Auto-Detecting Drilling Vibrations Through Intelligent 4IR Solution

Abstract

Downhole vibrations have a significant impact on downhole equipment performance, wellbore stability, and drilling efficiency. High levels of drillstring vibration often led to equipment failure, hole problems, energy loss, and reduced drilling performance. Detecting these vibrations through downhole sensors is costly and time-consuming. However, advancements in new technologies and machine learning present opportunities for intelligent data analysis and addressing complex technical problems. In this study, a successful application of a machine learning technique was proposed to auto-detect downhole vibrations during the drilling curve section using surface drilling data. The axial, torsional, and lateral drillstring vibration modes were detected using random forests (RF) machine learning models trained with real field data. The model was developed through comprehensive data-driven research, including data collection, preprocessing, analytics, model optimization, and performance evaluation. Overall, the developed machine learning model achieved high accuracy, with R values exceeding 0.87 and average absolute percentage error (AAPE) below 8.4% between actual readings and predictions. The proposed ML algorithm offers an intelligent solution for predicting drilling vibrations using only surface drilling parameters, eliminating the need for downhole sensors. Implementing this solution on drilling rigs enables real-time monitoring of vibrations and supports automated advisory systems. It provides valuable insights for directional drillers and drilling engineers, facilitating drilling optimization, and improved well planning.