A Novel Rock Drillability Characterization and Prediction Method Based on Drilling Big Data and Unsupervised Clustering Algorithm

Abstract

ABSTRACT Rock drillability measures the ability of rock to resist damage by the drill bit. Accurate evaluation of rock drillability is crucial for understanding formation properties and selecting appropriate drill bits and drilling parameters. However, many existing intelligent evaluation methods for rock drillability mainly rely on supervised learning algorithms, which require a large number of rock core samples for drillability analysis data as labels. Core acquisition is difficult and expensive. To address this issue, this study proposes a method for continuous formation drillability evaluation using logging big data and unsupervised clustering algorithms. Firstly, self-organizing mapping (SOM) neural network is used to cluster logging data. The formation is then divided into six drillability grades by analyzing the rate of penetration (ROP) distribution of each type of logging data. We used this method to grade the drillability of the test well formation. The results show that the ROP of the formation decreases gradually with the increase of drillability grade, which verproves the effectiveness of our method. INTRODUCTION Rock drillability is a crucial factor that determines drilling efficiency, reflecting how easily the rock breaks during drilling operations. Accurate evaluation of rock drillability can help drilling engineers select drilling tools and parameters to improve drilling efficiency (Kong, et al. 2022). The microdrill bit experimental method is a classic approach for evaluating drillability. It involves recording the time it takes for a miniature gear or PDC bit to drill 2.4mm under a specific drilling pressure and rotation speed condition, and calculating the formation drillability based on standard formulas and recorded time. (Zhang & Xue, 2019; Chen, et al. 2010). In the experimental method, the test rock core samples are not in the real underground environment, and the accuracy of the results is difficult to guarantee. Statistical analysis is another method for evaluating rock drillability, which predicts the drillability of rock by establishing a mathematical model that incorporates logging data, drilling parameters, and other factors that influence drillability (Andrews, et al.; Tang, et al.). However, a significant amount of rock core data is required to support the model and ensure accurate predictions.