Stuck pipe incidents remain as one of the major problems in the drilling industry. The incidents will lead to expensive loss time in daily spread cost, bottom hole assembly cost, sidetracking cost as well as fishing cost. The Wells Augmented Stuck Pipe (WASP) Indicator, a state-of-the-art machine learning technology that seamlessly integrates with PETRONAS existing technologies, is introduced as the stuck pipe prevention detection system for the company. Historical real-time drilling data and stuck pipe incidents reports between 2007 and 2019 are used for the development of machine learning models. The models utilize key drilling parameters such as hookload and equivalent circulating density (ECD) to predict and analyze trends to detect any signature pattern anomalies for various stuck pipe events. The prediction and alarm are displayed in real-time monitoring software to trigger the operation team for prompt intervention. The WASP solution has demonstrated proven outcomes using historical and live well with high confidence in detecting stuck pipe incidents due to differential sticking, hole cleaning, and wellbore geometry. The WASP Indicator is envisaged to provide the company with cutting edge advantages in the industry. It is expected that the system will reduce the identification period and improve the reaction time of the monitoring specialists in recognizing the stuck pipe symptoms and highlighting potential incidents. The system is also bringing value to the company via non-productive time (NPT) cost avoidance and identification of early onset of various stuck pipe events based on distinct mechanisms. With the system, the existing portfolio value can be enhanced via setting dynamic trends and models into historical experiences context. The WASP Indicator is aspired to be the forefront innovation that will leap through the norm and lead the region in a greater plan of drilling automation system.

Despite many drilling technology improvements during recent years, hole cleaning remains a significant challenge. The variation of equivalent circulation density (ECD) is a symptom of borehole instability. Therefore, the ability to accurately estimate ECD is a key consideration for preventing hole cleaning problems that may lead to a stuck pipe, and well pressure management more generally. In this work, we demonstrate a Machine Learning approach to estimating downhole ECD in real-time using a deep neural network. Surface measurements that are widely available from most rigs are used as the model inputs, hence less configuration information is required relative to hydraulic simulations for pressure loss. Mean Absolute Errors of ~0.3-0.4 ppg were achieved on 16 validation wells and 7 holdout wells (blind test); these wells were independent of those in the training data. Prediction errors often reflect offsets between reference and predicted values; however, even with these offsets, trends in ECD behavior can still be captured correctly. The model shows promise for real-time ECD monitoring purposes to complement existing numerical methods and downhole tools. Beyond real-time estimation, other applications could include forecasting ECD a short time ahead to provide early indications of hole cleaning issues; case studies obtained from a real-time monitoring centre where this approach is used are presented as part of this work. The software tool was capable of detecting such symptoms in advance, giving the driller opportunity to take preventive actions to avoid a potential stuck pipe.

Wellbore geometry stuck pipe mechanism occurs when the string and the well are incompatible with each other. This issue is commonly related to changes in hole diameter, angle, and direction associated with symptoms such as mobile/swelling formation, undergauged hole, key seating, ledges, and high doglegs. An internal study identified that many stuck pipe incidents were associated with mechanical sticking, specifically wellbore geometry sticking with high-cost impact, which warrants proactive prevention. Throughout this paper, we provide and demonstrate how machine learning solutions can foresee the potential stuck pipe related to wellbore geometry issues based on two signs: hookload signature and dogleg severity. The application is based on the Artificial Neural Network (ANN) approach that reads the surface parameters sequence of hookload real-time data and learns with historical wells data. Machine learning (ML) then determines how the hookload behaves for each type of activity (tripping and drilling). The machine learning predictions can then be streamed on a web-based application accessible to the operations and project team. The neural network design for hookload prediction while tripping in/out considers a drag when the string moves towards a region with doglegs severity higher than the threshold chosen based on engineering judgment. This paper also discusses applications beyond real-time estimation, such as predicting the trend of the few subsequent expected hook loads up to 6 to 10 stands ahead based on case studies from previous live wells obtained from the real-time monitoring center where the product is used. The output from the machine learning solution provided a basis for risk identification and further analysis by the monitoring specialist in a proactive intervention effort to prevent stuck pipe incidents. The implementation of applications described in this paper could detect an early symptom of wellbore geometry issue; hence proactive action can be taken to avoid a potential stuck pipe event.