An Artificial Intelligence Model to Synthesize Measurements While Drilling Sensors for Coiled Tubing Drilling

Abstract

Summary This paper presents a comprehensive methodology for developing an artificial intelligence (AI) model to synthesize downhole measurements as a digital backup for measurement while drilling (MWD) sensors, ensuring uninterrupted drilling in coiled tubing drilling operations (CTD). The MWD tool plays a pivotal role in CTD, acquiring critical measurements for safe drilling operations. These measurements are critical for decision-making, monitoring, and managing the drilling process. One significant challenge faced during CTD is the occurrence of sensor failures in MWD tools, which hinders the real-time assessment of downhole conditions. Such failures can lead to operational downtime due to the need to trip out the bottomhole assembly (BHA) for sensor replacement. To address this issue, an AI based model using synthesized downhole measurements as backup for MWD tools has been developed for CTD. The proposed AI model leverages custom regression models for each MWD sensor, using various machine learning techniques such as Random Forest, Gradient Boosted Trees, and more to predict sensor values when a sensor fails. The model is continuously updated with new data and uses predefined thresholds and AI-based models to detect sensor failures and assess the uncertainty of predictions. To evaluate the model's effectiveness, various machine learning models are compared using metrics such as mean absolute error (MAE), and r-squared score (R2). The results indicate high accuracy in predicting sensor data, even in the absence of failed sensors, for annulus pressure, pipe pressure, and weight on bit (WOB). This approach could reduce nonproductive time (NPT) and costs associated with sensor failures in CTD operations. It provides a robust framework for using AI to maintain uninterrupted drilling operations by synthesizing sensor data when needed, ensuring the seamless execution of drilling operations.