Evaluating Profitability of Hybrid Approach for Early Stuck-Sign Detection: Analyzing False Alarms and Quantifying Reduction in Nonproductive Time

Abstract

Abstract This study investigates the profitability of early sign detection of stuck pipes by considering the trade-off between reducing nonproductive time (NPT) through detection and increasing NPT through false alarms. Our objectives are to propose a hybrid approach combining physics-based knowledge and data science, analyze the timing of false alarms, improve the model to minimize false alarms, and evaluate profitability by quantifying NPT reduction. Our proposed detection method combines physics-based knowledge and data science techniques, specifically capturing torque and standpipe pressure increases. First, we enhance the method by refining the model equation, learning method, and anomaly calculation method. Second, using field data from multiple wells collected over several months, we analyze the timing of false alarms and further refine the model to reduce them. Finally, to evaluate the profitability of our method, we examine the frequency of stuck incidents, NPT owing to stuck incidents, and NPT owing to false positives, and then quantify the reduction in NPT, considering both true-positive and false-positive rates. By applying our method to 11 wells, we generated approximately 130 days of stuck risk output. By determining a threshold that detects 40% of the stuck signs, we identified 37 false alarms. Analysis of these false positives revealed that nine could potentially indicate stuck signs, 23 could be disregarded by the operator or filtered out through data preprocessing, and five persist as challenging cases for the current method. Based on these findings, we further enhanced the model to reduce false alarms and successfully reduced the count to 11. In addition, our profitability calculation, based on NPT and considering the trade-off between true positives and false positives, demonstrates the potential for a reduction of several hours per year. Furthermore, implementing stuck sign detection is expected to lead to cost savings associated with bottom hole assembly losses, fishing, and sidetrack operations. The novelty of our research lies in evaluating the profitability of early stuck sign detection by analyzing false alarms using multiple wells and long-term data. This analysis enables us to enhance the detection model and demonstrate the profitability of the proposed hybrid approach. Our study emphasizes the importance of considering both the true-positive and false-positive rates to enhance and evaluate the performance of early stuck sign detection methods.