Machine Learning Advisory System for Mitigating Downhole Vibrations for Horizontal Sections

Abstract

Horizontal drilling has become a widely adopted technique in the oil and gas industry due to its ability to maximize reservoir contact and increase hydrocarbon recovery. However, one of the major challenges encountered during horizontal drilling operations is downhole vibrations, which can lead to reduced drilling efficiency, equipment failures, and increased costs. To address this issue, a Machine Learning Advisory System (MLAS) can be implemented to monitor, analyze, predict, and guide to mitigate downhole vibrations in horizontal sections. This research explores the concept of MLAS for predicting and mitigating downhole vibrations, highlighting its potential benefits and key components. A Machine Learning (ML) approach, specifically an Artificial Neural Network (ANN), was employed to predict downhole vibrations through drilling horizontal sections. Artificial Neural Network (ANN) model that utilizes surface rig sensor data as inputs to accurately predict axial, lateral, and torsional vibrations during drilling operations. The study utilized a dataset consisting of 5000 measurements specifically collected from horizontal drilling sections. To evaluate the performance of the model, two metrics were employed. The optimized ANN model demonstrated exceptional accuracy, surpassing a correlation coefficient (R) of 0.97, and maintaining an average absolute percentage error below 2.6%. These results highlight the effectiveness of the developed ANN algorithm in accurately forecasting drilling vibrations solely based on surface drilling parameters. The potential to eliminate the need for downhole sensors makes this approach more cost-effective and efficient.