Page 213 - Machine Learning for Subsurface Characterization

Chapter 7

Deep neural network

architectures to approximate the
fluid-filled pore size

distributions of subsurface
geological formations

Siddharth Misra* and Hao Li †
*
Harold Vance Department of Petroleum Engineering, Texas A&M University, College Station,
†
TX, United States, The University of Oklahoma, Norman, OK, United States

Chapter outline
1 Introduction 184 4.2 VAE-NN architecture,
1.1 Log-based subsurface training, and testing 193
characterization 184 4.3 GAN-NN architecture,
1.2 Deep learning 185 training, and testing 195
1.3 NMR logging 186 4.4 VAEc-NN architecture,
2 Introduction to nuclear training, and testing 197
magnetic resonance (NMR) 4.5 LSTM architecture,
measurements 187 training, and testing 200
2.1 NMR relaxation 4.6 Training and testing the four
measurements 187 deep neural network models 202
2.2 Relationships between 5 Application of the VAE-NN
NMR T2 distribution and model 204
conventional logs 189 6 Application of the GAN-NN
3 Data acquisition and model 207
preprocessing 190 7 Application of the VAEc-NN
3.1 Data used in this chapter 190 model 210
3.2 Data preparation 192 8 Application of the LSTM
4 Neural network architectures network model 212
for the NMR T2 synthesis 192 9 Conclusions 214
4.1 Introduction to NMR T2 References 216
synthesis 192

Machine Learning for Subsurface Characterization. https://doi.org/10.1016/B978-0-12-817736-5.00007-7
© 2020 Elsevier Inc. All rights reserved. 183