Systems and methods for optimizing rate of penetration utilizing automated safeguards against drilling dysfunctions and incidents

Abstract

The disclosure describes methods and systems for optimizing the rate of penetration (ROP) in real-time during drilling operations for subterranean wells using a rotary drilling rig, while maintaining operational safety using one or more generalized safeguarding systems coupled to the ROP optimizer. The method can be used in either an advisory capacity, or as part of a rig automation system through integration with a computer-based rig control system. The method involves two broad classes of safeguarding systems, which can be implemented as computational modules that communicate with the main ROP optimization system. The first class of safeguarding systems are based on real-time drilling dysfunction detection to identify risks occurring during the drilling operations. The second class of systems are based on scenario modeling and analysis. This involves use of computational models to estimate the expected effect of changing drilling parameters with respect to one or more operational risks.

Description

TECHNICAL FIELD

[0001] The present disclosure relates generally to well drilling using a rotary drilling rig and well operations. More specifically, this disclosure relates to a method and a system using machine learning models to optimize a rate of penetration (ROP) of the drilling operation, to detect drilling dysfunctions or operational risks in real-time to allow mitigation, and to model scenarios with modified drilling parameters in order to prevent drilling dysfunctions from being induced by implementing recommendations for drilling parameters to increase ROP.BACKGROUND

[0002] Drilling and well operations in oil and gas wells are expensive activities. The typical operating spread costs can range from several tens to several hundred thousands of US dollars per day. Prevention of drilling incidents causing non-productive time, mitigating drilling dysfunctions that reduce drilling efficiency, and optimizing the rate of penetration to reduce bit-on-bottom time, all contribute to reducing well construction and operational costs. However, these objectives may be competing in many cases, for example increasing rate of penetration also increases the rate at which cuttings are generated, which increases the risk of poor hole cleaning and subsequent pack-offs and stuck pipe incidents potentially resulting in extensive non-productive time. Hence, effective drilling optimization requires a holistic approach balancing faster drilling with risk avoidance and mitigation, in order to minimize overall well construction time.

[0003] ROP optimization methodologies and systems utilizing machine learning or physics-based models in prior art are typically limited in their applicability to specific oilfields, formation types, particular intervals within wells, or in some cases by modelling assumptions made in their designs or requirements for downhole measurements as inputs, or use of specialized equipment. These limitations require models embedded in these systems to be customized and pre-trained on local offset well data, or in some cases specific depth ranges within these offset wells. Some prior disclosures presented methods where multiple models defined for specific depth ranges were required, further hindering applicability and substantially increasing the model management burden. Out-of-the-box usage on new operations is not possible under these scenarios, adding considerable (and expensive) pre-project work in order to utilize such systems, and preventing usage on exploration wells where no local offset data may be available. Furthermore, many prior methodologies consider only the goal of increasing drilling speed, without factoring in potential hazards and risks of drilling dysfunctions that may be exacerbated by high rates of penetration. In this disclosure, a methodology using only surface-measured data as inputs, and utilizing embedded machine learning models which a pre-trained and validated for generalizability against independent (out of sample) wells that replicate how the system would be used on unfamiliar fields or exploration wells. The disclosed methodology and system integrates multiple component systems for holistic optimization balancing drilling speed and operational safety, including those for drilling parameter advice intended for increasing ROP, scenario modelling for potential drilling hazards, real-time drilling dysfunction detection (for example relating to vibration), and modelling potential risks for equipment failure (such as downhole motors) based on the observed operational usage.SUMMARY

[0004] This summary is provided to introduce a selection of concepts that are further described below in the detailed description. This summary is not intended to identify indispensable features of the claimed subject matter, nor is it intended for use as an aid in limiting the scope of the claimed subject matter.

[0005] The present disclosure introduces a method and system, in the context of well construction using a rotary drilling rig, for optimizing drilling rate of penetration (ROP) while maintaining operational safety and preventing potential problems causing non-productive time, using only surface-measured time series data from the rig site as inputs to all relevant components or sub-systems. The method includes use of a core system for identifying drilling parameters that maximize expected ROP subject to a variety of limits, which may be configured ahead of operations or dynamically determined during the drilling process, or alternatively based on real-time risk detection methods. The method utilizes two classes of safeguarding sub-systems integrated with the core optimizer, here termed “reactive safeguards” and “preventive safeguards”.

[0006] Reactive safeguards monitor real-time or historical data for prompt detection of symptoms of drilling dysfunctions, and notify users or other software systems to allow critical mitigation steps to be taken. Preventive safeguards perform scenario modelling to estimate the potential impact of proposed drilling parameter changes on other key parameters relating to operational safety or drilling dysfunctions, in order to pro-actively avoid issues by imposing dynamic safety limits. Particular embodiments of both the reactive and preventive safeguards are targeted at mitigation or avoidance of specific drilling dysfunctions or operational hazards. The disclosed method and system is modular and extensible, allowing any number of safeguarding systems to be integrated with the core drilling optimization system, according to operational requirements.

[0007] All sub-systems and components that involve modelling (scenario based or risk detection) utilize generalized or “global” pre-trained machine learning models for estimations, which are not limited in their utility to specific fields, regions, formation types, BHA or bit types, or intervals within particular well. The aforementioned models are trained using only surface-measured data, given its routine availability in drilling operations, to support widespread usage of the system. The models' generalization capability and dependence only on highly available input data enables the disclosed system and its constituent models to be used directly on new and unfamiliar wells without the need for customization of models to specific fields, sets of wells or intervals within wells, which would otherwise greatly hinder the useability of the disclosed method and system.

[0008] Integration of a sub-system containing a model capable of estimating downhole Equivalent Circulating Density (ECD), based on the surface parameters measured by the rig sensors, allows the expected effect on ECD of changing drilling parameters and ROP to be estimated. Increases in ECD can be a symptom of poor hole cleaning, and are monitored for by practitioners in the field. The ECD estimation model utilized in the disclosed method and system is an updated version of the model described in the research publication SPE-208675-MS. Based on the expected changes in ECD modelled for different scenarios (candidate solutions), the risk of hole cleaning issues can be quantified, and a set of drilling parameters that are expected to keep downhole ECD within a configured limit can be selected from the set of possible candidate solutions.

[0009] A prior disclosure by the Company has described a method utilizing Machine Learning (ML) techniques for detecting rig events and states through time-frequency analysis and analysis of time-series data measured at the surface, including downlinking events (used for mud pulse telemetry) and signs of wave-induced heave on floating rigs, see WO 2024 / 057230. An extension to this methodology was used by the Company to build a system for detecting symptoms of torsional vibrations (stick-slip), using surface data as its inputs, and generating estimated probabilities that the targeted drilling dysfunction is presently occurring. This system is incorporated into the drilling optimization advisory system as a reactive safeguard.

[0010] Other embodiments can involve safeguards against different drilling dysfunctions, for example lateral and axial vibration modes (in addition to torsional vibration issues described in previous embodiment), bit-wear (through estimation of changing bit wear state during operations, either proactively or reactively) and other equipment failures (mud motors, rotary steerable systems, downhole logging tools, drill string buckling).

[0011] These and additional aspects of the present disclosure are set forth in the description that follows, and / or may be learned by a person having ordinary skill in the art by reading the material herein and / or practicing the principles described herein. At least some aspects of the present disclosure may be achieved via means recited in the attached claims.BRIEF DESCRIPTION OF THE DRAWINGS

[0012] FIG. 1 is a schematic showing how the remote systems and remote data store integrate with data feeds from the rig, and sensors gathering data from rig equipment.

[0013] FIG. 2 provides a high-level overview of ROP optimization system with safeguarding modules, and the systems links with the rig-site and remote data stores

[0014] FIG. 3 is a detailed schematic and workflow describing the methodology utilized by the ROP optimization system with two types of safeguarding modules, based on real-time dysfunction detection and scenario modelling to identify optimal drilling parameters within operational constraints.

[0015] FIG. 4 is a detailed schematic describing the process used by a drilling dysfunction detection module to estimate stick-slip (torsional vibration) symptom probabilities and make this information available to other software systems.

[0016] FIG. 5 is a pair of charts showing an example time interval of surface torque measurements, and the corresponding (aggregated) power spectrum that is used as an input to the machine learning model responsible for estimating a metric quantifying the severity of dysfunction symptoms, such as a probability.

[0017] FIG. 6 is a pair of charts showing an example time interval of surface rotary speed measurements, and the corresponding (aggregated) power spectrum that is used as inputs to the machine learning model responsible for estimating a metric quantifying the severity of dysfunction symptoms, such as a probability.

[0018] FIG. 7 is a set of charts showing an example time interval where the onset of stick-slip symptoms is visible, followed by severe torsional vibrations; the first two sub-plots contain time-series data from surface RPM and torque measurements, while the third sub-plot depicts the corresponding stick-slip probabilities (used to quantify dysfunction symptom severity) for the example time period. Shaded areas indicate regions with stick-slip symptoms.

[0019] FIG. 8 is an illustrative example of a screenshot from a real-time viewer application, integrated with a remote WITSML data store, that could be used to display the drilling parameter advice for ROP optimization generated by the disclosed invention, as well as stick-slip probabilities which can be plotted in time alongside the history of parameter recommendations.

[0020] FIG. 9 is a schematic describing the process used by a drilling dysfunction detection module for estimating risk of failure of a downhole mud motor, in the form of a probability or degradation index, and subsequently publishing this information for consumption by other systems or end-users.

[0021] FIG. 10 is a pair of charts plotting the system-estimated failure risk probability for a downhole positive displacement motor used during drilling operations. In both of these cases, a motor failure was reported during the time frames indicated in the latter time period of the plots.DETAILED DESCRIPTION

[0022] The present disclosure introduces a method and system, in the context of well construction using a rotary drilling rig and extending the drilled wellbore into a subterranean formation either onshore or offshore, for optimizing drilling ROP while maintaining operational safety and preventing problems causing non-productive time, using only surface-measured data from the rig site as inputs to all relevant components or sub-systems. The method includes use of a core system for identifying drilling parameters that maximize expected ROP subject to a variety of limits, which may be configured ahead of operations or dynamically determined during the drilling process, or alternatively based on real-time risk detection methods. Controllable drilling parameters in this context could include WOB, surface rotary speed and mud flow rates into the wellbore. The core system estimates the expected ROP using a pre-trained generalized machine learning model, which is not limited in scope to specific fields, formation types, drill bit or Bottom-Hole Assembly configurations, or particular intervals of certain wells. The method utilizes two classes of safeguarding sub-systems integrated with the core optimizer, here termed “reactive safeguards” and “preventive safeguards”.

[0023] The reactive safeguarding systems are complementary modules monitoring in real-time for risks or symptoms of drilling dysfunctions. While these can be used independently for standalone drilling dysfunction detection purposes, in the scope of the present disclosure, the reactive safeguards are used to ensure that drilling parameter recommendations generated by the ROP Optimization system do not compromise operational safety or pose risks to drilling equipment. This can be achieved through altering the behavior of the optimization system, for example by changing its objective from increasing ROP to mitigation of the detected dysfunction. One example dysfunction detection module that has been developed and tested with the ROP optimizer is focused on detection of torsional vibration (stick-slip) symptoms. In one possible embodiment, this could involve detecting symptoms of torsional vibration, and notifying monitoring staff to allow them to mitigate the issue separately, and pausing the optimization system's attempts to increase ROP while risk symptoms continue to be detected. In other embodiments, this may involve strategies to actively mitigate the risk through issuing appropriate drilling parameter advice, in the case of stick-slip by following industry-standard mitigation techniques involving reducing WOB and increasing drill string rotary speed.

[0024] Preventive safeguards are based on scenario modelling to estimate one or more risks associated with potential drilling parameter changes, ahead of implementing them, in order to proactively avoid entering such a situation. An example risk to avoid is poor hole cleaning, which may lead to pack-offs and stuck pipe incidents associated with extended periods of non-productive time. A symptom of poor hole cleaning commonly known by practitioners in the field is increased downhole Equivalent Circulating Density (ECD), as excess cuttings produced by drilling increase the average density of the circulating drilling fluid when in suspension. An embodiment of a preventive safeguard targeted at poor hole cleaning was implemented as part of the system described in this disclosure. This embodiment utilized a machine learning model for estimating downhole ECD, using only the surface-measured data and ROP as inputs. This ML model was pre-trained on a curated historical dataset, and not limited in scope (through assumptions or otherwise) to particular fields, regions, formations, or downhole equipment configurations. The ECD model can be used to assess the expected impact of candidate solutions proposing changes in drilling parameters (and corresponding ROP estimated by the ROP model) on the downhole ECD. If ECD is expected to increase above a configured safety threshold, then the candidate solution for this scenario can be filtered out by the optimizer; the optimal drilling parameters according to this logic will be those that maximize ROP within the safety limits. This method allows each scenario to be modelled ahead of implementing any drilling parameter changes, in order to prevent a drilling dysfunction from occurring, in this case poor hole cleaning or cuttings accumulation.

[0025] FIG. 1 shows a high-level representation of how the ROP optimization system of the present disclosure fits with respect to existing infrastructure, both at the rig site and remotely-located systems. At the rig site, parameters associated with various rig apparatus 101 are measured by a suite of sensors 102, which may include (but is not limited to) rotary torque and rotary speed associated with a top-drive system used to rotate the drill string, pressure and volumetric flow rate measurements associated with mud pumps, or weight measurements (hookload) associated with the hoisting system. Data from these sensors are collated via an Electronic Data Recorder (EDR) system 103, which is integrated with rig-based computers 105, which enable data to be displayed to rig site users 106 for operational monitoring purposes, and in some embodiments may be used to run the method and system of this disclosure locally, with the ROP optimization system's outputs optionally transmitted 107 to a remote data store 108.

[0026] With continue reference to FIG. 1, in an alternative embodiment, the drilling optimization system could be deployed in remote (“cloud-based”) infrastructure. In this embodiment, a data feed from the EDR 109 transmits data from the rig to a remote data store 108 (typically a WITSML system in the field of oil and gas well construction), from which one or more remote computer servers running the ROP optimization system 110 read data and generate advised drilling parameters for increasing ROP subject to operational safety constraints, which is transmitted back to the server hosting the remote data store. This process can be applied both to data updating in real-time, or to historical data. Monitoring specialists, office-based engineers or other relevant users not physically located at the rig site may then view the outputs of the optimization system using an appropriate display method 111 such as a real-time viewer software application connected to the remote data store 108.

[0027] With reference to FIG. 2, the ROP optimization system 110 reads surface measurements of drilling parameters directly from the rig 202 or remote data store 108 via a data feed connected to an Orchestrator module 205, which is responsible for preprocessing he received data and controlling when requests are made for advice on optimized drilling parameters. In one embodiment of the system, data is processed into an aggregated form according to discrete depth intervals of a fixed size, for example by calculating the mean of parameter values over the depth interval. Drilling parameter advice is then requested at the end of the latest depth interval, using this as contextual information, for the next depth interval about to be drilled. Here, the “baseline” drilling parameters are defined as those measured for the previous interval assumed to be kept fixed for the upcoming depth interval. The orchestrator is further responsible for incorporating information on possible drilling dysfunctions into the decision logic for requests. The method and system architecture readily enables modules monitoring for multiple types of drilling dysfunction to be integrated with the orchestrator 205, using an additional detector for a given type of drilling dysfunction. When drilling dysfunctions are detected, the system may be configured to alter its objective from maximizing ROP within safety constraints, to mitigating the drilling dysfunction. In its simplest form, alternative action may be to pause requests to the ROP optimizer until the monitored drilling dysfunction severity metrics fall below a level considered to represent operational risks. In other embodiments, the orchestrator could recommend a set of actions to mitigate the dysfunction, either from standard operating procedures, or alternatively through scenario modelling to estimate the potential impact of proposed changes.

[0028] With continued reference to FIG. 2, if the orchestrator 205 does not receive notification of a drilling dysfunction occurring, and drilling of a depth interval is completed, the pre-processed data will be sent in a request 207 for advised drilling parameters for the upcoming depth interval. Upon receiving the request, the ROP Optimizer Core 208 generates a set of advised drilling parameters based on the input data received. These advised drilling parameters are informed by and constrained according to one or more safeguarding systems performing scenario modelling 209, with the goal of preventing any potential drilling issues occurring from alternative drilling parameters and increased ROP. The ROP optimizer core then returns optimized drilling parameters 210 to the orchestrator 205, which is responsible for post-processing steps and publication of the system outputs relating to drilling parameters and possible dysfunction symptoms 211 to a remote data store, user displays 106 / 111 or in some embodiments to a rig control system 213.

[0029] With reference to FIG. 3 and further reference to the components of FIG. 2, the safeguarded ROP optimization methodology can be further broken down into a series of more detailed process steps. Upon receiving data from the drilling operations 203, the Orchestrator 205 executes the data pre-processing steps and decision logic with respect to whether drilling parameter advice should be requested 301. This involves performing a check for symptoms of drilling dysfunctions 302, based on information provided by a set of drilling dysfunction detectors 206 that independently monitor surface drilling parameters pertaining to a specific drilling dysfunction for risk symptoms, which in some embodiments might include vibrations (such as stick-slip) 303, or risk of tool damage or failure 304, for example relating to positive displacement motors, drill bits, rotary steerables, mud pumps or other equipment. When requesting drilling parameter advice, the optimizer sends pre-processed data from the operations 207 to the core optimization system 208.

[0030] With further reference to FIGS. 1-3 the optimizer then generates a set of potential candidate solutions 305, where candidate solutions are here defined as a set of alternative combinations of drilling parameters (such as WOB, rotary speed or flow rate), which are assessed with regard to whether they are expected to increase the ROP relative to the baseline drilling parameters, following generation of estimated ROP values using a dedicated ML model, where the candidate solutions are included as a subset of the model inputs. The set of candidate solutions is constrained by safe operating limits for rig equipment, which may include maximum limits on surface rotary speeds, weight-on-bit, or mud flow rates into the well. The values of the constraints may be referenced from a data store 306 or any other appropriate computer-readable method. Users of the optimization system may freely change these optimizer constraints based on their knowledge of the drilling operations, or any other requirements that are deemed appropriate. The optimizer will not recommend any drilling parameters outside of the chosen constraints, hence the space of allowed candidate solutions is reduced accordingly 307. After estimation of the expected ROP associated with a particular candidate solution 308, the set of candidate solutions may be further filtered down according to a configured limit on ROP 309, if required for operational reasons. Additional constraints can then be applied based on the preventive safeguards 310 that utilize the scenario modelling modules 209, which directly factor in the ROP values expected from the previous step 309.

[0031] With further reference to FIGS. 1-3, an example preventive safeguard used in one embodiment involves the integration of a sub-system containing a model capable of estimating downhole Equivalent Circulating Density (ECD), based on the surface parameters measured by the rig sensors 102, allows the expected effect on ECD of changing drilling parameters and ROP to be estimated 311. Increases in ECD can be a symptom of poor hole cleaning, and are monitored for by practitioners in the field. The ECD estimation model utilized in the disclosed method and system as a component is an updated version of the generalized machine learning model described in the research publication SPE-208675-MS [Robinson et al 2022 (b)]. Based on the expected changes in ECD modelled for different scenarios (candidate solutions) 311, the risk of hole cleaning issues can be quantified, and a set of drilling parameters that are expected to keep downhole ECD within a configured limit 312 while maximizing ROP can be selected from the set of possible candidate solutions 313. The candidate solution with the drilling parameter combination that is expected to maximize the ROP, subject to constraints 313, is then returned as an output of the optimization system 210 to the orchestrator 205, which applies post-processing steps 314 and sends optimized drilling parameter values with respect to ROP and drilling risks 211 either in an advisory capacity for displayed to human users 106 / 111, or alternatively as input data to a control system 213 if implemented as part of a closed-loop optimization system.

[0032] In an alternative embodiment, use of the ECD estimation model 311 could be extended to keep expected ECDs within a safe operating window between the formation pressure (to avoid influxes) and the fracture pressure (to avoid fracturing the formation and associated problems). This embodiment modifies process step 312 in FIG. 3, and pro-actively attempt to maintain ECD values in the defined safe (absolute) range, rather than simply limiting the impact of increased ROP on the downhole ECD. The defined safe range of ECD & pressure values could be sourced from other coupled modules, for example focused on pore pressure estimation or measurement, or focused on estimating fracture pressures based on formation strength characteristics.

[0033] With reference to FIG. 4, in one embodiment of a real-time dysfunction detector used as a reactive safeguard, the methodology uses time-series measurements for surface rotary torque and surface rotary speed; these are (separately) windowed in time to analyze recent observations, and then transformed into the frequency domain 402 using an appropriate transform method known to practitioners of digital signal processing, such as a Fast Fourier Transform or a wavelet transform, in order to obtain power spectra for each time series 403. The power spectra are then aggregated into a set of discrete (non-overlapping) frequency bands 404 by summing the spectral power densities within each band. From these, a vector is extracted for each time-series measurement (surface torque and rotary speed), where each vector element describes the aggregated spectral power in a particular frequency band. These vectors can then be concatenated into a single vector 405, which is sent in a request 406 to the dysfunction severity sub-module, which uses the vector as an input to a machine learning model that estimates a quantity related to the severity of the drilling dysfunction symptom 407. The model is general-purpose, and pre-trained on historical data without limitations in scope to specific regions, fields, formation types, drill bit types, or BHA configurations. Example quantities include an estimated probability that a risk symptom is present (generated using a binary classification model trained on a dataset with labelled intervals with vibration symptoms), or alternatively a symptom severity index (generated using a regression model). Probabilities may optionally be post-processed 408 into other forms of information, such as Boolean or ordinal variables, before being sent to the remote data store 109 or made available to other software systems 409, or displayed to rig-based users 106. In an embodiment where a probability is the severity metric, in situations where the probability exceeds a configured threshold, the orchestrator 205 controlling requests 207 to the ROP optimization system 208 can switch objective from increasing ROP to mitigating the dysfunction, and pause requests for drilling parameter advice until the probability falls back below the configured threshold.

[0034] For the embodiment of a dysfunction detector described in the previous paragraph, an example of an aggregated power spectrum for obtained from process step 404 is shown in FIG. 5. In the top sub-plot 501, a time-series of surface rotary torque measurements 502 within a sample 10 minute window is shown, although the length of this time window used in the method is configurable and may be longer or shorter. The sub-plot below shows the aggregated power spectrum 503 corresponding to the windowed torque time-series, calculated for the full resolvable spectral range, the bandwidth of which is determined by the Nyquist theorem at the upper limit, and the signal window length at the lower limit. Example frequency bands 504-506 are labelled in the chart.

[0035] With reference to FIGS. 5 and 6, FIG. 6 shows a sub-plot 601 with a windowed time-series of surface rotary speed measurements 602 corresponding to the rotary torque measurements 502 plotted in FIG. 5, measured during the same drilling operation. Similarly to FIG. 5, the lower sub-plot 603 shows the aggregated spectrum corresponding to 601, with example frequency bands 604 and 605 labelled illustratively for the reader. The aggregated spectral power densities for each band shown in 503 and 603 are combined in process step 405 into a form required by the machine learning model 407 estimating the severity of a possible symptom of the target drilling dysfunction. Alternative embodiments of the dysfunction detector could use other measurements in addition to surface torque and rotary speeds, depending on whether they improve the system's capability to detect the target dysfunction.

[0036] With reference to FIG. 7, example outputs from dysfunction severity model 407 implemented to estimate the probability of stick-slip symptoms being present are shown. The first two sub-plots show surface rotary speed 701 and torque 702 measurements from a real historical drilling operation over a time period spanning several hours, while the bottom sub-plot shows the model-generated stick-slip probability values over the same period 703. The extended time interval contains three shaded intervals with stick-slip symptoms of increasing severity. In the first shaded interval 704, the onset of stick-slip issues can be observed, with estimated probability values in the range 0.45-0.65. The following shaded intervals contain more severe examples of torsional vibration, with severe vibration symptoms emerging intermittently 705, followed by an extended period with severe vibration issues 706, where the estimated stick-slip probabilities were consistently in the range 0.8-1.0, indicating the system had a high degree of confidence that stick-slip was occurring. These are in contrast to the (non-shaded) intervals without torsional vibration issues, where very stick-slip probabilities were estimated, in the range 0.0-0.3. Based on these observations, an example probability threshold of 0.5 could be reasonably applied 707 in a dysfunction detection system employed as a reactive safeguard 303 for an ROP optimizer 208.

[0037] With reference to FIG. 8, in an embodiment where the disclosed system and method for ROP optimization is used in an advisory capacity, the advice generated can be displayed in a real-time viewer software application. An example screen capture obtained by implementing a display dedicated to ROP optimization in a third-party commercially-available application is shown in FIG. 8, which includes a vertically-oriented display with several tracks 802-807 containing curves representing the time-evolution of various drilling parameters and measurements with time-index 801. The contextual tracks 802 and 803 show drilling parameters where the meanings of most curves, included by not limited to standpipe pressure, measured depths, or rotary torque, are readily understood by a practitioner in the field. The third track from the left 804 shows observed ROP values, in addition to ROP values produced by the optimization system 208, namely the estimated ROP values for the candidate solution where drilling parameters are left unchanged (the baseline) and for the “optimized” candidate solution expected to maximize ROP within the constraints. Messages providing information related to the drilling operation are also shown in this track, such as notifications for when stick-slip symptoms are detected, or when on-bottom rotary drilling or sliding activities are started. The second and third tracks from the right show the history of drilling parameter recommendations for WOB 805 and rotary speed 806, in addition to the observed parameter values from the operations. The right-most track displays a history of stick-slip symptom probabilities 807, which quantify the perceived severity of the torsional vibrations at a given point in time. In this embodiment, parameter recommendations 805&806 and corresponding optimized ROP estimates 804 are not generated while stick-slip probabilities exceed a certain threshold, as determined by the orchestrator 205 which pauses requests to the optimization system for drilling parameter advice.

[0038] Other embodiments can involve safeguards against different drilling dysfunctions, for example lateral and axial vibration modes (in addition to torsional vibration issues described in previous embodiment), bit-wear (through estimation of changing bit wear state during operations) and other equipment failures 304 (mud motors, rotary steerables, downhole logging tools, drillstring buckling).

[0039] With reference to FIG. 9, in an embodiment where drilling takes place with a downhole mud motor, a safeguarding module estimating a numerical metric relating to risk of motor failure, for example a failure probability or a degradation index, may be integrated with the ROP optimization method and system as a real-time drilling dysfunction detector 304. This safeguard receives surface measurements from the rig or remote data store feed 203 for a variety of drilling parameters, which may include, but are not limited to, hookload, standpipe pressure, WOB, ROP, mud flow rates, mud densities, rotary torque, rotary speeds, bit and hole depths. These data may be optionally augmented with additional equipment characteristics, for example relating to the bit, motor power section, or motor efficiency at transferring energy from drilling fluid into additional bit torque. From the received data, a range of numerical features are calculated, encoding the recent usage history of the motor and operational context, current drilling parameters, and any risk symptoms tracked during the drilling run 901. Risk symptoms pertaining to the motor may include tracked instances of drilling into hard stringers, pressure spikes, stalling events, or vibration symptoms (in the context of this disclosure, information on these may be obtained from another integrated safeguarding module, namely the vibration safeguard 303). The calculated features 901 may be optionally further augmented by user-provided information on any past reported maintenance issues with the motor 902, for example through input fields in the user-interface used to control and configure the optimization system, or alternatively via reading this information from an external system. The (augmented) calculated features describing recent usage history are then used to update a set of cumulative features describing longer term usage history and features describing current drilling parameters which may have an impact on failure risk 903. After concatenating the various pre-processed numerical feature vectors together into a single vector 904 compatible with the ML model embedded into the safeguarding module, an estimate for a metric quantifying the motor degradation or failure risk is requested from the model 905. The ML model maps the input features onto an estimated probability of motor failure, or alternatively a degradation index 906, and returns this estimate to the equipment failure module 304. After post-processing steps 907 on the returned metric, the metric quantifying motor failure risk is then published (via a message broker, publish-subscribe system, RESTful API or an appropriate alternative) for other software systems to utilize in their logic 908.

[0040] With reference to FIG. 10, example outputs from a drilling dysfunction detection system targeting motor failures and degradation are shown for two different drilling operations, denoted motor failure case 11001, and failure case 21002. In these graphs, the time-evolution of the motor failure probability estimates are shown as a plotted curve 1003, leading up to reported failure incidents at the marked times for both case 11004 and case 21005. Failure risk probabilities may be used to create a categorical status describing the motor degradation state, with example thresholds indicated via dotted lines for the initial degradation phase 1006 above 0.2, and a phase with elevated risk of failure 1007 for probabilities above 0.6. These categorical status values may be utilized in operating procedures and used to inform decision-making related to the aggressiveness of drilling (by controlling drilling parameters) and maintenance plans or scheduling.Definitions

[0041] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. It will be apparent, however, to one skilled in the art, that the present invention may be practiced without some or all of these specific details. In other instances, well known process steps have not been described in detail in order not to unnecessarily obscure the present invention. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art. In case of conflict, the present document, including definitions, will control. Preferred methods and materials are described below, although methods and materials similar or equivalent to those described herein can be used in practice or testing of the present disclosure.

[0042] As noted herein, the disclosed embodiments have been presented for illustrative purposes only and are not limiting. Other embodiments are possible and are covered by the disclosure, which will be apparent from the teachings contained herein. Thus, the breadth and scope of the disclosure should not be limited by any of the above-described embodiments but should be defined only in accordance with claims supported by the present disclosure and their equivalents. Moreover, embodiments of the subject disclosure may include methods, compositions, systems and apparatuses / devices that may further include any and all elements from any other disclosed methods, compositions, systems, and devices. In other words, elements from one or another disclosed embodiments may be interchangeable with elements from other disclosed embodiments. Moreover, some further embodiments may be realized by combining one and / or another feature disclosed herein with methods, compositions, systems and devices, and one or more features thereof, disclosed in materials incorporated by reference. In addition, one or more features / elements of disclosed embodiments may be removed and still result in patentable subject matter (and thus, resulting in yet more embodiments of the subject disclosure). Furthermore, some embodiments correspond to methods, compositions, systems, and devices which specifically lack one and / or another element, structure, and / or steps (as applicable), as compared to teachings of the prior art, and therefore represent patentable subject matter and are distinguishable therefrom (i.e. claims directed to such embodiments may contain negative limitations to note the lack of one or more features prior art teachings).

[0043] For the recitation of numeric ranges herein, each intervening number there between with the same degree of precision is explicitly contemplated. For example, for the range of 6-9, the numbers 7 and 8 are contemplated in addition to 6 and 9, and for the range 6.0-7.0, the number 6.0, 6.1, 6.2, 6.3, 6.4, 6.5, 6.6, 6.7, 6.8, 6.9, and 7.0 are explicitly contemplated.

[0044] In addition, it should be understood that embodiments may include hardware, software, and electronic components or modules that, for purposes of discussion may be illustrated and described as if the majority of the components were implemented solely in hardware. However, one of ordinary skill in the art, and based on a reading of this detailed description, would recognize that, in at least one embodiment, the electronic-based aspects may be implemented in software (e.g., stored on non-transitory computer-readable medium) executable by one or more processing units, such as a microprocessor and / or application specific integrated circuits (“ASICs”). As such, it should be noted that a plurality of hardware and software-based devices, as well as a plurality of different structural components, may be utilized to implement the embodiments. For example, “servers” and “computing devices” described in the specification can include one or more processing units, one or more computer-readable medium modules, one or more input / output interface, and various connections (e.g., a system bus) connecting the components. In one embodiment, the software-based components can be containerized and deployed on Virtual Machines (Windows, Linux or similar) in a Data Center that are either cloud based (provided by Azure, AWS or similar) or by the user organization itself. In another embodiment, the software-based components can be deployed on a physical server.

[0045] Drilling parameters are defined as measured values of physical quantities relevant to the drilling operations, which include but are not limited to hookload, standpipe pressure, mud flow rates, mud densities (and other mud characteristics), rotary torque, rotary speed, block position, hole depth, bit depth, weight-on-bit (WOB), rate of penetration (ROP). These may be directly measured by sensors at the rig, or derived from one or more of the directly measured parameters, for example WOB is often calculated from the hookload. The indefinite articles “a” and “an,” as used herein in the specification and in the claims, unless clearly indicated to the contrary, should be understood to mean “at least one.”

[0046] The phrase “and / or,” as used herein in the specification and in the claims, should be understood to mean “either or both” of the elements so conjoined, i.e., elements that are conjunctively present in some cases and disjunctively present in other cases. Multiple elements listed with “and / or” should be construed in the same fashion, i.e., “one or more” of the elements so conjoined. Other elements may optionally be present other than the elements specifically identified by the “and / or” clause, whether related or unrelated to those elements specifically identified. Thus, as a non-limiting example, a reference to “A and / or B”, when used in conjunction with open-ended language such as “comprising” can refer, in one embodiment, to A only (optionally including elements other than B); in another embodiment, to B only (optionally including elements other than A); in yet another embodiment, to both A and B (optionally including other elements); etc.

[0047] As used herein in the specification and in the claims, “or” should be understood to have the same meaning as “and / or” as defined above. For example, when separating items in a list, “or” or “and / or” shall be interpreted as being inclusive, i.e., the inclusion of at least one, but also including more than one, of a number or list of elements, and, optionally, additional unlisted items. Only terms clearly indicated to the contrary, such as “only one of” or “exactly one of” or, when used in the claims, “consisting of,” will refer to the inclusion of exactly one element of a number or list of elements. In general, the term “or” as used herein shall only be interpreted as indicating exclusive alternatives (i.e. “one or the other but not both”) when preceded by terms of exclusivity, such as “either,”“one of,”“only one of,” or “exactly one of.”“Consisting essentially of,” when used in the claims, shall have its ordinary meaning as used in the field of patent law.

[0048] As used herein in the specification and in the claims, the phrase “at least one,” in reference to a list of one or more elements, should be understood to mean at least one element selected from any one or more of the elements in the list of elements, but not necessarily including at least one of each and every element specifically listed within the list of elements and not excluding any combinations of elements in the list of elements. This definition also allows that elements may optionally be present other than the elements specifically identified within the list of elements to which the phrase “at least one” refers, whether related or unrelated to those elements specifically identified. Thus, as a non-limiting example, “at least one of A and B” (or, equivalently, “at least one of A or B,” or, equivalently “at least one of A and / or B”) can refer, in one embodiment, to at least one, optionally including more than one, A, with no B present (and optionally including elements other than B); in another embodiment, to at least one, optionally including more than one, B, with no A present (and optionally including elements other than A); in yet another embodiment, to at least one, optionally including more than one, A, and at least one, optionally including more than one, B (and optionally including other elements); etc.

[0049] In the claims, as well as in the specification above, all transitional phrases such as “comprising,”“including,”“carrying,”“having,”“containing,”“involving,”“holding,”“composed of,” and the like are to be understood to be open-ended, i.e., to mean including but not limited to. Only the transitional phrases “consisting of” and “consisting essentially of” shall be closed or semi-closed transitional phrases, respectively, as set forth in the United States Patent Office Manual of Patent Examining Procedures, Section 2111.03.REFERENCES

[0050] The patents, patent applications, and other publications cited herein are incorporated by reference in their entirety as if each individual publication or patent application were specifically and individually set forth herein. In case of conflict between any incorporated reference and the present disclosure, the present disclosure controls.Patent ReferencesPublication NumberPublication DateTitleWO2023067391A12023 Apr. 27System and method for predicting andoptimizing drilling parametersWO 2024 / 0572302024 Mar. 21Frequency based rig analysisUS 2022 / 0137568 A12022 May 5Predictive models and multi-objectiveconstraint optimization algorithm to optimizedrilling parameters of a wellboreUS 2022 / 0307365 A12022 Sep. 29Method and system for rate of penetrationoptimization using artificial intelligencetechniquesU.S. Pat. No.2021 Dec. 7Performance index using frequency or11,193,364B1frequency-time domainUS 2020 / 0371495 A12020 Nov. 26Automated real-time hole cleaning efficiencyindicatorUS20230184107A12023 Jun. 15Machine-Learning based Rig-Site On-Demand Drilling Mud Characterization,Property Prediction, and OptimizationU.S. Pat. No.2006 May 16System and method for automatic drilling to7,044,239 B2maintain equivalent circulating density at apreferred valueU.S. Pat. No.2016 Mar. 15Drilling advisory systems and methods with9,285,794 B2decision trees for learning and applicationmodesU.S. Pat. No.2013 Sep. 3System and method for optimizing drilling8,527,249 B2speedU.S. Pat. No.2023 May 23Borehole cleaning monitoring and advisory11,655,690B2systemU.S. Pat. No.2022 Sep. 20Adjusting well tool operation to manipulate11,448,057 B2the rate - of - penetration (ROP) of a drill bitbased on multiple ROP projectionsUS 2024 / 0044241 A12024 Feb. 8Method and apparatus for steering a bit usinga quill and based on learned relationshipsU.S. Pat. No.2015 Mar. 10Methods to estimate downhole drilling8,977,523B2vibration amplitude from surfacemeasurementU.S. Pat. No.2019 Oct. 29System and method for mitigating stick-slip10,458,223B2U.S. Pat. No.2018 Aug. 21Method and apparatus for detecting downhole10,053,971B2torsional vibration based on measurement ofsurface torqueU.S. Pat. No.2021 Oct. 12Real-time modification of a slide drilling11,143,011B2segment based on continuous downhole dataU.S. Pat. No.2020 Jan. 14Methods and apparatus for reducing stick-slip10,533,407B2U.S. Pat. No.2018 Aug. 28Vibration detection in a drill string based on10,060,248B2multi-positioned sensorsNon-Patent ReferencesRobinson, T. S., Batruny, P., Gomes, D. et al. 2022 (a). “Successful Development and Deployment of a Global ROP Optimization Machine Learning Model”, Offshore Technology Conference Asia, Kuala Lumpur, Malaysia, March 2022. OTC-31680-MS. DOI: doi.org / 10.4043 / 31680-MS

[0052] Robinson, T. S., Gomes, D., Meor Hashim, M. M. H. et al. 2022 (b). “Real-time Estimation Of Downhole Equivalent Circulating Density (ECD) Using Machine Learning And Applications”. Paper presented at SPE / IADC International Drilling Conference and Exhibition, March 2022, Galveston, TX, USA. SPE-208675-MS. DOI: doi.org / 10.2118 / 208675-MS

[0053] Robinson, T. S., Revheim, O. 2023. “Automated Detection of Rig Events From Real-time Surface Data Using Spectral Analysis and Machine Learning”. Forthcoming paper at SPE / IADC International Drilling Conference and Exhibition, March 2023, Stavanger. SPE-212481-MS. DOI: doi.org / 10.2118 / 212481-MS

Claims

1. A method for optimizing drilling parameters during wellbore construction, comprising:receiving real-time surface measurement data during a drilling operation;generating, using a first machine learning model, a set of candidate drilling parameter combinations and corresponding predicted rate of penetration (ROP) values;analyzing the candidate combinations using at least one preventive safeguarding module that generates, using a second machine learning model, predicted risk metrics for implementing each candidate combination;selecting optimized drilling parameters by:maximizing predicted ROP when no drilling dysfunctions are detected and predicted risk metrics are below configured thresholds; orimplementing dysfunction mitigation procedures when drilling dysfunctions are detected or predicted risk metrics exceed configured thresholds; andoutputting the selected drilling parameters to at least one of a display interface, a data store, or a control system.

2. The method of claim 1, further comprising:monitoring the real-time surface measurement data using at least one reactive safeguarding module that:processes time windows of the surface measurement data to detect drilling dysfunction symptoms; andgenerates, using a third machine learning model, dysfunction severity metrics;wherein selecting the optimized drilling parameters further comprises:maximizing predicted ROP when no drilling dysfunctions are detected; orimplementing dysfunction mitigation procedures when drilling dysfunctions are detected.

3. The method of claim 2, wherein the first, second, and third machine learning models are generalized models trained on global datasets spanning multiple wells, fields, and equipment configurations.

4. The method of claim 1, wherein analyzing candidate combinations comprises:filtering combinations based on equipment operating limits;predicting equivalent circulating density (ECD) values using the second machine learning model;comparing predicted ECD values to at least one of:formation fracture pressure;formation pore pressure; ora maximum allowable increase from baseline ECD; andeliminating combinations exceeding the ECD thresholds.

5. The method of claim 2, wherein the processing of time windows of the surface measurement data comprises:applying frequency domain transforms to surface measurements;aggregating transformed data into discrete frequency bands;generating feature vectors from the aggregated frequency bands;inputting the feature vectors to the third machine learning model; andgenerating dysfunction probabilities as severity metrics.

6. The method of claim 5, wherein the surface measurements comprise at least:surface torque; rotary speed; weight on bit; hook load; flow rate; standpipe pressure; and depth.

7. The method of claim 1, wherein the drilling dysfunctions comprise at least one of: torsional vibrations; axial vibrations; lateral vibrations; bit wear; motor failures; poor hole cleaning; stuck pipe; or wellbore instability.

8. The method of claim 1, wherein implementing dysfunction mitigation procedures comprises:generating severity metrics based on detected drilling dysfunctions;pausing ROP optimization requests until severity metrics return below thresholds;applying predefined parameter adjustments based on dysfunction type; orgenerating new parameter recommendations targeting dysfunction reduction.

9. A system for optimizing drilling parameters during wellbore construction, comprising:at least one processor;a memory storing computer-readable instructions that, when executed by the at least one processor, cause the system to:receive real-time surface measurement data from sensors at a drilling rig;maintain an orchestrator module that:preprocesses the received data;controls timing of optimization requests; andmanages communication between system components;execute a core optimization module that generates candidate drilling parameter combinations using a first machine learning model;execute at least one preventive safeguarding module that performs scenario modeling using a second machine learning model;execute at least one reactive safeguarding module that detects drilling dysfunctions using a third machine learning model; andselect and output optimized drilling parameters based on combined outputs from the core optimization module and safeguarding modules.

10. The system of claim 9, further comprising:a data store containing:equipment operating limits;threshold configurations;historical dysfunction occurrences; andmodel parameters; andwherein the orchestrator module dynamically updates threshold values based on contents of the data store.

11. The system of claim 9, wherein the preventive safeguarding module comprises:an ECD estimation component that:receives candidate parameter combinations;predicts corresponding ECD values; andfilters combinations exceeding configured ECD limits.

12. The system of claim 9, wherein the reactive safeguarding module comprises:a signal processing component that:windows the real-time surface measurement data into time series data;performs frequency domain transforms;aggregates spectral power in discrete bands; andgenerates feature vectors for dysfunction detection.

13. The system of claim 9, further comprising:an interface component that enables runtime addition of new safeguarding modules through a standardized protocol defining:required inputs;expected outputs; andconfiguration parameters.

14. A method for detecting drilling dysfunctions during wellbore construction, comprising:receiving real-time surface measurement data comprising at least surface torque and rotary speed;processing sequential time windows of the measurement data by:applying frequency domain transforms to create power spectra;aggregating spectral power into discrete frequency bands; andgenerating feature vectors from the aggregated bands;inputting the feature vectors to a machine learning model trained to detect specific dysfunction types;generating dysfunction severity metrics using the machine learning model;comparing the severity metrics to configured thresholds; andoutputting dysfunction alerts when thresholds are exceeded.

15. The method of claim 14, wherein the dysfunction types comprise at least one of:torsional vibrations;axial vibrations;lateral vibrations;bit wear; ormotor failures.

16. The method of claim 14, further comprising:maintaining historical severity metrics;analyzing metric trends over time;updating threshold values based on the analysis; andgenerating predictive dysfunction warnings.

17. The method of claim 1, further comprising:maintaining an event log recording:detected dysfunctions;threshold violations;parameter recommendations;user overrides; andsystem state changes.

18. The method of claim 1, wherein the preventive and reactive safeguarding modules operate:asynchronously on different time scales;with configurable analysis window sizes; andwith independent processing threads.

19. The method of claim 14, wherein generating dysfunction severity metrics comprises:calculating probabilities of dysfunction occurrence;determining severity levels based on the probabilities; andclassifying operational states as normal, warning, or critical.

20. The method of claim 14, wherein the machine learning model is:pre-trained on a global dataset;not limited to specific fields or equipment configurations; andcapable of detecting dysfunctions without customization to local conditions.

Images