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

Abstract

The efficiency of drilling operations is determined by numerous aspects, including the particle size of the material being drilled. To achieve efficiency, drilling engineers must take into consideration the size, shape, and density of the cuttings generated during the drilling process. Ineffective drilling can result in increasing expenses and delays for projects involving the extraction of natural resources. The objective of this study is to enhance drilling efficiency by investigating the correlation between drilling parameters such as weight on bit, revolutions per minute, torque, and rate of penetration and features of particle size distribution such as mean particle size and coarseness index as well as mechanical specific energy (MSE). The influence of particle size on drilling has been evaluated through the application of machine learning techniques and comprehensive datasets. The study highlighted relationships between particle size characteristics and the effectiveness of drilling, offering valuable insights into the optimal particle size for tonalite formations that are bordered by mica gneiss. Three machine learning techniques were employed to determine the closest relationship between drilling characteristics and particle size, with the Random Forest approach exhibiting the strongest correlation. This technique may be employed to forecast the size attributes of particles for data points that are not available within the usual range of drilling parameters. This work successfully emphasizes the significance of particle size in drilling operations and showcases the practical use of machine learning in enhanced drilling efficiency.