Earthquake data interpretation system

Abstract

A method may include: accessing a trained machine model that is trained to analyze digital seismic data of a region relative to structural features of the geological region; analyzing at least a portion of the digital seismic data using the trained machine model to generate a result; and outputting the result as an indicator of a spatial location of the structural feature of the geological region.

Description

[0001] Related Applications

[0002] This application claims the benefit of and priority to U.S. Provisional Application Serial No. 62 / 683,011, filed on June 10, 2018, which is incorporated herein by reference. Background Art

[0003] In oil and gas exploration, interpretation is a process involving analyzing data to identify and locate various subsurface structures (e.g., horizons, faults, geological bodies, etc.) in a geological environment. Various types of structures (e.g., stratigraphic groups) may indicate hydrocarbon traps or flow channels that may be associated with one or more reservoirs (e.g., fluid reservoirs). In the field of resource extraction, enhancements to interpretation may allow for the construction of more accurate models of subsurface regions, which in turn may improve the characterization of subsurface regions for the purposes of resource extraction. The characterization of one or more subsurface regions in a geological environment may guide, for example, the execution of one or more operations (e.g., oilfield operations, etc.). As an example, a more accurate model of a subsurface region may make the trajectory of a drilling operation more accurate with respect to a wellbore, where the wellbore will have a trajectory that penetrates a reservoir, etc. Summary of the invention

[0004] A method may include: accessing a trained machine model, the trained machine model being trained to analyze digital seismic data of a region relative to structural features of a geological region; analyzing at least a portion of the digital seismic data using the trained machine model to generate a result; and outputting the result as an indicator of a spatial location of a structural feature of the geological region. A system may include: a processor; a memory, the memory being operatively coupled to the processor; and processor-executable instructions stored in the memory to instruct the system to: access a trained machine model, the trained machine model being trained to analyze digital seismic data of a region relative to structural features of a geological region; analyzing at least a portion of the digital seismic data using the trained machine model to generate a result; and outputting the result as an indicator of a spatial location of a structural feature of a geological region. One or more computer-readable storage media may include computer-executable instructions executable to instruct a computing system to: access a trained machine model, the trained machine model being trained to analyze digital seismic data of a region relative to structural features of a geological region; analyzing at least a portion of the digital seismic data using the trained machine model to generate a result; and outputting the result as an indicator of a spatial location of a structural feature of a geological region. Various other devices, systems, methods, etc. are also disclosed.

[0005] This Summary is provided to introduce a selection of concepts that are further described below in the Detailed Description. This Summary is not intended to identify key or essential features of the claimed subject matter, nor is it intended to be used to help limit the scope of the claimed subject matter. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] The features and advantages of the described implementations may be more readily understood by reference to the following description taken in conjunction with the accompanying drawings.

[0007] Figure 1 An exemplary system including various components for modeling a geological environment and various equipment associated with the geological environment is shown;

[0008] Figure 2 Examples of sedimentary basins, examples of methods, examples of formations, examples of wellbores, examples of wellbore tools, examples of conventions, and examples of systems are shown;

[0009] Figure 3 Examples of techniques by which data can be collected are shown;

[0010] Figure 4 Examples of equipment including an example of a downhole tool and an example of a borehole are shown;

[0011] Figure 5 An example of equipment including an example of a downhole tool is shown;

[0012] Figure 6 An example of forward modeling and inversion of an earth model with respect to seismic data and acoustic impedance is shown;

[0013] Figure 7 An example of a computational framework is shown;

[0014] Figure 8 An example of a method is shown;

[0015] Fig. 9 An exemplary graphical user interface is shown;

[0016] Fig.10 An example of a method is shown;

[0017] Fig.11 An example of a method is shown;

[0018] Fig.12 An example of a graphical user interface is shown;

[0019] Fig.13 An example of a graphical user interface is shown;

[0020] Fig.14 An example of a graphical user interface is shown;

[0021] Fig.15 An example of a graphical user interface is shown;

[0022] Fig.16 An example of a graphical user interface is shown;

[0023] Fig.17 An example of a graphical user interface is shown;

[0024] Fig.18 An example of a graphical user interface is shown;

[0025] Fig.19 An example of a graphical user interface is shown;

[0026] Fig. 20 An example of a graphical user interface is shown;

[0027] Fig.21 An example of a method is shown;

[0028] Fig. 22 An example of a graphical user interface is shown;

[0029] Fig.23 An example of a method is shown;

[0030] Fig.24 An example of a method is shown;

[0031] Fig.25 An example of a method is shown;

[0032] Fig.26 An example of a method is shown;

[0033] Fig. 27 An example of a method is shown;

[0034] Fig.28 Examples of methods and examples of neurons are shown;

[0035] Fig.29 An example of a method and an example of a graphical user interface are shown;

[0036] Fig.30 An example of a method and an example of a graphical user interface are shown;

[0037] Fig.31 An example of a method is shown;

[0038] Fig.32 An example of a method is shown;

[0039] Fig.33 An example of the method is shown as pseudo code;

[0040] Fig.34 An example of a method is shown; and

[0041] Fig.35 Exemplary components of systems and networked systems are shown. DETAILED DESCRIPTION

[0042] This description is not to be taken in a limiting sense but is merely intended to illustrate the general principles of implementations. Reference should be made to the issued claims to determine the scope of the described implementations.

[0043] Figure 1 An example of a system 100 is shown that includes various management components 110 to manage various aspects of a geological environment 150 (e.g., an environment including a sedimentary basin, a reservoir 151, one or more faults 153-1, one or more geological bodies 153-2, etc.). For example, the management components 110 may allow for direct or indirect management of sensing, drilling, injection, extraction, etc. with respect to the geological environment 150. In turn, further information about the geological environment 150 may be made available as feedback 160 (e.g., optionally as input to one or more of the management components 110).

[0044] exist Figure 1 In the example of FIG. 1 , management component 110 includes seismic data component 112, additional information component 114 (e.g., well / log data), processing component 116, simulation component 120, property component 130, analysis / visualization component 142, and workflow component 144. In operation, seismic data and other information provided by components 112 and 114 may be input to simulation component 120.

[0045] In an exemplary embodiment, simulation component 120 may rely on entity 122. Entity 122 may include earth entities or geological objects, such as wells, ground, bodies, reservoirs, etc. In system 100, entity 122 may include a virtual representation of an actual physical entity reconstructed for simulation purposes. Entity 122 may include an entity based on data collected via sensing, observation, etc. (e.g., seismic data 112 and other information 114). Entities may be characterized by one or more properties (e.g., a geometric pillar grid entity of an earth model may be characterized by a porosity property). These properties may represent one or more measurements (e.g., collected data), calculations, etc.

[0046] In an exemplary embodiment, simulation component 120 can operate in conjunction with a software framework, such as an object-based framework. In such a framework, entities can include entities based on predefined classes to facilitate modeling and simulation. An example of an object-based framework is the MICROSOFT.NET framework (Redmond, Washington), which provides a set of extensible object classes. In the .NET framework, object classes encapsulate modules of reusable code and associated data structures. Object classes can be used to instantiate object instances for use by programs, scripts, etc. For example, a wellbore class can define objects for representing a wellbore based on well data.

[0047] exist Figure 1 In the example of , the simulation component 120 can process the information to conform to one or more attributes specified by the attribute component 130, which can include an attribute library. Such processing can occur prior to input to the simulation component 120 (e.g., consider the processing component 116). As an example, the simulation component 120 can perform operations on the input information based on one or more attributes specified by the attribute component 130. In an exemplary embodiment, the simulation component 120 can construct one or more models of the geological environment 150, which can be relied upon to simulate the behavior of the geological environment 150 (e.g., in response to one or more actions, whether natural or artificial). Figure 1 In the example of , the analysis / visualization component 142 can allow interaction with the model or model-based results (e.g., simulation results, etc.). As an example, the output from the simulation component 120 can be input to one or more other workflows, as indicated by the workflow component 144.

[0048] As an example, the simulation component 120 may include one or more features of a simulator, such as the ECLIPSE Reservoir Simulator (Schlumberger Limited, Houston, Texas), the INTERSECT Reservoir Simulator (Schlumberger Limited, Houston, Texas), etc. As an example, the simulation component, simulator, etc. may include features that implement one or more gridless techniques (e.g., solve one or more equations, etc.). As an example, one or more reservoirs may be simulated with respect to one or more enhanced oil recovery techniques (e.g., considering thermal processes such as SAGD).

[0049] In an exemplary embodiment, the management component 110 may include features of a framework such as the PETREL seismic to simulation software framework (Schlumberger, Inc., Houston, Texas). The PETREL framework provides components that can optimize exploration and development operations. The PETREL framework includes a seismic to simulation software component that can output information for use in improving reservoir performance, such as 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. Such a framework can be considered an application and can be considered a data-driven application (e.g., where data is input for the purpose of modeling, simulation, etc.).

[0050] In an exemplary embodiment, various aspects of the management component 110 may include add-ons or plug-ins that operate according to the specifications of the framework environment. For example, a framework environment with a trade name of OCEAN Framework Environment (Schlumberger, Inc., Houston, Texas) allows the integration of add-ons (or plug-ins) into the PETREL framework workflow. The OCEAN Framework Environment utilizes .NET tools (Microsoft Corporation, Redmond, Washington) and provides a stable user-friendly interface for efficient development. In an exemplary embodiment, various components can be implemented as add-ons (or plug-ins) that comply with the specifications of the framework environment and operate according to the specifications of the framework environment (e.g., according to application programming interface (API) specifications, etc.).

[0051] As an example, the framework may be implemented within or operatively coupled to the DELFI cognitive exploration and production (E&P) environment (Schlumberger Ltd., Houston, Texas), which is a secure, cognitive, cloud-based collaborative environment that integrates data and workflows with digital technologies such as artificial intelligence and machine learning. As an example, such an environment may provide operations involving one or more computing frameworks. For example, various types of computing frameworks may be utilized within the environment, such as a drilling planning framework, a seismic to simulation framework (e.g., PETREL framework, Schlumberger Inc. of Houston, Texas), a measurement framework (e.g., TECHLOG framework, Schlumberger Inc. of Houston, Texas), a mechanical earth modeling (MEM) framework (PETROMOD framework, Schlumberger Inc. of Houston, Texas), an exploration risk, resource, and value assessment framework (e.g., GEOX, Schlumberger Inc. of Houston, Texas), a reservoir simulation framework (INTERSECT, Schlumberger Inc. of Houston, Texas), a surface facility framework (e.g., PIPESIM, Schlumberger Inc. of Houston, Texas), a production enhancement framework (MANGROVE framework, Schlumberger Inc. of Houston, Texas). As an example, one or more methods may be implemented at least in part via a framework (e.g., a computing framework) and / or an environment (e.g., a computing environment).

[0052] Figure 1 Also shown is an example of a framework 170, which includes a model simulation layer 180 and a framework service layer 190, a framework core layer 195, and a module layer 175. The framework 170 can include an OCEAN framework, wherein the model simulation layer 180 is a PETREL model-centric software package that hosts an OCEAN framework application. In an exemplary embodiment, the PETREL software can be considered a data-driven application. The PETREL software can include a framework for model building and visualization. As mentioned, the PETREL framework can be implemented in conjunction with the DELFI environment.

[0053] As an example, seismic data can be processed using a framework such as the OMEGA framework (Schlumberger, Inc., Houston, Texas). The OMEGA framework provides features that can be implemented for processing seismic data, for example, through prestack seismic interpretation and seismic inversion. The framework can be scalable so that it can be processed and imaged on a single workstation, on a large computing cluster, etc. As an example, one or more techniques, technologies, etc. described herein can be optionally implemented in conjunction with a framework (e.g., the OMEGA framework).

[0054] The framework for processing data may include features for 2D line and 3D seismic surveys. The module for processing seismic data may include features for prestack seismic interpretation (PSI), which may optionally be inserted into a framework (such as an OCEAN framework). A workflow may be specified to include processing via one or more frameworks, plug-ins, add-ons, etc. A workflow may include quantitative interpretation, which may include performing prestack and poststack seismic data conditioning, inversion (e.g., seismic to properties and properties to synthetic seismic), wedge modeling for thin layer analysis, amplitude to offset (AVO) and amplitude to angle (AVA) analysis, surveys, etc. As an example, a workflow may be intended to output rock properties based at least in part on the processing of seismic data. As an example, various types of data may be processed to provide one or more models (e.g., earth models). For example, consider the processing of one or more of seismic data, well data, electromagnetic and geomagnetic data, reservoir data, etc.

[0055] As an example, the framework may include features for implementing one or more grid generation techniques. For example, the framework may include an input component for receiving information from an interpretation of seismic data, one or more attributes based at least in part on the seismic data, well logging data, image data, etc. Such a framework may include a grid generation component that processes the input information (optionally in conjunction with other information) to generate a grid.

[0056] exist Figure 1 In the example of , model simulation layer 180 can provide domain objects 182, act as a data source 184, provide presentation 186, and provide various user interfaces 188. Presentation 186 can provide a graphical environment in which an application can display its data, while user interface 188 can provide a common look and feel for application user interface components.

[0057] As an example, domain objects 182 may include entity objects, property objects, and optionally other objects. Entity objects may be used to geometrically represent wells, ground, bodies, reservoirs, etc., while property objects may be used to provide property values ​​as well as data versions and display parameters. For example, an entity object may represent a well, where the property object provides well logging information as well as version information and display information (e.g., displaying the well as part of a model).

[0058] exist Figure 1In the example of , data can be stored in one or more data sources (or data stores, typically physical data storage devices), which can be located at the same or different physical sites and can be accessed via one or more networks. The model simulation layer 180 can be configured to model a project. In this way, a specific project can be stored, where the stored project information may include inputs, models, results, and cases. Therefore, when a modeling session is completed, the user can store the project. Later, the project can be accessed and restored using the model simulation layer 180, which can recreate instances of related domain objects.

[0059] exist Figure 1 In the example of, the geological environment 150 may include multiple layers (e.g., strata) including a reservoir 151 and one or more other features, such as a fault 153-1, a geological body 153-2, etc. As an example, the geological environment 150 may be equipped with any of a variety of sensors, detectors, actuators, etc. For example, equipment 152 may include communication circuits to receive and send information relative to one or more networks 155. Such information may include information associated with downhole equipment 154, which may be equipment used to collect information, assist in resource recovery, etc. Other equipment 156 may be located remote from the well site and include sensing, detection, transmission or other circuits. Such equipment may include storage and communication circuits to store and transmit data, instructions, etc. As an example, one or more satellites may be provided for purposes of communication, data collection, etc. For example, Figure 1 A satellite is shown in communication with a network 155 that may be configured for communications, noting that the satellite may additionally or alternatively include circuitry for imaging (eg, spatial, spectral, temporal, radiometric, etc.).

[0060] Figure 1 Also shown is a geological environment 150, which optionally includes equipment 157 and 158 associated with a well, the well including a substantially horizontal portion that may intersect one or more fractures 159. For example, consider a well in a shale formation that may include natural fractures, artificial fractures (e.g., hydraulic fractures), or a combination of natural fractures and artificial fractures. As an example, a well may be drilled into a laterally extending reservoir. In such examples, there may be lateral changes in properties, stresses, etc., where an assessment of such changes may help plan, operate, etc. to develop a laterally extending reservoir (e.g., via fracturing, injection, extraction, etc.). As an example, equipment 157 and / or 158 may include components, one or more systems, etc., for fracturing, seismic sensing, seismic data analysis, assessment of one or more fractures, etc.

[0061] As mentioned, the system 100 can be used to execute one or more workflows. A workflow can be a process that includes several work steps. The work steps can operate on data, for example, create new data, update existing data, etc. As an example, a workflow can operate on one or more inputs and produce one or more results, for example, based on one or more algorithms. As an example, the system can include a workflow editor for the creation, editing, execution, etc. of a workflow. In such examples, the workflow editor can provide a selection of one or more predefined work steps, one or more custom work steps, etc. As an example, a workflow can be a workflow that can be implemented in PETREL software, for example, the workflow operates on seismic data, one or more seismic attributes, etc. As an example, a workflow can be a process that can be implemented in an OCEAN framework. As an example, a workflow can include one or more work steps that access a module such as a plug-in (e.g., an external executable code, etc.).

[0062] Figure 2 An example of a sedimentary basin 210 (e.g., a geological environment), an example of a method 220 for model building (e.g., for a simulator, etc.), an example of a formation 230, an example of a wellbore 235 in the formation, an example of a convention 240, and an example of a system 250 are shown.

[0063] As an example, reservoir simulation, petroleum system modeling, etc. can be applied to characterize various types of subsurface environments, including Figure 1 One or more operations may be performed in an environment based at least in part on such representations of one or more subsurface environments (e.g., via acquired data, simulations, modeling, etc.).

[0064] exist Figure 2 In the present invention, a sedimentary basin 210 as a geological environment includes horizons, faults, one or more geological bodies and phases formed during a certain geological period. These features are distributed in two or three dimensions in space, for example, relative to a Cartesian coordinate system (e.g., x, y and z) or other coordinate systems (e.g., cylindrical, spherical, etc.). As shown, a model building method 220 includes a data acquisition block 224 and a model geometry block 228. Some data may be involved in building an initial model, and thereafter, the model may be optionally updated in response to model output, time changes, physical phenomena, additional data, etc. As an example, the data used for modeling may include one or more of the following: depth or thickness maps and fault geometry and timing from seismic, remote sensing, electromagnetic, gravity, outcrop and well logging data. In addition, the data may include depth and thickness maps derived from phase changes (e.g., due to seismic unconformity) assumed to follow a geological event ("iso" time), and the data may include lateral phase changes (e.g., due to lateral changes in sedimentation characteristics).

[0065] To continue modeling geological processes, data can be provided, for example, data such as geochemical data (e.g., temperature, kerogen type, organic abundance, etc.), time series data (e.g., from paleontology, radioactive dating, magnetic reversals, rock and fluid properties, etc.), and boundary condition data (e.g., heat flow history, ground temperature, paleowater depth, etc.).

[0066] In basin and petroleum system modeling, quantities such as temperature, pressure, and porosity distribution within sediments can be modeled, for example, by solving partial differential equations (PDEs) using one or more numerical techniques. Modeling can also model geometry with respect to time, for example, to account for changes resulting from geological events (e.g., deposition of material, erosion of material, displacement of material, etc.).

[0067] The modeling framework of trade name PETROMOD framework (Schlumberger Co., Ltd., Houston, Texas) includes the feature for inputting various types of information (e.g., earthquake, well, geology, etc.) to model the evolution of sedimentary basins. PETROMOD framework provides oil and gas system modeling by inputting various data (such as seismic data, well data and other geological data) for modeling the evolution of sedimentary basins, for example. PETROMOD framework can predict whether and how the reservoir is full of hydrocarbons, including the source and time of hydrocarbon generation, migration route, quantity, pore pressure and underground or surface conditions of hydrocarbon types. In combination with frameworks such as PETREL framework, workflows can be constructed to provide basin-to-prospect scale exploration solutions. Data exchange between frameworks can promote model building, data analysis (e.g., PETROMOD framework data analyzed using PETREL framework functions) and coupling of workflows.

[0068] like Figure 2 As shown, the formation 230 includes horizontal ground and various underground layers. As an example, the wellbore can be vertical. As another example, the wellbore can be deviated. Figure 2 In the example of , borehole 235 can be considered a vertical borehole, for example, where the z-axis extends downward perpendicular to the horizontal ground surface of formation 230. As an example, tool 237 can be positioned in the borehole, for example, to collect information. As mentioned, the borehole tool can be configured to collect electronic borehole images. For example, a full borehole formation microimager (FMI) tool (Schlumberger, Inc., Houston, Texas) can collect borehole image data. A data acquisition sequence for such a tool may include lowering the tool into the borehole with the acquisition pad closed, opening the pad and pressing the pad against the borehole wall, delivering current to the material defining the borehole while translating the tool in the borehole, and remotely sensing the current that is changed by interaction with the material.

[0069] As an example, the wellbore can be vertical, deviated and / or horizontal. As an example, the tool can be positioned to collect information in the horizontal portion of the wellbore. Analysis of such information can reveal caves, dissolution planes (e.g., dissolution along bedding planes), stress-related features, tilt events, etc. As an example, the tool can collect information that may help characterize fractured reservoirs, optionally where the fractures can be natural and / or artificial (e.g., hydraulic fractures). Such information can help with well completion, production stimulation treatments, etc. As an example, a framework such as the TECHLOG framework (Schlumberger, Inc., Houston, Texas) can be used to analyze the information collected by the tool.

[0070] As shown in the figure, the three-dimensional orientation of a plane can be defined by its inclination and direction. The inclination is the angle of inclination of a plane relative to a horizontal plane (e.g., an imaginary plane) measured in a vertical plane along a specific direction. The inclination can be defined by a magnitude (e.g., also called an angle or amount) and an azimuth (e.g., also called a direction). Figure 2 As shown in convention 240, the various angles Φ indicate, for example, angles of inclination downward from an imaginary horizontal plane (e.g., a flat upper surface); however, the inclination refers to the direction in which the inclination plane is inclined (e.g., which can be given in degrees, compass directions, etc.). Figure 2 Another feature shown in the convention of is strike, which is the orientation of the line resulting from the intersection of the dip plane and the horizontal plane (eg, treating the flat upper surface as an imaginary horizontal plane).

[0071] Certain additional terms related to dip and strike may be applied to the analysis, e.g., depending on the circumstances, the orientation of the data collected, etc. One term is "true dip" (see, e.g., Figure 2 Dip in the agreement 240 T True dip is the dip of a plane measured directly perpendicular to strike (e.g., see the direction north and labeled "strike" and the angle α 90 Another term is "apparent dip" (see, for example, Figure 2 Dip in the agreement 240 A The apparent tilt angle may be the tilt angle of a plane measured in any direction other than the direction of the true tilt angle (see, for example, Φ A , as the Dip for angle α A ); however, the apparent dip may be equal to the true dip (see, for example, Φ, as the angle α relative to the strike 90 Dip A =Dip T). In other words, where the term apparent inclination is used (e.g., in a method, analysis, algorithm, etc.), for a specific inclination plane, the value of the "apparent inclination" may be equal to the true inclination of the specific inclination plane.

[0072] like Figure 2 As shown in the convention 240 in FIG. 24, the dip of a plane as seen in a cross section perpendicular to the strike is the true dip (see, for example, the angle α with Φ as the angle relative to the strike). 90 Dip A =Dip T As indicated, the dip angle observed in the cross section in any other direction is the apparent dip angle (e.g., see the A In addition, Figure 2 As shown in convention 240, the apparent dip angle can be approximately 0 degrees (e.g., parallel to a horizontal surface, with the edge of the cutting plane traveling along the strike direction).

[0073] In terms of observing inclination in a wellbore, true inclination is observed in a well drilled vertically. In a well drilled at any other orientation (or deviation), the observed inclination is an apparent inclination (e.g., which is sometimes referred to as relative inclination). To determine the true inclination value of a plane observed in such a wellbore, as an example, a vector calculation (e.g., based on the wellbore deviation) may be applied to one or more apparent inclination values.

[0074] As mentioned, another term used in sedimentological interpretation from borehole images is “relative dip” (e.g., Dip R ). The value of the true dip measured from a borehole image in a rock deposited in a very quiet environment can be subtracted from the dip in the sandstone volume (e.g., using vector subtraction). In such examples, the resulting dip is called the relative dip and is used in interpreting the orientation of the sandstone volume.

[0075] As opposed to analysis, explanation, property, etc. (see, for example, Figure 1 Conventions such as convention 240 may be used by various blocks of system 100 of the present invention. As an example, various types of features (e.g., sedimentary bedding, faults and fractures, single-sided mountains, igneous rock walls and sills, metamorphic foliations, etc.) may be described in part by dip angles. As an example, dip angles may vary spatially as layers approach a geological body. For example, consider a salt body that may rise due to various forces (e.g., buoyancy, etc.). In such an example, as the salt body moves upward, the dip angle may tend to be upward.

[0076] The data-based interpretation may be directed to identifying and / or classifying one or more subsurface boundaries based at least in part on one or more dip parameters (e.g., angle or magnitude, azimuth, etc.). As an example, various types of features (e.g., sedimentary bedding, faults and fractures, single-sided mountains, igneous dykes and sills, metamorphic foliations, etc.) may be described at least in part by angle, at least in part by azimuth, etc.

[0077] As an example, equations for oil displacement and migration can be provided, which can be modeled and simulated, for example, relative to a period of time. Oil migration (e.g., primary migration or displacement) from source materials can include the use of a saturation model, wherein the migration saturation value controls the displacement. Determination of the secondary migration of oil (e.g., oil or gas) can include the use of the hydrodynamic potential energy of the fluid and the driving force that accounts for the promotion of fluid flow. Such forces can include buoyancy gradients, pore pressure gradients, and capillary pressure gradients.

[0078] like Figure 2 As shown, system 250 includes one or more information storage devices 252, one or more computers 254, one or more networks 260, and one or more sets of instructions 270. With respect to one or more computers 254, each computer may include one or more processors (e.g., or processing cores) 256 and a memory 258 for storing instructions (e.g., one or more of the one or more sets of instructions 270), which instructions may be executed, for example, by at least one of the one or more processors 256. As an example, a computer may include one or more network interfaces (e.g., wired or wireless), one or more graphics cards, a display interface (e.g., wired or wireless), etc. As an example, imaging (such as surface imaging) (e.g., satellite, geological, geophysical, etc.) may be stored, processed, transmitted, etc. As an example, data may include SAR data, GPS data, etc., and may be stored, for example, in one or more storage devices in storage device 252.

[0079] As an example, one or more sets of instructions 270 may include instructions (e.g., stored in memory 258) executable by one or more of processors 256 to instruct system 250 to perform various actions. As an example, system 250 may be configured such that one or more sets of instructions 270 provide for establishing Figure 1 As an example, one or more sets of instructions may be used to perform one or more methods, techniques, etc., and the one or more sets of instructions may be, for example, Figure 2 One or more of one or more sets of instructions 270.

[0080] As mentioned, seismic data can be collected and analyzed to better understand the underground structure of a geological environment. Reflection seismology is used in geophysics, for example, to estimate the properties of underground strata. As an example, reflection seismology can provide seismic data representing waves of elastic energy (e.g., as transmitted by P-waves and S-waves in a frequency range of about 1 Hz to about 100 Hz, or optionally less than about 1 Hz and / or optionally greater than about 100 Hz). Seismic data can be processed and interpreted, for example, to better understand the composition, fluid content, extent, and geometry of underground rocks.

[0081] Figure 3 An example of an acquisition technique 340 for acquiring seismic data (e.g., see data 360) is shown. As an example, the system can process the data acquired by the technique 340, for example, to allow for direct or indirect management of sensing, drilling, injection, extraction, etc. relative to the geological environment. In turn, further information about the geological environment can be made available as feedback (e.g., optionally as input to the system). As an example, the operation can involve a reservoir, such as a reservoir, present in the geological environment. As an example, the technique can provide information (e.g., as output) that can specify one or more position coordinates of a feature in the geological environment, one or more characteristics of a feature in the geological environment, etc.

[0082] exist Figure 3 , technology 340 can be implemented with respect to a geological environment 341. As shown, an energy source (e.g., a transmitter) 342 can transmit energy, where the energy propagates as waves that interact with the geological environment 341. As an example, the geological environment 341 can include a borehole 343, where one or more sensors (e.g., receivers) 344 can be positioned in the borehole 343. As an example, the energy transmitted by the energy source 342 can interact with a layer (e.g., a structure, an interface, etc.) 345 in the geological environment 341, so that a portion of the energy is reflected, and the reflected energy can then be sensed by one or more sensors in the sensors 344. This energy can be reflected as an upgoing first arrival wave (e.g., or a "first arrival" or "single" reflection wave). As an example, a portion of the transmitted energy can be reflected by more than one structure in the geological environment and is referred to as a multiple reflection wave (e.g., or a "multiple wave"). For example, the geological environment 341 is shown as including a layer 347 located below a ground layer 349. Given this environment and arrangement of source 342 and sensor(s) 344, energy may be sensed as being associated with a particular type of wave.

[0083] As examples, seismic data may include evidence of interlayer multiples from layer interfaces, evidence of multiples from water interfaces (e.g., the interface between a water base and the rock or sediment beneath it), or evidence of multiples from air-water interfaces, etc.

[0084] like Figure 3 As shown, the collected data 360 may include data associated with the downlink direct wave, the reflected uplink first arrival wave, the downlink multiple reflection wave, and the reflected uplink multiple reflection wave. The collected data 360 is also shown along the time axis and the depth axis. As indicated, in a manner that depends at least in part on the characteristics of the medium in the geological environment 341, the wave propagates a certain distance at a certain speed, so that there may be a relationship between time and space. Therefore, the time information associated with the sensed energy can allow the spatial relationship of the layers, interfaces, structures, etc. in the geological environment to be understood.

[0085] Figure 3 A schematic diagram 370 illustrating various types of waves (including P, SV and SH waves) is also shown. As an example, a P wave can be an elastic body wave or an acoustic wave, in which particles oscillate in the direction of propagation of the wave. As an example, a P wave incident on an interface (e.g., with non-vertical incidence, etc.) can produce a reflected and transmitted S wave (e.g., a "converted" wave). As an example, an S wave or shear wave can be, for example, an elastic body wave, in which particles oscillate perpendicular to the direction of propagation of the wave. S waves can be generated by a seismic energy source (e.g., in addition to an air gun). As an example, an S wave can be converted into a P wave. S waves tend to travel slower than P waves and do not travel through fluids that do not support shear. Generally speaking, the recording of S waves involves the use of one or more receivers operatively coupled to the earth (e.g., capable of receiving shear forces relative to time). As an example, the interpretation of S waves can allow, for example, by cross-plotting P wave velocities and S wave velocities and / or by other techniques to determine rock properties, such as fracture density and orientation, Poisson's ratio and rock type. As examples of parameters that can characterize the anisotropy of a medium (eg, seismic anisotropy, etc.), consider the Thomsen parameters ε, δ, and γ.

[0086] Figure 3 Also shown is an example of ocean harvesting technology 380 where a vessel can tow various equipment. Figure 3As shown, below the air-water interface 381, there is a seafloor or ocean floor surface 385, and further below, there are one or more other surfaces 386. The vessel can tow an energy source 382 and a sensor 384, which can include, for example, at least one seismic detector and a hydrophone. As an example, a seismic detector can be a sensor configured for seismic acquisition on land and / or at sea, which can detect the velocity generated by seismic waves and, for example, can convert movement into electrical pulses. As an example, a seismic detector can be configured to detect movement in a single direction. As an example, a seismic detector can be configured to detect movement in a vertical direction. As an example, three mutually orthogonal seismic detectors can be used in combination to collect so-called 3C seismic data. As an example, a hydrophone can be a sensor configured to detect seismic energy in the form of underwater pressure changes during marine seismic acquisition. As an example, a hydrophone can be positioned along one or more pipe strings to form one or more streamers, which can be towed by a seismic survey vessel (e.g., or deployed in a borehole). Thus, in Figure 3 In an example, at least one geophone can provide motion detection, and a hydrophone can provide pressure detection. As an example, data (eg, analog and / or digital) can be transmitted via equipment, such as for processing, etc.

[0087] As an example, a method may include analyzing hydrophone responses and vertical geophone responses, which may help improve PZ summation, for example, by reducing receiver ghosts (e.g., deghosts) and / or free surface multi-noise contamination. As an example, a ghost may be defined as a reflection of a wavefield (e.g., receiver ghost, source ghost, etc.) reflected from a water surface (e.g., water and air interface) located above a receiver, source, etc. As an example, a receiver may experience a delay between an upgoing wavefield and its downgoing ghost, which may depend on the depth of the receiver.

[0088] As an example, a surface marine cable (e.g., a streamer) can be or include a buoyant assembly of wires that connect sensors and can relay seismic data to a recording seismic survey vessel. As an example, a multi-streamer vessel can tow more than one streamer cable to increase the amount of data collected in a single pass. As an example, for example, a marine seismic survey vessel can be about 75m long and travel at a speed of about 5 knots, while towing an array of airguns and a streamer containing sensors, which can be located, for example, about a few meters below the water surface. A so-called tail buoy can help staff locate the end of the streamer. As an example, the airgun can be activated periodically, such as at intervals of about 25m (e.g., at intervals of about 10 seconds), where the resulting sound waves propagate into the earth, which can be reflected back to the sensors on the streamer by one or more rock layers, and can then be relayed to the equipment on the tugboat as a signal (e.g., data, information, etc.).

[0089] As examples, seismic surveys may be land surveys, well surveys, a combination of well surveys and another type of survey, marine surveys, etc. With respect to marine surveys, although ship-based techniques are shown as examples, other examples may be or may include the use of seafloor equipment (e.g., one or more of an ocean bottom node (OBN) and an ocean bottom cable (OBC)).

[0090] As an example, a seismic survey can generate seismic data, which can be in the form of seismic traces. Such data can be in the spatial domain or in the spatial and temporal domains (e.g., considering time, frequency, etc.). Seismic data can include markers of one or more events. An event can be defined as the appearance of seismic data, such as diffraction, reflection, refraction, or other similar features produced by the arrival of seismic energy. As an example, an event can be a single swing within a trace (e.g., amplitude versus time or distance), or a consistent arrangement of several swings on several traces. As an example, an event in a seismic profile can represent a geological interface, such as a fault, an unconformity, or a lithology change.

[0091] As an example, the seabed (e.g., the ocean floor) can be an event. In such an example, where a marine survey generates seismic traces with seabed events, the seismic traces can be analyzed to determine a surface representing the seabed. Such an event can be the first event in a trace, with one or more additional events representing one or more structures beneath the seabed. In a land survey, for example, the first event can be determined via trace analysis, where changes in amplitude with respect to time (e.g., or depth) indicate a reflector, etc. A series of events can be ordered, where, for example, it can be desired that the order remain consistent across a region.

[0092] As an example, inversion techniques may be applied to generate models of the Earth's subsurface regions. Such techniques are intended to reproduce layer models, where interfaces between layers represent reflectors that give rise to corresponding events. As an example, one or more types of data may be received and used to solve an inversion problem for an output model (e.g., a reflectivity model, an impedance model, a fluid flow model, etc.).

[0093] exist Figure 3In the example of FIG. 390 , a collection device 392 is shown, which transmits energy from a source (e.g., a transmitter) and receives the reflected energy via one or more sensors (e.g., receivers) arranged in a row along the main line (inline) direction. Because the area includes a layer 393 and, for example, a geological body 395, the energy emitted by the transmitter of the collection device 392 can be reflected from the layer 393 and the geological body 395. Evidence of such reflections can be found in the collected traces (e.g., as events, etc.). Regarding the portion of the trace 396, the received energy can be discretized by an analog-to-digital converter (ADC) operating at a certain sampling rate. For example, the collection device 392 can convert the energy signal sensed by the sensor Q into digital samples at a rate of approximately one sample per 4ms. Given the speed of sound in one or more media, the sample rate can be converted into an approximate distance. For example, the speed of sound in rock can be on the order of about 5km per second. Therefore, a sample time interval of approximately 4ms will correspond to a sample "depth" interval of approximately 10 meters (e.g., assuming a path length from the source to the boundary and from the boundary to the sensor). As an example, a trace may be about 4 seconds in duration; thus, for a sampling rate of one sample per about 4 ms, such a trace would include about 1000 samples, with samples acquired later corresponding to deeper reflective boundaries. If the 4 second trace duration of the previous example is divided by 2 (e.g., to account for reflections), then for vertically aligned sources and sensors, the deepest boundary depth may be estimated to be about 10 km (e.g., assuming a sound speed of about 5 km per second).

[0094] 4D seismic surveys involve the acquisition of 3D seismic data at different times over a particular area. This approach can allow for the evaluation of changes in hydrocarbon-producing reservoirs with respect to time. As an example, changes in one or more of fluid position and saturation, pressure, and temperature can be observed. 4D seismic data can be considered a form of time-lapse seismic data.

[0095] As an example, seismic surveys and / or other data acquisition can be used in onshore and / or offshore geological environments. With respect to offshore, as mentioned, streamers, seabed cables, nodes and / or other equipment can be utilized. As an example, nodes can be used as a substitute and / or supplement to seabed cables, which have been installed in several oil fields to acquire 4D seismic data. Nodes can be deployed to acquire seismic data (e.g., 4D seismic data) and can be retrieved after acquiring the seismic data. As an example, 4D seismic surveys may require one or more processes intended to achieve data repeatability. 4D surveys can include two phases: a baseline survey phase and a monitoring survey phase.

[0096] As an example, seismic data may be processed in a technique known as "depth imaging" to form an image (eg, a depth image) of reflection amplitudes of a particular target structure (eg, a geological subsurface region of interest) in the depth domain.

[0097] As an example, seismic data may be processed to obtain an elastic model related to the elastic properties of a geological subsurface region. For example, elastic properties such as density, compression (P) impedance, compression velocity (v p ) and shear velocity (v s ), anisotropy, etc. As an example, an elastic model can provide various insights into the lithology, reservoir quality, fluids, etc. of the survey area.

[0098] Figure 4 An example of a wellsite system 400 is shown (e.g., at a wellsite that may be located onshore or offshore). As shown, the wellsite system 400 may include: a mud tank 401 for storing mud and other materials (e.g., where the mud may be drilling fluid); a suction line 403 serving as an inlet for a mud pump 404 for pumping mud from the mud tank 401 to a vibrating hose 406; a winch 407 for winching one or more drilling wires 412; a riser 408 for receiving mud from the vibrating hose 406; and a kelly hose 409 for receiving mud from the riser 408. ; one or more goosenecks 410; a traveling block 411; a crown block 413 for carrying the traveling block 411 via one or more drilling wire ropes 412; a derrick 414; a kelly 418 or a top drive 440; a kelly bushing 419; a rotary table 420; a drill floor 421; a bell-mouth nipple 422; one or more blowout preventers (BOPs) 423; a drill string 425; a drill bit 426; a casing head 427; and a flow pipe 428 for conveying mud and other materials to, for example, a mud tank 401.

[0099] exist Figure 4 In the exemplary system of FIG. 4 , a wellbore 432 is formed in a subsurface formation 430 by rotary drilling; it should be noted that various exemplary embodiments may also use directional drilling.

[0100] like Figure 4 As shown in the example of FIG. 4 , a drill string 425 is suspended in a wellbore 432 and has a drill string assembly 450 including a drill bit 426 at its lower end. As an example, the drill string assembly 450 may be a bottom hole assembly (BHA).

[0101] The wellsite system 400 can provide for the operation of the drill string 425 and other operations. As shown, the wellsite system 400 includes a platform 411 and a derrick 414 positioned above a wellbore 432. As mentioned, the wellsite system 400 can include a rotary table 420, wherein the drill string 425 passes through an opening in the rotary table 420.

[0102] like Figure 4 As shown in the example of , the wellsite system 400 may include a kelly 418 and associated components, or a top drive 440 and associated components. Regarding the example of a kelly, the kelly 418 may be a square or hexagonal metal / alloy rod with holes drilled therein for use as a mud flow path. The kelly 418 may be used to transfer rotational motion from the rotary table 420 to the drill string 425 via the kelly bushing 419, while allowing the drill string 425 to be lowered or raised during rotation. The kelly 418 may pass through the kelly bushing 419, which may be driven by the rotary table 420. As an example, the rotary table 420 may include a main bushing that is operatively coupled to the kelly bushing 419, such that rotation of the rotary table 420 may rotate the kelly bushing 419 and thereby rotate the kelly 418. The kelly bushing 419 may include an inner profile that matches the outer profile of the kelly 418 (eg, square, hexagonal, etc.); however, it is slightly larger in size so that the kelly 418 can move freely up and down within the kelly bushing 419 .

[0103] With respect to the top drive example, the top drive 440 can provide the functions performed by the kelly and rotary table. The top drive 440 can rotate the drill string 425. As an example, the top drive 440 may include one or more (e.g., electric and / or hydraulic) motors connected to a short pipe section called a hollow shaft using an appropriate transmission, which in turn can be screwed into a saver sub or the drill string 425 itself. The top drive 440 can be suspended on a traveling block 411 so that the rotating mechanism can move freely up and down along the derrick 414. As an example, the top drive 440 can allow drilling to be performed using more single columns than the kelly / rotary table approach.

[0104] exist Figure 4 In the example of, mud tank 401 can store mud, and mud can be one or more types of drilling fluids. As an example, a wellbore can be drilled to produce fluids, inject fluids, or both (e.g., hydrocarbons, minerals, water, etc.).

[0105] exist Figure 4In the example of , a drill string 425 (e.g., including one or more downhole tools) can be composed of a series of pipes threaded together to form a long pipe with a drill bit 426 at its lower end. As the drill string 425 enters the wellbore for drilling, mud can be pumped from a mud tank 401 (e.g., or other source) via pipelines 406, 408, and 409 to a port of the kelly 418, or, for example, to a port of the top drive 440, by a pump 404, prior to or at some point in time coinciding with the drilling. The mud can then flow through a channel (e.g., or multiple channels) in the drill string 425 and out of a port located on the drill bit 426 (e.g., see directional arrows). As the mud leaves the drill string 425 via the port in the drill bit 426, the mud can circulate upward through the annulus region between the outer surface of the drill string 425 and the surrounding wellbore wall (e.g., open hole, casing, etc.), as shown by the directional arrows. In this manner, the mud lubricates the drill bit 426 and carries heat (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 401, for example, for recirculation (e.g., by processing to remove cuttings, etc.).

[0106] The mud pumped into the drill string 425 by the pump 404 can form a mud cake attached to the wellbore after leaving the drill string 425, which, among other functions, can reduce the friction between the drill string 425 and the surrounding wellbore wall (e.g., wellbore, casing, etc.). The reduction in friction can facilitate the advancement or retraction of the drill string 425. During the drilling operation, the entire drill string 425 can be pulled out of the wellbore and optionally replaced, for example, with a new or sharp drill bit, a smaller diameter drill string, etc. As mentioned, the action of pulling the drill string out of the wellbore or replacing the drill string in the wellbore is called tripping. Depending on the tripping direction, the tripping can be called up-tripping or out-tripping or down-tripping or in-tripping.

[0107] As an example, consider drilling down, where when the drill bit 426 of the drill string 425 reaches the bottom of the wellbore, the pumping of mud begins to lubricate the drill bit 426 for drilling purposes to expand the wellbore. As mentioned, the mud can be pumped into the channel of the drill string 425 by the pump 404, and when filling the channel, the mud can be used as a transmission medium for transmitting energy (e.g., energy that can encode information like mud pulse telemetry).

[0108] As an example, mud pulse telemetry equipment may include a downhole device configured to implement pressure changes in the mud to generate one or more acoustic waves based on which information can be modulated. In such examples, information from the downhole equipment (e.g., one or more modules of the drill string 425) can be transmitted up to the wellhead device, which can relay such information to other equipment for processing, control, etc.

[0109] As an example, the telemetry equipment may operate by transmitting energy through the drill string 425 itself. For example, consider a signal generator that transmits a coded energy signal to the drill string 425, and a repeater that may receive such energy and relay it for further transmission of the coded energy signal (e.g., information, etc.).

[0110] As an example, the drill string 425 may be equipped with a telemetry device 452, the telemetry device comprising: a rotatable drive shaft; a turbine wheel mechanically coupled to the drive shaft so that mud can rotate the turbine wheel; a modulator rotor mechanically coupled to the drive shaft so that rotation of the turbine wheel causes the modulator rotor to rotate; a modulator stator mounted adjacent or proximate to the modulator rotor so that rotation of the modulator rotor relative to the modulator stator generates pressure pulses in the mud; and a controllable brake for selectively braking rotation of the modulator rotor to modulate the pressure pulses. In such an example, an alternator may be coupled to the drive shaft, wherein the alternator includes at least one stator winding electrically coupled to a control circuit to selectively short-circuit the at least one stator winding to electromagnetically brake the alternator, thereby selectively braking rotation of the modulator rotor to modulate pressure pulses in the mud.

[0111] exist Figure 4 In an example of, the wellhead control and / or data acquisition system 462 may include circuitry for sensing pressure pulses generated by the telemetry equipment 452 and (e.g.) transmitting the sensed pressure pulses or information derived therefrom for processing, control, etc.

[0112] The components 450 of the illustrated example include a logging while drilling (LWD) module 454 , a measurement while drilling (MWD) module 456 , optional modules 458 , a rotary steerable system and motor 460 (RSS), and the drill bit 426 .

[0113] The LWD module 454 may be housed in a suitable type of drill collar and may include one or more selected types of logging tools. It should also be understood that more than one LWD and / or MWD module may be employed, for example, as represented by module 456 of the drill string assembly 450. Where the location of an LWD module is mentioned, by way of example, it may refer to a module at the location of the LWD module 454, module 456, etc. The LWD module may include capabilities for measuring, processing, and storing information, as well as the ability to communicate with surface equipment. In the example shown, the LWD module 454 may include a seismic measurement device.

[0114] The MWD module 456 may be housed in a suitable type of drill collar and may include one or more devices for measuring characteristics of the drill string 425 and the drill bit 426. As an example, the MWD tool 454 may include equipment for generating electricity, for example, to power various components of the drill string 425. As an example, the MWD tool 454 may include telemetry equipment 452, for example, where a turbine wheel may generate electricity from the flow of mud; it will be appreciated that other power sources and / or battery systems may be used to power the various components. As an example, the MWD module 456 may include one or more of the following types of measurement devices: a bit pressure measurement device, a torque measurement device, a vibration measurement device, an impact measurement device, a stick-slip measurement device, a direction measurement device, and an inclination measurement device.

[0115] With respect to the RSS 460, various types of suitable rotary steerable tool configurations may be used. For example, the RSS may include a substantially non-rotating (or slowly rotating) housing that employs blades that engage the wellbore wall. The engagement of the blades with the wellbore wall is intended to be eccentric to the tool body, thereby pointing or pushing the drill bit in a desired direction when drilling. During drilling, a rotating shaft deployed in the housing transmits rotational power and axial weight-on-bit to the drill bit. An accelerometer and magnetometer array may be deployed in the housing and, therefore, does not rotate or rotates slowly relative to the wellbore wall. As an example, an RSS such as the POWERDRIVE rotary steerable system (Schlumberger, Inc., Houston, Texas) may rotate fully with the drill string (e.g., the housing rotates with the drill string). As an example, the RSS may utilize an internal guide mechanism that may operate without requiring contact with the wellbore wall and may enable the tool body to rotate fully with the drill string. As an example, the RSS may include features that provide for the use of mud-driven blades (or pads) that contact the wellbore wall. As such a system rotates in the wellbore, the extension of the blades (or pads) can be adjusted quickly and continuously. As an example, the RSS may include and utilize a lower guide portion that engages an upper portion at a swivel. Such a swivel may be actively tilted via a piston so as to change the angle of the lower portion relative to the upper portion and maintain the desired drilling direction as the BHA rotates in the wellbore. As an example, one or more accelerometers and magnetometer groups may rotate with the drill string, or may alternatively be deployed in an internal roll-stable housing so that they remain substantially stationary (in a bias phase) or rotate slowly relative to the wellbore (in a neutral phase). In order to drill the desired curvature, the bias phase and the neutral phase may alternate at a predetermined ratio, referred to as the steering ratio (SR), during drilling.

[0116] As an example, the deviation of the borehole can be achieved in part by using a downhole motor and / or turbine. With respect to the motor, for example, the drill string may include a positive displacement motor (PDM). Deflection can also be achieved by using a rotary steerable system (RSS).

[0117] Figure 4 Also shown are some examples of the types of wellbores that can be drilled. For example, consider a slanted straight wellbore 472, an S-shaped wellbore 474, a deeply deviated wellbore 476, and a horizontal wellbore 478.

[0118] As an example, the drilling operation may include directional drilling, where, for example, at least a portion of the well includes a curved axis. For example, consider a radius defining a curvature, where the inclination relative to the vertical may vary until an angle between about 30 degrees and about 60 degrees is reached, or, for example, an angle of about 90 degrees or possibly greater than about 90 degrees is reached.

[0119] As an example, a directional well can include a variety of shapes, each of which can be designed to meet specific operational requirements. As an example, when information is communicated to a drilling engineer, a drilling process can be performed based on the information. As an example, the inclination and / or direction can be modified based on information received during the drilling process.

[0120] As an example, the system may be a steering system and include equipment for performing methods such as geosteering. As an example, the steering system may include a PDM or turbine located at the bottom of the drill string, which is just above the drill bit and a bent joint may be installed. As an example, above the PDM, MWD equipment and / or LWD equipment that 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.) may be installed. For the latter, the LWD equipment may send various types of data of interest to the ground, including, for example, geological data (e.g., gamma ray logging, resistivity, density, and sonic logging, etc.).

[0121] The coupling of sensors that provide information about the progress of the well trajectory in real time or near real time with one or more well logs that, for example, characterize the formation from a geological perspective can allow for geosteering methods to be implemented. Such methods may include navigating the subsurface environment, for example, to follow a desired route to reach a desired target or targets.

[0122] As an example, the drill string may include an azimuthal density neutron (ADN) tool for measuring density and porosity; an MWD tool for measuring inclination, azimuth, and shock; a compensated dual resistivity (CDR) tool for measuring resistivity and gamma-ray related phenomena; one or more variable diameter stabilizers; one or more curved joints; and a geosteering tool, which may include a motor and, optionally, equipment for measuring and / or responding to one or more of inclination, resistivity, and gamma-ray related phenomena.

[0123] As an example, geosteering may include intentional directional control of a wellbore based on downhole geological logging measurements in a manner intended to maintain the directional wellbore within a desired area, zone (e.g., a pay zone), etc. As an example, geosteering may include directing the wellbore to maintain the wellbore within a particular well section of a reservoir, for example, to minimize breakthrough of gas and / or water, and, for example, to maximize economic production of the well that includes the wellbore.

[0124] Reference again Figure 4 , the wellsite system 400 may include one or more sensors 464, which are operatively coupled to the control and / or data acquisition system 462. 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 that are not within a distance of about one hundred meters from the wellsite system 400. As an example, one or more sensors may be located at a compensation wellsite, where the wellsite system 400 and the compensation wellsite are in a common oil and gas field (e.g., an oil field and / or a gas field).

[0125] As an example, one or more sensors 464 may be provided for tracking drill rods, tracking movement of at least a portion of a drill string, etc.

[0126] As an example, the system 400 may include one or more sensors 466 that can sense and / or transmit signals to a fluid conduit, such as a drilling fluid conduit (e.g., a drilling mud conduit). For example, in the system 400, one or more sensors 466 may be operatively coupled to a portion of the riser 408 through which mud flows. As an example, a downhole tool may generate pulses that may pass through the mud and be sensed by one or more of the one or more sensors 466. In such examples, the downhole tool may include associated circuitry, such as an encoding circuit that may encode the signal, such as to reduce the requirements for transmission. As an example, the circuitry located at the surface may include a decoding circuit to decode the encoded information transmitted at least in part via mud pulse telemetry. As an example, the circuitry located at the surface may include an encoder circuit and / or a decoder circuit, and the downhole circuitry may include an encoder circuit and / or a decoder circuit. As an example, the system 400 may include a transmitter that may generate a signal that may be transmitted downhole via mud (e.g., drilling fluid) as a transmission medium.

[0127] As an example, a method may include seismic while drilling. For example, consider the use of the SEISMICVISION seismic while drilling equipment (Schlumberger, Inc., Houston, Texas), which can provide time-depth-velocity information in real time during drilling operations. Seismic while drilling can include obtaining borehole seismic measurements (e.g., optionally including real-time calibration shots, interval velocity data, etc.), which helps reduce uncertainty in front of the drill bit (e.g., from hundreds of feet or more). As an example, real-time waveforms can be transmitted to the surface via MWD telemetry for quality control of downhole data, etc. In seismic while drilling, real-time waveform resolution and sufficient length can provide advanced vertical seismic profile (VSP) processing. Seismic while drilling can be combined with one or more other technologies, which can facilitate structural identification (e.g., interpretation), model building, etc. Seismic while drilling can be used, for example, to guide drilling according to a planned trajectory and / or determine when or where to deviate from a planned trajectory (e.g., update one or more target locations, etc.).

[0128] Figure 5 An example of an environment 501 is shown that includes an underground portion 503 in which a drilling rig 510 is positioned at a surface location above a borehole 520. Figure 5 In the example of FIG. 5 , various cable service equipment may be operated to perform one or more cable services, including, for example, collecting data from one or more locations within borehole 520 .

[0129] exist Figure 5In the example of , the borehole 520 includes drill pipe 522, casing shoe, cable side entry sub (CSES) 523, wet connector adapter 526, and open hole section 528. For example, the borehole 520 can be a vertical borehole or a deviated borehole, wherein one or more portions of the borehole can be vertical, and one or more portions of the borehole can be deviated, including substantially horizontal.

[0130] exist Figure 5 In the example of , CSES 523 includes a cable clamp 525, an isolation seal assembly 527, and a check valve 529. These components can provide for the insertion of a logging cable 530, which includes a portion 532 that continues outside the drill pipe 522 to be inserted into the drill pipe 522, so that at least a portion 534 of the logging cable continues inside the drill pipe 522. Figure 5 In the example of FIG. 5 , the logging cable 530 continues through the wet connector adapter 526 and into the open hole section 528 to the logging string 540 .

[0131] like Figure 5 As shown in the example of FIG. 5 , a logging vehicle 550 (e.g., a cable service vehicle) can deploy a cable 530 under the control of a system 560. Figure 5 As shown in the example of , the system 560 may include one or more processors 562, a memory 564 operatively coupled to at least one of the one or more processors 562, instructions 566 that may be stored, for example, in the memory 564, and one or more interfaces 568. As an example, the system 560 may include one or more processor-readable media, the one or more processor-readable media including processor-executable instructions that can be executed by at least one of the one or more processors 562 to enable the system 560 to control one or more aspects of the equipment of the logging string 540 and / or the logging vehicle 550. In such an example, the memory 564 may be or include one or more processor-readable media, wherein the processor-executable instructions may be or include instructions. As an example, the processor-readable medium may be a computer-readable storage medium, which is neither a signal nor a carrier wave.

[0132] Figure 5 Also shown is a battery 570 that can be operatively coupled to the system 560, for example, to power the system 560. As an example, the battery 570 can be a backup battery that operates when another power source is not available to power the system 560 (e.g., via the generator of the well testing vehicle 550, a separate generator, a power line, etc.). As an example, the battery 570 can be operatively coupled to a network, which can be a cloud network. As an example, the battery 570 can include smart battery circuitry and can be operatively coupled to one or more pieces of equipment via an SMBus or other type of bus.

[0133] As an example, system 560 can be operatively coupled to client layer 580. Figure 5 In the example of , the client layer 580 may include features that allow access and interaction via one or more private networks 582, one or more mobile platforms and / or mobile networks 584, and via a "cloud" 586, which can be considered to include distributed equipment that forms a network (such as a network among multiple networks). As an example, the system 560 may include circuits for establishing multiple connections (e.g., sessions). As an example, the connection may be via one or more types of networks. As an example, the connection may be a client-server type connection, where the system 560 is used as a server in a client-server architecture. For example, a client can log in to the system 560, in which multiple clients can optionally be processed simultaneously.

[0134] As an example, a seismic workflow can provide processing of microseismic data as a type of seismic data. Microseismic monitoring (e.g., a type of seismic survey) provides a valuable tool for real-time evaluation of hydraulic fracture treatments and can be used to plan and manage reservoir development. Microseismic event locations, source characteristics, and attributes can provide estimates of hydraulic fracture geometry, which can be evaluated relative to completion plans and expected fracture growth. Attributes derived from microseismic events (such as fracture azimuth, height and length, location, and complexity) can be used to determine fracture coverage and effective stimulation volumes of reservoir targets, as well as for diagnosing under-stimulated sections of reservoirs and for planning re-stimulation of under-produced perforations and wells. Microseismic event locations can also help avoid hazards (e.g., faults, karsts, aquifers, etc.) during stimulation. As an example, a method may include, as part of a seismic interpretation workflow, at least in part based on microseismic interpretation to modify one or more treatment plans and operations.

[0135] An integrated workflow utilizing multi-scale, multi-domain measurements and microseismic interpretations can allow optimization of hydraulic fracturing treatments to increase production. This integrated completion planning workflow can use a wide variety of information about geology (e.g., lithology, stress differentials, natural fractures, structural or sedimentary tilts, faults) and associated rock properties (e.g., noise, slowness, anisotropy, attenuation) to improve hydraulic fracturing operations to result in improved hydraulic fracture stimulation, completion planning, and well placement, and thereby increase production. As an example, microseismic event locations and attributes can be integrated and compared with treatment pressure records, proppant concentrations, and injection rates to better perform field operations.

[0136] Figure 1 , Figure 2 , Figure 3 , Figure 4 and Figure 5Various examples of equipment in various environment examples are shown. As an example, one or more workflows may be implemented to perform operations using equipment in one or more environments. As an example, a workflow may be intended to understand an environment. As an example, a workflow may include performing a seismic survey, which may be land-based, sea-based (e.g., ships, ocean floors, etc.), or land-based and sea-based. As an example, a seismic survey may include an acquisition geometry in which a receiver and / or source is positioned according to the acquisition geometry. As an example, a seismic survey may be performed using one or more receivers and / or one or more sources positioned in an underground environment (e.g., in a borehole). As an example, a workflow may include acquiring various types of data, which may include seismic data as a type of data and one or more other types of geophysical data, which may include imaging data (e.g., borehole imaging, satellite imaging, drone imaging, etc.).

[0137] As an example, a workflow may be intended to drill into an environment, for example, to form a borehole defined by the surrounding earth (e.g., rock, fluid, etc.). As an example, a workflow may be intended to collect data from a downhole tool disposed in a borehole, wherein such data may be collected via a drilling tool (e.g., as part of a bottom hole assembly) and / or a wireline tool. As an example, a workflow may be intended to support a borehole, for example, via casing. As an example, a workflow may be intended to fracture an environment, for example, via injection of a fluid. As an example, a workflow may be intended to produce a fluid from an environment via a borehole. As an example, a workflow may utilize one or more frameworks that operate at least in part via a computer (e.g., a computing device, a computing system, etc.).

[0138] Figure 6 An example of forward modeling 610 and an example of inversion 630 (e.g., inversion or reversal) are shown. As shown, forward modeling 610 proceeds from an earth model of acoustic impedance and an input wavelet to synthetic seismic traces, while inversion 630 proceeds from recorded seismic traces to estimated wavelets and an earth model of acoustic impedance. As an example, forward modeling can take a model of formation properties (e.g., acoustic impedance that can be obtained from well logging) and combine this information with seismic wavelengths (e.g., pulses) to output one or more synthetic seismic traces, while inversion can start with recorded seismic traces, taking into account the effects of estimated wavelets (e.g., pulses) to generate acoustic impedance values ​​at a series of time points (e.g., depths).

[0139] As an example, one method may employ amplitude inversion. For example, an amplitude inversion method may receive the arrival times and amplitudes of reflected seismic waves at multiple reflection points to solve for the relative impedance of the formation defined by the imaged reflectors. This method may be a form of seismic inversion for reservoir characterization, which may help generate models of rock properties.

[0140] As an example, the inversion process can begin with forward modeling, for example, to provide a layer model with estimated formation depth, thickness, density, and velocity, which can be based, for example, at least in part, on information such as well log information. The model can account for compressional wave velocity and density, which can be used for inversion of P waves or acoustic impedance. As an example, the model can account for shear velocity and, for example, solve for S waves or elastic impedance. As an example, the model can be combined with seismic wavelets (e.g., pulses) to generate synthetic seismic traces.

[0141] Inversion may aim to generate a "best fit" model by, for example, iterating between forward modeling and inversion while trying to minimize the difference between one or more synthetic traces and the actual seismic data.

[0142] As an example, a framework such as the ISIS Inversion Framework (Schlumberger Inc. of Houston, Texas) may be implemented to perform the inversion. As an example, a framework such as the Linearized Orthogonal Inversion Framework (Schlumberger Inc. of Houston, Texas) may be implemented to perform the inversion.

[0143] As mentioned above, with respect to seismic data, forward modeling may include receiving an earth model of acoustic impedance and input wavelets to synthetic seismic traces, while inverse may include going from recorded seismic traces to estimated wavelets and an earth model of acoustic impedance.

[0144] As an example, another method of forward modeling and inversion can be for measurements collected at least in part via a downhole tool, where such measurements can include one or more of different types of measurements, which can be referred to as multi-physics measurements. As an example, the multi-physics measurements can include logging while drilling (LWD) measurements and / or wireline measurements. As an example, a method can include joint rock physics inversion (e.g., reversal) for interpreting multi-physics logging while drilling (LWD) measurements and / or wireline (WL) measurements.

[0145] As an example, a method may include estimating static and / or dynamic formation properties based on various logging while drilling (LWD) measurements (e.g., including pressure, resistivity, sonic, and nuclear data) and / or wireline (WL) measurements, which may provide at least formation parameters that characterize the formation. As an example, where the method is performed during drilling, the LWD measurements may be used in a joint inversion to output formation parameters (e.g., formation parameter values) that may be used to guide drilling (e.g., avoid stuck pipe, reduce one or more types of formation damage, etc.).

[0146] In petroleum exploration and development, formation evaluation is performed to interpret data collected from a drilled wellbore to provide information about the geological formation and / or one or more in situ fluids that can be used to evaluate the production properties of the reservoir rock penetrated by the wellbore.

[0147] As an example, data used for formation evaluation may include one or more of core data, mud logging data, wireline logging data (e.g., wireline data), and LWD data, where the LWD data may be a source for some or all types of formation evaluation (e.g., particularly when wireline acquisition is operationally difficult and / or economically not feasible).

[0148] Regarding the types of measurements, these measurements may include, for example, one or more of resistivity, gamma ray, density, neutron porosity, spectroscopy, sigma, magnetic resonance, elastic wave, pressure, and sample data (e.g., which may be acquired while drilling to enable timely quantitative formation evaluation).

[0149] Table 1 below shows some examples of data that may be referred to as "well log" data associated with petrophysical and petrophysical property calculations and analyses.

[0150] Table 1. Example of well logging data

[0151]

[0152] Information from one or more interpretations can be used in one or more ways with a system, which can be a well construction ecosystem. For example, seismic data can be acquired and interpreted and used to generate one or more models (e.g., earth models) for the purpose of construction and / or operation of one or more wells.

[0153] Figure 7 An example of a computing framework 700 is shown, which may include one or more processors and memory and, for example, one or more interfaces. Figure 7 The computational framework of the OMEGA framework (Schlumberger Ltd., Houston, Texas) may include one or more features of the OMEGA framework, which includes a finite difference modeling (FDMOD) feature for two-way wavefield extrapolation modeling to generate synthetic shot gathers with and without multiples. The FDMOD feature can generate synthetic shot gathers using full 3D, two-way wavefield extrapolation modeling, which can be matched with the wavefield extrapolation logic used by reverse time migration (RTM). The model can be specified as velocity on a dense 3D grid, and optionally as anisotropy, dip, and variable density.

[0154] like Figure 7As shown, the computing framework 700 includes RTM, FDMOD, adaptive beam migration (ABM), Gaussian bag migration (Gaussian PM), depth processing (e.g., Kirchhoff prestack depth migration (KPSDM), tomography (Tomo)), time processing (e.g., Kirchhoff prestack time migration (KPSTM), general surface multiple wave prediction (GSMP), extended interlayer multiple wave prediction (XIMP)), framework basic features, desktop features (e.g., GUI, etc.) and features of development tools.

[0155] Framework 700 may include features for geophysical data processing. Framework 700 may allow processing of various types of data, such as one or more of the following: land, ocean, and transition zone data; time and depth data; 2D, 3D, and 4D surveys; isotropic and anisotropic (TTI and VTI) velocity fields; and multi-component data.

[0156] The framework 700 may allow seismic, electromagnetic, microseismic and / or vertical seismic profile (VSP) data to be converted into actionable information, for example, to perform one or more actions in the field for purposes such as resource production. The framework 700 may extend workflows into reservoir characterization and earth modeling. For example, the framework 700 may extend geophysical data processing into reservoir modeling by integrating with the PETREL framework via an earth model building (EMB) tool, which enables various deep imaging workflows, including model building, editing and updating, deep tomography QC, residual time difference analysis, and volume common image point (CIP) selection QC. Such functionality, combined with the framework's deep tomography and migration algorithms, may produce accurate and precise images of the subsurface. The framework 700 may provide support for field to final imaging, prestack seismic interpretation, and quantitative interpretation from exploration to development.

[0157] As an example, the FDMOD component can be instantiated via one or more CPUs and / or one or more GPUs for one or more purposes. For example, consider using FDMOD to generate synthetic shot gathers using full 3D bidirectional wavefield extrapolation modeling, the same wavefield extrapolation logic matches the logic used by reverse time migration (RTM). FDMOD can model various aspects and effects of wave propagation. The output from FDMOD can be or include synthetic shot gathers, which include direct waves, first arrivals, surface multiple waves, and interlayer multiple waves. The model can be specified as velocity on a dense 3D grid, and optionally as anisotropy, dip, and variable density. As an example, the survey design can be modeled to ensure the quality of the seismic survey, which can take into account the structural complexity of the model. This method can assess the degree to which the target area is illuminated. This method can be part of a quality control process (e.g., a task) as part of a seismic workflow. As an example, the size of the FDMOD method can be specified, which can be a model size (e.g., a grid cell model size). Such parameters can be used to determine the resources that will be allocated to perform FDMOD-related processing tasks. For example, a relationship between model size and CPU, GPU, etc. may be established for the purpose of generating results within a desired amount of time, which may be part of the planning (eg, scheduling) of a seismic interpretation workflow.

[0158] As an example, as survey data becomes available, interpretation tasks may be performed to build, adjust, etc. one or more models of the geological environment. For example, considering a vessel that transmits a portion of acquired data while at sea and a portion of acquired data while in port, this may include physically unloading one or more storage devices and transporting such one or more storage devices to an onshore site that includes equipment operatively coupled to one or more networks (e.g., cables, etc.). As the data is available, there are options for tasks to be performed.

[0159] Various embodiments of the present disclosure can provide systems, methods and computer-readable storage media for interpreting data into a reservoir characterization workflow. In certain embodiments, the method can reduce the time spent interpreting in reservoir characterization studies, while improving quality and productivity, while reducing costs. When utilizing one or more automatic interpretation techniques, reservoir characterization can be more accurate.

[0160] As an example, a method may include applying machine learning and artificial intelligence to seismic surface extraction. In various cases, for various regions of the Earth, a large amount of training data is not available for a particular surface. In this case, the available data may include some semi-human interpreted data. For acceptable learning to generate a trained system, reliable training data and minimal user input (e.g., human time, resources, etc.) are desired.

[0161] As an example, the method may implement an ensemble approach in which tracking and learning gradually expands the amount of available training data and improves one or more models for distinct features of the subsurface surface, such as indicated in traces of seismic data (see, e.g., Figure 3 Trace 396, which is given as amplitude relative to time in 1D (note that time can be a proxy for depth or distance).

[0162] As an example, the method for training a machine model can operate in an interactive manner, where, for example, a user can select or input training information to train the machine model, or where the user selects an automated technique for selecting training information to train the machine model. As an example, the training information can include location information of one or more events. For example, a user can select a point on a visual representation of seismic data (where the point represents an event), or an automated technique can analyze the seismic data for one or more points representing one or more events.

[0163] As an example, the method may start with as few as a single point selected by a user who interacts with a graphical user interface (GUI) presented to a display of a seismic data interpretation system (e.g., a computing framework). In such examples, the user may interact with the system via a mouse, finger, voice command, stylus, etc., so that a point may be registered relative to seismic data (e.g., a seismic cube in three dimensions), where the point may be associated with a trace, for example. In such examples, whether the trace is specified relative to time or another dimension, the point may be determined as a point corresponding to the trace. In such examples, the point may correspond to an amplitude defined by a dimension along the trace (e.g., as an event, etc.). As an example, a point may be associated with a window, which may extend in one or more directions. In a 1D scenario, for a trace defined relative to time, the window may extend to a time less and / or longer than the time of a particular specific point. Since the trace may have a known position given a seismic survey geometry, the position of the trace may be known or approximately estimated based on information from the seismic survey geometry (e.g., the position of a receiver receiving seismic energy represented in the form of an amplitude versus time 1D trace, etc.). As an example, an interpreter may visually associate the selected points with features in the subsurface environment, such as reflectors or other geological features.In this manner, a seismic survey may be transformed into a model of the subsurface environment being surveyed.

[0164] As an example, given at least one picked point, the method can involve subsequent robust tracking to provide additional training data while generating a more robust machine learning model to identify the specific surface (e.g., the surface of a geological feature) to which the point corresponds. In such examples, the method can construct an acceptable amount of training data from minimal user input (e.g., at least one picked point).

[0165] In various seismic interpretation workflows, seismic surface tracking involves many parameters that need to be fine-tuned according to the surface to be tracked. As an example, a seismic interpretation system using machine learning and artificial intelligence can learn to automatically identify distinctive features of the surface through self-tuning (e.g., learning), which can save a considerable amount of work for the interpreter. For example, consider a self-tuning method that uses fewer parameters than a manual tuning method that uses many parameters. The self-tuning method can allow the interpreter to continue various interpretation tasks, optionally with quick corrections, without having to navigate to the parameter field of the GUI to tune, readjust, etc. a list of many parameters. In addition, a method may include generating and outputting confidence information that may allow a user to determine the quality of the interpretation. For example, consider a method that includes generating a confidence cube as a type of interpretation attribute, which may be visualized by visually presenting the confidence cube (e.g., volume, slide, line, point, etc.) to a display. In such examples, a user may select one or more portions of the underlying seismic data for acceptance, correction, etc. based on such a visual presentation, wherein the user has an indication of confidence and therefore quality (e.g., good or bad).

[0166] As mentioned, a method may include starting by receiving one or more points indicating a surface to be identified and extracted from seismic data. Such one or more points and seismic data provide a first basis for extracting training data for machine learning. Since machine learning can involve both positive examples and negative examples, amplitude data around points in the example data can be extracted as positive examples. Negative examples can be extracted, for example, around other locations other than the location indicated by the input point. For example, a metric can be utilized that is a distance metric or a time metric (e.g., about trace data) based on geological knowledge about an area of ​​the earth so that negative examples may be sufficiently different from positive examples.

[0167] As an example, to minimize the chance of false negatives, negative examples can be taken from other locations along the same trace from which the positive examples were extracted. An example in this case can be amplitude values ​​around a specified location in a seismic cube. Examples can also include attribute data derived from the seismic data (e.g., one or more seismic attributes, other attributes, etc.).

[0168] Given positive and negative data, machine learning or statistical learning can be applied to learn a model M (e.g., a machine model) to predict whether an exemplary seismic data profile represents a surface of interest. In this case, the profile can mean the surrounding data. As an example, the learned model (e.g., a trained model) can produce an output in the range [0, 1] or other ranges indicating the confidence that the provided example data is the expected surface.

[0169] As mentioned, confidence values ​​can be stored in a form corresponding to an array of seismic data. As an example, a point can include one or more confidence values, which can provide directionality and / or other metrics. As an example, a method can include calculating information based on one or more types of data that can be other than seismic data. For example, consider cable data, measurement while drilling / logging data, etc. In such examples, one or more metrics can be calculated based on such data, wherein confidence can be determined at least in part based on at least one of the one or more metrics. As an example, based on data collected from a wellbore (or, for example, a lithology, fluid, etc.), based on data collected from the wellbore that can be evaluated using seismic data, consider confidence about distinct features and confidence about location. In this way, one type of confidence can be associated with a seismic distinct feature, and another type of confidence can be associated with at least one other type of information.

[0170] As an example, the generated model can be applied to one or more purposes, such as being suitable for one or more workflows. For example, consider using the above model M as a quality metric for surface tracking (e.g., seismic surface tracking), as a "detector" in seismic data, and / or as a converter to generate confidence display volumes / 2D lines as a further interpretation aid.

[0171] Used as a quality metric for surface tracking, model M may provide a basis for determining whether the surface tracking algorithm can confidently track further on the surface once it has started, or whether stopping is advisable (eg, end of surface, change in seismic signature, etc.).

[0172] As an example, where 4D seismic survey data is available, one or more distinctive features, models, etc. may be reapplied. For example, consider utilizing a method involving training based on a baseline seismic survey at time T0 and then acquiring seismic data via a monitoring seismic survey at time T1. In such examples, training / learning from the baseline seismic survey data may be applied to interpret the monitoring seismic survey data, and / or new training / learning may be applied to the monitoring seismic survey data, wherein the information generated thereby (e.g., about distinctive features, models, etc.) may be compared with the information generated from the baseline seismic survey. In such examples, the comparison between the information may provide an understanding of physical changes that may have occurred in the geological region relative to the time survey. In such examples, the final interpretation results and the information generated by using a seismic interpretation system may be compared, and the seismic interpretation system utilizes machine learning and artificial intelligence to automatically evaluate the seismic data.

[0173] With regard to use as a "detector" scanning through the seismic volume, the model M can detect and suggest locations in the volume that are likely to be the same surface as one or more starting points. Used as a translator of visual aids, the model M can guide and facilitate further interpretation, as it can be used to highlight places in the seismic data that the model M indicates are likely to be the same surface, and thus can be selected next, for example.

[0174] As an example, the model M can be an artificial neural network model, a statistical model, or another type of numerical computational structure that can capture the seismically distinctive features of the surface with sufficient accuracy to distinguish it from one or more other surfaces in the seismic data. The accuracy of the model can depend on the surface selected, the amount of training data, the quality of the seismic data, and the choice of method for estimating M.

[0175] Various examples are described herein, including examples of radial basis function (RBF) methods. An example of an RBF may be an RBF that includes features suitable for processing seismic traces. For example, consider an RBF that includes a heavy tail or a fat tail. A fat-tailed distribution may be defined as a probability distribution type that exhibits a large skewness or kurtosis relative to a normal distribution or an exponential distribution.

[0176] As an example, the RBF may include one or more decay terms. For example, consider quadratic decay. In such an example, the confidence may decrease as the distance from the distribution increases, and may converge toward a certain value (e.g., consider convergence toward 0.5, etc.). As an example, consider the following equation as a function of decay with dimension d (e.g., divided by (1+d 2 )):

[0177]

[0178] As an example, a radial basis function can include many skewed distributions for each class in a classification problem. However, 1D observations about Gaussian versus heavy-tailed density functions are applicable to more than one kernel function per class and multidimensional problems. With Gaussian functions, high levels of confidence can be assigned to outliers that are far from previously observed data. However, this is less likely for heavy-tailed density functions. Heavy-tailed radial basis density functions (RBFs or RBDFs) based on existing seismic data observations tend to be more robust than Gaussian functions to make very inaccurate predictions.

[0179] For tracking, the model M can be refined and improved by relearning based on tracking points and extracting new / extended training data based on both starting points and tracking points. This approach can be, for example, a continuous or step-wise process.

[0180] Figure 8 An example of a method 800 is shown that includes various blocks that may be implemented by a seismic interpretation system, which may be a computing framework. Figure 8 As shown, method 800 includes a receiving block 810 for receiving one or more selected points (e.g., by one or more interpreters, one or more machines, etc.), an extraction block 820 for extracting training data based on the one or more selected points, an execution block 830 for performing learning of digital distinctive features (e.g., seismic distinctive features in digital seismic data, which can be in the form of a model), an output block 840 for outputting a prediction having at least a corresponding digital distinctive feature confidence score, a tracking block 845 for tracking to expand the point set (e.g., locally and / or otherwise), a generation and output block 850 for generating and outputting a seismic cube (e.g., or other type of data structure) for generating and outputting confidence scores, a scanning detection block 860 for scanning one or more detection points above a score threshold (e.g., an adjustable confidence score threshold, etc.), and a presentation block 870 for presenting one or more detection points to a display, wherein the presentation can be based on the application of the score threshold (e.g., above, below, within, within its range, etc.).

[0181] As indicated in method 800, one parameter may be a score threshold. As an example, another parameter may be a window for extracting training data based on one or more selected points, for example. As yet another example, a parameter may relate to "negative" examples, which may ensure that the "negative" examples are sufficiently different from the "positive" examples (e.g., of the selected points). As an example, a method may be implemented in which both positive and negative points are selected as examples. As with respect to Figure 8As explained by the example of , method 800 can be implemented with few parameters and can be implemented in a way that the user does not have to set the parameters. For example, the window can be automatically determined by default or otherwise, the negative example can be automatically determined, and the score threshold can be automatically determined, or, for example, can be a graphical control on a graphical user interface presented to a display, and the user can adjust the value, range, etc. of the score threshold in the display. For example, consider a slider control in which a user can adjust a slider up or down to set a score threshold, where the computing framework automatically presents information to the display based on the movement of the slider. In such examples, the computer can present one or more surfaces, parts thereof, etc. to the display using points corresponding to confidence levels above or below the score threshold.

[0182] like Figure 8 As shown in the example of , execution block 830 may include performing learning using one or more techniques, which may include RBF, neural network (NN), tree and / or one or more other techniques. In this way, the learning block for learning based on digital distinctive features can apply one or more different machine learning techniques. As mentioned, RBF can be utilized, which may include one or more heavy-tailed RBFs, etc.

[0183] Various experiments utilized a method implemented via a seismic interpretation system, the method comprising receiving data from a seismic cube S and a set of points P as input, the set of points P representing locations where surfaces are located in the seismic data. Given the seismic data and points, the following method actions include:

[0184] 1. Extract training data

[0185] 2. Learn the model M (e.g., radial basis function model):

[0186] a. Output QC metrics about model accuracy

[0187] 3. Start tracking from P and use M to evaluate the data from S, where if you want to track:

[0188] a. Output a point set P* representing the tracked points, where P is a subset of it.

[0189] 4. Create the output volume S* by converting S to [0,1] data using M

[0190] 5. Detect the entire S* position using the score compared to a certain threshold calculated in action 2 above:

[0191] a. Create a point set P' indicating these locations

[0192] b.P' may come from a random subset of the traces to save time and space

[0193] Fig. 9 Exemplary graphical user interfaces (GUIs) 910 and 930 are shown for an exemplary extraction of training data based on surface points (see circles in GUI 910). GUI 910 shows picking points on a surface as represented in seismic data acquired for a geological region, wherein a window for extracting a trace profile is also shown, wherein such a profile may be referred to as a positive profile or positive example; however, GUI 930 shows a negative example, wherein a trace of another portion of the trace is extracted as in GUI 910, wherein the trace window covers the portion of the trace not covered in the window of GUI 910. As indicated, the center of the window of the negative example is located at a position above the center of the window of the positive example (e.g., earlier time, shallower depth, etc.). As an example, a method may include positive points and negative points, wherein, for example, the negative points may be referred to as avoidance points.

[0194] As an example, the extraction of training data can be performed by taking a sub-image around each point in P (e.g., seismic trace data around each point in P), which can be a positive example. In such an example, since the surface point may not fall exactly on the location of a voxel in a seismic cube or a pixel in a seismic profile (e.g., a seismic data array, a seismic image, a seismic volume, etc.), the amplitude value can be extracted with sub-sample accuracy using interpolation. However, negative examples can be extracted from different locations along a common trace with the positive examples in a manner similar to the positive examples. Therefore, a single trace can provide one or more negative examples and one or more positive examples. As an example, a negative example can be obtained from a trace that does not include a corresponding positive example; however, the method of obtaining negative examples and positive examples from a common trace can increase computational efficiency when the trace is stored in a memory of a storage device as a 1D data structure (e.g., amplitude versus time or depth). For example, consider a vector in which a vector can be loaded into memory when a point is selected and in which the vector can be a source of positive examples and one or more negative examples (e.g., as a portion of a vector accessed and loaded into RAM, etc.).

[0195] When learning a model M to make predictions, as mentioned, one or more approaches may be taken. Various experiments include implementing radial basis functions to compute predictions. This approach falls within the realm of instance-based learning, where K nearest neighbors is an example of an implementation. In nearest neighbor classification, one approach may include storing observed examples in a database and making predictions to find K nearest neighbors and generate predictions based on a weighted distance function. This approach may require considerable storage space and considerable computing resources to produce results.

[0196] Another approach may involve linear classification, where each class has one centroid, and where the prediction is a function of the distance to the class centroid. A radial basis function may include selecting class centroids based on the training data, where each class may have more than one centroid. Thus, a radial basis function may represent high-dimensional nonlinear manifolds that a linear classifier does not represent, while also being able to reduce storage and / or computational requirements, for example by having less data to store and compare than a nearest neighbor classifier.

[0197] As an example, the selected method for creating M may be a RBF method or optionally a random forest (tree) or an artificial neural network (NN) or another method (see, for example, Figure 8 ).

[0198] As an example, more than one method may be implemented. As an example, a method may include selecting a method for creating a model M, where the model M may be used to generate a confidence value, e.g., in the range [0,1] given a new example. This method may be implemented in single or multiple uses in one or more workflows. As mentioned, in tracking, if the match of the seismic representation of a candidate surface to the selected surface exceeds a threshold, the point set P may be gradually expanded to a point set P* by locally looking at nearby traces. Using a priority queue, a workflow may involve tracking the best matching new points first.

[0199] As mentioned, the workflow may include the use of confidence surfaces or confidence volumes based on machine learning outputs, such as visualization and / or quality control of one or more tracked surfaces. Given a "confidence attribute" that can be calculated for a tracked surface, a user can compare one or more methods for one or more selected points (e.g., positive examples) and optionally for one or more negative examples. If a surface has a consistent signal on one or more portions thereof, the visualization presented to the display may highlight it. If one or more portions (e.g., multiple portions) of a surface have one or more low confidence scores, this may indicate that such one or more portions in the seismic data may not follow a geological surface. As an example, a confidence attribute may be stored as a confidence attribute cube (e.g., volume), a confidence attribute portion (e.g., slice), or in another form. As an example, a confidence attribute may be considered to be a seismic attribute of a specific type that can be used for one or more purposes.

[0200] As an example, if a tracker performs unacceptably in tracking within one or more regions due to similar seismic events nearby (e.g., reflections, etc.), the method may include inputting an additional set of points representing one or more surfaces that are not being tracked. Such a set of points may then be given additional emphasis in learning, e.g., as part of a negative example, to reduce the risk of making such a mistake again in tracking. Such a set of points may have confidence scores altered so that the additional points receive a lower score than if they were not given additional emphasis.

[0201] Fig.10 An example of method 1000 is shown, which includes selecting positive and negative examples and using machine learning to generate a model that can be used to analyze seismic data and to make predictions within such seismic data about whether portions of the seismic data correspond to geological features. Although various examples involve selecting (e.g., picking and evaluating) surfaces, as an example, features can include seismic survey features such as artifacts (such as acquisition geometry artifacts). As an example, a seismic interpretation system can include an artifact (e.g., a type of noise) identifier and a surface identifier. In such examples, a user can pick points representing artifacts and pick points representing surfaces. This method can help reduce one or more artifacts and enhance one or more surfaces.

[0202] same, Fig.10 The exemplary method 1000 of FIG. 1 shows extracting training data from earthquakes as input points, where positive examples are extracted as sub-images around the input points and negative examples are extracted as sub-images elsewhere. These examples can be used for machine learning to obtain a prediction model. For example, a workflow can include the selection of points, which can be of various types (e.g., positive, negative, etc.) and generating a prediction model as a machine model via machine learning.

[0203] Fig.11 An example of a method 1100 for radial basis function (RBF) learning is shown. As shown, negative examples and positive examples can be compressed into smaller sets of positive prototype patterns and negative prototype patterns. Given a prototype pattern, the confidence of each class (negative or positive) can be a multimodal relative density function (MRDF) composed of individual density functions around each prototype pattern. The prototype density can be based on a distance function between an example and a prototype pattern. Although distance is mentioned, dimensions can generally be used, where such dimensions can be distance, time, or other metrics.

[0204] Fig.12An example of a graphical user interface (GUI) 1200 including a seismic profile is shown with four picked points in white. As examples, the picked points may be points picked by one or more users via visual interpretation of seismic data presented to a display, and / or may be points picked by one or more machines via execution of an algorithm that automatically selects a plurality of points as representative of structural features to be tracked.

[0205] Fig.13 An example of a graphical user interface (GUI) 1300 including displayed confidence levels based on four picked points is shown. In the table at the top left, the confidence levels are indicated, which may be colored to make high confidence information stand out from dark or grayscale information. Fig.13 In the example above, white is used to indicate a high confidence level.

[0206] Fig.14 An example of a graphical user interface (GUI) 1400 is shown that shows tracking points starting from four picked points and guided by a confidence volume (eg, a confidence score volume). Fig.14 In the example of , the tracking stops at the fault or the tracking crosses the fault, thus following the confidence score. Fig.14 In an example of a fault, a fault is a discontinuity that can be "tracked across" relative to a surface. As an example, a method may include parameters such as a "jump" parameter that can encounter a discontinuity and identify a jump to the surface on the other side of the discontinuity, make a jump, etc. In such examples, a jump may be specified as an "up" jump, a "down" jump, or an up-down jump. For example, where information is available about how a layer may have shifted during a fault, a method may implement a jump method that jumps up or down based on such information. This method may speed up "jumping" across a discontinuity and improve tracking.

[0207] As an example, a method may include performing an analysis on seismic data in an iterative manner that starts from a selected point in an adjacent manner, or a method may include analyzing seismic data in a manner that does not follow an adjacent method but is intended to analyze a region (possibly an entire region). For example, a method may implement a continuity parameter so that the surface will have continuity, optionally with jumps that account for interruptions. Or, for example, a method may be implemented without considering continuity, and the method may evaluate seismic data in a region, which may be an entire seismic cube. In the latter case, processing may optionally be performed in parallel to, for example, calculate confidence information for points in the seismic cube voxel by voxel to generate a confidence cube. Such a confidence cube may then be presented to a display where a user may interact with a GUI to adjust a confidence threshold (e.g., or range), and / or where a user may implement one or more tracking routines that account for neighborhoods, adjacencies, etc. to create one or more continuous structures (e.g., surfaces or portions of surfaces). As an example, this method may be multi-directional and / or multi-seed, and may be implemented optionally with serial processing, parallel processing, or a combination of serial and parallel processing. With respect to tracking, an example of tracking is referred to as ant tracking. Ant tracking can be an automatic tracking or semi-automatic tracking that can be intended to facilitate the interpretation of multi-dimensional seismic data. However, as mentioned, tracking may encounter errors in the event of faults and / or other stratigraphic changes. With respect to the calculation of confidence cubes, this approach can help identify one or more surfaces and / or portions of a common surface without encountering problems that ant tracking methods may encounter at faults or other types of interruptions. In the confidence cube approach, a user can view a presentation and then select portions of a cube based at least in part on confidence values, and then such portions are positively assigned to a common surface that may be discontinuous on one or more types of interruptions in a geological region imaged by reflection seismology, and optionally supplemented via one or more other data acquisition techniques (e.g., cable, MWD, LWD, imaging, etc.).

[0208] exist Fig.14 In the exemplary GUI 1400 of FIG. 14 , voxels can be seen within a seismic cube or a confidence cube, and the voxels can be assigned colors corresponding to confidence values ​​calculated based at least in part on an implementation of a radial basis function method.

[0209] Fig.15 An example of a graphical user interface (GUI) 1500 is shown in which tracking has been implemented such that tracking points are guided by confidence scores, as displayed by the extracted confidences.

[0210] Fig.16An example of a graphical user interface (GUI) 1600 including a seismic profile is shown, wherein one starting point is a white filled circle which is Fig.12 , Fig.13 , Fig.14 and Fig.15 The previous examples are different surfaces.

[0211] Fig.17 An example of a graphical user interface (GUI) 1700 is shown, which includes using the confidence score as a guide to Fig.16 A starting point of is presented in white to present the tracking results. In GUI 1700, the confidence score is displayed on the point. Again, for this experiment, note how the tracking stops at the fault.

[0212] Fig.18 An example of a graphical user interface (GUI) 1800 is shown. Fig.18 In the example of , the surface is tracked from a starting point represented as a white filled circle, following the amplitude peaks, without using confidence scores for guidance. As an example, a method may include using amplitude peaks and / or confidence scores.

[0213] Fig.19 An example of a graphical user interface (GUI) 1900 including a tracking surface with confidence scores is shown. Fig.19 In the example of , the confidence score reveals one area with high confidence, but a larger area with low confidence. For the interpreter, such a GUI 1900 with its presentation of information generated by the seismic interpretation system may be a signal to warrant a closer inspection, possibly with a reinterpretation. As an example, the confidence score may be used as a metric to consider the quality of the tracking. For example, consider ant tracking, where one may implement a method that may include Figure 8 The method of various actions of ant tracking can be used to evaluate the quality of ant tracking. As an example, the tracking algorithm can be implemented in a way to determine confidence information (e.g., based on RBF, etc.) so that the algorithm can be evaluated and / or guided.

[0214] Fig. 20 An example of a graphical user interface (GUI) 2000 is shown that displays confidence scores on a cross section, revealing where a surface may be re-tracked because of an indicated error, and where re-tracking may reduce the error.

[0215] Fig.21An example of a method 2100 including machine learning using a neural network is shown. In such a method, instead of learning prototype patterns, as with radial basis functions, class information is captured in the neural network weights. For example, a trained network can generate and output a value in, for example, the range [0, 1] from an input image (e.g., a seismic data image), which can be viewed as a neural network confidence score.

[0216] Fig. 22 An example of a graphical user interface (GUI) 2200 is shown, which includes a presentation using a machine learning prediction model as a detector, thereby returning the highest confidence trace by trace, and when the confidence score is higher than a pre-calculated threshold (e.g., predetermined, etc.), the result gives a dense set of points at the surface. Regarding false positives, these can be regarded as points away from the main surface (e.g., the desired surface to be explained). These points can be processed by changing the confidence score threshold or, for example, by geometric filtering. Regarding geometric filtering, this method can utilize a neighborhood exclusion method, a dual trace identification exclusion method, a region selection method, etc. Regarding neighborhood exclusion, extreme points within a depth or time range can be used to draw a boundary around the identified surface, where points outside (above and / or below) the depth or time range can be selected for exclusion (filtering out). Regarding the dual trace identification exclusion method, a criterion can be specified so that the trace is to include a single instance of the surface so that a single trace does not produce multiple instances of the desired surface. In such methods, neighboring criteria can be utilized to ensure that outliers are excluded rather than the desired surface. With respect to the region selection method, a user may interact with the GUI 2100 to select points in space that form a region in which surface identifiers within the region are retained and those outside the region are excluded.

[0217] Regarding various types of frameworks including features of networks (e.g., kernel-based neural networks, etc.), consider the TENSORFLOW framework (Google LLC, Mountain View, California), which is an open source software library for data flow programming, which includes a symbolic mathematics library that can be implemented for machine learning applications that may include neural networks and kernel methods (e.g., with linear models). As an example, the CAFFE framework can be implemented, which is a deep learning framework developed by the Berkeley AI Research (BAIR) (University of California, Berkeley, California). As an example, consider the SCIKIT-LEARN library or platform (e.g., scikit-learn), which utilizes the PYTHON programming language (e.g., with features for neural networks, kernel methods, etc.). As an example, a framework such as the APOLLO AI framework (APOLLO.AI GmbH, Germany) can be utilized.

[0218] Fig.23 , Fig.24 , Fig.25 and Fig.26 Examples of methods 2300, 2400, 2500, and 2600 utilizing various techniques are shown, which are illustrated via graphs. Fig.23 As shown, method 2300 uses a Gaussian function as the RBF, which provides 2 classes and steep boundaries, where the flat top of class 1 does not indicate the center of the class 1 Gaussian distribution. Fig.24 As shown, method 2400 uses a heavy-tailed quadratic decay as the RBF, which provides 2 classes and smooth boundaries, where the top of class 1 indicates the center of the distribution of class 1 (see also the bottom of class 2). Fig.25 As shown, on a large scale, the method 2500 with a Gaussian function as the RBF provides a flat top that is continuous along the dimension x. Fig.26 As shown, on a large scale, the method 2600 with heavy-tail decay as the RBF provides a decreasing confidence with increasing distance (dimension) from the Class 1 distribution, where this decreasing confidence can be reduced to a certain value (e.g., at Fig.26 converges to a value of about 0.5 in the example).

[0219] As mentioned, one approach can utilize a nearest neighbor method, such as a K-nearest neighbor method. An example of an instance-based method is the K-nearest neighbor algorithm, which assumes that an instance corresponds to a point R in n-dimensional space. n As an example, the neighbors of an instance can be defined according to standard Euclidean distances (e.g., metric, dimension, etc.). More precisely, any instance x can be described by a feature vector:

[0220] <a1(x),a2(x),...a n (x)

[0221] where a r (x) represents the value of the rth attribute of instance x. In this method, two instances x i and x j The distance between them can be defined as d(x i , x j ),in:

[0222]

[0223] As explained, a radial basis function (RBF) is an approximation function. As an example, an RBF can be represented by a linear combination of many local kernel functions. As an example, the learned hypothesis can be a function of the following form:

[0224]

[0225] Where each x u is an instance from X, and the kernel function K can be defined u (d(x u , x)), so that it increases with the distance d(x u , x) increases and decreases. Above, k can be a user-provided constant that specifies a number of kernel functions to include, or it can be a machine-provided number that can be customized for a particular seismic dataset (e.g., region of the Earth, acquisition technique, etc.).

[0226] Although is a global approximation to f(x), but from K u (d(x u The contribution of each term in the term , x) can be localized to the point x u As an example, a method may include transforming each function K u (d(x u , x)) is selected as the point x u A Gaussian function centered at , with a certain amount of variance, is expressed as As mentioned, one or more other types of methods may be utilized, such as heavy-tailed methods. An example of a Gaussian kernel function is as follows:

[0227]

[0228] In the SCIKIT-LEARN library written for the PYTHON implementation, you can call the function rbf_kernel. The function rbf_kernel calculates the RBF kernel between two vectors, where the kernel is defined as:

[0229]

[0230] where x and y are the input vectors. As explained, the variance can be used, for example, consider setting γ equal to σ -2 , so that the kernel is variance σ 2 Gaussian kernel.

[0231] In SCIKIT-LEARN, the rbf_kernel function is specified as follows:

[0232] sklearn.metrics.pairwise.rbf_kernel(X,Y=None,gamma=None)

[0233] The parameters are:

[0234] X : array of shape (n_samples_X, n_features)

[0235] Y : array of shape (n_samples_Y, n_features)

[0236] gamma : float, default is None

[0237] If None, it defaults to 1.0 / n_features

[0238] where the call to the function returns:

[0239] kernel_matrix : array of shape (n_samples_X, n_samples_Y)

[0240] The SCIKIT-LEARN library also provides the following class: sklearn.gaussian_process.kernels with RBF(length_scale=1.0,length_scale_bounds=(1e-05,100000.0)). RBF() provides the kernel given by:

[0241] k(x_i,x_j)=exp(-1 / 2d(x_i / length_scale,x_j / length_scale)^2)

[0242] The parameters are:

[0243] length_scale : float element or array with shape (n_features), default: 1.0; length scale of the kernel, where if float element, an isotropic kernel is used, and if array, an anisotropic kernel is used, where each dimension of l defines the length scale of the corresponding feature dimension; and

[0244] length_scale_bounds: Pair of float elements >= 0, default: (1e-5, 1e5), lower and upper bounds for length_scale.

[0245] The SCIKIT-LEARN library provides the definition of custom kernels, which can be passed in methods as parameters, etc. The SCIKIT-LEARN library also includes RBFSampler, which provides an approximation to the RBF kernel using random Fourier features. Specifically, RBFSampler approximates the feature map of the RBF kernel by a Monte Carlo approximation of its Fourier transform.

[0246] As explained, the RBF kernel includes a square term in the form of an exponential, which may be referred to as a "squared exponential" kernel. In the PYTHON programming language, using the SCIKIT-LEARN library, for example, the kernel may be set as follows: kernel = 1.0*RBF([1.0]), or kernel = 1.0*RBF([1.0, 1.0]), to provide an isotropic kernel classification and an anisotropic kernel classification, respectively. With respect to the anisotropic kernel, different length scales may be assigned to different feature dimensions.

[0247] As an example, the method can be implemented using a library such as the SCIKIT-LEARN library. As an example, the method can be implemented using a programming language such as the PYTHON programming language. As an example, one or more features in a library such as the SCIKIT-LEARN library can be used to train a machine model and / or utilize a trained machine model (e.g., RBF, neural network, etc.).

[0248] Fig. 27 An example of a method 2700 including a network that includes features associated with the RBF method is shown. With respect to the method 2700, the function can be viewed as describing a two-layer network, where the first layer units calculate various K u (d(x u , x)), and the second layer computes linear combinations of these first layer unit values. Although a two-layer network is shown, networks with more than two layers can be utilized.

[0249] exist Fig. 27 In the example, the network associated with method 2700 may be an RBF network, wherein each hidden unit generates a signal generated by a certain instance x. u The activation is determined by a Gaussian function centered at y (e.g., or another function or functions). In such examples, the activation can be close to zero (e.g., or other null value) unless the input x is close to x u .exist Fig. 27 In the exemplary method 2700, the output unit produces a linear combination of the hidden unit activations.

[0250] As an example, given a set of training examples, an RBF network can be trained in a two-stage process. In such an example, first, the number of hidden units k can be determined, where each hidden unit u defines its kernel function K by choosing u (d(x u , x)) u and to define; secondly, the weights w can be trained, for example, using a global error criterion u To maximize the network's fit to the training data.

[0251] As an example, a method may include minimizing the squared error over k nearest neighbors as follows:

[0252]

[0253] Since the kernel function can be kept fixed during the second stage, the linear weight values ​​w can be trained u , which can improve the efficiency of the method.

[0254] When the number of training examples may be large, a method for selecting an appropriate number of hidden units or kernel functions may be to select a set of kernel functions that is smaller than the number of training examples. As an example, a set of kernel functions may be distributed so that the centers are evenly spaced in the instance space X.

[0255] As an example, the network can be viewed as a smooth linear combination of many local approximations to the target function. The benefit of RBF networks can be efficient training, because RBF networks can be trained more efficiently than feedforward networks trained with back-propagation. The efficiency comes from the fact that the input and output layers of the RBF are trained separately.

[0256] With respect to the neuronal approach, the biological definition of a neuron may be a cell, such as in the brain, whose primary function is the collection, processing and dissemination of electrical signals.

[0257] Fig.28 An example of a method 2800 is shown that may utilize a neural network that includes neurons, such as exemplary neuron 2850. With respect to neuron 2850, it is shown as a schematic diagram in a mathematical sense, where the neuron includes an input function, an activation function, and an output; it is noted that a neural network may include multiple such neurons.

[0258] Regarding the mathematical description of a neuron 2850 (e.g., as a unit), the output activation of the unit can be defined as where a j is the output activation of unit j, and w j，i is the weight on the link from unit j to this unit.

[0259] A neural network can be defined as a set of machine learning algorithms (e.g., logistic regression) that are combined to simulate biological neural activity. As an example, a unit can be trained to provide a trained machine model suitable for performing one or more tasks. Types of neural networks include acyclic or feedforward networks and cyclic or recursive networks. The difference between a feedforward network and a recursive network is that a feedforward network represents a function of its current input, while a recursive network feeds its output back into its own input. This means that the activation levels of the neural network form a dynamic system that can reach a stable state or exhibit oscillatory or even chaotic behavior. Recursive neural networks can support short-term memory, which can be used for machine learning, such as the ability to remember inputs and learn from them.

[0260] about Fig.28 Method 2800, which may include utilizing a multi-layer feedforward neural network. As an example, the input component may take a value varying from 0 to 1, and for example, neurons without input may output the same value, which may be used for a shift in the activation function, which is referred to as a "bias" (e.g., see the bias weight of neuron 2850).

[0261] As an example, the neural network can be a fully connected neural network having three or more hidden layers, where each hidden layer can include, for example, at least about 10 units. For example, consider a fully connected neural network having an input layer, three hidden layers, and an output layer, where each of the hidden layers includes at least about 10 units (e.g., neurons). Such a neural network can be trained (e.g., via weight assignment, etc.) to predict confidence information (e.g., confidence scores, etc.) of seismic data (e.g., points, pixels, voxels, etc.) associated with structural features that can be represented as events in the seismic data (e.g., "wiggles" in the seismic data).

[0262] With respect to training, a method may include augmenting the data. For example, one or more methods may be employed to generate more data for training, where the data may be based on a smaller set of real data and / or synthetic data (e.g., as may be generated using an earth model, etc.). As an example, the amount of training data may be less for a kernel-based method than for a neural network-based method. As an example, a computational framework may utilize a kernel-based method by default, particularly where a user has access to a limited amount of training data. In such examples, there may be an option to utilize a neural network-based method, for example, where a user has access to an appropriate amount of training data (e.g., real, augmented real, synthetic, etc.).

[0263] As explained with respect to neuron 2850, a weight may be assigned to a particular input having a real value, where the value may indicate a relevance. Each input may be multiplied by its weight. As explained with respect to neuron 2850, an activation function may be a mathematical function for mapping a weighted input to an output. As an example, the output value may be relative to a scale such as from 0 to 1, from -1 to +1, from 0 to 100, etc. When the last component is used for output, it receives one or more results from one or more activation functions.

[0264] like Fig.28 As shown, the layered structure of a neural network takes one or more components as input and produces one or more results from the output. The layers between the input and output are called hidden layers because they can be hidden (e.g., not directly observed) when the neural network is run. By adding hidden layers, the hypothesis space that the network can represent can be expanded.

[0265] As an example, a neural network may include an output indicating whether a point is likely to be a point of a structural feature, which may be a structural feature to be tracked. As an example, in the case of a multi-event approach, a neural network may output multiple indications that may match a point to an event. As an example, an output node of a neural network may be a value or may be a binary (e.g., yes or no) type of output. As an example, a neural network may include one or more output nodes representing a value such as a confidence value (e.g., considering a range from 0 to 1 or from 0 to 100). In a single event approach, as an example, a node corresponding to a confidence value may be activated, or as an example, a node may have a numerical value corresponding to a confidence value. As an example, a neural network may be trained and have an architecture that provides one or more types of outputs for analyzing seismic data, which may be in digital form (e.g., vectors as traces, 2D arrays as images / slices, 3D arrays as volumes, etc.).

[0266] Precision and recall can be based on the understanding and measurement of relevance. Precision (positive predictive value) is the fraction of relevant instances among the retrieved instances, while recall (sensitivity) is the fraction of relevant instances that have been retrieved over the total amount of relevant instances. Precision can provide an indication of the usefulness of the search results, while recall can provide an indication of the completeness of the results. A high precision can produce an algorithm that returns results that are significantly more relevant than irrelevant results. A high recall means that the algorithm returns most relevant results.

[0267] As an example, these results can be combined into a single metric called the F-measure. The F-measure is a weighted harmonic mean of precision and recall (or the Matthews correlation coefficient, which is the geometric mean of chance-corrected variation). The traditional F-measure or balanced F-score can be given as:

[0268]

[0269] The F-measure is approximately the average between precision and recall (when they are close).

[0270] Exemplary methods may be based on one or more of a variety of techniques within machine learning. For example, a method may include one or more of a radial basis function (RBF) and a neural network (NN) as options.

[0271] Fig.29 An example of a method 2900 that can utilize a graphical user interface 2910 presented to a display is shown, wherein the GUI 2910 can be part of a computing framework. For example, consider a processor executable instruction that can be executed via a PETREL framework during a seismic interpretation workflow. Such instructions can provide local and / or remote presentation. For example, in a cloud architecture, a computing framework can be executed using cloud-based resources, wherein a user interacts with such resources using a client device (e.g., a tablet computer, a laptop computer, a workstation, etc.). In such examples, the client device can be operatively coupled to a network for receiving information and transmitting information to the computing framework, for example, via a browser application. In such examples, the interaction can utilize one or more types of formats, such as a JAVASCRIPT OBJECT NOTATION format (JSON).

[0272] JSON is a lightweight data exchange format based on a subset of the JAVASCRTIP programming language (Standard ECMA-262 Third Edition, December 1999). JSON is a text format that is language-independent but uses conventions familiar to programmers of the C family of languages ​​(including C, C++, C#, JAVA, JAVASCRIPT, PERL, PYTHON, etc.). Such properties make JSON a useful data exchange language.

[0273] JSON is built on two structures: collections of name / value pairs (implemented, for example, as objects, records, structures, dictionaries, hash tables, keyed lists, or associative arrays); and ordered lists of values ​​(implemented, for example, as arrays, vectors, lists, or sequences).

[0274] As an example, the method can be implemented at least in part as a microservice. As an example, a microservice can utilize JSON format to transmit both requests and responses. In such examples, data can be optionally sent using REST, where information is converted to JSON format. As an example, the method can be implemented using DELFI environment, PETREL framework, etc. As an example, the method can be implemented using one or more application programming interfaces (APIs). For example, consider a method utilizing an API-based microservice, where resources can be far away from the user's device. In such examples, the device can include a browser application that can transmit information as one or more API calls to a remote computing system (e.g., cloud-based, etc.) and receive information generated by the remote computing system in response, the remote computing system can include one or more machine models that are trained, can be trained, etc. For example, consider a user watching a display of a client device, where seismic data is presented as a seismic image. In such examples, a user can interact with a GUI to transmit an API call associated with a microservice. Such a call can be to train a machine model using selected points representing structural features in seismic data (e.g., from a survey of a geological region). In response, the client device may receive an indication from the microservice that the trained machine model is ready, for example, for analyzing at least a portion of the seismic data to generate a result (e.g., a confidence result, a tracking result, a confidence and tracking result, etc.).

[0275] Fig.29 Method 2900 can be a workflow generation method that interacts with GUI 2910 to generate a workflow that can be executed to, for example, use one or more machine learning techniques to help interpret seismic data. As an example, one or more features of GUI 2910 can be associated with one or more microservices (e.g., provided via one or more microservices). As shown, a user can provide a name and description of the workflow to be generated. GUI 2910 can include features for selecting various types of utilities, operations (Ops), and processes. As shown, the MLTrackerWL operation is selected, where a panel of GUI 2910 specifies how the MLTrackerWF will execute, for example, to track structural features in seismic data.

[0276] like Fig.29As shown in the example of , GUI 2910 includes a cube field for selecting a seismic volume (e.g., a seismic cube), which may be stored locally and / or remotely; note that there may be a seismic slice (e.g., a 2D dataset) option for selecting a seismic slice. GUI 2910 also includes a point field for entering a plurality of points and a machine learning technique option, which may be a default RBF or a neural network selected via the use of a graphical control. As mentioned, this default kernel-based approach may be beneficial to users where the user has access to a limited amount of training data in the event that a kernel model is to be trained (e.g., as part of a workflow). As mentioned, in the event that training is to be performed and in the event that the user has access to a sufficient amount of training data, the user may select the "UseNN" option. As an example, in the event that a trained neural network is available, the user may select the "UseNN" option for a workflow. GUI 2910 also includes graphical controls for points to avoid, as well as for negative examples (e.g., see Fig.10 , Fig.11 and Fig.21 ) field of the points to be avoided for the destination.

[0277] As an example, GUI 2910 may include options for adjusting one or more machine learning techniques. For example, consider various types of kernels that may be available and / or customizable (e.g., via SCIKIT-LEARN libraries, etc.). GUI 2910 may provide one or more graphical controls that enable selection, adjustment, etc. of one or more kernels. For example, consider selecting a Gaussian kernel, a non-Gaussian kernel, adjusting tail parameters, adjusting attenuation parameters, adjusting length parameters, adjusting isotropic parameters, adjusting anisotropic parameters, etc., which may be performed using knowledge of seismic surveys and / or knowledge of one or more structural features in a region.

[0278] As an example, a user can generate a workflow via interaction with a GUI to process tracking of the seabed as the first event in a series of traces from a seismic survey. In such an example, a user can execute a workflow to track the seabed as a structural feature. Once the seabed is fully characterized via the workflow, the user can proceed to another structural feature, for example, as the second event in a series of traces. This method can be referred to as an iterative stratigraphic method, in which a workflow can be designed for a specific event in a series of stratigraphic events (e.g., layers, etc.). This method can optionally customize the kernel in a manner that is most suitable for a specific event, recognizing that events may exhibit differences. For example, an event below the seabed may be different from an event on the seabed (e.g., a trace "swing" of an event below the seabed may be different from a trace "swing" of an event on the seabed).

[0279] As an example, a workflow may involve the selection of multiple events, e.g., consider selecting five events, where the workflow starts with the strongest event of the five events or starts with the highest event of the five events. In such methods, the workflow may analyze the seismic volume and determine which traces indicate "strong" events (e.g., amplitude, gradient, minimum, maximum, etc. of the trace data). As an example, for the seafloor, it may be the first strong event of the trace starting from the top of the trace. The workflow may optionally be automated starting from the seafloor in a manner that does not involve manual input, as the seafloor (reflector) may be determined in a machine-based approach.

[0280] As an example, GUI 2910 may include options for processing multiple events in parallel and / or serially. For example, consider a method for processing two events simultaneously, wherein the two events may be separated by a material that is a relatively homogeneous layer. For example, a first event in the event may be an upper interface between the two layers, and a second event in the event may be a lower interface between the two layers, wherein the relatively homogeneous layer is disposed between the two interfaces. In such examples, the thickness of the layer may provide a time and / or distance interval between the two events. As an example, one or more machine models may utilize a thickness that may be an approximate thickness that optionally has a minimum and a maximum value. This approach may be used for quality control and / or otherwise improve the quality of tracking of two events. As an example, in an iterative approach, wherein a structural feature as an event has been tracked to provide a surface (e.g., an interface), this information may be utilized in subsequent iterations to facilitate tracking of lower events of another surface. Information about layers, such as information about one or more cores, may be utilized. For example, consider utilizing an anisotropic kernel, where the lateral and vertical dimensions can be customized based on knowledge of the thickness of a layer with upper and lower events, where the workflow is intended to track the upper and / or lower events as different events. In cases where the layer thickness is relatively small, the vertical dimension of the kernel can be reduced, and in cases where the layer thickness is relatively large, the vertical dimension of the kernel can be increased (e.g., or an isotropic kernel can be utilized, etc.). As an example, a method can include manually and / or automatically selecting and / or adjusting a kernel to improve tracking of events present in seismic data (e.g., seismic traces). As an example, a GUI can provide interaction with a computational framework, one or more libraries, etc., to provide flexibility in kernel selection, adjustment, etc.

[0281] As an example, a method of performing simultaneous tracking of multiple structural features (e.g., multiple events) may utilize a kernel-based approach because the computational requirements (e.g., training requirements, etc.) may be less than the computational requirements of a neural network-based approach. In such an example, multiple machine models may be trained and utilized, where each of the multiple machine models is trained to a specific structural feature (e.g., a specific event).

[0282] As an example, a tracking method may be referred to as an extension method because the method may track via an extension of points. As mentioned, a method may calculate a confidence score to generate, for example, a confidence score slice, a confidence score volume, etc. as a type of seismic attribute. Such seismic attributes are based at least in part on the things on which the measurement confidence is based. For example, where one or more data points are provided as an indication of a certain type of feature, a confidence score may be calculated relative to the feature. As mentioned, a feature may be a structural feature (e.g., ground, horizon, conformity, unconformity, geological body, etc.).

[0283] exist Fig.29 In an example, GUI 2910 includes an automatic generation of a workflow graphic control that can be initiated via user interaction so that a specified workflow is automatically generated, which can then be executed (e.g., automatically and / or via user interaction). As an example, GUI 2910 can provide a batch mode in which a series of workflows can be generated and then executed serially and / or in parallel without user interaction. As an example, during batch execution, a graphical user interface can present the results to a display in which a user can optionally visualize tracking in real time. In such examples, the GUI can provide operations such as pausing, terminating, adjusting, etc., of a workflow based on a user's evaluation of the presented results.

[0284] Fig.30 An example of a method 3000 that can utilize a GUI 3010 is shown. As shown, the GUI 3010 specifies various aspects of an MLTrackerWF workflow (WF). The parameter name column includes Cube, Point, UseNN, UseAvoidPoints, and Avoid, which are some examples. The GUI 3010 includes an Apply control, an OK control, and a Cancel control. The OK control can be activated to cause the specified workflow to be generated and the GUI 3010 to be closed, as well as the workflow to be executed and the results to be presented to a display.

[0285] Fig.31 An example of a method 3100 is shown, which involves a kernel-based method, such as an RBF-based method. Fig.31It is shown how to process seed points (e.g., pick points or otherwise selected points) for seeding methods, etc. As shown, ellipses are used as data samples to represent traces, and the data samples can be raw or processed seismic data samples. Decisions can be used to determine whether the trace presents a maximum or minimum value at a specific time or depth or other dimension (e.g., point). In the case where the decision is "yes", method 3100 can continue to a calculation block for calculating a confidence score. The confidence score can then be compared to a threshold, where if the confidence score is greater than (e.g., or greater than or equal to) the threshold, method 3100 can proceed to a positioning block for positioning a point in a priority queue (e.g., as the highest value, etc.). Method 3100 can then continue appropriately in a loop considering other points.

[0286] exist Fig.31 In the example of , the seed point can be compared with surrounding seeds in the interval [-5,5], where if the seed is a maximum or minimum value, the seed will calculate a confidence score to determine whether it is greater than a predetermined value (e.g., 0.5, etc.). And, if it is greater than the predetermined value, method 3100 can place the seed in a priority queue, and method 3100 can continue to the next seed (e.g., in the case where there are multiple seeds, multiple seeds are created, etc.).

[0287] Fig.31 Also shown is an example of an expansion process that can be performed as part of method 3100. As shown from left to right, a placed seed point with its potential subpoints can be defined, where one of the subpoints becomes a first expansion point (red) and generates a definition of its surrounding subpoints. Similarly, a second expansion can be performed with another of the potential subpoints to define a corresponding subpoint. Fig.31 In the example of , the L-shaped box area identifies the seed points that can be placed in the queue for evaluation.

[0288] Fig.32 An example of a method 3200 is shown that includes output of confidence values ​​as a result of a predictive model (e.g., generating digital distinctive features as a result of training a trained model, etc.). As mentioned, the predictive model can be a kernel-based model or a neural network-based model. As an example, a method can generate confidence trace samples, confidence pixels, confidence voxels, etc. For example, consider a method that generates a confidence trace for an input trace, generates an array of confidence pixels for a seismic image (e.g., a seismic slice), and generates an array of confidence voxels for a seismic volume.

[0289] As an example, a method may include one or more inputs of selecting data, which may be seismic cubes and points to be received and / or accessed, which may be data structures stored in a memory, the data structures being generated via interaction with a graphical user interface that presents a visualization including pixel values ​​to a display based on seismic data (e.g., digital seismic data such as acquired for a geological region including various subsurface structures (e.g., reflectors, interfaces, horizons, etc.). The method may include generating a confidence value (e.g., a confidence score) using a machine model, which may be a predictive model.

[0290] Fig.33 An example of method 3300 is shown in the form of example pseudo code. Method 3300 may include Fig.31 For example, consider using a trained machine model to evaluate a point. Fig.33 As shown, method 3300 may include learning a model (e.g., model learning or model training) using input points and surrounding data as positive and negative examples, and evaluating points in a window (e.g., a vertical window) using the learned model (e.g., a trained model), where such points may represent a minimum or maximum value. Fig.31 In the example of , the window is shown as +5 and -5 from the selected point, where positive and negative values ​​can represent samples, pixels, voxels, etc. having corresponding values ​​(e.g., amplitude values, etc.). Fig.33 As shown, a model and a threshold (e.g., a predetermined confidence value) can be used to evaluate a point (e.g., a minimum / maximum value). For example, the model evaluation can be a model prediction, where the prediction is a confidence value associated with a structural feature. In such methods, a higher confidence value means that the evaluated point is more likely to be associated with the structural feature. Fig.32 As shown, a predictive model (e.g., a trained machine model) can be used to generate values ​​for seismic data points (e.g., samples, pixels, voxels, etc.), where the values ​​are associated with structural features that can be represented as events in the seismic data.

[0291] As explained, a trained machine model can be trained for specific structural features represented by events in seismic traces acquired by seismic survey equipment. Such a trained machine model can output confidence information that can be directly related to whether a point is likely to be a point representing a structural feature.

[0292] The interpretation results of the acquired data can be used to construct a model that is a geological unit model (e.g., a grid unit model, a node model, etc.). For example, consider a geological unit model that includes units with a length and width size of several meters by several meters and a height of about 0.1 meters to several meters. As an example, the geological unit model may include more than 10,000 units, more than 100,000 units, more than 1,000,000 units, etc. As an example, the geological unit model can be grouped into layers, where the layers can be grouped into stratigraphic intervals. Structural features can be classified as objects, which can be one-dimensional or multi-dimensional (e.g., 1D, 2D, 2.5D, or 3D).

[0293] Fig.34 An example of a method 3400 is shown, which includes: a receiving block 3410 for receiving one (or more) selected points of digital seismic data of a geological region (e.g., selected points from selection using a visual presentation of digital seismic data based on the geological region, selected points via an automated structural feature identification technique, etc.); an extraction block 3420 for extracting training data based on the one (or more) selected points; an execution block 3430 for performing machine learning of a machine model based on the training data to generate a trained machine model; an analysis block 3440 for analyzing at least a portion of the digital seismic data using the trained machine model to generate a result; and an output block 3450 for outputting the result. The method 3400 may include an execution block 3460 for performing one or more actions based at least in part on the result, wherein the one or more actions include issuing a signal to a device.

[0294] like Fig.34 As shown in the example of , method 3432 may start with an access block 3436 for accessing a trained machine model. For example, method 3432 may include: according to access block 3436, accessing a trained machine model that is trained to analyze digital seismic data of a geological region relative to structural features of the region; according to analysis block 3440, analyzing at least a portion of the digital seismic data using the trained machine model to generate a result; and according to output block 3450, outputting the result as an indicator of the spatial location of the structural feature of the geological region. Method 3432 may include an execution block 3460 for performing one or more actions based at least in part on the result, wherein the one or more actions include issuing a signal to a device.

[0295] As an example, a method may include animation. For example, consider performing an Fig.34The method of 3400, wherein the method iteratively generates information, wherein a visualization can be presented in one or more iterations so that a surface can be seen to extend from a selected point. In such examples, a GUI can be presented that includes a visualization that can be interacted with the GUI to, for example, avoid one or more areas, select one or more other points, etc.

[0296] As an example, one or more computer-readable storage media may include computer-executable instructions executable to instruct a computing system to perform a method, such as the methods described herein.

[0297] exist Fig.34 In the example of , method 3400 is shown relative to computer readable storage media blocks (CRM blocks) 3411, 3421, 3431, 3441, 3451 and 3461; and method 3432 is shown relative to CRM blocks 3437, 3441, 3451 and 3461. Such blocks can be part of a system such as a seismic interpretation system. Such blocks can include instructions that can be executed by one or more processors, which can be one or more processors of a computing framework, system, computer, etc. The computer readable medium can be a non-signal, non-carrier and non-transitory computer readable storage medium. For example, a computer readable medium can be a physical memory component that can store information in a digital format. In the case where the CRM is part of a computing system or is otherwise operatively coupled to a computing system, the system can be executed Fig.34 A dedicated system for method 3400 and / or method 3432 of the present invention. Such a system can improve the accuracy of the interpretation process of the subsurface area of ​​the earth. Such a process can include presenting one or more images to a display, presenting one or more graphics to a display, etc. In such examples, the image can be a ground area, a subsurface area, or a combination of a ground area and a subsurface area (e.g., as in an overlap of a simulated area and a subsurface area). In such examples, the graphic can be a vector graphic presentation, which can have a model defined by a spatial position corresponding to the subsurface area of ​​the earth, and the model can be in one or more coordinate systems and / or reference systems (e.g., actual, previous time, future time, etc.).

[0298] As an example, a method may include: receiving a selected point from a visualization presentation of digital seismic data based on a geological region; extracting training data based on the selected point; performing machine learning of a model based on the training data; analyzing at least a portion of the digital seismic data based on the model to generate a result; and outputting the result. This method may include presenting a visualization to a display, wherein the digital seismic data of the geological region was acquired by equipment that receives seismic energy signals and digitizes the seismic energy signals to generate digital seismic data.

[0299] As an example, a method may include: accessing a trained machine model that is trained to analyze digital seismic data of a region relative to structural features of the geological region; analyzing at least a portion of the digital seismic data using the trained machine model to generate a result; and outputting the result as an indicator of a spatial location of the structural features of the geological region. In such an example, the method may include training to generate the trained machine model, wherein the training includes: receiving selected points of the digital seismic data for the geological region; extracting training data based on the selected points; and performing machine learning on the machine model based on the training data to generate the trained machine model.

[0300] As an example, the trained machine model can be a trained neural network model. As an example, a method can include training a neural network model to generate a trained neural network model. As an example, the trained machine model can be a trained kernel-based model. As an example, a method can include training a kernel-based model to generate a trained kernel-based model. For example, the kernel can be a radial basis function kernel or another type of kernel. For example, as mentioned, various types of kernels can be accessed via a library, or can be accessed programmatically or otherwise, for example. As mentioned, the kernel can be isotropic or anisotropic. As mentioned, the kernel can have a decay characteristic, which can be a tail characteristic (e.g., a heavy tail or a fat tail, etc.). As mentioned, the neural network can be a fully connected neural network having several hidden layers (e.g., about three to five hidden layers, etc.), wherein each hidden layer includes at least about 10 units (e.g., neurons).

[0301] As an example, outputting the results may include enhancing the visualization by presenting at least one confidence indicator indicating the confidence of the correspondence between the trained machine model (e.g., the features upon which the training model is based, etc.) and at least a portion of the digital seismic data. Such a confidence indicator may be associated with a feature (such as a surface desired to be interpreted).

[0302] As an example, digital seismic data may include trace data specified with respect to time and / or trace data specified with respect to distance.As an example, seismic attribute data may be used as a form of digital seismic data (eg, processed raw seismic data, etc.).

[0303] As an example, machine learning can utilize radial basis functions. As an example, such radial basis functions can include heavy tails. As an example, the radial basis function can include a decay term that decays relative to a dimension (e.g., "d"). As an example, the dimension can be a time dimension or a distance dimension.

[0304] As an example, a method can utilize digital seismic data of a seismic cube. In such an example, a result can be a confidence cube corresponding to at least a portion of the seismic cube. In such an example, a method can include performing tracking on the confidence cube.

[0305] As an example, the analysis may include parallel processing that processes portions of the seismic cube in parallel relative to the model to generate results for each portion of the seismic cube. In such an example, a method may include performing tracking on at least one of the results for at least one corresponding portion of the seismic cube. As an example, the results may include a confidence sub-cube (e.g., corresponding to the seismic sub-cube).

[0306] As an example, the results may include a confidence result, and the method may include analyzing the confidence result via a confidence score threshold. In such an example, a method may include comparing the analysis to the results of the tracking process to at least assess the quality of the tracking process.

[0307] As an example, a method may be implemented by a system that is a computing system operatively coupled to one or more displays, where such a method may include presenting at least one graphical user interface to the display.

[0308] As an example, a system may include: a processor; a memory operatively coupled to the processor; and processor-executable instructions stored in the memory to instruct the system to: receive selected points from a visual presentation of digital seismic data based on a geological region; extract training data based on the selected points; perform machine learning of a model based on the training data; analyze at least a portion of the digital seismic data based on the model to generate a result; and output the result.

[0309] As an example, a system may include: a processor; a memory operatively coupled to the processor; and processor-executable instructions stored in the memory to instruct the system to: access a trained machine model, the trained machine model being trained to analyze digital seismic data of a region relative to structural features of the geological region; analyze at least a portion of the digital seismic data using the trained machine model to generate a result; and output the result as an indicator of a spatial location of a structural feature of the geological region. As an example, such a system may include processor-executable instructions stored in the memory to instruct the system to perform training to generate a trained machine model, wherein the training utilizes instructions to receive selected points of digital seismic data for the geological region; extract training data based on the selected points; and perform machine learning on the machine model based on the training data to generate the trained machine model.

[0310] As an example, one or more computer-readable storage media may include computer-executable instructions that are executable to instruct a computing system to: access a trained machine model that is trained to analyze digital seismic data of a region relative to structural features of the geological region; analyze at least a portion of the digital seismic data using the trained machine model to generate a result; and output the result as an indicator of the spatial location of the structural features of the geological region.

[0311] As an example, one or more computer-readable storage media may include computer-executable instructions executable to instruct a computing system to perform one or more of the methods described herein, or a portion thereof.

[0312] As an example, a workflow may be associated with various computer readable media (CRM) blocks. Such blocks typically include instructions suitable for execution by one or more processors (or cores) to instruct a computing device or system to perform one or more actions. As an example, a single medium may be configured with instructions to at least partially allow the various actions of a workflow to be performed. As an example, a computer readable medium (CRM) may be a non-transitory, non-carrier, and non-signal computer readable storage medium. As an example, a block may be provided as one or more sets of instructions, such as Figure 2 One or more sets of instructions 270 for system 250.

[0313] Fig.35 Components of an example of a computing system 3500 and an example of a networking system 3510 are shown. System 3500 includes one or more processors 3502, memory and / or storage components 3504, one or more input and / or output devices 3506, and bus 3508. In an exemplary embodiment, instructions can be stored in one or more computer-readable media (e.g., memory / storage components 3504). Such instructions can be read by one or more processors (e.g., one or more processors 3502) via a communication bus (e.g., bus 3508), which can be wired or wireless. The one or more processors can execute such instructions to (in whole or in part) implement one or more properties (e.g., as part of a method). A user can view output from a process and interact with the process via an I / O device (e.g., device 3506). In an exemplary embodiment, a computer-readable medium can be a storage component, such as a physical memory storage device, such as a chip, a chip on a package, a memory card, etc. (e.g., a computer-readable storage medium).

[0314] In an exemplary embodiment, the components may be distributed in, for example, a network system 3510. The network system 3510 includes components 3522-1, 3522-2, 3522-3, ..., 3522-N. For example, component 3522-1 may include one or more processors 3502, and one or more components 3522-3 may include a memory accessible by one or more processors 3502. In addition, one or more components 3502-2 may include an I / O device for display and optionally interacting with the method. The network may be or include the Internet, an intranet, a cellular network, a satellite network, etc.

[0315] 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., operable via IEEE 802.11, ETSI GSM, BLUETOOTH, satellite, etc.). As an example, the mobile device may include components such as a main processor, a memory, a display, a display graphics circuit (e.g., optionally including a touch and gesture circuit), a SIM slot, an audio / video circuit, a motion processing circuit (e.g., an accelerometer, a gyroscope), a wireless LAN circuit, a smart card circuit, a transmitter circuit, a GPS circuit, and a battery. As an example, the mobile device may be configured as a mobile phone, a tablet computer, etc. As an example, the method may be implemented using a mobile device (e.g., in whole or in part). As an example, the system may include one or more mobile devices.

[0316] As an example, the system may be a distributed environment, such as a so-called "cloud" environment, in which various devices, components, etc. interact for purposes of data storage, communication, computing, etc. As an example, a device or system may include one or more components for communicating information via one or more of the Internet (e.g., in the case of communicating via one or more Internet protocols), a cellular network, a satellite network, etc. As an example, the method may be implemented in a distributed environment (e.g., in whole or in part as a cloud-based service).

[0317] As an example, information can be input from a display (e.g., consider a touch screen), output to a display, or both. As an example, information can be output to a projector, a laser device, a printer, etc. so that the information can be viewed. 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 substances that can be output to build a 3D object. For example, data can be provided to a 3D printer to build a 3D representation of an underground stratum. As an example, layers (e.g., horizons, etc.) can be built in 3D, geological bodies, etc. can be built in 3D. As an example, wellbores, fractures, etc. can be built in 3D (e.g., as positive structures, as negative structures, etc.).

[0318] Although only a few exemplary embodiments are described in detail above, it should be readily apparent to those skilled in the art that many modifications may be made in the exemplary embodiments. Therefore, all such modifications are intended to be included within the scope of the present disclosure as defined by the following claims. In the claims, means-plus-function clauses are intended to cover structures described herein that perform the functions described, not just structural equivalents, but also equivalent structures. Thus, although nails and screws may not be structural equivalents because nails use cylindrical surfaces to fasten wooden parts together, while screws use helical surfaces, nails and screws may be equivalent structures in the context of fastening wooden parts. It is not the express intent of the applicant to invoke 35 U.S.C. § 112, paragraph 6, to impose any limitation on any of the claims herein, except where a claim expressly uses the term "device for..." together with those limitations of the associated functions.

Claims

1. A method (3432) comprising: receiving (81v) selected input points within digital seismic data for a geological region corresponding to structural features of the geological region; extracting (820) training data from the digital seismic data based on the selected input point, wherein the training data includes positive training data extracted near the input point on a trace and negative training data extracted from another portion of the same trace; performing (830) machine learning on a machine model based on the training data to generate a trained machine model, wherein for a given input, the trained machine model predicts a confidence value for the given input associated with the structural feature, wherein the trained machine model comprises a trained kernel-based model, wherein the trained kernel-based model comprises a radial basis function, wherein the radial basis function comprises a heavy tail and / or comprises a decay term that decays with respect to a dimension, wherein the dimension comprises a time dimension or a distance dimension; accessing a trained machine model, the trained machine model being trained to analyze digital seismic data of a region relative to structural characteristics of the geological region (3436); predicting a confidence value for the digital seismic data associated with the structural feature using the trained machine model, wherein a higher confidence value indicates a more likely association between a portion of the digital seismic data and the structural feature; and A result is generated based on the confidence value indicating the spatial location of the structural feature of the geological region. 2 . The method of claim 1 , comprising presenting a visualization comprising at least one confidence indicator based at least in part on the confidence value.

3. The method of claim 1, wherein the digital seismic data comprises trace data specified with respect to time, or wherein the digital seismic data comprises trace data specified with respect to distance.

4. The method of claim 1, wherein the trained machine model comprises a trained neural network model, wherein the trained neural network model is trained using positive training data and negative training data about the structural features so as to predict a confidence value of a given input associated with the structural features for a given input.

5. The method of claim 1, wherein the digital seismic data comprises digital seismic data of a seismic cube.

6. The method of claim 5, wherein the results include a confidence cube corresponding to at least a portion of the seismic cube. The method of claim 6 , comprising performing tracking on the confidence cube.

8. A method as claimed in claim 5, wherein predicting confidence values ​​using a trained machine model includes parallel processing, wherein the parallel processing processes the portions of the seismic cube in parallel relative to the trained machine model to generate results for each of the portions of the seismic cube, and optionally includes performing tracking on at least one of the results for at least one corresponding portion of the seismic cube.

9. The method of claim 1, comprising analyzing the confidence value via a confidence score threshold.

10. The method of claim 9, comprising comparing the analysis with a tracking process of the structural feature to at least assess the quality of the tracking process based at least on the confidence value; If the quality drops below a threshold, stopping the tracking process; And the tracking process is stopped at the fault or the fault is jumped over to continue the tracking process of the structural feature.

11. The method of claim 1, comprising presenting at least one graphical user interface to a display.

12. A system (250), comprising: processor(256); a memory (258) operatively coupled to the processor; as well as Processor executable instructions (270) stored in the memory to instruct the system to perform the method of any one of claims 1-11.

13. One or more computer-readable storage media comprising computer-executable instructions executable to instruct a computing system to perform the method of any one of claims 1 to 11.

Images