Loading...
Thumbnail Image
Publication

Unconfined compressive strength prediction using drilling parameters and analyzing feature importance through principal components analysis

Kaya, Muhammed Esad
Research Projects
Organizational Units
Journal Issue
Embargo Expires
Abstract
Knowledge of geomechanical properties is beneficial if not essential for drilling and completion operations in the oil and gas industry. The Unconfined Compressive Strength (UCS) is the maximum compressive force applied to cylindrical rock samples without breaking under unconfined conditions. Unconfined Compressive Strength (UCS) is one of the key criteria to ensure safe, efficient, and successful drilling operations, and estimation of UCS is vital to avoid wellbore stability problems that are inversely correlated with the pace of drilling operations. Furthermore, UCS is an essential input to ensure the success of completion operations such as acidizing and fracturing. Different methods are available to estimate UCS. The common practice to estimate UCS is to conduct experiments with a laboratory testing setup. These laboratory experiments are considered the most accurate way to measure UCS, but they are destructive, time-consuming, and expensive. Alternatively, empirical equations are derived to estimate UCS from well-logging tool readings. These empirical equations are generally derived from physical properties such as interval transit time, porosity, and Young’s modulus. However, most of these equations are not generic, and their applicability for other formation types is limited. The limitations of existing methods to estimate UCS promoted the development of data-driven solutions to estimate UCS. The data-driven methods include but are not limited to basic regression, machine learning, and deep learning algorithms. Data-driven methods to identify patterns in the data to estimate geomechanical parameters are considered to be implemented for drilling operations. This study proposes methods to assist safe and successful drilling operations while eliminating the need for coring, saving a vast amount of time and money by estimating UCS from drilling parameters instantaneously. The goal is to develop a machine-learning algorithm to analyze and process high-frequency data to estimate UCS instantaneously while drilling, allowing safer and more efficient drilling operations. The drilling data used to train, validate, and test the machine learning model is re-purposed from data collected during drilling in a previous study. The algorithm consists of a data processing method called Principal Component Analysis (PCA) to indicate the importance of each parameter by quantifying their variance contribution. Random Forest machine learning algorithm is utilized to build a regression model to estimate UCS. The regression model developed uses Depth, Rotation per Minute (RPM), Weight-on-Bit (WOB), Torque, Rate of Penetration (ROP), Mechanical Specific Energy (MSE), and Normalized Formation Penetration Index (N-FPI) as an input to estimate UCS. The blind data split from the original data set is used to confirm the veracity and applicability of the algorithm. The tests conducted on the final model indicated that the algorithm built is flexible enough to adjust to different conditions and formations to deliver accurate estimations. With a proper allocation of computational power and high-quality drilling data, the algorithm built can be trained to estimate UCS instantaneously while drilling.
Associated Publications
Rights
Copyright of the original work is retained by the author.
Embedded videos