Machine Learning and Artificial Intelligence in Geosciences

Opis

Advances in Geophysics, Volume 61 - Machine Learning and Artificial Intelligence in Geosciences, the latest release in this highly-respected publication in the field of geophysics, contains new chapters on a variety of topics, including a historical review on the development of machine learning, machine learning to investigate fault rupture on various scales, a review on machine learning techniques to describe fractured media, signal augmentation to improve the generalization of deep neural networks, deep generator priors for Bayesian seismic inversion, as well as a review on homogenization for seismology, and more.

Provides high-level reviews of the latest innovations in geophysics
Written by recognized experts in the field
Presents an essential publication for researchers in all fields of geophysics

Spis treści

Front Cover 2
Machine Learning in Geosciences 5
Copyright 6
Contents 7
Contributors 9
Preface 11
- References 12
Chapter One: 70 years of machine learning in geoscience in review 13
- 1. Historic machine learning in geoscience 15
  - 1.1. Expert systems to knowledge-driven AI 16
  - 1.2. Neural networks 18
  - 1.3. Kriging and Gaussian processes 20
- 2. Contemporary machine learning in geoscience 22
  - 2.1. Modern machine learning tools 23
  - 2.2. Support-vector machines 24

Front Cover 2
Machine Learning in Geosciences 5
Copyright 6
Contents 7
Contributors 9
Preface 11
- References 12
Chapter One: 70 years of machine learning in geoscience in review 13
- 1. Historic machine learning in geoscience 15
  - 1.1. Expert systems to knowledge-driven AI 16
  - 1.2. Neural networks 18
  - 1.3. Kriging and Gaussian processes 20
- 2. Contemporary machine learning in geoscience 22
  - 2.1. Modern machine learning tools 23
  - 2.2. Support-vector machines 24
  - 2.3. Random forests 29
  - 2.4. Modern deep learning 32
  - 2.5. Neural network architectures 40
  - 2.6. Convolutional neural network architectures 40
  - 2.7. Generative adversarial networks 45
  - 2.8. Recurrent neural network architectures 47
  - 2.9. The state of ML on geoscience 49
- References 55
Chapter Two: Machine learning and fault rupture: A review 69
- 1. Introduction 69
- 2. Machine learning: A shallow dive 73
  - 2.1. Learning tasks, performance, and experience 73
  - 2.2. Learning capacity 75
  - 2.3. Geophysical data 77
- 3. Laboratory studies 78
  - 3.1. Laboratory geodesy 79
  - 3.2. Laboratory seismology 83
- 4. Field studies 87
  - 4.1. Techniques 90
    - 4.1.1. Earthquake catalog building 90
      - 4.1.1.1. Earthquake detection 91
      - 4.1.1.2. Phase picking and polarity determination 94
      - 4.1.1.3. Event association 96
      - 4.1.1.4. Event location 97
    - 4.1.2. Seismic waveform denoising and enhancing 98
    - 4.1.3. Tectonic tremor detection 101
    - 4.1.4. Fault slip inversion 101
    - 4.1.5. Automatic detection of geodetic deformation 102
  - 4.2. Applications 103
    - 4.2.1. Early warning 103
    - 4.2.2. Induced seismicity 105
    - 4.2.3. Earthquake catalog forecasting 106
  - 5. Conclusion 107
  - Acknowledgments 109
  - References 109
- Chapter Three: Machine learning techniques for fractured media 121
  - 1. Introduction 121
  - 2. Preliminaries 125
    - 2.1. Governing equations 125
    - 2.2. DFN to graph mappings 128
      - 2.2.1. Topology map 130
      - 2.2.2. Pipe-network map 131
      - 2.2.3. Flow topology graph (FTG) construction 132
    - 4. Pruned DFN as a reduced-order model 134
      - 4.1. Existence of backbones 134
    - 3. Graph as a DFN reduced-order model 134
    - 5. Machine learning methods for backbone identification 136
      - 5.1. Fracture-classification: Labeling by FTG membership 137
      - 5.2. Fracture-classification: Labeling by mass flux 142
      - 5.3. Path-classification: Labeling by FTG membership 148
        
        5.3.1. Logistic regression 151
        
        5.3.1.1. Random forest 152
        
        6. Further scope for ML in fractured media 156
        
        References 156
        
        Further reading 162
      - Chapter Four: Seismic signal augmentation to improve generalization of deep neural networks 163
        
        1. Introduction 163
        
        2. Benchmark data and training procedure 166
        
        3. Augmentations 168
        
        3.1. Random shift 168
        
        3.2. Superimposing events 171
        
        3.3. Superposing noise 173
        
        3.4. False positive noise 175
        
        3.5. Channel dropout 176
        
        3.6. Resampling 178
        
        3.7. Augmentation for synthetic data generation 179
        
        4. Discussion 182
        
        5. Conclusions 185
        
        Acknowledgments 186
        
        References 186
      - Chapter Five: Deep generator priors for Bayesian seismic inversion 191
        
        1. Introduction 191
        
        2. Methodology 195
        
        2.1. Bayesian inference 195
        
        2.2. Deep GAN prior model generator 196
        
        2.3. Sample the posterior PDF by pCN 198
        
        3. Seismic inversion applications 199
        
        3.1. Traveltime tomography 199
        
        3.2. Full waveform inversion 200
        
        4. Numerical examples 200
        
        4.1. Training the network architecture 200
        
        4.1.1. Network architectures 200
        
        4.1.2. Training 201
        
        4.2. Quantitative performance evaluation of different generators 203
        
        4.2.1. Comparison of generated samples 203
        
        4.2.2. Quantitative evaluation criterion 203
        
        4.3. Statistical inversion with Overthrust models 208
        
        4.3.1. Traveltime Tomography example 210
        
        4.3.2. FWI example 211
        
        4.4. Statistical inversion with Sigsbee models 215
        
        4.4.1. Traveltime tomography example 215
        
        4.4.2. FWI example 218
        
        5. Conclusions and discussion 224
        
        References 226
        
        Acknowledgments 226
        
        Chapter Six: An introduction to the two-scale homogenization method for seismology 229
        
        1. Introduction 230
        
        2. Mathematical notions and notations 234
        
        2.3. Linear filtering and Fourier domain 235
        
        2.3.1. Arithmetic and harmonic averages 238
        
        2.1. Periodic functions 235
        
        2.2. Two-variable functions 235
        
        3. A numerical introduction to the subject 238
        
        3.1. A simple example in 1-D 239
        
        3.2. Small scales heterogeneities and solution smoothness 242
        
        3.3. The Backus solution 243
        
        3.4. Beyond the Backus solution 247
        
        4. Two-scale homogenization: the 1-D periodic case 249
        
        4.1. Resolution of the homogenization problem 252
        
        4.1.1. Resolution of the homogenized equations, step 1 252
        
        4.1.2. Resolution of the homogenized equations, step 2 253
        
        4.1.3. Resolution of the homogenized equations, step 3 254
        
        4.1.4. Analytical solution to the cell problem 256
        
        4.1.5. Strain leading order 257
        
        4.1.6. Summary 257
        
        4.2. Convergence theorem 258
        
        4.3. External point sources 260
        
        4.4. An example 261
        
        5. Two-scale homogenization: The 1-D nonperiodic case 264
        
        5.1. The Backus solution 264
        
        5.2. Naive solutions to the two-scale homogenization periodicity limitation 265
        
        5.3. New functional spaces 266
        
        5.4. Cell properties in the nonperiodic case 269
        
        5.5. Nonperiodic homogenized equations and resolution 269
        
        5.6. Construction of the cell properties E 272
        
        5.6.1. Direct construction 272
        
        5.6.2. Implicit construction 273
        
        5.7. Two examples 275
        
        5.7.1. 1-D bar with random properties 275
        
        5.7.2. 1-D bar with random properties but no smooth variations 277
        
        6. Two-scale homogenization: Higher dimensions 278
        
        6.1. The 2-D and 3-D elastic case 278
        
        6.1.1. Total anisotropy and isotropic projection 278
        
        6.1.2. Asymptotic expansion 279
        
        6.1.3. Series of equations 280
        
        6.1.4. Resolution 280
        
        6.1.5. Construction of the effective elastic tensor and density 281
        
        6.2. The 2-D and 3-D acoustic case 282
        
        7. What we skipped 285
        
        7.1. Spatial filters 285
        
        7.2. Boundary conditions 286
        
        7.3. Other aspects to address 287
        
        8. Examples of applications 288
        
        8.1. 2-D refracted waves 288
        
        8.2. A solid with a fluid inclusion 292
        
        8.3. A failing case: Helmholtz resonators 294
        
        8.4. A 3-D geological model 297
        
        8.5. Rotational receivers corrector 302
        
        8.6. Full waveform inversion (FWI) 304
        
        9. Discussion and conclusions 311
        
        Acknowledgments 314
        
        References 314
        
        Back Cover 319

Pokaż więcej

Opis

Spis treści

Recenzje (0)