Page 126 - Machine Learning for Subsurface Characterization

Chapter 4

Stacked neural network

architecture to model the
multifrequency conductivity/

permittivity responses of
subsurface shale formations

Siddharth Misra* and Jiabo He †,a
*
Harold Vance Department of Petroleum Engineering, Texas A&M University, College Station,
†
TX, United States, School of Computing and Information Systems, University of Melbourne,
Parkville, VIC, Australia

Chapter outline
1 Introduction 103 3 Results 115
2 Method 106 3.1 Sensitivity analysis 117
2.1 Data preparation 106 3.2 Generalization capability
2.2 Methodology for the of the DD log synthesis
dielectric dispersion (DD) using the SNN model 121
log synthesis 106 3.3 Petrophysical and statistical
2.3 Evaluation metric/measure controls on the DD log
for log-synthesis model 109 synthesis using the SNN
2.4 Data preprocessing 112 model 123
2.5 ANN models for dielectric 4 Conclusions 126
dispersion log generation 113 References 127

1 Introduction
Electromagnetic (EM) properties, such as electrical conductivity, dielectric per-
mittivity, and magnetic permeability, are dispersive in nature, such that the EM
properties are functions of the operating frequency of the externally applied EM
field. Such frequency dependence is because the polarization phenomenon in a
material does not change instantaneously with the applied EM field. EM

a
Formerly at the University of Oklahoma, Norman, OK, United States
Machine Learning for Subsurface Characterization. https://doi.org/10.1016/B978-0-12-817736-5.00004-1
© 2020 Elsevier Inc. All rights reserved. 103