Machine Predict Lithologies Using Wireline logs

1. Introduction

1.1 What is machine learning?

Machine Learning (ML) is the method of applying programmatic and statistical computer technology to analyse large datasets, and as a result, uncover new understandings. It is a focussed sub-section of Artificial Intelligence (AI), where a more extensive set of predictive modelling tools enable a computer program to learn by generalising from examples. For more information read https://www.golder.com/insights/machine-learning-and-the-growing-use-cases-for-geoscience-practices/.

1.2 Problem Statement:

Identifying subsurface lithologies or rock types is essential for all geoscientists in order to explore our subsurface resources, particularly in the oil and gas industry. Lithology refers to the type of rock that forms the subsurface, and it is classified into, for instance, sandstone, claystone, marl, limestone, and dolomite. Several subsurface data can be used to identify lithologies, such as wireline logs petrophysical data. However, it is often a tedious, repetitive and time-consuming task. This project will predict lithology from petrophysical logs using machine learning techniques (classification) and add to the solution of these problems since these logs are direct proxies of lithology.