Drilling incident remediation framework

Abstract

One approach may include: receiving a description of an event occurring at a well site; extracting a failure mode from the description using a fine-tuned large language model (LLM); identifying matching failure modes from historical data processed using the fine-tuned LLM, wherein the matching failure mode is associated with one or more remedial actions that successfully resolved the matching failure mode; and outputting one or more remedial actions for implementation at the well site.

Description

[0001] Related applications

[0002] This application claims priority and benefit to U.S. Provisional Application No. 63 / 591,297, filed October 18, 2023, the entire contents of which are incorporated herein by reference. Background Technology

[0003] A reservoir can be a subsurface stratum, characterized at least in part by its porosity and fluid permeability. As an example, a reservoir can be part of a basin (such as a sedimentary basin). A basin can be a depression in which sediments accumulate (e.g., caused by plate tectonics, subsidence, etc.). As an example, when source rocks are present in combination with appropriate burial depth and duration, petroleum systems can be developed within a basin, which can form reservoirs containing hydrocarbon fluids (e.g., oil, gas, etc.).

[0004] In oil and gas exploration, interpretation involves analyzing data to identify and locate various subsurface structures (e.g., strata, faults, geological bodies, etc.) within a geological environment. Various types of structures (e.g., stratigraphic structures) can indicate hydrocarbon traps or flow channels, and may be associated with one or more reservoirs (e.g., fluid reservoirs). In the field of resource extraction, enhanced interpretation can allow for the construction of more accurate models of subsurface areas, which in turn improves the characterization of these areas for resource extraction purposes. The characterization of one or more subsurface areas within a geological environment can guide the execution of, for example, one or more operations (e.g., field operations, etc.). As an example, a plan can be based on a model of the subsurface area, where the plan can specify how drilling operations can accurately construct boreholes based on trajectories penetrating reservoirs, etc., through which fluids can be generated (e.g., as well completions, etc.). As an example, one or more computational frameworks, systems, etc., can be used to execute one or more workflows for analysis, acquisition, model building, control, etc., for exploration, interpretation, drilling, injection, fracturing, production, etc. Summary of the Invention

[0005] One approach may include receiving a description of events occurring at a well site and extracting failure modes from the description using a fine-tuned large language model (LLM). The approach may also include identifying matching failure modes from historical data processed using the fine-tuned LLM. Matching failure modes may be associated with one or more remedial actions that successfully resolved the matching failure modes. The approach may also include outputting one or more remedial actions to be implemented at the well site. Various other apparatuses, systems, methods, etc., are also disclosed.

[0006] This summary is provided to introduce some concepts that will be further described in the following detailed description. This summary is not intended to identify key or essential features of the claimed subject matter, nor is it intended to help limit the scope of the claimed subject matter. Attached Figure Description

[0007] The following detailed description refers to the accompanying drawings. The convenient features and advantages of the described embodiments can be more easily understood by referring to the following description taken in conjunction with the accompanying drawings.

[0008] Figure 1 An example of the system is shown;

[0009] Figure 2 An example of the system is shown;

[0010] Figure 3 An example of the system is shown;

[0011] Figure 4 An example of the system is shown;

[0012] Figure 5 An example of the architecture is shown;

[0013] Figure 6 An example of a graph is shown;

[0014] Figure 7 An example of the description is shown;

[0015] Figure 8 An example of the description is shown;

[0016] Figure 9 An example of the system is shown;

[0017] Figure 10 Examples of methods and systems are shown; and

[0018] Figure 11 An example of the system is shown. Detailed Implementation

[0019] This embodiment should not be considered limiting, but is made only for the purpose of describing the general principles of the embodiment. The scope of the described embodiment should be determined with reference to the published claims.

[0020] Figure 1 An example of a system 100 including a workspace framework 110 is shown, which can provide instantiation, rendering, and interaction with a graphical user interface (GUI) 120. Figure 1In the example, GUI 120 may include graphical controls for a computing framework (e.g., an application, etc.) 121, a project 122, a visualization feature 123, one or more other features 124, data access 125, and data storage 126.

[0021] exist Figure 1 In the example, the workspace framework 110 can be customized for a specific geological environment (such as example geological environment 150). For example, geological environment 150 may include multiple layers (e.g., strata) containing reservoir 151 and may be penetrated by fault 153. As an example, geological environment 150 may be equipped with various sensors, detectors, actuators, etc. For example, device 152 may include communication circuitry for receiving and transmitting information about one or more networks 155. Such information may include information associated with downhole device 154, which may be a device for acquiring information, assisting in resource recovery, etc. Other devices 156 may be located remotely from the well site and include sensing, detection, transmission, or other circuitry. Such devices may include storage and communication circuitry for storing and transmitting data, instructions, etc. As an example, one or more satellites may be provided for communication, data acquisition, and other purposes. For example, Figure 1 A satellite communicating with network 155 is shown. The satellite can be configured for communication. Note that the satellite may additionally or alternatively include circuitry for imaging (e.g., spatial, spectral, temporal, radiometric, etc.).

[0022] Figure 1 The geological environment 150 is also shown as optionally including equipment 157 and 158 associated with a well, the well comprising a substantially horizontal portion that may intersect with one or more fractures 159. For example, consider a well in a shale formation, which may include natural fractures, artificial fractures (e.g., hydraulic fractures), or a combination of natural and artificial fractures. As an example, drilling may be performed for a laterally extending reservoir. In such an example, there may be lateral variations in properties, stresses, etc., where assessment of such variations can aid in planning, operations, etc., to develop the laterally extending reservoir (e.g., through fracturing, injection, extraction, etc.). As an example, equipment 157 and / or 158 may include components, a system, multiple systems, etc., for fracturing, seismic sensing, seismic data analysis, assessment of one or more fractures, etc.

[0023] exist Figure 1 In the examples, GUI 120 shows some examples of computing frameworks, including DRILLPLAN, DRILLOPS, PETREL, TECHLOG, PETROMOD, ECLIPSE, PIPESIM, and INTERSECT frameworks (SLB, Houston, Texas).

[0024] The DRILLPLAN framework provides digital well construction planning and includes features for automating repetitive tasks and validating workflows, enabling the rapid generation of improved quality drilling procedures (e.g., digital drilling plans, etc.) while ensuring consistency.

[0025] The DRILLOPS framework enables the execution of digital drilling plans and ensures adherence to those plans, while providing goal-based automation. The DRILLOPS framework can automatically generate activity plans for individual operations, whether they are monitored and / or controlled on the drilling rig or in town. Automation can leverage data analytics and learning systems to assist and optimize tasks, such as setting the ROP (Recovery Point of Operation) for the drill string. Preset menus for automatable drilling tasks can be presented, and using data analytics and models, plans can be executed in a manner that achieves specified objectives, where, for example, measurements can be used for calibration. The DRILLOPS framework provides the flexibility to dynamically modify and reschedule activities, for example, based on real-time assessments of various factors, such as equipment, personnel, and supplies. Well construction activities (e.g., tripping, drilling, cementing, etc.) can be continuously monitored and dynamically updated using feedback from operational activities. The DRILLOPS framework can provide various levels of automation based on plans and / or rescheduling (e.g., via the DRILLPLAN framework), feedback, and more.

[0026] The PETREL framework can be part of a DELFI environment used in earth sciences and geoengineering, for example, to analyze subsurface data from exploration to fluid production from reservoirs. The DELFI cognitive exploration and production (E&P) environment (SLB, Houston, Texas), referred to herein as the DELFI environment or DELFI framework, is a secure, cognitive, cloud-based collaborative environment that integrates data and workflows with digital technologies such as artificial intelligence and machine learning.

[0027] The PETREL framework provides components that allow for the optimization of various exploration, development, and production operations. The PETREL framework includes seismic simulation software components that can output information for improving reservoir performance, for example, by increasing asset team productivity. By using such a framework, various professionals (e.g., geophysicists, geologists, and reservoir engineers) can develop collaborative workflows and integrate operations to streamline processes (e.g., regarding one or more geological environments). Such a framework can be considered an application (e.g., executable using one or more devices) and can also be considered a data-driven application (e.g., where data is input for modeling, simulation, etc.).

[0028] The TECHLOG framework can handle and process field and laboratory data from various geological environments (e.g., deepwater exploration, shale, etc.). The TECHLOG framework can construct wellbore data for analysis, planning, and other purposes.

[0029] The PETROMOD framework provides petroleum system modeling capabilities, allowing the combination of one or more seismic, well, and geological information to model the evolution of sedimentary basins. The PETROMOD framework can predict whether and how reservoirs become filled with hydrocarbons, including the source and timing of hydrocarbon formation, migration routes, quantities, and types of hydrocarbons under subsurface or surface conditions.

[0030] The ECLIPSE framework provides a reservoir simulator (e.g., as a computational framework) employing numerical solutions for rapidly and accurately predicting the dynamic behavior of various reservoir types and development scenarios.

[0031] The INTERSECT framework provides a high-resolution reservoir simulator for simulating detailed geological features and quantifying uncertainties. For example, by creating accurate production scenarios and integrating precise models of surface facilities and field operations, the INTERSECT framework can produce reliable results that are continuously updated through real-time data exchange (e.g., from one or more types of data acquisition devices in the field that can acquire data during one or more types of field operations). The INTERSECT framework can provide completion configurations for complex wells that can be built in the field, detailed enhanced oil recovery (EOR) formulas that can be implemented in the field, analysis of the application of steam injection and other thermal EOR technologies for field implementation, advanced production control in terms of reservoir coupling and flexible field management, and the flexibility to write custom solutions to improve modeling and field management control. Like other example frameworks, the INTERSECT framework can be used as part of the DELFI environment, for example, for rapid simulation of multiple concurrent scenarios. For example, workflows can utilize one or more of the on-demand reservoir simulation features of the DELFI environment.

[0032] The aforementioned DELFI environment provides workflows with various features related to underground analysis, planning, construction, and production, as illustrated in workspace frame 110. Figure 1 As shown, the output from the workspace frame 110 can be used to guide, control, or otherwise manage one or more processes in the geological environment 150, and the feedback 160 can be received via one or more interfaces in one or more forms (e.g., data on operating conditions, equipment conditions, environmental conditions, etc.).

[0033] As an example, the workflow can proceed to a geological and geophysical (“G&G”) service provider, which can generate well trajectories, which may involve the execution of one or more G&G frameworks (e.g., consider the PETREL framework, etc.).

[0034] exist Figure 1 In the example, visualization feature 123 can be implemented via workspace framework 110, for example, to perform one or more tasks associated with subsurface areas, planned operations, construction wells and / or surface fluid networks, and reservoir exploitation.

[0035] As an example, the visualization process can be implemented to suit one or more of a variety of features that can be adapted to one or more web applications. For example, templates may involve using JavaScript Object Notation (JSON) and / or one or more other languages / formats. As an example, a framework may include one or more converters. For example, consider a JSON to Python converter and / or a Python to JSON converter. Such an approach can provide compatibility with one or more instruction sets regarding devices, frameworks, etc.

[0036] As an example, a visualization feature can provide visualizations of various Earth models, properties, etc., in one or more dimensions. As an example, a visualization feature can provide rendering of information in multiple dimensions, which may optionally include multi-resolution rendering. In such an example, the rendered information may be associated with one or more frames and / or one or more data stores. As an example, a visualization feature may include one or more control features for controlling equipment, which may include, for example, field equipment capable of performing one or more field operations. As an example, a workflow may utilize one or more frames to generate information that can be used to control one or more types of field equipment (e.g., drilling equipment, cable equipment, fracturing equipment, etc.).

[0037] Regarding reservoir models potentially suitable for simulator use, consider the acquisition of seismic data via reflection seismology, which can be used in geophysics, for example, to estimate the properties of subsurface strata. As an example, reflection seismology can provide seismic data representing elastic energy waves (e.g., transmitted by P-waves and S-waves, in a frequency range of approximately 1 Hz to approximately 100 Hz). Seismic data can be processed and interpreted, for example, to better understand the composition, fluid content, extent, and geometry of subsurface rocks. Such interpretation results can be used to plan, simulate, execute, etc., one or more operations for producing fluids from a reservoir (e.g., reservoir rocks, etc.).

[0038] As an example, the model can be a simulated version of a geological environment. As an example, the simulator can include features for simulating physical phenomena in a geological environment, at least partially based on one or more models. A simulator, such as a reservoir simulator, can simulate fluid flow in a geological environment, at least partially based on a model that can be generated via a framework receiving seismic data. The simulator can be a computerized system (e.g., a computing system) that can use one or more processors to execute instructions to solve a set of equations describing physical phenomena subject to various constraints. In such an example, the set of equations can be spatially defined (e.g., numerically discretized) according to a spatial model including rock layers, geological bodies, etc., which have corresponding locations that can be interpreted based on seismic and / or other data. The spatial model can be a cell-based model, where cells are defined by a grid (e.g., a mesh). Cells in a cell-based model can represent physical regions or volumes in a geological environment, where cells can be assigned physical properties (e.g., permeability, fluid properties, etc.) that may be closely related to one or more physical phenomena (e.g., fluid volume, fluid flow rate, pressure, etc.). The reservoir simulation model can be a cell-based spatial model.

[0039] Although Figure 1 Several simulators are shown in the examples, but additionally or alternatively, one or more other simulators may be utilized. For example, consider the VISAGE geomechanical simulator (SLB, Houston, Texas) or the PIPESIM network simulator (SLB, Houston, Texas). As an example, the KINTEX framework (SLB, Houston, Texas) can be used as a well service plugin for, for example, the PETREL framework. The KINETIX framework can provide multi-stage completion and stimulus design, as well as production assessment, for both conventional and unconventional reservoirs. As an example, the VISAGE geomechanical simulator and the KINETIX framework can be operatively coupled (e.g., for running stimulus stress processes, etc.).

[0040] As an example, a workflow can utilize one or more types of data from one or more processes (e.g., formation modeling, basin modeling, well completion design, drilling, production, injection, etc.). As an example, one or more tools can provide data that can be used in one or more workflows that can implement one or more frameworks.

[0041] exist Figure 1 In the example, drilling can be performed in geological environment 150, for example, to access reservoir 151, which can be accessed from land or sea. Figure 1In this configuration, downhole equipment 154 may be, for example, part of a bottom hole assembly (BHA). The BHA is used for drilling. Downhole equipment 154 can transmit information to and receive instructions and information from surface equipment. During well construction, various operations (such as cementing, cable assessment, testing, etc.) can be performed. In such embodiments, data collected by tools and sensors and used for purposes such as reservoir characterization can be collected and transmitted.

[0042] A well may include a substantially horizontal portion (e.g., a transverse portion) that may intersect one or more fractures. For example, a well in a shale formation may pass through natural fractures, man-made fractures (e.g., hydraulic fractures), or a combination thereof. Such wells can be constructed using directional drilling techniques as described herein. However, these same techniques can be used in conjunction with other types of directional wells (e.g., deviated wells, S-shaped wells, deep inclined wells, etc.) and are not limited to horizontal wells.

[0043] Figure 2 An example of a well site system 200 is shown (e.g., at a well site that may be onshore or offshore). As shown, the well site system 200 may include: a mud tank 201 for holding mud and other materials (e.g., the mud may be drilling fluid); a suction line 203 serving as an inlet for a mud pump 204, which pumps mud from the mud tank 201 to a vibratory hose 206; a winch 207 for winding one or more drill lines 212; a riser 208 for receiving mud from the vibratory hose 206; and a kelly hose 209 for receiving mud from the riser 208. One or more gooseneck pipes 210; traveling block 211; overhead crane 213 for transporting traveling block 211 via one or more drill ropes 212; derrick 214; kelly 218 or top drive 240; kelly drive casing 219; rotary table 220; drill rig 221; bell sub 222; one or more blowout preventers (BOPs) 223; drill string 225; drill bit 226; casing head 227; and flow pipe 228 for conveying mud and other materials to, for example, mud tank 201.

[0044] exist Figure 2 In the exemplary system, a borehole 232 is formed in the underground formation 230 by rotary drilling; note that various example embodiments may also use one or more directional drilling techniques, devices, etc.

[0045] like Figure 2 As shown in the example, drill string 225 is suspended within borehole 232 and has drill string assembly 250, which includes drill bit 226 at its lower end. As an example, drill string assembly 250 may be a bottomhole assembly (BHA).

[0046] The well site system 200 can provide operation of the drill string 225 and other operations. As shown, the well site system 200 includes a traveling block 211 and a derrick 214 positioned above the borehole 232. As described above, the well site system 200 may include a rotary table 220 through which the drill string 225 passes.

[0047] like Figure 2 As illustrated in the example, well site system 200 may include a crisscross drill pipe 218 and associated components, or a top drive 240 and associated components. Regarding the crisscross drill pipe example, the crisscross drill pipe 218 may be a square or hexagonal metal / alloy bar with holes drilled in it to serve as a mud flow path. The crisscross drill pipe 218 can be used to transmit rotational motion from rotary table 220 to drill string 225 via crisscross drill pipe drive casing 219, while allowing drill string 225 to be lowered or raised during rotation. The crisscross drill pipe 218 may pass through the crisscross drill pipe drive casing 219, which may be driven by rotary table 220. As an example, rotary table 220 may include a main bushing operatively coupled to the crisscross drill pipe drive casing 219, such that rotation of rotary table 220 can rotate the crisscross drill pipe drive casing 219 and thus rotate the crisscross drill pipe 218. The square drill pipe drive sleeve 219 may include an internal profile that matches the external profile (e.g., square, hexagonal, etc.) of the square drill pipe 218; however, it has a slightly larger size so that the square drill pipe 218 can move freely up and down within the square drill pipe drive sleeve 219.

[0048] Regarding the top drive example, top drive 240 can provide functionality performed by the kauri and rotary table. Top drive 240 can rotate drill string 225. As an example, top drive 240 may include one or more motors (e.g., electric and / or hydraulic motors) connected via suitable transmissions to a short section of tubing called a casing shaft, which can in turn screw into a protective joint or the drill string 225 itself. Top drive 240 may be suspended on a traveling block 211, thus allowing the rotating mechanism to move freely up and down along derrick 214. As an example, top drive 240 may allow drilling to be performed using more joint supports than the kauri / rotary table method.

[0049] exist Figure 2 In this example, mud tank 201 can contain mud, which can be one or more types of drilling fluid. As an example, a wellbore can be drilled to produce fluid, inject fluid, or both (e.g., hydrocarbons, minerals, water, etc.).

[0050] exist Figure 2In this example, drill string 225 (e.g., including one or more downhole tools) may consist of a series of pipes threaded together to form a long tube with a drill bit 226 at its lower end. As drill string 225 advances into the wellbore for drilling, at some point before or simultaneously with drilling, mud may be pumped by pump 204 from mud tank 201 (e.g., or other source) via lines 206, 208, and 209 to a port of kelly 218, or, for example, to a port of top drive 240. The mud may then flow through channels (e.g., or multiple channels) in drill string 225 and out of the port located on drill bit 226 (see, for example, directional arrows). As the mud exits drill string 225 through the port in drill bit 226, it may then circulate upwards through an annular region between the outer surface of drill string 225 and the surrounding wall (e.g., open borehole, casing, etc.), as indicated by the directional arrows. In this way, the mud lubricates the drill bit 226 and carries heat energy (e.g., friction or other energy) and formation cuttings to the surface, where the mud (e.g., and cuttings) can be returned to the mud tank 201, for example for recycling (e.g., by treatment to remove cuttings, etc.).

[0051] The mud pumped into the drill string 225 by pump 204 forms a mud cake lining the wellbore after leaving the drill string 225. This reduces friction between the drill string 225 and the surrounding walls (e.g., borehole, casing, etc.), among other functions. Reduced friction facilitates the advance or retraction of the drill string 225. During drilling operations, the entire drill string 225 can be pulled out of the wellbore and optionally replaced, for example, with a new or sharper drill bit, a smaller diameter drill string, etc. As described above, the action of pulling the drill string out of the hole or putting it back into the hole is called tripping. Depending on the direction of travel, the travel may be referred to as an upward travel, outward travel, downward travel, or inward travel.

[0052] As an example, consider a downstroke where, when the drill bit 226 of the drill string 225 reaches the bottom of the wellbore, mud is pumped to lubricate the drill bit 226 in order to drill and enlarge the wellbore. As described above, the mud can be pumped by pump 204 into the channels of the drill string 225, and while filling the channels, the mud can be used as a medium for transmitting energy, for example, energy that can encode information, as in mud pulse telemetry.

[0053] As an example, a mud pulse telemetry device may include a downhole device configured to influence pressure changes in the mud to generate one or more acoustic waves that can modulate information. In such an example, information from downhole equipment (e.g., one or more modules of drill string 225) can be transmitted to a surface device that can relay such information to other devices for processing, control, etc.

[0054] As an example, the telemetry device can operate via energy transmission through the drill string 225 itself. For example, consider a signal generator that transmits coded energy signals to the drill string 225 and a repeater that can receive such energy and repeat it to further transmit coded energy signals (e.g., information, etc.).

[0055] As an example, drill string 225 may be equipped with telemetry device 252, which includes: a rotatable drive shaft; a turbine impeller mechanically coupled to the drive shaft such that mud can cause the turbine impeller to rotate; a modulator rotor mechanically coupled to the drive shaft such that rotation of the turbine impeller causes rotation of the modulator rotor; a modulator stator mounted adjacent to or near the modulator rotor such that rotation of the modulator rotor relative to the modulator stator generates pressure pulses in the mud; and a controllable brake for selectively braking the rotation of the modulator rotor to modulate the pressure pulses. In such an example, an alternator may be coupled to the aforementioned drive shaft, wherein the alternator includes at least one stator winding electrically coupled to a control circuit to selectively short-circuit at least one stator winding to electromagnetically brake the alternator, thereby selectively braking the rotation of the modulator rotor to modulate the pressure pulses in the mud.

[0056] exist Figure 2 In one example, the wellhead control and / or data acquisition system 262 may include circuitry to sense pressure pulses generated by the telemetry device 252, and, for example, transmit the sensed pressure pulses or information derived therefrom for processing, control, etc.

[0057] The example component 250 shown includes a logging-while-drilling (LWD) module 254, a measurement-while-drilling (MWD) module 256, an optional module 258, a rotary steerable system (RSS) and / or a motor 260, and a drill bit 226. Such components or modules may be referred to as tools, and the drill string may include multiple tools.

[0058] As for RSS, it relates to techniques used in directional drilling. Directional drilling involves drilling into the ground to create a deviated hole, such that the hole's trajectory is not vertical; instead, the trajectory deviates from the vertical along one or more sections of the hole. As an example, consider a target located at a lateral distance from a surface location where the drilling rig can be placed. In such an example, drilling could begin from a vertical section and then deviate from that section, causing the hole to aim at and eventually reach the target. Directional drilling can be implemented where the target may not be accessible from a vertical position on the Earth's surface, where there are materials in the Earth that could hinder drilling or otherwise harm it (e.g., consider salt domes), where the formation extends laterally (e.g., consider relatively thin but laterally extending reservoirs), where multiple boreholes need to be drilled from a single surface borehole, where decompression wells are required, and so on.

[0059] One approach to directional drilling involves mud motors; however, mud motors can present several challenges depending on factors such as the rate of penetration (ROP) and the transfer of weight to the drill bit due to friction (e.g., pressure on drill bit, WOB). A mud motor can be a positive displacement motor (PDM) that operates to drive the drill bit (e.g., during directional drilling). The PDM operates as drilling fluid is pumped through it, converting the hydraulic power of the drilling fluid into mechanical power to rotate the drill bit.

[0060] As an example, a mud motor (e.g., a PDM) can operate in different modes, which can include rotary mode and sliding mode. Sliding mode involves drilling with a mud motor that rotates the drill bit downhole without rotating the drill string from the surface. This operation can be performed when the BHA is already equipped with a bent-end or bent-shell mud motor, or both, for directional drilling. Sliding can be used to build and control or adjust the hole angle. In directional drilling, the pointing of the drill bit can be achieved via a bent-end and a measuring device for determining the direction of offset, the bent-end being able to have a relatively small angular offset from the axis of the drill string. Without rotating the drill string, the drill bit can rotate with the flow of mud through the mud motor to drill in the direction it is pointing. With a steerable motor, the entire drill string can rotate to drill in a straight line rather than at an angle when the desired wellbore orientation is obtained. By controlling the number of holes drilled in sliding mode relative to the number drilled in rotary mode, the wellbore trajectory can be controlled quite precisely.

[0061] As an example, PDM can operate in a combined rotary mode, where the drill string bit is rotated by rotating the entire drill string using surface equipment (e.g., rotary table, top drive, etc.), and the drill string bit is rotated using drilling fluid. In this example, the surface RPM (SRPM) can be determined using surface equipment, and the downhole RPM of the mud motor can be determined using various factors related to drilling fluid flow rate, mud motor type, etc. As an example, in combined rotary mode, assuming the SRPM and mud motor RPM are in the same direction, the bit RPM can be determined or estimated as the sum of the SRPM and mud motor RPM.

[0062] As an example, when the drill string is not rotating from the ground, the PDM mud motor can operate in a so-called slippery mode. In such an example, the drill bit RPM can be determined or estimated based on the mud motor's RPM.

[0063] RSS (Resistant Side Controller) can be used for directional drilling where there is continuous rotation from surface equipment, which can mitigate slippage of the steerable motor (e.g., PDM). RSS can be deployed when drilling directionally (e.g., deviated, horizontal, or extended reach wells). RSS can be designed to minimize interaction with the borehole wall, which can help maintain borehole quality. RSS may be designed to apply a relatively consistent lateral force, similar to a stabilizer, which rotates with the drill string or orients the drill bit in the desired direction while continuously rotating at the same number of rotations per minute as the drill string.

[0064] LWD module 254 can be housed in a suitable type of drill collar and may contain one or more selected types of logging tools. It should also be understood that more than one LWD and / or MWD module may be used. When referring to the location of an LWD module, by way of example, it may refer to the module located at the position of LWD module 254, MWD module 256, etc. An LWD module may include capabilities for measuring, processing, and storing information, as well as for communicating with surface equipment. In the example shown, LWD module 254 may include a seismic measurement device.

[0065] MWD module 256 can be housed in a suitable type of drill collar and may include one or more devices for measuring the characteristics of drill string 225 and drill bit 226. As an example, MWD module 256 may include devices for generating electricity, for example, to power various components of drill string 225. As an example, MWD module 256 may include telemetry device 252, for example, where a turbine impeller generates electricity via mud flow; it should be understood that other power sources and / or battery systems may be used to power the various components. As an example, MWD module 256 may include one or more measuring devices of the following types: weight-on-bit, torque, vibration, impact, stick-slip, direction, and tilt.

[0066] Figure 2 Some examples of the types of holes that can be drilled are also shown. For example, consider slanted hole 272, S-shaped hole 274, deep slanted hole 276, and horizontal hole 278.

[0067] As an example, drilling operations may include directional drilling, where, for example, at least a portion of the well includes a curved axis. For instance, consider defining a radius of curvature, where the inclination relative to the vertical direction can vary up to an angle between about 30 degrees and about 60 degrees, or, for example, an angle of about 90 degrees or possibly greater than about 90 degrees.

[0068] As an example, directional wells can include several shapes, each designed to meet specific operational requirements. As an example, when information is relayed to a drilling engineer, that information can be used to perform the drilling process. As an example, the inclination and / or direction can be modified based on information received during the drilling process.

[0069] As an example, borehole deviation can be achieved in part by using downhole motors and / or turbines. Regarding motors, for example, the drill string may include a positive displacement motor (PDM).

[0070] As an example, the system may be a steerable system and include equipment for performing methods such as geosteering. As mentioned above, the steerable system may be or include an RSS (Resistant Support System). As an example, the steerable system may include a PDM (Pile Driver Deposition) or turbine located on the lower portion of the drill string, with a bend mount directly above the drill bit. As an example, above the PDM, an MWD (Modular Detection and Control) device and / or LWD (Low-Wave Detection) device may be mounted to provide real-time or near-real-time data of interest (e.g., inclination, direction, pressure, temperature, actual weight on the drill bit, torque stress, etc.). For the latter, the LWD device can transmit various types of data of interest to the surface, including, for example, geological data (e.g., gamma-ray logging, resistivity, density, and sonic logging, etc.).

[0071] The coupling of sensors that provide real-time or near-real-time information about the progress of a well trajectory with one or more logs, such as those characterizing strata from a geological perspective, can enable geologically guided approaches. These approaches can include navigating subsurface environments, for example, to follow a desired route to reach one or more desired targets.

[0072] As an example, a drill string may include an azimuth density neutron (ADN) tool for measuring density and porosity; a MWD tool for measuring dip, azimuth, and impact; a compensated dual resistivity (CDR) tool for measuring resistivity and gamma-ray related phenomena; one or more variable-size stabilizers; one or more bend joints; and a geological guidance tool that may include a motor and, optionally, devices for measuring and / or responding to one or more of dip, resistivity, and gamma-ray related phenomena.

[0073] As an example, geological steering can include the intentional directional control of a wellbore based on the results of downhole geological logging measurements, with the aim of keeping the directional wellbore within a desired area, zone (e.g., production area), etc. As an example, geological steering can include guiding the wellbore to keep it within a specific portion of the reservoir, for example, to minimize gas and / or water breakthrough, and for example, to maximize economic production from the well that includes the wellbore.

[0074] Refer again Figure 2Well site system 200 may include one or more sensors 264 operatively coupled to control and / or data acquisition system 262. As an example, one or more sensors may be located at a surface location. As an example, one or more sensors may be located at a downhole location. As an example, one or more sensors may be located at one or more remote locations not within approximately one hundred meters of well site system 200. As an example, one or more sensors may be located at an adjacent well site, wherein well site system 200 and the adjacent well site are located in a common oil field (e.g., an oil field and / or a gas field).

[0075] As an example, one or more of sensors 264 can be provided for tracking the movement of the pipe, tracking at least a portion of the drill string, etc. For example, consider tracking safety, position, velocity, acceleration, momentum, etc. As an example, tracking can be performed for one or more of the following: tripping in / out operations, drilling rate while drilling, non-productive time (NPT), invisible loss time (ILT), production time, rig status, etc.

[0076] Regarding NPT, it can be used as a metric for one or more reasons, such as considering it as an indicator of cost-effectiveness and successful drilling operations. NPT can be caused by one or more of a variety of reasons, which may include, for example, unpredictable situations, unexpected events, one or more types of technical problems, etc. Regarding ILT, it can relate to one or more routine drilling operations and can be characterized as the difference between the actual operation duration and the best practice target, which can be considered "invisible" in various cases, for example, if it does not appear in the regular morning reports.

[0077] As an example, system 200 may include one or more sensors 266 that can sense signals and / or transmit signals to a fluid conduit such as a drilling fluid conduit (e.g., a drilling mud conduit). For example, in system 200, one or more sensors 266 may be operatively coupled to a portion of riser 208 through which mud flows. As an example, a downhole tool may generate pulses that can travel through the mud and be sensed by one or more of the one or more sensors 266. In such an example, the downhole tool may include associated circuitry, such as encoding circuitry that can encode signals, for example, to reduce the need for transmission. As an example, surface circuitry may include decoding circuitry to decode encoded information that is at least partially transmitted via mud pulse telemetry. As an example, surface circuitry may include encoder circuitry and / or decoder circuitry, and downhole circuitry may include encoder circuitry and / or decoder circuitry. As an example, system 200 may include a transmitter that can generate signals that can be transmitted downhole via mud (e.g., drilling fluid) as a transmission medium.

[0078] As an example, one or more portions of the drill string may become stuck. The term "stuck" can refer to varying degrees of inability to move or remove one or more portions of the drill string from the borehole. As an example, in a stuck state, the tube may be rotated or lowered back into the borehole, or, for example, in a stuck state, axial movement of the drill string within the borehole may be impossible, although some degree of rotation is possible. As an example, in a stuck state, at least a portion of the drill string may be impossible to move axially and rotationally.

[0079] The term "stuck pipe" can refer to a portion of the drill string that cannot rotate or move axially. As an example, a condition known as "differential stuck pipe" can be a situation where the drill string cannot move along the axis of the borehole (e.g., rotate or reciprocate). Differential stuck pipe can occur when high contact forces caused by low reservoir pressure, high borehole pressure, or both are applied over a sufficiently large area of ​​the drill string. Differential stuck pipe can have both time and financial costs.

[0080] As an example, stuck pipe force can be the product of the pressure differential between the wellbore and the reservoir and the area over which the pressure differential acts. This means that a relatively low pressure differential (delta P) applied over a large working area can be just as effective at causing stuck pipe as a high pressure differential applied over a small area.

[0081] As an example, a condition known as "mechanical stuck" can occur when the movement of the drill string is restricted or prevented by a mechanism other than differential pressure stuck. Mechanical stuck can be caused by one or more of the following: debris in the hole, abnormal wellbore geometry, cement, keyway, or drill cuttings accumulation in the annulus.

[0082] As explained, a well site system may include various types of equipment for handling fluids such as drilling fluids (e.g., mud). As explained, drilling fluids may provide one or more functions (e.g., lubrication, transport of cut materials, etc.).

[0083] Drilling fluids can consist of various liquid and / or gaseous fluids, as well as mixtures of fluids and solids (e.g., as solid suspensions, mixtures of liquids, gases, and solids, and emulsions), and can be used in a variety of operations to drill holes into the ground. Drilling fluids can be classified using one or more classification schemes. For example, consider water-based mud (WBM), oil-based mud (OBM), non-water-based mud (NQBM), gas-based mud (e.g., pneumatic mud, etc.) (GBM), etc.

[0084] As an example, drilling fluid may leak into the formation and / or reservoir fluids may enter the drilling fluid. Therefore, one or more functions of the drilling fluid may be impaired due to changes in the drilling fluid. For instance, if the density of the drilling fluid is altered by introducing reservoir fluids, the drilling fluid may reduce its ability to transport cuttings to the surface. As a result, cuttings may accumulate in the annulus between the drill string and the borehole wall or casing, which may increase the risk of stuck pipe (e.g., stuck casing). To address changes in the drilling fluid, one or more actions can be taken, such as considering adding one or more components to the drilling fluid, adding additional drilling fluid, etc.

[0085] The terms circulating loss (loop leakage) or circulating leakage (circulating leakage) can refer to, for example, the loss of drilling fluid to the formation when the hydrostatic head pressure of the drilling fluid column exceeds the formation pressure. This fluid loss can be broadly classified as seepage loss, partial loss, or catastrophic loss, and each type of loss is treated differently depending on the risk to equipment, materials, borehole quality, drilling fluid properties, personnel, etc.

[0086] The inflow of formation fluids (e.g., reservoir fluids, etc.) can include events known as well kicks. A well kick can be defined as the inflow of formation fluids into the borehole during drilling operations. A well kick may be physically caused by the pressure in the borehole being less than the pressure of the formation fluids, thus inducing the flow. This situation, where the borehole pressure is lower than the formation pressure, can occur in various ways. For example, if the mud weight is too low, the hydrostatic pressure exerted on the formation by the fluid string may be insufficient to hold the formation fluids in the formation. This can happen if the mud weight suddenly becomes lighter or is initially not up to standard, or if the drilled formation has a higher pressure than expected. This type of well kick can be called an underbalanced well kick. As another example, consider a well kick that can occur if dynamic and transient fluid pressure effects (e.g., due to the movement of the drill string or casing) effectively reduce the pressure in the borehole to below the formation pressure. This type of well kick can be called an induced well kick.

[0087] Other phenomena that can occur during drilling operations include pumping and surges. Pumping involves reducing the pressure in the borehole by moving pipes, cable tools, or sealing the borehole with rubber cups. If the pressure is sufficiently reduced, reservoir fluids can flow into the borehole and to the surface. Pumping is often harmful during drilling operations because it can lead to well kicks and borehole stability problems. Regarding surges, consider the example of the drill string being pulled out (e.g., tripping out (POOH)), where the upward movement of the drill string causes friction between the drill string and the drilling fluid. In a surge, the pressure inside the borehole may decrease due to the surge effect; note that the opposite effect can occur when the drill string is tripped up or down (e.g., running down (RIH)), as downward movement can cause an increase in pressure (e.g., the pumping effect).

[0088] As explained, the drill string may include a mud motor that is rotary driven by the flow of drilling fluid. In this drilling mode, the properties of the drilling fluid can affect the performance of the mud motor. For example, density (e.g., mud weight) can affect how much energy the mud motor can deliver to the drill bit for a given drilling fluid flow rate.

[0089] Regarding the risk of a stuck pipe or stuck drill pipe event, as explained, one or more actions can be taken. For example, consider adding acid as a remedy to resolve a stuck pipe event or reduce the risk of a stuck drill pipe event. In such an instance, multiple barrels of acid may be added to the drilling fluid, which is circulated downhole to the annular area between the drill string and the borehole wall to "dissolve" the material causing the stuck pipe or stuck drill pipe risk. While the addition of acid is mentioned, it can be one of a series of actions that can be taken, where each action may have associated benefits and harms. As for the harms, these may include NPT, ILT, costs, further remedial measures (e.g., the effects of acid on one or more additives in the drilling fluid), etc. Therefore, in the event of an event or an increased risk of an event, one or more actions may be implemented strategically to resolve the event or otherwise reduce the risk in order to maintain adherence to the plan.

[0090] Figure 3 An example of a well site system 300 is shown, specifically, Figure 3 Block diagrams of well site system 300 and system 370 are shown in approximate side view and approximate plan view.

[0091] exist Figure 3 In the example, well site system 300 may include a compartment 310, a rotary table 322, a winch 324, a mast derrick 326 (e.g., optionally carrying a top drive, etc.), a mud tank 330 (e.g., having one or more pumps, one or more vibrating screens, etc.), one or more pump houses 340, a boiler house 342, an HPU house 344 (e.g., having a rig fuel tank, etc.), a combination house 348 (e.g., having one or more generators, etc.), piping 362, a catwalk 364, a flare 368, etc. Such equipment may include one or more associated functions and / or one or more associated operational risks, which may be risks related to time, resources, and / or people.

[0092] like Figure 3As illustrated in the example, well site system 300 may include system 370, which includes one or more processors 372, memory 374 operatively coupled to at least one of the processors 372, instructions 376 that may be stored, for example, in memory 374, and one or more interfaces 378. As an example, system 370 may include one or more processor-readable media comprising processor-executable instructions executable by at least one of the processors 372 to cause system 370 to control one or more aspects of well site system 300. In such an example, memory 374 may be or include one or more processor-readable media, wherein processor-executable instructions may be or include instructions. As an example, processor-readable media may be computer-readable storage media that is not a signal and not a carrier wave.

[0093] Figure 3 A battery 380 is also shown, which can be operatively coupled to system 370, for example, to power system 370. As an example, battery 380 can be a backup battery that operates when another power source is unavailable to power system 370. As an example, battery 380 can be operatively coupled to a network, which can be a cloud network. As an example, battery 380 can include smart battery circuitry and can be operatively coupled to one or more devices via SMBus or other types of buses.

[0094] exist Figure 3 In the example, service 390 is shown to be available, for example, via a cloud platform. Such services may include data service 392, query service 394, and drilling service 396. As an example, service 390 could be such as Figure 1 This is part of system 100 (e.g., consider planning services and / or operational services). As an example, service 390 may include one or more services for directional drilling (e.g., consider a computational framework, which can provide one or more services that utilize real-time data to estimate one or more parameters).

[0095] As an example, system 370 can be used to generate one or more penetration rate drilling parameter values, which can be used, for example, to control one or more drilling operations.

[0096] Figure 4An example of a system 400 is shown, comprising off-site equipment 401 (e.g., remote) and on-site equipment 402 (e.g., local). As shown, off-site equipment 401 may include a drilling operations framework 410, a drilling planning framework 420, and a database 430, and on-site equipment 402 may include a controller 440 that can receive real-time data and output recommendations such as control commands to control the on-site equipment. In such an example, drilling operations framework 410 may provide steering tables, execution parameters, etc., and drilling planning framework 420 may provide steering response and statistical evaluations. As shown, controller 440 may output information to drilling operations framework 410 and receive information from drilling planning framework 420. System 400 may include planning generation features for real-time planning generation during the drilling operations execution phase and / or planning generation during the planning phase. System 400 may be used for one or more types of drilling (e.g., rotary, mud motor, RSS, ABSS, etc.). System 400 may have an operational loop that may include at least one real-time loop providing control of the equipment to perform drilling operations.

[0097] Systems such as System 400 can utilize various functions and constraints to generate plans that can provide single or multiple target targeting. As explained, plans designed to provide drilling operations for multi-well configurations can be generated. As explained, the plan can be a digital plan that can be used to instruct one or more controllers, such as an automated drilling rig controller, which can control one or more pieces of equipment (e.g., winches, top drives, one or more drilling fluid pumps, etc.). As an example, an automated drilling rig can control the energy delivered to the drill bit via one or more pieces of equipment, where the drill bit crushes and / or cuts rock to extend the borehole. As an example, an automated drilling rig can be controlled to focus on minimizing or otherwise reducing mechanical specific energy (MSE) and maximizing or otherwise increasing the rate of drilling (ROP).

[0098] As an example, a framework can be used to facilitate control over one or more field operations at a well site, providing one or more actions to address events, event risks, etc. For instance, consider a framework that allows engineers to interact with it via one or more graphical user interfaces (GUIs) and / or one or more other types of user interfaces (UIs) (e.g., microphone, QR code, etc.). In such an example, an engineer could input information about events such as pipe jamming, well leakage, etc. In response, the framework can transform the received input into prompts or queries submitted to one or more large language models (LLMs), which can be enhanced with embeddings from historical drilling events and applied remedies. In response, the framework can generate outputs such as, for example, a ranked list of various historical events and remedies based on their validity, which can be presented to the engineer as recommendations, such as text summaries of remedies, rendered graphical controls that can be actuated to request field actions (e.g., by one or more pieces of equipment, by an automated borehole rig, etc.).

[0099] As an example, the framework can provide machine-to-machine interaction, such as for monitoring, control, and notification. As another example, machine-to-machine methods can provide natural language accompanying machine instructions, where, for example, natural language can be generated and translated into machine instructions, and / or machine instructions can be generated and translated into natural language. In such examples, the user interface can be used for Human-Machine Loop (HITL), which can be used for regulatory compliance, monitoring, control, and other purposes. In various situations, while basic automation can be achieved to control field operations, regulatory and / or practical issues (e.g., regarding equipment, drilling, environment, etc.) may require HITL.

[0100] As an example, a framework can leverage one or more generative artificial intelligence (AI) techniques. For instance, a generative AI model can provide a form of AI that primarily creates new text, images, videos, audio, or other content based on training data used to train the generative AI model. Large language models (LLMs) can be considered a form of generative AI that can focus on understanding text input (e.g., using natural language processing) and can be used to create outputs (e.g., human-understandable text) based on given input. LLMs can be viewed as a subset of generative AI for language-related tasks, which can be adapted to assist humans, machines, and both humans and machines.

[0101] A large language model (LLM) can be a language model known for its ability to achieve general language understanding and generation (e.g., as a generative model, etc.). An LLM acquires this capability by learning parameters (e.g., determining parameter values, etc.) using a relatively large amount of data during training. An LLM can be one or more artificial neural networks (e.g., consider a transformer, etc.) and can be trained and / or pre-trained using one or more types of learning (e.g., self-supervised learning, semi-supervised learning, unsupervised learning, etc.).

[0102] Regarding the transformer, it can include a block architecture, where, for example, multiple transformer blocks can be arranged as and / or referred to as layers. For example, a transformer may include self-attention layers, feedforward layers, and normalization layers, which can operate to process the input to predict the output during inference. As an example, layers can be stacked to create deeper transformers.

[0103] As an example, autoregressive language models (e.g., AR LLMs) can operate by taking input text and repeatedly predicting the next lexical unit or word. As an example, LLMs can be tailored for specific domains, for example. As an example, LLMs such as Generative Pre-trained Transformer (GPT) 3 (GPT-3) can perform cue engineering. As an example, LLMs can acquire representational knowledge about the inherent syntax, semantics, and ontology in human language corpora; note that LLMs may also acquire inaccuracies and biases present in the corpus. As an example, one or more generative AI models can, for example, utilize encoder-decoder architectures to provide encoding, decoding, or both. Regarding encoders, consider implementations for tasks that can understand language, such as classification and sentiment analysis (e.g., consider Bidirectional Encoder Representation (BERT) models from transformers, etc.). Regarding decoders, consider implementations for tasks such as generating language and / or other content (e.g., consider GPT-3 models, etc.). Regarding encoder-decoder architectures, such architectures can be implemented to understand and generate content. For example, consider translation and generalization (e.g., consider T5 text-to-text converter methods, etc.). As explained, the framework can provide translations such as from machine language to machine language, from human language to human language, from machine language to human language, and from human language to machine language.

[0104] Figure 5An example architecture 500 of GPT is shown; note that one or more features of architecture 500 can be used in LLMs and / or other generative AIs suitable for use within the framework. As explained, generative AIs can leverage transformer blocks, where each transformer block can include various components. As an example, the base GPT model can be further adapted to produce more targeted systems for specific tasks and / or subject domains. Techniques for such adaptation can include additional fine-tuning (e.g., adjustments beyond the base model, etc.), some form of prompt-engineering, etc. As an example, an LLM can be a chatbot-type LLM. For example, consider the OpenAI ChatGPT LLM, an online chat interface driven by instructions trained in a manner similar to InstructGPT that tunes a language model. Other chatbots can include features of GPT-4 (OpenAI), Bard (e.g., the LaMDA family of dialogue-trained language models, PaLM, etc.) (Google, MountainView, California), etc.

[0105] As an example, LLM Meta (LLaMA) can be utilized, which includes a transformer architecture; note some architectural differences compared to GPT-3. For instance, LLMaMA utilizes the SwiGLU activation function instead of ReLU, uses rotated position embeddings instead of absolute position embeddings, and uses root mean square layer normalization instead of standard layer normalization. Furthermore, the context length can be increased from 2K (Llama 1) tokens to 4K (Llama 2) tokens.

[0106] Figure 6 An example graph 600 shows the cumulative nonproductive time (NPT) for each failure mode in field operations. Specifically, failure modes (such as, for example, stuck pipe, tool jamming, lost circulation, downhole (DH) equipment mechanical integrity, borehole problems, and cementing) can include one or more types of events and / or be associated with one or more types of events. As shown in Figure 600, stuck pipe and / or tool jamming can cumulatively result in hundreds of days of NPT. NPT can be time, but resources, energy, emissions, risks, etc., can also be associated with one or more failure modes. For example, failure modes may lead to increased carbon emissions (e.g., greenhouse gas emissions), fuel consumption, risks to people, risks to borehole integrity, etc. Although NPT is shown in graph 600, as mentioned above, ILT can be tracked. As an example, the framework can provide an assessment of NPT and / or ILT, which can provide controls for improved field operations.

[0107] As an example, during field operations, one or more individuals can enter a description of an event, followed by a description of one or more remedial and / or containment options that can be taken and / or are being taken. For instance, consider a daily drilling report, which may include one or more descriptions of one or more events, event risks, etc.

[0108] As an example, a Daily Drilling Report (DDR) may include one or more of the following: the name of the well; the latitude and longitude of the well location; the water depth at the well; the depth drilled; the work performed; the lithology of the formation penetrated; details of hydrocarbon indications; an overview of the materials used; drilling fluid losses; stuck pipe (e.g., stuck tubing, stuck tools, etc.); a summary of leak tests; the geometry of the wellbore; the results of explorations conducted in the wellbore; and estimated daily and cumulative well costs. While various inputs may involve some human-machine interface (HMI) interactions, as explained, the framework can provide machine-to-machine interactions (e.g., via one or more machine-to-machine interfaces (MMIs)).

[0109] Figure 7 An example of a portion of the DDR 700 is shown, including a description of out-of-range (OOS) problems, outlining various measures to address them. For example, consider circulation, pumping material, etc., which can occur with or without rotation of the drill string and / or other tool strings. As an example, the description may include a physical description of one or more of the following: volumetric rate, volume, type of plugging slurry, type of measure, etc. As an example, one or more such descriptions may be entered via an HMI and / or MMI. In various instances, when utilizing an HMI, it may be in a relatively harsh environment, potentially due to noise, equipment, environmental conditions, etc. In various instances, people may wear PPE, making data entry more challenging. For example, consider using a keyboard while wearing PPE gloves and / or simultaneously monitoring equipment, people, materials, etc. In various instances, typesetting problems may arise, which can be handled by one or more features of the frame (e.g., in an automatic, semi-automatic, etc.)

[0110] Figure 8 Examples of one or more parts of one or more DDR 800 are shown, as indicated by the markings regarding hole problems, particularly circulation losses (e.g., well leakage or circulation loss). Various measures can be described as shown in the figure. Figure 8In the example, the description may include a description of the characteristics of the well leakage event (e.g., considering severity between approximately 180 bbl / h and 260 bbl / h), followed by remedial measures (e.g., “100 barrels of 200 ppb LCM plugging slurry injected into OH via PBL sub (first time)”). As an example, the framework could provide such data on the event and one or more measures to address problems where such events occur at the same well site and / or one or more other well sites.

[0111] Because each well and its associated operations may have some characteristics that differ from one or more other wells and their associated operations, teams (e.g., new teams, shift teams, etc.) may need time to become familiar with the well and / or a portion of it, and how to control the associated operations, especially when facing events or the risks associated with those events. For example, a team's learning curve may require approximately three to four events before the team learns the most effective ways to manage that type of event or its risks. However, this learning can be affected by personnel turnover. For example, consider replacements due to shift changes, transfers, etc., involving one or more individuals within the team. Furthermore, various entities may tend to have organizational silos, where learning occurring in one silo does not propagate to one or more other silos, which may prevent lessons learned in a specific operational environment from being transferred for future application.

[0112] Standard operating procedures (SOPs), local practices, and successful case studies of incident response are often quite extensive (e.g., thousands of instances, which may be documented in one or more knowledge bases, scientific publications, regulatory reports, etc.). Such data, or portions thereof, can be accessed for the purpose of training one or more models of the framework, allowing the framework to effectively shorten the learning curve and reduce the impact of NPT; note that, based on... Figure 6 As shown in Figure 600 and elsewhere, the duration of NPT for human teams at the beginning of the learning curve is often almost twice the duration at the end of the learning curve. As an example, a framework can be operable to reduce NPT at the start of field operations and, for example, can continue to reduce NPT as field operations continue, with NPT being reduced more effectively because, for example, the framework can provide one or more measures early in the event, thereby appropriately addressing the occurrence and / or risk of such an event at a later time. For example, given available resources (e.g., materials, equipment, personnel, etc.), the framework can provide one or more measures suitable for the well site, where individuals on one or more teams (e.g., due to shifts, reassignments, etc.) can more easily understand the characteristics of the event and its solutions, which in turn can reduce NPT and / or one or more other metrics (e.g., emissions, fuel use, risk, etc.).

[0113] Figure 9 An example of a system 900 including various features 902, 904, and 906 is shown. As illustrated, feature 902 may include features for generating a fine-tuned LLM 920, for example, using named entity definitions 912, one or more vocabularies 914 (e.g., SLB energy vocabulary, etc.), and one or more general models 916 (e.g., one or more general LLMs). Regarding feature 904, these may provide processing of DDR 942, where the fine-tuned LLM 920 may be prompted to extract fault modes, remedies, and results 944. In such an example, the fine-tuned LLM 920 may output contextualized event and remedy descriptions 946, which may be processed, for example, using a sentence transformer 948 to generate labeled embeddings 950. Regarding feature 906, these may include an interface for receiving drilling reports, events, and / or risks 962 (e.g., as input via a UI, automatically generated, semi-automatically generated, etc.). As shown in the figure, received data (e.g., data, descriptions, etc.) can be used to prompt the LLM 920 of the 964 fine-tuning to extract one or more fault modes, which can be one or more contextualized events using description 966. Given one or more descriptions, the fine-tuning LLM 920 can be operated using tagged embeddings 950 to generate outputs 968 that may include remedial recommendations, which can be associated with control commands that can be implemented to control field equipment, etc.

[0114] like Figure 9 As shown in the example, one or more criteria 980 can be used to rank remedial recommendations so that a ranked list and / or other presentation of one or more remedial recommendations can be reviewed for implementation and / or implemented automatically (e.g., consider receiving input via a UI, which allows selection of one remedial recommendation for implementation, automatic implementation based on ranking criteria, etc.). In such an example, the one or more criteria that can be used for ranking may include one or more criteria associated with the well site, such as equipment, one or more controllers, materials, personnel, etc. For example, if the well site does not have a sufficient supply of a particular material specified in the remedial recommendation, this could be a criterion that lowers the ranking of that remedial recommendation. As another example, if emissions are being monitored and emissions are close to limits, the ranking of a remedial recommendation that might push emissions above the limits can be lowered. As an example, one or more criteria may be real-time, such as regarding the supply of materials, available energy, emission levels, etc. As an example, criteria may involve time, such as NPT and / or other time periods. As an example, ranking may depend on a reduction in NPT.

[0115] As explained, a workflow may include fine-tuning a generic LLM, for example, enriching an LLM with field-specific definitions (e.g., drilling, etc.) using a series of questions / answers. As an example, a workflow may involve translating or otherwise transforming descriptions from available DDRs into a suitable language, such as English (e.g., optionally, using a natural language processor, etc.). As an example, a workflow may include extracting event descriptions and / or remedial descriptions (e.g., descriptions of remediation and / or containment measures, characteristics, etc.). As an example, a workflow may include performing sentence embedding extraction. For example, a generic LLM may be trained (e.g., pre-trained) using a language such as English, where a specific type of structure (e.g., sentences, etc.) can be expected. In such an example, information from the DDR can be transformed into this type of structure (e.g., sentences, etc.).

[0116] As an example, system 900 can utilize sentence transformer 948 to extract various keywords (e.g., related terms, etc.), where such keywords can be used during the tagging process to generate labeled embeddings 950. As an example, a specific event description that can be characterized as a failure mode can be represented as a vector, for example, based on labeled embeddings generated using the sentence transformer. As an example, the workflow can utilize indexing, where keywords are indexed for the purpose of finding one or more similar descriptions (e.g., DDR, etc.) within a fine-tuned LLM. As an example, the workflow may include transforming event descriptions (see, for example, DR / event block 962) into vectors, such that said vectors can be used to efficiently perform searches against other vectors to identify one or more similar events, which, as explained, may have associated remedies (remedies) leading to a successful outcome. While various examples focus on successful outcomes, the system can provide outputs generating information about unsuccessful outcomes, which can be used, for example, as part of a warning technique to alert operators (e.g., operations teams) which measures are unlikely to provide a successful outcome. As an example, the system (e.g., a framework, etc.) can provide the generation of permitted and prohibited measures, which, as explained, can be used for the control of field equipment. In such a method, a certain level of automatic control can be achieved, with safety protection measures, such as preventing unauthorized actions from being implemented. These measures can be individual, a series of, or parallel measures. As an example, the system can be or includes a controller.

[0117] As explained, System 900 can utilize a vector-based approach, where historical data (e.g., DDR 942) is transformed into vectors using a fine-tuned LLM, and a description is transformed into a vector using a fine-tuned LLM, which can be used to identify one or more relevant historical events in the historical data (e.g., transformed into vectors). Since historical data can be processed to obtain remedies and results, these remedies and results can be returned to consider resolving the risk of events or incidents associated with the description, which can be a fresh description (e.g., a description from a few minutes ago) of one or more physical phenomena occurring at the well site (e.g., during ongoing field operations, etc.). As explained, this approach can provide a relatively rapid resolution of events, allowing for appropriate management of NPT and / or one or more other metrics (e.g., energy, materials, emissions, etc.).

[0118] As explained, workflows can include LLMs fine-tuned using well drilling definitions. For example, a generic LLM might be trained based on publicly available information, and when prompted with specific drilling information containing unknown definitions such as "LCM" or "HIVIS PILL," the model might fail to correctly understand the context of the query. Therefore, fine-tuning could involve providing a set of questions and answers to a generic (e.g., generalized) LLM to adjust (tune) the LLM using new information. Some examples of questions and answers are shown in Table 1 below.

[0119] Table 1. Some examples of questions and answers

[0120]

[0121] As shown in Table 1, various specialized terms can be introduced, which may relate to one or more specific organizations. As an example, a fine-tuned LLM can be tailored to a specific organization or multiple organizations. As an example, the framework can provide a language-based preliminary search to automatically select a fine-tuned LLM. For example, if the DDR includes terms indicating a specific organization, the framework can select a fine-tuned LLM associated with that specific organization (e.g., fine-tuned using that term). As explained, Standard Operating Procedures (SOPs) can be implemented in the field, where, for example, the SOPs may be different and / or the same for one or more organizations. Since SOPs can be part of remedial work and / or incident detection, a fine-tuned LLM can be SOP-specific.

[0122] As an example, the framework can operate using one or more information sources, which may include privately available definition sources, such as from one or more drilling reporting systems and / or one or more oil and gas-specific vocabularies. As explained, the generic LLM can be adapted to better understand drilling domain language and / or one or more field operations domain languages.

[0123] Regarding drilling event summaries, consider including incremental DDR information on how the event unfolded and how the operations team handled it. In various situations, reviewing historical data and identifying what happened, what or which remedial actions were taken, and which were successful can be challenging. As explained, a fine-tuned LLM can provide processing of historical data (e.g., DDR, etc.) that allows it to leverage such knowledge in real-time (e.g., or near real-time) to generate outputs (e.g., remediation recommendations) using one or more descriptions specific to the well currently being drilled. As explained, a fine-tuned LLM can be used to summarize cascading descriptions of events. As explained, a fine-tuned LLM can be prompted to use descriptions as input to extract failure modes, remedial actions, and outcomes.

[0124] As an example of description, consider the following text:

[0125] Circulate the acid solution. - Close the blowout preventer and circulate the solution through the throttling manifold as needed in a controlled manner. - Dispose of 280 barrels of contaminated sludge and acid… Pick up and assemble the 4-11 / 16 inch retrieval bucket assembly. This includes: - 4-11 / 16 inch lip rails - 4-11 / 16 inch bowl (equipped with a W / 3-1 / 8 inch auger grab, mill control, and packer) - 2 x 4-11 / 16 inch extensions - Top subpump - Outlet time / O (3-1 / 2 inch IF Box-XT-39) - Monitor the well on the retrieval tank. Hold a pre-operation safety meeting with all personnel involved in the acid washing operation. Address the stuck tubing. Lift the tubing and confirm it is free. Install ground piping + 2x FOSV + side inlet connector. -P / Pressure test ground piping to 5000 psi, normal. … The 5-7 / 8 inch VORTEX bottomhole assembly was pulled from 12,046 feet to 11,380 feet (40 feet inside the 7-inch liner). - Obstruction occurred at 11,862 feet and 11,745 feet, which was resolved by moving the tubing string. - Obstruction occurred at 11,841 feet, which was resolved by slightly backreaming. The 4-11 / 16 inch retrieval chute is being lowered into the well using a 5-1 / 2 inch drill pipe. The well condition is being monitored from the tripping chute. Retrieval operations were conducted using a 4-11 / 16 inch retrieval cartridge assembly. - Flow rate: 37 gallons / minute; riser pressure: 320 psi. - A staged lowering force was applied, increasing to 5 klbs and then increasing to 335 klbs (40 klb overpulling). No clear indication of catching the object was observed, and the riser pressure remained unchanged. - The object was re-probeed and observed to be free to descend; it had fallen. - The object was lowered freely to 12,021 feet, where it encountered confinement. - The lowering continued, and a 5 klb drag was observed at 12,148 feet (the 4-11 / 16 inch retrieval cartridge assembly was at the bottom of the well at 12,360 feet). Pumping at 37 gallons / minute, the pressure increased to 420 psi. - Pumping at 37 gallons / minute, the pressure increased to 420 psi. - 4-11 / 16 inch fishing tubing assembly was moved to the bottom of the well, lowered to 6K lbs and raised to 335K lbs. - Pressure increase was observed at a flow rate of 26 gallons / minute, riser pressure: 410 psi. Depressurization was performed. - The tubing string was checked for freeness and pumped at @ 3 SPM, riser pressure: 440 psi. The debris was engaged. - Pressurization was increased to 1000 psi according to the WL procedure to confirm debris engagement, which was normal. - Mud return was monitored at the tripping and running-out point during circulation. Measuring instrument 0 MRM / h, radiation source integrity, normal. … At a controlled rate, retrieve the 4-11 / 16 inch fishing tubing assembly from the bottom of the well to 11,210 feet (the end of the 4-11 / 16 inch fishing tubing assembly). - Monitor well conditions using the tripping chuck and tripping log. - Wet tripping of the tubing string. ... Rinse and ream the borehole from 11,420 feet down to the bottom at 12,360 feet in the open hole. - From: 260 gallons / min; Riser pressure: 1,660 psi - TDS RPM: 100 rpm; Torque: 3.3-4K FT-LB. - Perform two reams and backholes at the 1,750-12,050 foot clamping interval. - Perform a clamping test on each bracket before connection. Depth

[0126] 1MIN 2MIN 3MIN Torque: 11488 ft 6.7K 6.8K 6.8K ft-lb 11502 ft 6.7K 6.8K 7.0K ft-lb 11677 ft 6.8K 7.2K 7.2K ft-lb 11771 ft 6.7K 6.8K 6.8K ft-lb 11864 ft 6.7K 7.1K 9.6K ft-lb 11958 ft 6.7K 6.9K 7.1K ft-lb 12053 ft 7.0K 6.8K 7.0K ft-lb 12148 ft 7.0K 6.9K 7.0K ft-lb 12241 ft 6.2K 7.3K 7.1K ft-lb 12337 ft 6.9K 6.8K 7.0K Feet to pounds. Flushing ground-level pipelines. - Flushing throttling manifolds and throttling pipes. At 11,840 feet, the stuck tubing was addressed in two directions. - Upward jerk and pull up to 500 klbs (210 klbs overpulling). - Downward jerk and apply a right-hand torque of 12 kft-lbs and depress the entire weight of the tubing. - Unsuccessful. - Simultaneously install cementing equipment and surface lines for acid treatment operations. Pressure test the surface lines to 5,000 psi. Normal. Address the stuck drill string and release it. - Allow the acid reaction to proceed for 30 minutes before moving the tubing. - Pump 5 barrels of fresh acid into the formation every 30 minutes. Observe a loss of 30 barrels at the end of the acid treatment operation. Replenish the annulus; the fluid level should remain stable. Run the 5-7 / 8 inch cleaning bottomhole assembly from 11,840 feet to 11,300 feet (120 feet inside the 7-inch liner). - Open hole condition is good. Address stuck pipe. Inject 40 barrels of 20% HCl as follows: - Pump 180 barrels of 125 PCF mud at the front. - Pump 20 barrels of 105 PCF mud. Switch to cementing assembly; - Pump 40 barrels of 20% HCl acid. - Pump 50 barrels of 105 PCF mud containing bridging material as a placebo. - Switch to drill rig pumps: B / O 2x FOSV and side inlet sub. - Displace 20% HCl acid with a total of 139 barrels. Retain 20 barrels of acid in the string. - Compress the string (lower by 50 klbs) and apply 9 kft-lb torque to the string. Unsuccessful attempt to resolve stuck pipe at the bottom of the well. - Shocking upwards and pulling up to 520K lbs (200-220K lbs overpulling), unsuccessful. - Shocking downwards and applying maximum lowering weight, applying 9K ft-lb torque to the tubing string. - Displace 1.5 buckets of acid into the open hole section every 30 minutes. Lowering a 4-11 / 16 inch fishing rig into the well using a 4-inch drill pipe stand, from 11 feet to 3128 feet. - Monitoring well conditions on the tripping hopper. Using a 4-11 / 16 inch fishing canister, the fishing assembly was attached to a 5-1 / 2 inch x 4 inch drill pipe stand and lowered from 3600 feet to 11395 feet. - Well conditions were monitored on the tripping chuck. Surface tubing + 2 FOSVs + side inlet subs were assembled. The 4-11 / 16 inch fishing canister assembly was pulled out from 1500 feet to 1000 feet at a controlled rate. - Well conditions were monitored via the tripping chuck and tripping log. - Tripping rate: 3 min slip to slip. - Wet tripping. - Cable gauge 0 MREM / h, radiation source integrity, normal. Replace the hanger with a 5-1 / 2 inch drill pipe. A sudden 25klb overpulsation was observed at 11832 feet, immediately followed by free pull. Free pull continued for 15 feet to 11817 feet. To confirm free passage, the jack was lowered back into the well. At the same point, a 12klb descent block was observed. The tubing was moved and free passage was achieved. The tubing was lowered and a 12klb descent block was encountered at 11835 feet. Rotation was applied, and free passage was achieved several times between 11840 and 11800 feet. Rotation was stopped. A descent block was encountered at 1184 feet when the jack was lowered back into the well. The tubing was moved, with no progress. The pump was started from 11840 to 11780 feet for back-reaming. Torque was stable: + / - 3.9kft-lb, no issues. To clear the area below 11840 feet, slips were placed 5 feet above the rotary table and connected. Differential pressure sticking was observed after connection. - Sticking point inclination + / - 49 degrees. Circulate and adjust mud properties. - Flow rate: 260 gallons / minute; riser pressure: 1660 psi - Up / down swivel: 60 / 100-120 rpm; Torque: 3.3-4 ft-lb. - No increment / no leakage - Recorded parameters: Lift weight: 290 lb, Lower weight: 275 lb, TDS speed: 5 / 13 rpm; Torque: 3.5 ft-lb - Mud weight in / out: 107 PCF. Disconnect the ground pipelines. Circulate at 11,395 feet before entering the open hole section. Flow rate: 200-160 gallons / minute; riser pressure: 440 psi. Recorded parameters: lifting weight: 280 psi, lowering weight: 271 psi, torque: 2.7 psi @ 10 rpm. Pull out the wellbore and drop the retrieval logging string. - 100% of the dropped material was recovered. - The traction tool was retrieved in good condition. - The remaining retrieved logging tools were retrieved in good condition. - The radioactive source has been recovered. - Well conditions were monitored on the tripping rig. - The WL measuring instrument showed 0 MREM / hour, and the radioactive source integrity was normal. A pre-operation safety meeting was held with all personnel involved in the acid treatment operation. Unsuccessful attempts to resolve stuck pipe at the bottom of the well. - Upward jerk and pull up to 400K lbs (100-110K lbs overpulling), unsuccessful. - Lowering 50K lbs and applying a maximum torque of 9K lb-ft, unsuccessful. - Simultaneously installing cementing equipment and surface pipelines for acid treatment operations. Flow inspection, well stationary. Activate ECRD (Electronic Controlled Release Device). - Release the cable from the cable cap. Raise the cable to the surface. - Remove the cable surface tools. - Monitor well conditions from the trip hopper. HPJSM was performed before assembling the salvage unit. Pump acid into the casing section. - Verify free passage by rotating the tubing string back and forth. Retrieve the 4-11 / 16 inch fishing rig assembly from 1500 feet at a controlled rate; in progress. - Monitor well conditions using the tripping rig and tripping log. - Tripping rate: 3 min slip to slip. - Wet tripping. - Cable measuring instrument 0 MREM / hour; radiation source integrity normal. Circulate and adjust mud properties. - Total gas 1.1-2%, CO2 29%. - Flow rate: 260 gallons / minute; riser pressure: 1550 psi - No increment / no leakage - Mud weight in / out: 107 PCF. Unsuccessfully dealt with cable string jamming. - Moved the cable downwards but the tool did not move. - Pulled to the cable's maximum safe tension (MSP) (13984 LBF) unsuccessfully. - Pulled the auger downwards unsuccessfully. - Pulled the auger upwards unsuccessfully. - Maintained the tool under tension unsuccessfully. - Monitored well conditions on the tripping hopper. - Conducted a hydrogen sulfide drill, with a good response. The drill string got stuck at 12,046 feet after connection. Handling the stuck drill. - Applying torque to 9,000 lb-ft, the drill string failed to move. - Multiple upward jogs, the drill string did not release. - R / B stand-up and single connection. - Unrestricted circulation, differential pressure stuck. - Jogs working, no problem. Reamed one single drill before connection (torque: 4.1-4.5 lb-ft). Tracked the inclination, the result exceeded acceptable standards. Reamed and back-reamed 11 feet (torque: 4.0-4.3 lb-ft). Repeated inclination, normal. Reamed one single drill (torque: 3.9-4.2 lb-ft) and connected. Address stuck pipe. Inject 100 barrels of 20% HCl acid plugging slurry as follows: Drill rig pump; - 180 barrels of 125 PCF weighted plugging slurry - 20 barrels of 125 PCF high-viscosity weighted plugging slurry. Switch to cementing equipment; - 100 barrels of 20% HCl + VDA acid plugging slurry. Switch to drill rig pump: B / O FOSV + side inlet sub - 20 barrels of 125 PCF high-viscosity weighted plugging slurry - 30 barrels of 105 PCF mud, containing 30 lbs / barrel of bridging material - Replace the acid with 105 PCF mud at a drill string capacity of 99 barrels. - Place 40 barrels of 20% HCl VDA acid in the open hole section, keeping 60 barrels in the tubing string. Place 50 barrels of CASING GO plugging grout at the bottom of the well - Flow rate: 260 gallons / minute; riser pressure: 1650 psi. - Up / down reaming: 60 / 120 rpm; torque: 3.5 ft-lb. Recirculate the acid solution. - Close the blowout preventer and circulate the acid solution through the throttling manifold in a controlled manner as required. - Discharge 180 barrels of contaminated sludge and acid. Circulate and adjust mud properties. - Total gas 1.1-2%, CO2 29%. Raise the 4-11 / 16 inch fishing can assembly from 1000 feet to the surface at a controlled rate. - Monitor well conditions via the tripping can and tripping log. - Tripping rate: 3 MIN SLIP to SLIP. - Wet tripping. - Logging material engaged when the fishing can assembly is pulled out. - Cable measuring instrument 0 MREM / hour, radioactive source integrity normal. Immersion Time: - Allow 30 minutes for acid reaction before moving the tubing string. - Lower the tubing string to a weight of 150 kg and apply a torque of 9 kg ft-lb. - After the 30-minute immersion time, drain 10 barrels from the drill bit to refresh the acid in the open hole section. 20 barrels of acid were observed lost at the end of the acid treatment operation. The annulus was replenished by tripping the drill string, and the fluid level stabilized. Handling stuck pipe: Inject 55 barrels of 20% HCl as follows: Drill rig pump; - Pre-pump 160 barrels of 125 PCF mud. - Pump in 20 barrels of high-viscosity 125 PCF mud. Switch to cementing equipment; - Pump in 55 barrels of 20% HCl acid. - Pump in 20 barrels of 125 PCF high-viscosity mud at the rear. - Pump in 30 barrels of 107 PCF mud containing bridging material. - Switch to drill rig pump: B / O 2x FOSV and side inlet sub. - Displace the 20% HCl acid with a total of 118 barrels of 107 PCF mud. - Place 20 barrels of HCl acid in the open hole section and retain 35 barrels of acid in the tubing string. Circulate and adjust mud properties. - Mud weight in / out: 105 PCF. Check D&meter tools and measure without drilling the wellbore. Circulate back acid. - Flow rate: 210 gallons / minute, riser pressure: 2480 psi. - Close the blowout preventer on the last 200 barrels and circulate in a controlled manner through the choke manifold. A 4-11 / 16 inch retrieval rig was lowered into the open hole section to retrieve the rig to the top of the debris at 11,716 feet. - Flow rate: 50 gallons / minute, riser pressure: 320 psi. - Recorded lift-up weight: 284 kg, lower-down weight: 275 kg. Immersion time: - Allow acid reaction for 30 minutes before moving the tubing string. - Lower the tubing string with a weight of 50 kg and apply a torque of 9 kg ft-lb. After the 30-minute immersion time, drain 5 buckets from the drill bit to refresh the acid in the open hole section.

[0127] The aforementioned text can be sequential in its generation, and therefore can be associated with time, measured depth, etc. As an example, one or more of time and measured depth can be used to associate text and / or text strings with drilling rig field data, for example, to determine drilling rig status, action timing (e.g., start, stop, duration, etc.), action conditions, confirmation of one or more actions, control instructions, etc. As shown, various acronyms, technical terms, numbers, etc., can be used. As an example, the term "bucket" can be represented as "bucket" or "bbl" or one or more other forms. As an example, shorthand terms such as " / n" can appear, which can be specific to the operating team, organization, etc. In the above text, the term "acid" appears together with the term "HCl," which can be hydrochloric acid (HCl). As an example, fine-tuning can be performed to process text such as the example text described above using a fine-tuned LLM.

[0128] As an example, the aforementioned sample text is used as input to a fine-tuned LLM, which is able to extract specific remedies that lead to the resolution of the event:

[0129] Failure mode: Pipe jamming (differential pressure jamming)

[0130] Remedial measures:

[0131] 1. Switch to the cementing unit and pump 55 barrels of acid, 20% HCl.

[0132] 2. Replace the 20% HCl acid solution with a total of 118 barrels of 107 PCF mud.

[0133] 3. Place 20 barrels of HCl acid in the open hole section and keep 35 barrels of acid inside the tubing string.

[0134] 4. Circulate and adjust the mud. Mud weight in / out: 105 pcf.

[0135] 5. Check the D&meter tool and take measurements when drilling is not in progress.

[0136] Result: The stuck pipe was successfully removed and drilling operations resumed.

[0137] As indicated, a fine-tuned LLM can extract the event (stuck pipe), the cause of the stuck pipe or the classification of the stuck pipe (differential stuck pipe as a failure mode, etc.), a list of remedial actions (e.g., remedies), and the result of "drilling operation recovery" (the stuck pipe was successfully removed). Such a method can provide a link between the description and / or the action and time. In such an example, NPT and / or other times can be determined, which can be used, for example, for sequencing, etc. As an example, the aforementioned remedial actions can be compared with one or more SOPs, which may include one or more differential stuck pipe SOPs. In this method, the fine-tuned LLM can generate outputs that can provide an assessment of compliance with one or more SOPs and / or improvements to one or more SOPs (e.g., for a specific type of field operation, well type, formation type, drilling fluid type, etc.).

[0138] As an example, a system capable of extracting specific remedies that lead to the resolution of an event can also provide identification of control instructions that may have been implemented to execute one or more of the specific remedies. For example, consider accessing rig field data that may relate to certain equipment, specific control commands, etc. In such an example, timestamps can be used to note the potential time difference between the recording time of the extracted specific remedy (e.g., as in a report) and the issuance and / or implementation time of the control instructions (e.g., commands, notices, etc.). As an example, the system can provide isolation of instances of rig field data corresponding to specific remedies, such as those indicated in reports, etc. In such an example, the system can provide the ability to examine one or more extracted remedies against the rig field data to confirm their execution and / or their timing (e.g., start, stop, duration, etc.). In such an example, the rig field data can provide information on one or more rig states, conditions, etc., that may be closely related to the execution of one or more extracted remedies. In this approach, the system can add rig field data in a contextual manner, which can also provide the generation of instructions regarding the timing and / or sequence of measures (e.g., serial, parallel, if-then, etc.). As explained, this type of instruction can provide control over equipment to perform one or more types of field operations.

[0139] Such as about Figure 9 As explained by System 900, contextualized event descriptions and remedial descriptions (e.g., remedial descriptions) can be used to generate embeddings, which can allow users of the drilling reporting system to quickly find remedial measures that have resolved similar issues on one or more historical wells (e.g., neighboring wells, etc.). As explained, the system can provide associations and / or translations of information, for example, in human-readable form, machine-readable form (e.g., processor-executable form, etc.).

[0140] As an example, the framework can receive a description of an event or event risk and can generate output based on successful remedies for events and / or event risks in previous instances. This risk might be at a neighboring well that was drilled at least partially when the event or event risk occurred and was successfully remedied. As explained, the framework can generate output for more than a single successful remedy, which can be used for wells currently being drilled, wells already drilled, etc. As an example, the system can utilize one or more DDRs of a currently drilled well to improve the performance of a fine-tuned LLM, which can be applied to that well or one or more other wells currently being drilled.

[0141] Figure 10Examples of method 1000 and system 1090 are shown. As illustrated, method 1000 may include: a receiving block 1010 for receiving a description of an event occurring at the well site; an extraction block 1020 for extracting fault modes from the description using a fine-tuned large language model (LLM); an identification block 1030 for identifying matching fault modes from historical data processed using the fine-tuned LLM, wherein the matching fault modes are associated with one or more remedial actions that successfully resolve the matching fault modes; and an output block 1040 for outputting one or more remedial actions to be implemented at the well site. As an example, method 1000 may include an implementation block for automatically, semi-automatically, etc., implementing at least one of one or more remedial actions (e.g., remedies, etc.). As explained, the output may include control instructions, may be associated with control instructions, and may be adapted to be converted into control instructions, etc. As explained, the output may include safety protections, such as specifying one or more measures that should not be performed, one or more measures that may lead to unfavorable conditions, etc.

[0142] Figure 10 Various computer-readable medium (CRM) blocks 1011, 1021, 1031, and 1041 are also shown. Such blocks may include instructions executable by one or more processors, which may be one or more processors of a computing framework, system, computer, etc. A computer-readable medium may be a computer-readable storage medium that is not a signal, not a carrier wave, and is non-transitory. For example, a computer-readable medium may be a physical memory component capable of storing information in a digital format.

[0143] exist Figure 10 In this example, system 1090 includes one or more information storage devices 1091, one or more computers 1092, one or more networks 1095, and instructions 1096. Regarding the one or more computers 1092, each computer may include one or more processors (e.g., processing cores) 1093 and memory 1094 for storing, for example, instructions 1096 executable by at least one of the one or more processors. As an example, the computers may include one or more network interfaces (e.g., wired or wireless), one or more graphics cards, display interfaces (e.g., wired or wireless), etc. System 1090 may be specifically configured to execute... Figure 10 Method 1000 is one or more parts of it.

[0144] As an example, a computational framework may include a solver, which may be implemented via executable instructions. For instance, consider a computational framework including a processor and processor-accessible memory, where executable instructions may be stored in memory and accessed by the processor for execution, causing the computational framework to perform one or more actions. Such a computational framework may include one or more interfaces for receiving information and / or for outputting information, which may include parameter values, instructions, etc. As an example, a computational framework may be part of a controller. As an example, a computational framework may be part of a system.

[0145] As examples, various systems, methods, etc., can implement one or more ML models. Regarding the types of ML models, consider one or more of the following: Support Vector Machine (SVM) models, k-Nearest Neighbor (KNN) models, ensemble classifier models, neural network (NN) models, etc. As examples, machine learning models can be deep learning models (e.g., deep Boltzmann machines, deep belief networks, convolutional neural networks, stacked autoencoders, etc.), ensemble models (e.g., random forests, gradient boosting machines, bootstrap aggregation, AdaBoost, stacked generalization, gradient boosting regression trees, etc.), neural network models (e.g., radial basis function networks, perceptrons, backpropagation, Hopfield networks, etc.), regularization models (e.g., ridge regression, minimum absolute shrinkage and selection operators, elastic networks, minimum angular regression), rule-based system models (e.g., cubist, single-rule, zero-rule, repeated incremental pruning for error reduction), and regression models (e.g., linear regression, ordinary least squares regression, stepwise regression, multivariate adaptive regression splines, local estimation scatter plot smoothing, logistic regression, etc.). Bayesian models (e.g., Naive Bayes, Average Dependency Estimator, Bayesian Belief Network, Gaussian Naive Bayes, Multinomial Naive Bayes, Bayesian Networks), decision tree models (e.g., Classification and Regression Trees, Iterative Dichotomy 3, C4.5, C5.0, Chi-square Automatic Interaction Detection, Decision Stumps, Conditional Decision Trees, M5), dimensionality reduction models (e.g., Principal Component Analysis, Partial Least Squares Regression, Sammon Mapping, Multidimensional Scaling Analysis, Projective Pursuit, Principal Component Regression, Partial Least Squares Discriminant Analysis, Mixed Discriminant Analysis, Quadratic Discriminant Analysis, Regularized Discriminant Analysis, Flexible Discriminant Analysis, Linear Discriminant Analysis, etc.), instance models (e.g., k-Nearest Neighbors, Learning Vector Quantization, Self-Organizing Map, Locally Weighted Learning, etc.), clustering models (e.g., k-means, k-median, Expectation Maximization, Hierarchical Clustering, etc.), etc.

[0146] As an example, the system can utilize one or more recurrent neural networks (RNNs). One type of RNN is called Long Short-Term Memory (LSTM), which can be a unit or component that can be located in one or more layers (e.g., one or more units). An LSTM component can be an artificial neural network (ANN) designed to recognize patterns in a sequence of data (e.g., time-series data). When time-series data is provided, LSTMs consider both time and sequence, allowing them to include a temporal dimension. For example, consider using one or more RNNs to process temporal data from one or more sources, optionally combined with spatial data. This approach can identify temporal patterns, which can be used for prediction (e.g., one or more patterns about future times, etc.).

[0147] As an example, the TENSORFLOW framework (Google LLC, Mountain View, California) can be implemented. It is an open-source software library for dataflow programming, including a symbolic mathematics library, which can be implemented for machine learning applications that may include neural networks. As an example, the CAFFE framework can be implemented, a DL framework developed by Berkeley AI Research (BAIR) (University of California, Berkeley, California). As another example, consider utilizing the SCIKIT platform (e.g., scikit-learn) of the Python programming language. As an example, frameworks such as the APOLLO AI framework (APOLLO.AI GmbH, Germany) can be used. As mentioned above, frameworks such as the PYTORCH framework can also be used.

[0148] As an example, training methods can include various actions that can be performed on the dataset to train the ML model. As another example, the dataset can be split into training and testing data, where the testing data can be provided for evaluation. One approach could include cross-validation of parameters and optimal parameters, which can be provided for model training.

[0149] The TENSORFLOW framework can run on multiple CPUs and GPUs (with optional CUDA (NVIDIA Corp., Santa Clara, California) and SYCL (Khronos Group Inc., Beaverton, Oregon) extensions for general-purpose computing on graphics processing units (GPUs). TENSORFLOW is available on 64-bit Linux, macOS (Apple Inc., Cupertino, California), Windows (Microsoft Corp., Redmond, Washington), and mobile computing platforms including Android (Google LLC, Mountain View, California) and iOS (Apple Inc.) operating systems.

[0150] TENSORFLOW computation can be represented as a stateful data flow graph; note that the name TENSORFLOW derives from the operation performed by this neural network on a multidimensional data array. Such an array can be called a "tensor".

[0151] As an example, a method may include: receiving a description of an event occurring at the well site; extracting a failure mode from the description using a fine-tuned large language model (LLM); identifying matching failure modes from historical data processed using the fine-tuned LLM, wherein the matching failure mode is associated with one or more remedial actions that successfully resolved the matching failure mode; and outputting one or more remedial actions for implementation at the well site. In such an example, the event may be a drilling fluid event and / or another type of event. Regarding a drilling fluid event, it may be a circulation loss event. Regarding a stuck pipe event, it may be a differential pressure stuck pipe event and / or another type of stuck pipe event. As an example, a failure mode may be an event and / or a cause of an event, wherein such an event and / or cause exists in historical data, for example, along with one or more remedial actions and indications of the success and / or failure of one or more of the one or more remedial actions that resolved the failure mode.

[0152] As an example, the approach could include generating a fine-tuned LLM. In such an example, generating a fine-tuned LLM could include utilizing a set of specialized questions and answers. For example, consider a set of specialized questions and answers that might include field operation terminology and definitions of field operation terminology. As an example, a brand (e.g., a trademark) could be a term that can be associated with one or more specific items (such as, for example, chemicals). In such an example, one or more other brands could be possible alternatives. As an example, a fine-tuned LLM could provide differentiation between brands, chemicals, etc. (e.g., within the context of the description).

[0153] As an example, a method may include extracting fault modes at least in part by generating vectors. In such an example, the method may include identifying matching fault modes at least in part by comparing vectors with existing vectors, where, for example, the existing vectors are generated using a fine-tuned LLM. For example, the fine-tuned LLM may extract information from a current description relating to a problem (e.g., an event) occurring at the well site, and from historical descriptions, where vector-based methods can be utilized during the matching process. As an example, existing vectors may be generated at least in part by applying a sentence transformer to the output of the fine-tuned LLM, and existing vectors may be further generated, for example, using labeled embeddings based on the output of the sentence transformer.

[0154] As an example, a fine-tuned LLM may include at least a portion of a generative pre-trained transformer (GPT) architecture. As another example, the GPT architecture may be modified for one or more purposes.

[0155] As an example, historical data may include daily drilling reports (DDR).

[0156] As an example, one approach could include identifying multiple instances in historical data that match a failure pattern. In such an example, the approach could include sorting the multiple instances according to one or more criteria, and / or sorting the multiple instances based on the proximity of the matches.

[0157] As an example, the matching failure mode can be an exact match or a closest match. As explained, a description entered by one operations team may differ from one entered by another operations team for one or more reasons. These reasons could be due to organizational differences, personnel, materials, brands, equipment, formation, location, regulations, etc.

[0158] As explained, humans may face challenges in finding matches between current and past problems. As explained, a drilling incident remediation framework can find one or more matches within a timeframe that can accelerate field operations (e.g., reduce NPT, etc.). For example, consider real-time or near-real-time timeframes where drilling reporting systems can continuously evaluate event descriptions, etc., and where a matching process (e.g., as a background process) can be implemented to continuously generate matches. In such an approach, the framework may have one or more remediations ready for recommendation and / or implementation (e.g., automatically, semi-automatically, etc.) when an event (e.g., a problem) occurs or its risk is sufficiently high.

[0159] As an example, equipment at the well site can be operated using one or more levels of automation. As another example, a drilling incident recovery framework can report on the historical actions taken for recovery and / or the possible levels of automation achieved for the actions taken for recovery, noting that some types of actions can be automated while others may require some degree of operator involvement (e.g., acquiring materials).

[0160] As an example, the system may include one or more processors; at least one of the processors has processor-accessible memory; processor-executable instructions stored in the memory and executable to instruct the system to: receive a description of an event occurring at the well site; extract a failure mode from the description using a fine-tuned large language model (LLM); identify matching failure modes from historical data processed using the fine-tuned LLM, wherein the matching failure mode is associated with one or more remedies that successfully resolve the matching failure mode; and output one or more remedies for implementation at the well site.

[0161] As an example, one or more computer-readable storage media may include processor-executable instructions to instruct a computing system to: receive a description of an event occurring at a well site; extract a failure mode from the description using a fine-tuned large language model (LLM); identify a matching failure mode from historical data processed using the fine-tuned LLM, wherein the matching failure mode is associated with one or more remedies that successfully resolve the matching failure mode; and output the one or more remedies for implementation at the well site.

[0162] As an example, a computer program product may include computer-executable instructions to instruct a computing system to perform one or more methods, such as one or more of the methods described herein (e.g., partially, wholly, and / or in various combinations).

[0163] In some embodiments, one or more methods may be performed by a computing system. Figure 11An example of system 1100 is shown, which may include one or more computing systems 1101-1, 1101-2, 1101-3, and 1101-4. These computing systems may be operatively coupled via one or more networks 1109, which may include wired and / or wireless networks. As shown, system 1100 may include one or more other components 1108.

[0164] As an example, the system may include a standalone computer system or an arrangement of distributed computer systems. Figure 11 In the example, computer system 1101-1 may include one or more modules 1102, which may be or include processor-executable instructions, for example, executable to perform various tasks (e.g., receiving information, requesting information, processing information, simulating, outputting information, etc.).

[0165] As an example, the module can be executed independently or in coordination with one or more processors 1104 operatively coupled to one or more storage media 1106 (e.g., via wired, wireless, etc.). As an example, one or more of the processors 1104 may be operatively coupled to at least one of one or more network interfaces 1107. In such an example, the computer system 1101-1 can send and / or receive information, for example, via one or more networks 1109 (e.g., consider one or more of the Internet, private networks, cellular networks, satellite networks, etc.). As shown, one or more other components 1108 may be included in the computer system 1101-1.

[0166] As an example, computer system 1101-1 may receive information from and / or send information to one or more other devices, which may be or include one or more of, for example, computer system 1101-2. The devices may be located in a physical location different from the physical location of computer system 1101-1. As an example, the location may be, for example, a processing facility location, a data center location (e.g., a server cluster), a drilling rig location, a well site location, a downhole location, etc.

[0167] As an example, a processor may be or include a microprocessor, a microcontroller, a processor module or subsystem, a programmable integrated circuit, a programmable gate array, or another control or computing device.

[0168] As an example, storage medium 1106 may be implemented as one or more computer-readable or machine-readable storage media. As an example, storage may be distributed within multiple internal and / or external enclosures of the computing system and / or additional computing systems, and / or distributed across multiple internal and / or external enclosures of the computing system and / or additional computing systems.

[0169] As an example, one or more storage media may include one or more different forms of memory, including semiconductor memory devices (such as dynamic or static random access memory (DRAM or SRAM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM) and flash memory), magnetic disks (such as fixed disks, floppy disks and removable disks), other magnetic media (including magnetic tape), optical media (such as optical discs (CDs) or digital video discs (DVDs), Blu-ray discs), or other types of optical memory, or other types of storage devices.

[0170] As an example, one or more storage media may be located in a machine that runs machine-readable instructions, or at a remote site from which machine-readable instructions can be downloaded over a network for execution.

[0171] As an example, various components of a system, such as a computer system, can be implemented in hardware, software, or a combination of both (e.g., including firmware), including one or more signal processing and / or application-specific integrated circuits.

[0172] As an example, the system may include a processing device, which may be or include a general-purpose processor or a special-purpose chip (e.g., a chipset), such as an ASIC, FPGA, PLD, or other suitable device.

[0173] As an example, the device may be a mobile device including one or more network interfaces for information communication. For example, the mobile device may include a wireless network interface (e.g., operating via IEEE 802.11, ETSI GSM, Bluetooth, satellite, etc.). As an example, the mobile device may include components such as a main processor, memory, display, display graphics circuitry (e.g., optionally including touch and gesture circuitry), SIM slot, audio / video circuitry, motion processing circuitry (e.g., accelerometer, gyroscope), wireless LAN circuitry, smart card circuitry, transmitter circuitry, GPS circuitry, and a battery. As an example, the mobile device may be configured as a cellular phone, tablet computer, etc. As an example, a method may be implemented using a mobile device (e.g., wholly or partially). As an example, a system may include one or more mobile devices.

[0174] As an example, the system can be a distributed environment, such as a so-called "cloud" environment, in which various devices, components, etc., interact for purposes such as data storage, communication, and computing. As an example, a device or system may include one or more components for transmitting information via one or more of the Internet (e.g., where communication occurs via one or more Internet protocols), cellular networks, satellite networks, etc. As an example, a method can be implemented in a distributed environment (e.g., wholly or partially as a cloud-based service).

[0175] As an example, information can be input from a display (e.g., consider a touchscreen), output to a display, or both. As an example, information can be output to a projector, laser device, printer, etc., making the information viewable. As an example, information can be output stereoscopically or holographically. Regarding printers, consider 2D or 3D printers. As an example, a 3D printer can include one or more materials that can be output to construct 3D objects. For example, data can be provided to a 3D printer to construct a 3D representation of underground strata. As an example, layers (e.g., stratigraphic layers), geological bodies, etc., can be constructed in 3D. As an example, holes, cracks, etc. (e.g., as positive structures, as negative structures, etc.) can be constructed in 3D.

[0176] Although only a few examples have been described in detail above, those skilled in the art will readily understand that many modifications can be made to the examples. Therefore, all such modifications are intended to be included within the scope of this disclosure as defined by the following claims. In the claims, the device plus function clause is intended to cover structures described herein as performing said functions, and includes not only structural equivalents but also equivalent structures. Thus, although nails and screws may not be structural equivalents because nails use a cylindrical surface to hold wooden parts together while screws use a helical surface, in the context of fastening wooden parts, nails and screws can be equivalent structures.

Claims

1. A method (1000) comprising: Receive a description of the events that occurred at the well site (1010). Fault modes were extracted from the description using a finely tuned large language model (LLM) (1020). Identify matching failure patterns from historical data processed using the fine-tuned LLM, wherein the matching failure patterns are associated with one or more remedies that successfully resolve the matching failure patterns (1030). and Output one or more of the remedial measures for implementation at the well site (1040).

2. The method of claim 1, wherein the event includes a drilling fluid event, and optionally, the drilling fluid event includes a circulation loss event.

3. The method according to claim 1 or 2, wherein the event includes a stuck drill event, and optionally, the stuck drill event includes a differential stuck drill event.

4. The method of any one of the preceding claims, comprising generating the fine-tuned LLM, optionally wherein, Generating the fine-tuned LLM involves using a set of specialized questions and answers, and optionally, the set of specialized questions and answers includes field operation terminology and definitions of the field operation terminology.

5. The method according to any one of the preceding claims, wherein, Extracting the fault mode includes generating a vector.

6. The method according to any one of the preceding claims, wherein, Identifying the matching failure mode involves comparing the vector with existing vectors.

7. The method of any of the preceding claims, wherein, The existing vectors were generated using the fine-tuned LLM.

8. The method of any of the preceding claims, wherein, The existing vector is generated by applying a sentence transformer to the output of the fine-tuned LLM, and is further generated using labeled embeddings based on the output of the sentence transformer.

9. The method of any of the preceding claims, wherein, The fine-tuned LLM includes at least a portion of the Generative Pre-trained Transformer (GPT) architecture.

10. The method of any of the preceding claims, wherein, The historical data includes daily drilling reports (DDR).

11. The method according to any one of the preceding claims, comprising identifying multiple instances of the matching failure mode in the historical data.

12. The method according to any one of the preceding claims, comprising sorting the plurality of instances according to one or more criteria and / or based on the proximity of the matches.

13. The method of any of the preceding claims, wherein, The matching failure modes include exact match or closest match.

14. A system (1090) comprising: One or more processors (1093); A memory (1094) that can be accessed by at least one of the one or more processors; The processor can execute instructions (1096) stored in the memory and executable to instruct the system: Receive a description of the events that occurred at the well site (1011). Fault modes were extracted from the description using a finely tuned large language model (LLM) (1021). Identify matching failure patterns from historical data processed using the fine-tuned LLM, wherein the matching failure patterns are associated with one or more remedies that successfully resolve the matching failure patterns (1031). and Output one or more of the remedial measures for implementation at the well site (1041).

15. A computer program product comprising computer-executable instructions for instructing a computing system to perform the method according to any one of claims 1 to 13.

Images