Comparison of Machine Learning Models for Predicting Particle Size Parameters from Drilling Data

Abstract

Drilling efficiency is the capacity of a drilling operation to extract natural resources from subterranean reservoirs with maximum effectiveness. Drilling efficiency is crucial to the success of extraction operations, as it influences every aspect of the cost of the operation to the quantity and quality of the extracted natural resources. The effectiveness of a drilling operation is greatly affected by several factors, one of which is the particle size of the material being drilled. Drilling engineers must pay close attention to particle size characteristics, which characterize the size, shape, and density of the cuttings produced during drilling. Efficiency in drilling significantly impactsthe drilling process, which in turn affects the extraction technique. Since drilling is a complicated process used to recover precious natural resources, ineffective drilling will add additional expense and time to the project. The influence of particle size variables in drilling can be evaluated using machine learning by analyzing large datasets of drilling parameters and particle properties. In this study, machine learning algorithms are used to examine the relationship between particle size parameters and drilling performance, yielding insights into the optimal particle size parameters for a wide range of formations and drilling scenarios. To determine the characteristics of drilling and particle size most closely connected, three distinct machine learning techniques are used in this investigation. Among these methods, Random Forest shows the strongest correlation between these traits. This approach is used to anticipate particle size parameters for absent data points within the typical range for drilling parameters. The study can be used to evaluate the drilling performance and efficiency. A similar strategy can be implemented in formations with identical geological factors using the most applicable machine learning model.