Machine Learning Based Predictive Models for UCS and Young's Modulus of the Dakota Sand Using Schmidt Hammer Rebound

Abstract

ABSTRACT The Inyan Kara formation is a fluvial sandstone saline aquifer in the Williston basin that gained a lot of interest in the recent years due to it being the main saltwater disposal formation and one of the most promising CO2 storage formations in the basin. However very little is known about the geomechanical properties of this formation given the limited number of cored wells and other necessary data through this interval. Therefore, this study used a combination of well log data, routine geomechanical core data experiments and Schmidt hammer measurements to create a robust machine learning model that accurately predicts the deformability and strength of the Inyan Kara formation (i.e., the uniaxial confined stress (UCS) and Young's modulus) from Schmidt hammer rebound. The model performance was then compared to published empirical correlations and proved to be very accurate with near perfect predictions. Our model uses data from the Williston basin and therefore would be more suitable and perform much better on data from the same basin than the existing empirical relationships. INTRODUCTION The Inyan Kara formation in the Williston basin has gained a lot of interest in the recent years given the rising Drilling problems caused by the saltwater disposal activity in this formation (Basu et al., 2019) and the interest in this formation as a potential CO2 storage formation. However, this formation is a saline aquifer located at an average depth ranging between 4000 to 6000 feet with no potential hydrocarbon resources. Therefore, wireline logs and core data in this formation are very scarce (Basu et al., 2020) and this is the primary motivation for this study which aims to generate a machine learning model that correlates the Schmidt hammer rebound to the young's modulus and unconfined compressive strength of the different lithologies in this formation which are parameters usually measured through destructive laboratory tests. Given the importance of the mechanical properties of the rocks for geomechanical modeling (Chellal et al., 2022) as it relates to CO2 geological storage, wellbore stability and other problems related to saltwater disposal. This highlights the importance of finding a machine learning model that rely on non-destructive testing methods like the Schmidt hammer rebound to estimate the deformability and strength of the rock defined by young's modulus (YM) and unconfined compressive strength (UCS) (Malki et al., 2023).