This paper presents a generalized workflow for well integrity risk assessment that can be applied to both active and abandoned wells in any region. Wells with compromised integrity can release greenhouse gases (GHG) into the atmosphere and contaminate groundwater and soil at the surface. Well integrity risk assessment, therefore, plays a critical role in managing oil and gas wells throughout their entire lifecycle, including the abandonment stage. This study presents a data-driven workflow for projecting wells onto a risk contour map, utilizing a leaking risk gradient as an indicator. The methodology calls for utilizing various feature-rich datasets typically obtained from multiple sources. The dataset used in this study for the purpose of demonstration is a publicly available dataset consisting of 23067 wells in the San Juan Basin in the State of New Mexico. The inputs considered for the risk assessment model based on machine learning (ML) included well longitude, latitude, total injected or produced fluid throughout the well's lifetime, distance to seismic activity, well lifespan (age), TVD, surface casing bottom depth, production casing bottom depth, completion length, oil price, sacks of cement used to cement the surface casing, abandonment date, cement thickness, and target geologic reservoir formation. Four ML models (artificial neural network, support vector machine, random forest classifier, and extreme gradient boosting classifier) were trained, with the artificial neural network classifier showing the best performance with an accuracy of 75%. This artificial neural network model was then re-trained for regression, which allowed manual threshold setting. This increased the prediction accuracy to 85%. A leaking risk contour map with risk gradient indicators was created by applying the artificial neural network regression model. This approach provides an essential tool for guidance on wells with compromised well integrity that need to be permanently abandoned or re-abandoned, and wells with sufficient long-term well integrity that may be suitable for re-purposing. Reducing uncertainties associated with well integrity and generating a risk contour map provides valuable insight for asset management of wells to be either abandoned or repurposed and empowers decision-makers to prioritize resources efficiently. This paper presents a reproducible, data-driven method for assessing well integrity risk and identifying key features for such studies.

Abstract Knowledge of the subsurface geothermal gradient is important in geothermal and high-pressure/high-temperature (HPHT) drilling and can substantially influence downhole equipment reliability, wellbore integrity, and production optimization. Conventional geothermal gradient estimation methods typically rely on dedicated temperature logging operations, which introduce additional cost and operational complexity. This paper presents a hybrid framework for real-time estimation of the geothermal gradient using data acquired during routine well construction activities. The method leverages measurements of pump rate, surface mud inlet temperature, surface mud return temperature, and/or bottomhole circulating temperature (BHCT), together with well contextual data of drillstring configuration, casing/cement program, fluid and formation properties to develop a physics-based finite-volume thermal model. An iterative optimization routine is applied to continuously update the geothermal gradient as the well deepens. The proposed approach is validated using open-source datasets from the Utah FORGE site and additional European geothermal field operations. Results demonstrate reliable performance across geothermal gradients ranging from 16.5 to 43.9 °F/kft (30 to 80 °C/km), encompassing most practical geothermal and HPHT environments. In addition, the framework shows robustness to missing or degraded temperature measurements and achieves computational speeds 10-100x faster than real time, enabling deployment for real-time rig-site geothermal gradient prediction and operational decision support.

Summary This paper describes the results of a field test, conducted at a test rig, of a hole cleaning and wellbore stability monitoring system. This system encompasses a laser-based sensor and digital tools designed to process the collected data in real time. Additionally, this paper outlines the method for interpreting the processed data, including the combination of the estimated volume of cuttings recovered at the surface and the expected volume obtained with a transient cuttings transport model. The test was conducted while drilling three borehole sections at a test location in Germany. The preparation for the test included the installation of an auger system to divert the cuttings from three active shale shakers to the cuttings sensor. This sensor is equipped with a high-definition camera and a laser profile scanner, enabling the collection of both 2D images and point-cloud data (PCD) of the cuttings stream. These data were then fed into a digital tool capable of estimating the solids’ volume and morphological characteristics, identifying cavings and their shape and size, and providing an evaluation of the wellbore condition. The test revealed that the system can be successfully deployed in the field with a simple rig-up procedure. Additionally, it showed that the digital tool and the underlying machine-learning techniques can accurately estimate the solids’ volume and their morphological characteristics. Furthermore, a minor discrepancy between the measured volume and the expected volume according to a transient cuttings transport model showed that the sensor can reliably be used to assess hole cleaning sufficiency in real time. During these tests, artificial cavings were introduced to evaluate the system’s capability of cavings detection and characterization, the latter involving the shape of cavings, which is related to the wellbore failure mechanism that created them. The results indicate that the system can be used to evaluate the in-depth nature of wellbore stability in real time. This, in turn, will allow for targeted corrective action to deal with stability-related operational problems. Finally, we also identified limitations of the current system, such as the generation of nonoptimal segmentation results when the cuttings’ size ranges from medium to coarse sand size (0.25–1 mm). Based on these limitations, we have defined a plan for future improvement, further described in detail in this paper. Previous field testing of the sensor only evaluated the sensor’s readiness for deployment and the software’s capability to reliably measure bulk volume—largely without correction for void space. The tests presented in this paper are more comprehensive. They evaluated the system’s capabilities to measure the corrected solids’ volume, their morphological characteristics, and the integration of a transient cuttings model into the system for a holistic assessment of the well condition.

Abstract With advances in sensing technology, high frequency torsional oscillations (HFTOs) have been validated to be a significant cause of bottom hole assembly (BHA) component damage. There is, however, a shortage of understanding on the nature of HFTO origination in the drill bit and its transference to the BHA. This study presents a novel approach to integrate bit-rock interaction (BRI) and BHA models to investigate the emergence, propagation, and dynamic impacts of HFTOs. A high-fidelity BRI model was first developed to simulate the cutting loads on the polycrystalline diamond cutters (PDC) during the penetration of heterogeneous formations. The frequency spectrum from the BRI simulation was validated against bit vibration data collected in lab tests. The output from the BRI model is then provided as input to the BHA finite element model for HFTO mode analysis. Dynamic responses of the BHA, especially at signature BRI and BHA resonance frequencies, are adopted as key metrics to evaluate the HFTO damage risks of different BHA components. For BRI modeling, each rotation of the bit is discretized into hundreds of steps and Fourier transformation is utilized to generate the spectrum of the cutting load (2000 Hz in this study). The BHA was axially discretized at intervals of 0.1 ft for torsional FE modeling, capturing changes in BHA radial dimensions, components, and mechanical properties. Then, the transfer function from the bit to different BHA locations (dynamic responses) is obtained for a frequency range up to 1000 Hz. Analysis indicates that torsional BHA resonance and mode shapes are critical to determining the instantaneous HFTO strength and long-term HFTO-related fatigue damage in different BHA components. Validation against lab data confirms that the BRI model captures the appropriate BRI load frequencies. Combining physics-based high-resolution BRI and BHA models, this work delivers a novel and robust dynamic characterization of HFTO origination and propagation. The analysis framework technically explains the HFTO risk differences between bit-BHA designs and provide guidance for bit selection, optimization of bit/BHA designs, and development of effective HFTO mitigation strategies. This paper thereby extends the industry's understanding of HFTO and provides a modeling framework for its mitigation.

Abstract Geothermal drilling operations are highly sensitive to coupled well temperature and pressure dynamics that influence well integrity, fluid properties, downhole tool reliability, and overall drilling efficiency. This paper presents a control-oriented well thermodynamics model and a temperature-pressure controller developed for an 11,500-ft geothermal test well. The model captures dominant heat-transfer between the mud in the drillstring, annulus and surrounding formation through a lumped-parameter formulation. Validation against high-fidelity thermal-hydraulics simulators confirms that the reduced-order model can accurately predict the transient and equilibrium well temperature profiles within minutes. Based on this model, proportional-integral-derivative (PID) controllers were designed and tuned for both inlet mud temperature and mud flow rate across a wide operating range (70°F to 150°F and 30 gpm to 400 gpm). Simulation results show that the integrated managed-temperature-managed-pressure drilling (MTD-MPD) framework can stabilize bottomhole circulating temperature (BHCT) within 1°F of the target in less than 30 minutes while maintaining bottomhole pressure (BHP) fluctuations within 20 psi. The combined MTD-MPD system can compensate for transient thermal-hydraulic coupling effects, ensuring safe BHCT regulation and stable BHP control. The proposed methodology effectively addresses key challenges in geothermal and other high-temperature drilling, providing a robust, efficient, and scalable foundation for real-time downhole temperature-pressure management.

Abstract While multiple solutions exist to predict stuck pipe incidents, most fall short in preventing such incidents due to limitations in accuracy, interpretability, general applicability and/or timeliness of incident warnings. This paper proposes a novel method for real-time sticking risk assessment that is interpretable, universally applicable, and comprehensive, evaluating all primary sticking mechanisms. The approach employs multiple hybrid agents (combining machine-learning and physics-based models), and incorporates real-time simulation of key drilling components, including hole cleaning, pipe-borehole interaction, and rock mechanical behavior. The proposed model comprises hybrid physics + AI agents that operate in real time, processing both sensor signals and contextual (non-streaming) data to compute a sticking risk index. This index reflects the presence of historical sticking data signatures and physics-based indicators of various sticking mechanisms. These indicators rely on simulations of the drilling operation—specifically of the cuttings transport process, pipe-borehole interaction, and rock mechanical behavior around the borehole—to identify conditions that increase the risk of sticking, such as insufficient hole cleaning or borehole instability, and to detect early symptoms like elevated torque and increased drag after connections. The initial evaluation of the model was conducted using historical data from wells in Utah FORGE and the Gulf of Mexico. Results showed that the model could anticipate sticking incidents—by at least 22 minutes—involving various mechanisms, including geometric sticking and annular pack-off. Its performance was evaluated using modified versions of recall, precision, and F1-score, yielding encouraging scores for each metric. Additionally, the model successfully identified key contributing factors behind elevated sticking risks, serving as an interpretable tool to guide preventative actions. These results suggest the following: First, the model has general applicability and is not limited to a specific well type or geographic region. Second, modeling the associated uncertainty through a fuzzy perspective of stuck pipe further enhances interpretability by allowing the severity of the risk to be assessed. Third, combining anomalous pattern recognition—to identify previously observed signs of sticking—with physics-based evaluation of multiple mechanisms, such as insufficient hole cleaning and induced shear rock failure, proved essential to accurately estimating the sticking risk and maintaining general applicability. Finally, the model can serve as the real-time guiding backbone of a fully automated drilling solution by evaluating current risk levels and simulating the outcome of different preventative actions. The proposed method is the first to evaluate all primary sticking mechanisms—including annular pack-off from borehole instability and insufficient hole cleaning—by combining physics- and machine learning-based models in a multi-agent system. It assesses sticking risk thoroughly and in real time while identifying the contributing factors, offering critical insights for effective prevention. Additionally, the model can support the development of autonomous drilling systems informed by AI—a key objective actively pursued by the industry.

Drilling and other well construction operations in high-temperature geothermal wells face a fundamental challenge: preventing downhole tool failure caused by exceeding temperature limits. Tripping into such wells needs to be staged to lower the possibility of thermal tool damage. This study investigates the bottomhole assembly (BHA) temperature evolution, cooling effectiveness, and operational design of staged trip-in practices in geothermal and other high-temperature wells. A thermo-hydraulic modeling framework is developed, combining a full-well finite volume model (FVM) with a lumped BHA-wellbore model, to capture transient well thermodynamics during drilling and staged trip-in operations. Model validation using Utah Forge Well 16B(78)-32 data shows that the root mean square error (RMSE) of bit/BHA temperature prediction ranges from 4 degrees F (2.2 degrees C) to 8 degrees F (4.4 degrees C). Sensitivity analyses demonstrate that the maximum stage length remains under 4-5 stands when tripping into wellbores with near-field formation temperatures in the range of 250 degrees F (121 degrees C) to 320 degrees F (160 degrees C) unless significant well geometry or mud property changes occur. The only strategy that consistently extends downhole sensor survivability beyond 8-10 stands is BHA external thermal insulation. Simulation results demonstrate that adding a field-proven 0.15 in (3.8 mm) coating with thermal conductivity of 9 BTU.in/hr/ft2/degrees F (1.30 W/m/K) can reduce BHA temperatures by up to 30 degrees F (17 degrees C), compared to unprotected configurations under these downhole conditions. The modeling and analysis can also help identify scenarios where staged circulation is insufficient and continuous circulation (i.e., circulation while making connections) is required to maintain safe tripping BHA temperatures. These findings provide practical and insightful guidance for the design of effective cooling strategies during geothermal and high-temperature oil and gas well drilling and tripping operations, ensuring safer and more efficient operations in extreme downhole thermal environments with a lowered risk of BHA component failure.

Drillstring vibrations (axial, torsional, and lateral vibrations) occurring at various frequencies can significantly impact drilling performance and drilling equipment lifespan. Controlling excessive and (self-)sustained vibrations has become more challenging with increasing well depth and complexity. These vibrations primarily arise from the dynamic interactions between the drill bit, drillstring, wellbore, and surrounding formations. The ability to promptly identify and effectively control these vibrations presents a valuable opportunity for optimizing drilling operations and lowering drilling costs.This paper provides a comprehensive review of the dynamics of tri-axial drillstring vibrations, their modeling techniques, as well as tools and strategies to practically detect, analyze, and control them. Specifically, the axial, torsional, and lateral drilling vibration modes exhibit distinct patterns but are strongly interdependent due to structural and environmental coupling. Various modeling approaches, including lumped parameter method (LPM), finite element method (FEM), multibody dynamics (MBD), and continuum mechanics (CONT), have been developed to better understand detrimental drilling dysfunctions since the 1960s. Both state-of-practice tools adopted by the industry and emerging vibration mitigation strategies proposed by researchers are examined here and divided into three categories: 1) Passive drilling tools and shock subs; 2) Physics-/experience-based drilling optimization; 3) Proactive drilling tools and vibration controllers. Structures, mechanisms, and effectiveness of these technologies are analyzed and compared, highlighting their practical applications, limitations, and potential for improvement. Discussions and recommendations are presented to achieve more efficient and reliable drilling operations and guide the development of future proactive drilling vibration mitigation technologies, with the ultimate goal of achieving vibration-free drilling.

Abstract Existing methods for evaluating hole cleaning and borehole stability during drilling are often limited, relying either on model-based predictions (e.g., of cuttings accumulation downhole) with infrequent and often inaccurate surface validation, or on surface measurements of returning solids without support from robust cuttings-transport or borehole-stability models. This paper presents a system that integrates advanced modeling with a state-of-the-art sensor to diagnose potential borehole instability and insufficient hole cleaning, offering key information to assist in the prevention of stuck pipe incidents. The proposed system encompasses three main components. The first is a laser- and video-based sensor that collects 2D and 3D data off the stream of solids (cuttings, cavings) coming out of the well in real time. The second component is a digital tool that uses state-of-the-art artificial intelligence techniques to transform the collected data into information relevant for borehole condition evaluation, such as recovered cuttings volume and size distribution. The third component is a set of physics-based models—for cuttings transport and rock mechanical behavior—that provides a real-time baseline to assess whether there is an excess or deficiency of cuttings recovered at the surface and/or indications of rock failure, thereby supporting the diagnosis of poor hole cleaning and/or borehole instability. The system was tested by evaluating hole cleaning conditions in two wells where the sensor was deployed. Two tests were conducted. The first assessed the system by comparing the measured volume of cuttings against the expected volume and contrasting the resulting hole cleaning evaluation with observed drilling conditions (i.e., the presence or absence of stuck pipe indicators). The second involved deploying the integrated digital tool to evaluate its real-time applicability. This case relied on a simulated real-time feed of sensor data into the digital tool and the transmission of results via an application programming interface. The tests demonstrated that the system accurately identified hole cleaning conditions in both wells. They also confirmed that the system components—including the digital tool, which processes sensor data and physics-based simulations—can be deployed together to generate a holistic, real-time assessment of borehole conditions. Furthermore, the system serves as a foundational component for fully automated solutions for hole cleaning and borehole stability management, offering the potential to significantly reduce the occurrence of costly incidents such as stuck pipe and casing run failure. This work presents the first automatic, fully integrated system capable of providing a comprehensive real-time evaluation of hole cleaning and borehole stability. It combines an accurate and direct measurement of cuttings volume with a reliable estimate of the expected volume, thereby supporting stuck pipe prevention. More importantly, the system represents a key advancement toward fully automated hole cleaning and borehole stability management, as well as autonomous drilling—goals that are actively pursued by the drilling industry.

Traditionally, evaluating a borehole's readiness to accept casing has relied on a manual, limited, and often biased analysis of an insufficient amount of data. This approach frequently results in casing run failures, leading to high nonproductive time and cost. In this paper, we propose a digital tool to automate this analysis, providing the drilling team with a comprehensive, timely evaluation of the risk of casing run failure and an interpretation of that risk. The tool developed discretizes the hole section under analysis into small intervals. It then utilizes both time-series data (preferably 1 Hz) and contextual data to derive features associated with casing runnability for each interval. These features include the potential for cuttings accumulation, the presence of borehole undulations or excessive hole curvature, differences in rigidity between the casing and drillstring, and other factors further described in this paper. The tool then uses three models. The first model determines whether the casing run is likely to succeed; the second determines whether each interval is likely to cause severe restrictions during the casing run; and the third explains the predictions of the second model. The tool was constructed with data from 52 hole sections from deepwater wells in the Gulf of Mexico/America (GOM) and subsequently tested on five additional sections, comparing the actual casing runs with the tool's predictions. This test revealed that the model not only provides an accurate assessment of the risk of casing run failure-capturing actual risky intervals while avoiding spurious predictions-but also offers a meaningful interpretation of this risk. It identifies the location of high-risk intervals along the wellbore and the potential causes of such risks. Implementing this new interpretable tool can help reduce the frequency of casing run failures and the associated costs. Additionally, it can help avoid risks in subsequent hole sections caused by setting the casing above the planned depth, like the risk of getting stuck in the rathole or drilling the next section with a reduced kick tolerance. Ultimately, the drilling team is enabled to make better-informed decisions before and during the casing run. They can decide whether a conditioning trip is required and/ or whether to outfit the casing with reaming/drilling features. Furthermore, they can avoid inefficiencies caused by implementing overly conservative measures, like superfluous hole conditioning trips. The novelty of this tool is twofold. First, it is the first tool that can indicate the location of risk along the wellbore in addition to providing a risk evaluation. Second, it integrates data-driven and physics-based models, in a hybrid approach implemented only to a limited extent in past studies. This tool represents a significant advancement over the traditional approach, not only in the volume of data it processes but also in its accuracy and interpretability.

In drilling, the coupling between drillstring dynamics and drilling fluid hydraulics is instrumental to the excitation of vibrations in the drilling system and the applicable boundary conditions for such excitation. However, most of the existing drillstring models fail to capture the dynamic-hydraulic interactions and their impact on drilling performance. To address these challenges, comprehensive computational fluid dynamics (CFD) simulations were conducted with the finite volume method (FVM) to estimate the hydraulic forces on drillstrings under diverse conditions. Coupled lateral-torsional movements of a drill pipe were reconstructed and approximated as the moving boundaries of the CFD model. With the hybrid/dynamic mesh, the proposed CFD model can effectively calculate the real-time annulus fluid velocity field and hydraulic forces with improved accuracy. The simulations span a wide range of drilling scenarios, exploring different drillstring whirling and rotational frequencies, with varying flow rates of Newtonian and non- Newtonian fluids in 2D and 3D domains. The analysis results provide a detailed view of how the drillstring dynamics can affect the drilling fluid hydraulics and can be utilized to guide and improve drilling operations in the field.

Abstract Drill pipe (DP) damages, caused by excessive loads and suboptimal drilling operations, can lead to undesired drilling dysfunctions, extra non-productive time, and increasing pipe repair expenses. Identifying the root causes of DP damages and their correlation to the drilling programs can help optimize the well/drillstring design as well as the drilling operation and ultimately reduce the risk of severe drillstring failure. In this study, over 300 DP inspection reports (over 170,000 joints of 5" and 5.5" DPs) and drilling data from over 1,000 wells are investigated to summarize the average connection damage rates and DP thickness loss. Damage decomposition among thread, seal, hardband, reface, damage beyond repair (DBR), and thickness loss are concluded for different pipe sizes, formations, and contractors. Furthermore, operation envelope, mud type, vertical/lateral footage, lithology, drillstring structure, and other well designs are selected as essential features for the damage-operation correlation analysis. Cases of severely damaged strings are investigated as well. Based on the analysis results, a comprehensive understanding of DP damage rates and mechanisms has been achieved to practically optimize drilling jobs, reduce drillstring failure risks, and cut drilling expenses. A DP analysis automation pipeline is also built to make the analysis results evergreen.

Abnormal pressure loss during a drilling operation signals a failure in the fluid hydraulics circulating system. This could be due to a washout in the drill pipe (either in the body or connection), bottom hole assembly (BHA) component or connection, downhole tool failure, surface mud pump problem, or losses into the formation. This paper describes a machine learning approach to accurately flag abnormal pressure losses and identify the root cause. An operational procedure to prevent twist offs once abnormal pressure loss is flagged is also outlined. The primary method used to flag abnormal pressure loss is to compare the standpipe pressure to a statistically modeled pressure calibrated using past data from the same well. Using the standpipe pressure, statistically modeled pressure, weight on bit, RPM, and flow in, an abnormal pressure loss belief is then calculated. The belief tracks standpipe pressure trends when WOB, RPM, and flow-in signals are relatively constant. Once an abnormal pressure loss belief has been identified, further analysis is performed using contextual data, such as survey information and BHA components, to identify the root cause. This study included learning from over 100 wells on which the approach was deployed, and abnormal pressure loss was flagged. Washouts due to drill pipe and BHA connection failures and BHA failures, such as seal failures, have been shown to periodically result in abnormal pressure drops throughout a BHA run. Factors such as wellbore geometry may cause the seal/joint to open or close during the BHA run. This is reflected as drops in pressure under stable drilling parameters, such as weight on bit, RPM, and flow in, followed by a pressure increase back to normal. These periodic abnormal pressure decreases continue until the drill pipe or BHA is changed or fully twisted off. However, for drill pipe body washouts, the pressure tends to bleed off/decrease over a longer period in a more consistent manner. The ability to identify a washout in the drill pipe or BHA can be immensely helpful in preventing a twist off thereby eliminating significant non-productive time. This paper outlines a novel automatic abnormal pressure loss alert system, a process for differentiating between drill pipe body, drill pipe/BHA connection and tool failures, and an operational procedure to handle such alerts when they are flagged.

Stuck pipe events continue to be a major cause of nonproductive time (NPT) in well construction operations. Considerable efforts have been made in the past to construct prediction models and early warning systems to prevent stuck pipe incidents. This trend has intensified in recent years with the increased accessibility of artificial intelligence (AI) tools. This paper presents a comprehensive review of existing models and early- warning systems and proposes guidelines for future improvements. In this paper, we review existing prediction approaches on their merits and shortcomings, investigating five key aspects of the approaches: (1) the time- frequency and spatial bias of the data with which the models are constructed, (2) the variable space, (3) the modeling approach, (4) the assessment of the model's performance, and (5) the model's facility to provide intuitive and interpretable outputs. The analysis of these aspects is combined with advancements in anomaly detection across other relevant domains to construct guidelines for the improvement of real- time stuck pipe prediction. Existing solutions for stuck pipe prediction face numerous challenges, allowing this problem to remain unsolved in the broad scope of progressing drilling automation. In our analysis, we looked at notable approaches, including decentralized sticking prediction, sophisticated data- driven models coupled with explanation tools, and data- driven models coupled with physics- based simulations (hybrid sticking predictors). However, even these sophisticated approaches face challenges associated with general, nonspecific applicability, robustness, and interpretability. While the best approaches tackle some of these challenges, they often fail to address all of them simultaneously. Furthermore, we found that there is no standardized method for assessing model performance or for conducting comparative studies. This lack of standardization leads to an unclear ranking of (the merits and shortcomings of) existing prediction models. Finally, we encountered cases where unavailable information (i.e., information that would not be available when the model is deployed in the field for actual stuck pipe prediction) was used in the models' construction phase (referred to here as "data leakage"). These findings, along with good practices in anomaly detection, are compiled in the form of guidelines for the construction of improved stuck pipe prediction models. This paper is the first to comprehensively analyze existing methods for stuck pipe prediction and provide guidelines for future improvements to arrive at more universally applicable, real- time, robust, and interpretable stuck pipe prediction. The application of these guidelines is not limited to stuck pipe prediction and can be used for predictive modeling of other types of drilling abnormalities, such as lost circulation, drilling dysfunctions, and so on. Additionally, these guidelines can be leveraged in any drilling and well construction application, whether it is for oil and gas recovery, geothermal energy, or carbon storage.

Managed Pressure Drilling (MPD) enhances drilling efficiency and safety by enabling precise control of wellbore pressure. However, in the drilling of high-pressure-high-temperature (HPHT) and geothermal wells, downhole tool failure due to elevated temperatures necessitates active and efficient management of downhole temperature. To address this challenge, a physics-based pressure-temperature management framework integrating a Reduced Drift Flux Model (RDFM), thermodynamic equations, and proactive control algorithms is developed in this work to capture the transient behavior of both downhole pressure and temperature and achieve real-time control of these parameters. The Sparse Identification of Nonlinear Dynamics (SINDy) method is applied to synthetic datasets generated from a high-resolution RDFM, preserving the underlying pressure-temperature physics. Furthermore, a multivariable control framework based on Model Predictive Control (MPC) is developed using the reduced-order model to simultaneously regulate bottomhole pressure (BHP) and maintain the desired bottomhole circulating temperature (BHCT) by manipulating choke opening, mud flow rates, and inlet mud temperature. Simulation results demonstrate that the proposed control framework can effectively maintain the BHP within the narrow pressure window while satisfying the downhole temperature constraints.

Abstract The identification of the well/rig state in time is a key component in the construction of accurate risk-assessment, event-detection, and efficiency-tracking tools in drilling and production operations. Traditionally, this state identification has relied on insufficient rule-based systems, which often results in inaccurate predictions and leads to non-reliable risk-assessment tools, imprecise event detection, and biased estimated efficiency. This paper compares three state-of-the-art, scalable methods for automatically identifying the well/rig state and presents three use cases in drilling and production stages. Identifying the well/rig state is a time-series multi-class classification problem, in which the data is collected at high frequency (typically 0.02–1 Hz) with sensors installed inside the well, on the wellhead, or on the equipment intervening the well, such as a rig or a coiled-tubing unit. This paper presents three solutions to this classification problem, namely a first-order logic inference system, a recurrent neural network (RNN) classifier, and a transformer-based classifier. We implement and compare these methods in three applications in drilling and production operations, including the detection of stuck pipe incidents and pressure trend abnormality. The models were evaluated on a withheld, pre-labeled test dataset consisting of 75 hours of 1-Hz drilling data and 1072 h of 0.17-Hz production data. This evaluation showed that the transformer-based classifier outperformed the other three methods in all three applications. Additionally, we observed that the deep learning-based classifiers were only slightly more computationally expensive than inference systems, making all models suitable for real-time prediction. In the test, we also investigated the value of accurate well/rig state identification in the assessment of drilling and well-integrity risks. The test revealed that the lack of accuracy in this identification task can severely bias the risk assessment, for instance, by overestimating the risk of pipe sticking and generating spurious predictions of abnormal pressure trends. This, in turn, can result in unnecessary, costly preventative measures. The applicability of the analyzed methods is not limited to the examples here provided; they are applicable to any task requiring the classification of time-series data and thus, can be employed in other well operations/stages and applications of event detection. The novelty of this paper is twofold. First, it is the first study to compare various methods for automatically identifying well/rig state, including one not evaluated before—a transformer model. Second, it is the first to propose an automated engine for well state identification during the production stage. This paper also analyzes the advantages and shortcomings of well/rig state identification models through three different applications.

Abstract Coupled tri-axial drillstring vibrations are widely recognized as major causes of compromised drilling efficiency and safety. Excessive and self-sustaining drillstring oscillations, occurring at different frequencies, can result in premature drilling system failures, bit wear/damage, compromised hole quality, reduced rate-of-penetration (ROP), and non-productive time (NPT). In this study, a novel distributed in-bottom-hole-assembly (in-BHA) drilling vibration control system is proposed to robustly and efficiently suppress coupled drilling vibrations once they are detected. Multi-input-multi-output (MIMO) in-BHA controllers with local state-/output-feedback were designed following robust and optimal control theories. Lower drillstring dynamics and downhole drilling parameters are utilized to simultaneously control the at-bit weight-on-bit (WOB) and rotation speed (RPM). Ideally, the proposed control system requires no communication with surface once deployed, which significantly accelerates system control response action and expands the control bandwidth. Comprehensive simulations are performed to illustrate that the proposed controllers outperform existing drilling vibration mitigation technologies in terms of robustness, stabilization time, energy consumption, and actuation requirements.

Abstract Controlled Mud Level (CML) offers enabling managed pressure drilling (MPD) technology for drilling deepwater wells with narrow pressure windows. CML technology is now being extended with the addition of Surface Back Pressure (SBP) technology, creating a "dual MPD" technology. However, modeling CML operations in combination with SBP functionality and control has been difficult due to the lack of robust hydraulic models. This study presents a CML multi-phase drift-flux model with added SBP functionality, incorporating such features as the subsea pump module (SPM) operation, mud return line (MRL), SBP choke, and an active control device (ACD) placed in the marine riser. The developed hydraulic model significantly extends an existing model for SBP Managed Pressure Drilling (MPD) operations to include CML operations. In this model, the riser is subdivided into a lower riser section below the ACD, an upper riser section located above the ACD, and the section associated with the mud return line (MRL). The SPM, which features automatic speed control, acts as a sink upon the lower riser, adjusting the riser level to regulate the target bottom hole pressure PBHP. The built-in ACD can be used to restrict flow between riser sections and can be used to regulate pressure from the lower riser section for SBP application and influx management. A surface choke connected to the MRL outlet allows the system to adjust PBHP by adding SBP. Future work will extend the model to kick management. The use of SPM and MRL sub models in the hydraulic model have resulted in accurate replication of CML and CML+SBP drilling operations under both dynamic and static conditions. Simulations furthermore show that switching to CML with SBP application during static periods optimizes connection procedures, thus reducing connection time and enhancing overall operational efficiency. These results showcase the model's potential to significantly enhance the planning and execution of deepwater dual MPD operations, expanding opportunities for drilling deeper, lower-margin wells. This work represents the first integrated multi-phase hydraulic model that seamlessly integrates CML and CML+SBP modes, with the capability to transition seamlessly between them. Furthermore, it is the first to fully integrate and reflect the performance of a CML+SBP MPD system, with the integration of the SPM, the MRL, the surface choke, and an ACD.

Drilling is essential to extract subsurface hydrocarbons and geothermal energy, or store waste fluids/gases underground. Drilling efficiency directly affects the economics of well construction, and is to a large extent determined by the non-productive time (NPT) associated with downhole tool failure, bit wear/damage, unexpected drillstring vibrations, and insufficient surface-to-bottom energy transfer. Various models have been developed since the 1960s to understand and mitigate undesired drilling system oscillations to ensure efficient drilling operations and help optimize the well design. Significant recent advancements in drilling engineering technology, sensor development, and data science make it possible to develop and apply a faster and more comprehensive drilling system dynamic model that can be used for real-time drilling optimization and automation. In this study, a novel control-oriented physic-based drillstring dynamic modeling framework is developed using 3D field-consistent corotational beam elements with an updated Lagrangian formulation. The structure and element kinematics are firstly derived in global and local coordinate systems respectively, based on which the system dynamics are formulated by integrating the element stiffness, damping, and mass matrices. Different drilling boundary conditions and forms of external loads, including bit-rock interaction, wellbore-drillstring contact, stabilizers, bottom hole assembly (BHA) eccentricities, and gravity/buoyancy are defined within the modeling framework. The numerical accuracy of the model with linear/quintic beam elements is discussed and verified in four different static/dynamic loading cases, which are typically used for beam analysis. Simulations of different drilling scenarios, including dynamics during the pipe connection process, bit on/off bottom process, and pipe rocking during slide drilling, are conducted for an actual L-shape well configuration with comparison to field data. The simulated drillstring tri-axial vibrations exhibit diverse characteristics and patterns in terms of oscillation amplitudes, frequencies, and modes at varying depths, generating valuable understanding of potential drilling dysfunctions and insights for real-time drilling optimization and automation.

In the exploitation of subsurface hydrocarbon and geothermal energy, more than 50% of total expenses are typically attributed to drilling and well completion. Coupled triaxial vibrations in the drilling system, excited by various mechanical, hydraulic, and geological sources, are major causes of premature downhole tool failures, excessive bit damage, compromised hole quality, reduced rate of penetration (ROP), and increased nonproductive time. Drillstring dynamic models, shock absorption tools, and controllers have been developed since the 1960s to understand and mitigate detrimental vibrations in the drilling system, aiming for smoother and more efficient operations. With advancements in measurement/logging while drilling technologies, rig equipment, mud pulse/wired pipe telemetry, and the digitalization of drilling engineering, a foundation has been established for proactive drilling vibration detection, analysis, and control to achieve efficient and safe drilling. In this study, a reduced-order axial-torsional drillstring dynamic model with soft string lateral contacts is first developed. Nonlinear drilling boundary conditions and inputs, such as bit-rock interactions (BRIs) with torque-on-bit (TOB) velocity-weakening effects, wellbore-drillstring contacts/friction, drillstring gravity, mud buoyancy, and more, are defined within the modeling framework. Equilibrium states and linearization of the model are derived for any given weight-on-bit (WOB) and rotational speed (RPM) setpoints, 3D well trajectories, and drillstring dimensions. A linear-quadratic-integral (LQI) controller is developed and tested for drillstring vibration suppression using a 6,562 ft (2000 m) drillstring model. In comparison to commercially available controllers for stick/slip mitigation, which are designed based on impedance matching and only address the decoupled drillstring torsional dynamics, the proposed state-feedback LQI controller emphasizes the coupling between the axial and torsional drillstring dynamics by simultaneously controlling WOB and RPM. Simulation results demonstrate that the LQI controller can suppress stick/slip and stabilize downhole WOB within 7 seconds, while a commercial drilling controller takes more than 35 seconds to achieve the same. Additionally, valuable insights and potential future directions for nonlinear drillstring vibration mitigation and drilling controller design are provided. With its fast drilling dysfunction stabilization time, small downhole RPM/WOB overshoots, minimal rig control inputs, robustness against drilling nonlinearities, and high efficiency, the proposed multi-input-multi-output state-feedback drilling controller holds great potential for application in proactive drilling vibration control, drilling automation, and real-time drilling optimization.