Exploring the Landscape: Unveiling Dominant Approaches, Benefits and Potential Pitfalls in Utilizing Diverse Machine Learning Methods for Facies Classification – A Review

Authors

  • Ofoh Ifeyinwa Juliana Department of Geology, Federal University of Technology, Owerri, Imo State, Nigeria
  • Onyekuru Samuel Okechukwu Professor, Department of Geology, Federal University of Technology, Owerri, Imo State, Nigeria
  • Ikoro Diugo Associate Professor, Department of Geology, Federal University of Technology, Owerri, Imo State, Nigeria
  • Opara Alexander Iheanyichukwu Professor, Department of Geology, Federal University of Technology, Owerri, Imo State, Nigeria

Keywords:

Artificial Neural Network, Core data, Facies classification, Geosciences, Machine Learning, Reinforcement learning, Seismic data, Supervised learning, Well log data, Unsupervised learning

Abstract

Facies classification in geosciences has witnessed a remarkable transformation with the integration of diverse machine learning (ML) methods. An overview of the dominant approaches, benefits and potential pitfalls encountered in the application of different ML techniques for facies classification is presented in this review paper. ML technologies like convolutional Neural Networks (CNNs), Support Vector Machines (SVMs), Random Forests (RFs), and Deep Learning architectures have revolutionized the way geoscientists interpret subsurface reservoirs using well logs, seismic data, and core samples. The benefits of deploying Machine Learning techniques in facies classification range from enhanced speed and accuracy of interpretation, through facilitating the extraction of valuable geological insights from complex datasets, to flexible handling of multi-modal data, thus allowing for the combining many data sources to improve classification accuracy and enhance informed decision-making in exploration and development projects. However, the use of ML methods in facies classification are inherent to potential pitfalls and significant challenges, some of which include sparse data availability and poor data quality. Robust model training necessitates large, labeled datasets that are often costly and time-consuming to curate. In addition, model interpretability remains a major concern, since the 'black-box' nature of some ML algorithms can hinder geoscientists' ability to understand and trust the results. Overfitting, model generalization issues, and the risk of biases in training data are additional problems that must be addressed.

Downloads

Download data is not yet available.

Downloads

Published

25-03-2024

Issue

Section

Articles

How to Cite

[1]
O. I. Juliana, O. S. Okechukwu, I. Diugo, and O. A. Iheanyichukwu, “Exploring the Landscape: Unveiling Dominant Approaches, Benefits and Potential Pitfalls in Utilizing Diverse Machine Learning Methods for Facies Classification – A Review”, IJRESM, vol. 7, no. 3, pp. 62–71, Mar. 2024, Accessed: Dec. 14, 2024. [Online]. Available: https://journal.ijresm.com/index.php/ijresm/article/view/2967