A data-driven risk assessment approach for electronic boards used in oil well drilling operations

Abstract

To assist subject matter experts in investigating electronic failures of drilling tools, an innovative risk assessment approach for oil well drilling operations is developed that relies on synthetic time-series data to emulate environmental factors encountered downhole, explicitly focusing on temperature, shock, and vibration. The approach involves utilizing load cycle counting to extract meaningful features from each environmental channel measured by the drilling tool. The results from experiments with features related to dwell periods (dwell time and dwell damage) and load cycles (cycle means and cycle ranges) show a significant correlation between load cycle features and the risk label. Subsequently, a tree-based machine learning model is trained to label drilling operations based on synthetic data. Several models have been trained initially with comparable results. However, the advantage of using a tree-based model, specifically extra trees, is explainability and the stochastic aspect, which translates into model robustness when applied to real data. Preliminary results from a case study indicate that this new approach is highly effective in categorizing environmental risks associated with drilling operations. This risk assessment method can significantly enhance the decision-making process in investigating electronic board failures by offering reliable decision support.