric to evaluate the log-synthesis performance of the shallow-learning regression models
Shallow-learning regression models
Ordinary least squares (OLS) model
Partial least squares (PLS) model
Least absolute shrinkage and selection operator (LASSO) model
ElasticNet model
Multivariate adaptive regression splines (MARS) model
Artificial neural network (ANN) model
Clustering techniques
K-means clustering
Gaussian mixture model clustering
Hierarchical clustering
DBSCAN clustering
SOM followed by K-means clustering
Results
Prediction performances of shallow-learning regression models
Comparison of prediction performances of shallow-learning regression models in Well 1
Performance of clustering-based reliability of sonic-log synthesis
Conclusions
References
6
Index construction, dimensionality reduction, and clustering techniques for the identification of flow units i ...
Introduction
Geology of the shale formation
Literature survey
Objectives
Properties influencing EOR efficiency of light-hydrocarbon injection
Methodology to generate the ranking (R) index
Description of the R-index
Calculation of the R-index
Methodology to generate the microscopic displacement (MD) index
Description of the MD-index
Calculation of the MD-index
Step 1: NMR decomposition using factor analysis
Step 2: Determination of pore volumes of various fluid phases represented by the decomposed factors
Step 3: Correction of miscible, free-oil volume for pore confinement effect
Step 4: Compute the MD-index
Methodology to generate the K-means clustering (KC) index
Description of the KC-index
Calculation of the KC-index
Limitations
Conclusions
References
7
Deep neural network architectures to approximate the fluid-filled pore size distributions of subsurface geolog ...
Introduction
Log-based subsurface characterization
Deep learning
NMR logging
Introduction to nuclear magnetic resonance (NMR) measurements
NMR relaxation measurements
Relationships between NMR T2 distribution and conventional logs
Data acquisition and preprocessing
Data used in this chapter
Data preparation
Neural network architectures for the NMR T2 synthesis
Introduction to NMR T2 synthesis
VAE-NN architecture, training, and testing
GAN-NN architecture, training, and testing
VAEc-NN architecture, training, and testing
LSTM architecture, training, and testing
Training and testing the four deep neural network models
Application of the VAE-NN model
Application of the GAN-NN model
Application of the VAEc-NN model
Application of the LSTM network model
Conclusions
References
8
Comparative study of shallow and deep machine learning models for synthesizing in situ NMR T2 distributions
Introduction
Dataset
Shallow-learning models
Ordinary least squares
Least absolute shrinkage and selection operator
ElasticNet
Support vector regressor
k-Nearest neighbor regressor
Artificial neural network
Comparisons of the test accuracy and computational time of the shallow-learning models
Deep learning models
Variational autoencoder assisted neural network
Generative adversarial network assisted neural network
Variational autoencoder with convolutional layer assisted neural network
Encoder-decoder long short-term memory network
Comparisons of the accuracy and computational time of the deep learning models
Cross validation
Comparison of the performances of the deep and shallow models
Discussions
Conclusions
References
9
Noninvasive fracture characterization based on the classification of sonic wave travel times
Introduction
Mechanical discontinuities
Characterization of discontinuities
Machine learning for characterization of discontinuities
Objective
Assumptions and limitations of the proposed data-driven fracture characterization method
Significance and relevance of the proposed data-driven fracture characterization method
Fast-marching method (FMM)
Introduction
Validation
Numerical model of homogeneous material at various spatial discretization
Material with large contrasts in compressional wave velocity
Material with parallel discontinuities
Material with smoothly varying velocity distribution
Fast-marching simulation of compressional wavefront travel time for materials containing discontinuities
Methodology for developing data-driven model for the noninvasive characterization of static mechanical discontinuities ...
Classification methods implemented for the proposed fracture characterization workflow
K-nearest neighbors (KNN) classifier
Support vector machine (SVM) classifier
Decision tree (DT) classifier
Random forest (RF) classifier
AdaBoost classifier
NaΓ―ve Bayes (NB) classifier
Artificial neural network (ANN) classifier
Voting classifier
Results for the classification-based noninvasive characterization of static mechanical discontinuities in materials
Characterization of material containing static discontinuities of various dispersions around the primary orientation
Background
Model accuracy
Characterization of material containing static discontinuities of various primary orientations
Background
Model accuracy
Sensor/receiver importance
Characterization of material containing static discontinuities of various spatial distributions
Background
Model accuracy
Sensor importance
Acknowledgments
References
Further reading
10
Machine learning assisted segmentation of scanning electron microscopy images of organic-rich shales with fea ...
Introduction
Method
SEM images
Workflow
Feature extraction
Gaussian blur (one feature)
Difference of Gaussians (one feature)
Sobel operator (one feature)
Hessian matrix (three features)
Wavelet transforms (six features)
Local information (three features)
Other features investigated for this study
Training dataset
Classification of the feature vectors
Results and discussions
Four-component segmentation
Multilabel probability-based segmentation
Performance on testing dataset
Feature ranking using permutation importance
Deployment of the segmentation method on new, unseen SEM images
Conclusions
Acknowledgments
References
11
Generalization of machine learning assisted segmentation of scanning electron microscopy images of organic-ri ...
Introduction
Method
Image preprocessing
Segmentation performance evaluation
Training dataset
Hyperparameter optimization
Testing dataset
Results and discussion
Task 1: Training and testing on Map-1
Task 2: Training on Map-1 and testing on Map-2
Task 3: Training and testing on Map-2
Task 4: Training on Map-2 and testing on Map-1
Task 5: Training on both Map-1 and Map-2 and testing on Map-1
Task 6: Training on both Map-1 and Map-2 and testing on Map-2
Feature ranking
Conclusions
Acknowledgments
References
12
Characterization of subsurface hydrocarbon/water saturation by processing subsurface electromagnetic logs usi ...
Introduction
Error-minimization methods
Subsurface characterization problem
Method
Relevant EM logging tools
Relevant log interpretation models
Proposed deterministic inversion of EM logs
Introduction
Modified Levenberg-Marquardt nonlinear inversion
Limitations and assumptions in the proposed multifrequency-EM log processing using modified Levenberg-Marquardt algo ...
Results and discussion
Sensitivity of the integrated WS, SMD, and CRI model
Application of the proposed log processing to synthetic data
Application of the proposed log processing to the 9 EM logs acquired in the upper Wolfcamp formation
Petrophysical interpretation of the inversion-derived estimates in the upper Wolfcamp formation
Error analysis for the inversion-derived estimates in the upper Wolfcamp formation
Conclusions
References
13
Characterization of subsurface hydrocarbon/water saturation using Markov-chain Monte Carlo stochastic inversi ...
Introduction
Saturation estimation in shale gas reservoir
Electromagnetic (EM) logging tools
Conventional EM-log-interpretation models
Estimation of water/hydrocarbon saturation based on the inversion of electromagnetic (EM) logs
Method
Mechanistic clay-pyrite interfacial-polarization model
Proposed inversion-based interpretation method
Bayesian formulation of the parameter estimation problem
Bayesian framework:
Prior models:
Likelihood function:
Metropolis-Hastings sampling algorithms
Proposal distribution:
Acceptance probability:
Convergence monitoring
Limitations and assumptions in the proposed inversion-based interpretation method
Results and discussion
Application of the MCMC-based stochastic inversion to synthetic layers
Application of the MCMC-based stochastic inversion to process broadband EM dispersion logs acquired in a shale gas f ...
Petrophysical interpretation and log analysis of the inversion-derived estimates in the shale gas formation
Conclusions
References
Index
A
B
C
D
E
F
G
H
I
J
K
L
M
N
O
P
R
S
T
U
V
W
Back Cover