METHOD, COMPUTER-READABLE NON-TRANSIENTIAL STORAGE MEDIUM, AND APPARATUS

Abstract

The present disclosure provides techniques to identify material properties from measured physical response data in ways that are more efficient than conventional numerical inversion. Systems and techniques of the present disclosure may use machine learning (ML) techniques to generate mappings that map specific physical responses to specific material properties based on knowledge of factors that may be inherent to certain types of sensing equipment. Data generated by a ML computer model may be constrained, altered, or filtered to more closely correspond to characteristics known to be associated with a certain type of sensing equipment. Such a constrained model may identify fewer possible results as compared to an unconstrained model, thereby, solving problems that may be encountered when using ML techniques to identify material properties from measured responses.

Description

1 / 35 “METHOD, COMPUTER-READABLE NON-TRANSIENTIAL STORAGE MEDIUM, AND APPARATUS” CROSS-REFERENCE TO RELATED REQUESTS

[0001] This application claims the benefit of U.S. Non-Provisional Application No. 18 / 389,077, filed November 13, 2023, which claims the benefit of U.S. Provisional Application No. 63 / 469,191, filed May 26, 2023, which is incorporated herein by reference. TECHNICAL FIELD

[0002] This disclosure is geared toward conducting assessments based on data detected in a drilling well. More specifically, this disclosure is directed toward restricting portions of computer-modeled data when assessments are made that compare computer-modeled data with data measured in a drilling well. BACKGROUND

[0003] Tools are frequently employed in drilling well environments for a variety of purposes. In some cases, acoustic sensors may be deployed in a drilling well when assessments are made related to the proper progress of a drilling well operation. In other cases, electromagnetic sensors or nuclear magnetic resonance sensors may be deployed in a drilling well when determinations are made related to structures or material properties in the Earth. Assessments made with sensing data are important for determining how a drilling well should be managed. The faster and more accurate these determinations are, the faster and Petition 870250077104, dated 08 / 29 / 2025, page 22 / 69 2 / 35 efficiently the operator of a drilling well will be able to perform operations carried out in it. BRIEF DESCRIPTION OF THE DRAWINGS

[0004] In order to describe the manner in which the aforementioned and other advantages and features of disclosure can be obtained, a more particular description of the principles briefly described above will be provided by reference to specific embodiments thereof, which are illustrated in the accompanying drawings. With the understanding that these drawings represent only exemplary embodiments of disclosure and should therefore not be considered limiting of its scope, the principles in this document are described and explained with additional specificity and detail by use of the accompanying drawings, in which:

[0005] Figure 1A is a schematic diagram of an example of a logging environment during drilling (LWD), according to various aspects of the technology in question.

[0006] Figure 1B is a schematic diagram of an example of a cable profiling environment, according to various aspects of the technology in question.

[0007] Figure 2 illustrates actions that can be performed by a machine learning process that can be used to refine the machine learning process, according to various aspects of the technology in question.

[0008] Figure 3 illustrates other actions that can be performed by a machine learning process when the machine learning process is refined.

[0009] Figure 4 illustrates a flow of actions that can be performed during a learning training stage. Petition 870250077104, dated 08 / 29 / 2025, page 23 / 69 3 / 35 of machine learning and during a machine learning application stage, according to various aspects of the technology in question.

[0010] Figure 5 illustrates a series of actions that can be performed when a machine learning computer model performs evaluations that compare measured data with data calculated or otherwise determined by the machine learning computer model, according to various aspects of the technology in question.

[0011] Figure 6 illustrates an example architecture of a computing device that can implement the various technologies and techniques described here. DETAILED DESCRIPTION

[0012] In the current wellbore logging process, for example, using tools that include acoustic, electromagnetic, and nuclear magnetic tools, inversion is one of the processes that can be used to identify material properties that exist in a subsurface environment from data collected using these tools. Machine learning is an artificial intelligence technology that can be used to map an input to an output. When performing tasks in the oil and gas sector, machine learning can be used to replace advanced modeling that identifies a numerical solution, but machine learning is not suitable for solving problems that inversion is commonly used to solve. What this means is that, conventionally, machine learning can be used to map material properties to physical responses, but machine learning Petition 870250077104, dated 08 / 29 / 2025, page 24 / 69 4 / 35 The machine was not successful in mapping physical responses to material properties. One reason for this is that attempts to use machine learning to map physical responses to material properties are not stable because this mapping does not result in a one-to-one mapping.

[0013] For a mapping of physical responses to material properties to be stable, it may be necessary to identify a one-to-one correspondence of a physical property to a material property. The rules associated with this one-to-one mapping may require that a computer model generate only a response that potentially corresponds to a measured physical response. Such a rule may specify a requirement to identify when data from a machine learning process sufficiently maps physical responses to material properties. This requirement may include identifying that any errors associated with comparing physical properties to material properties are within a threshold level. One reason why machine learning processes may not be stable could be related to the fact that measured or detected data and calculated data may not have compatible bandwidths.For example, data that includes detected acoustic or vibration information may not include a frequency range below a first frequency or above a second frequency, where the data for a machine learning (ML) computer model may not be as restricted. This means that the data identified by the operation of a model... Petition 870250077104, dated 08 / 29 / 2025, page 25 / 69 5 / 35 Machine learning (ML) computers may have a higher bandwidth than the bandwidth of the measured data. The methods in this disclosure can therefore compensate for these differences by restricting the results generated by the computational ML model. This means that a bandpass or other filtering function can be performed to filter data generated by the ML computer model to match an expected bandwidth of the detected data. This bandpass filtering function can suppress outputs from the ML computer model that are associated with frequencies lower than the first frequency and / or frequencies higher than the second frequency. This filtering process may be able to generate a one-to-one mapping of physical responses to material properties in cases where, without the filtering process, a resulting mapping is a one-to-many (1 to N) mapping.

[0014] A machine learning (ML) model, or ML computer model, can be implemented using one or more processors, a computer network, or as a neural network. This model operates based on algorithms that use data inputs and desired outputs for training. The ML model may engage in a recursive process, repeatedly conducting evaluations and updating factors included in an equation or coefficients of equations, based on the data and calculations. This iterative process continues until the ML model's results converge to a resolution, effectively addressing the problem it was tasked with solving. After that, the trained ML computer model can be deployed to Petition 870250077104, dated 08 / 29 / 2025, page 26 / 69 6 / 35 application or forecast.

[0015] Other functions that can be performed to limit or update the data identified using a computational ML model or to update the computational ML model include restricting the evaluations to one or more material classifications or material properties. Additionally or alternatively, calculations can be performed to identify gradients that can be used or incorporated into the computer ML model. In these cases, one or more processors can execute instructions from the computer model so that functions or actions of the computer model can be performed.

[0016] One reason for replacing conventional inversion with machine learning is to identify the material properties of a drill well more quickly than would be possible using conventional inversion techniques. Sensing equipment that can be used to acquire data from which material properties are identified can be used to shorten the well development process or can be used to improve well drilling operations. By modeling underground structures more quickly, one or more types of well drilling operations can be initiated more rapidly. This can allow hydrocarbons to be extracted at scheduled rates, carbon dioxide (CO2) to be sequestered according to a plan, or a hydraulic fracturing process to be implemented more efficiently.For example, an acoustic source can be used to generate images based on how acoustic energy is reflected from it. Petition 870250077104, dated 08 / 29 / 2025, page 27 / 69 7 / 35 returns to a sensor or is absorbed by materials impacted by the acoustic energy. An array of ultrasonic imaging sensors is an example of an acoustic sensing device. An example of an electromagnetic sensing device is a device that includes transmitting coils and receiving coils.

[0017] Regardless of the type of detection equipment used, these tools are designed to make measurements that are converted into acquired data sets. These acquired data sets can be used as input for processes that identify the physical and / or material properties of materials based on knowledge of physics and how the physical and / or material properties of the materials are affected by the detection equipment.

[0018] The disclosure now turns to Figures 1A-B, which provide a brief introductory description of some systems that can be used to practice the concepts, methods, and techniques disclosed here. A more detailed description of the methods and systems for implementing the enhanced similarity processing techniques of the disclosed technology will follow.

[0019] Figure 1A shows an illustrative formation logging environment (LWD). Drilling platform 2 supports the derrick 4 with the drive block 6 to raise and lower the drill string 8. Kelly 10 supports the drill string 8 as it is lowered through the rotary table 12. The drill bit 14 is driven by a downhole motor and / or by the rotation of the drill string 8. As the drill bit 14 rotates, it Petition 870250077104, dated 08 / 29 / 2025, page 28 / 69 8 / 35 creates a wellbore 16 that passes through several formations 18. The pump 20 circulates the drilling fluid through a feed pipe 22 to the Kelly 10, through the inside of the drill string 8, through holes in the drill bit 14, back to the surface through the ring around the drill string 8 and into the holding tank 24. The drilling fluid carries the cuttings from the wellbore to the tank 24 and helps maintain the integrity of the wellbore.

[0020] The downhole tool 26 may take the form of a drill collar (i.e., a thick-walled tube that provides weight and rigidity to aid the drilling process) or other arrangements known in the art. Furthermore, the downhole tool 26 may include acoustic logging tools (e.g., sonic, ultrasonic, etc.) and / or corresponding components, integrated into the downhole assembly near the drill bit 14. In this way, as the drill bit 14 extends the wellbore through the formations, the downhole assembly (e.g., the acoustic logging tool) can collect acoustic logging data. For example, acoustic logging tools may include transmitters (e.g., monopole, dipole, quadrupole, etc.) to generate and transmit acoustic signals / waves to the wellbore environment.These acoustic signals subsequently propagate within and along the well and surrounding formation, creating acoustic signal responses or waveforms, which are received / recorded by uniformly spaced receivers. These receivers can be arranged in an array and can be evenly spaced to facilitate... Petition 870250077104, dated 08 / 29 / 2025, p. 29 / 69 9 / 35 captures and processes acoustic response signals at specific intervals. The acoustic response signals are further analyzed to determine the properties and / or characteristics of the well and the adjacent formation.

[0021] For communication purposes, a downhole telemetry sub 28 may be included in the downhole assembly to transfer measurement data to the surface receiver 30 and receive commands from the surface. Mud pulse telemetry is a common telemetry technique for transferring tool measurements to surface receivers and receiving commands from the surface, but other telemetry techniques may also be used. In some embodiments, the telemetry sub 28 may store log data for later retrieval at the surface when the log assembly is recovered.

[0022] On the surface, the surface receiver 30 can receive the uplink signal from the downhole telemetry sub 28 and can communicate the signal to the data acquisition module 32. The module 32 may include one or more processors, storage media, input devices, output devices, software and the like, as described in detail in Figures 2A and 2B. The module 32 may collect, store and / or process the data received from the tool 26, as described in this document.

[0023] Figure 1B illustrates how a tool used to collect data can be lowered into a drilling well. Figure 1B includes many of the same elements discussed in relation to Figure 1B. For example, Figure 1B includes platform 2, tower 4, block 6, and turntable 12. Petition 870250077104, dated 08 / 29 / 2025, page 30 / 69 10 / 35 which are included in Figure 1A. At various times during the drilling process, the drill string 8 shown in Figure 1A may be removed from the wellbore and the downhole tool 34 may then be lowered into the drill hole 16 of Figure 1A. After the drill string 8 has been removed, logging operations may be conducted using a downhole tool 34 (i.e., a sensing instrument probe) suspended by a carrier 42. In one or more embodiments, the carrier 42 may be a cable with conductors to carry power to the tool and telemetry from the tool to the surface. The downhole tool 34 may have cushions and / or centering springs to keep the tool close to the centerline of the wellbore or to tilt the tool toward the wall of the wellbore as the tool is moved up or down the well.

[0024] The downhole tool 34 may include an acoustic or sonic recording instrument that collects acoustic recording data within the wellbore 16. A recording installation 44 includes a computer system, which may be used to collect, store and / or process measurements collected by the recording tool 34. In one or more cases, the transport 42 may include at least one of the wires, conductive or non-conductive cables (e.g., steel cable, etc.) attached to the downhole tool 34. The transport 42 may include tubular transports such as spiral tubes, pipe ropes or a downhole tractor. The downhole tool 34 may have a local power source such as batteries, a downhole generator. Petition 870250077104, dated 08 / 29 / 2025, page 31 / 69 11 / 35 well or something similar. When employing non-conductive cable, coiled tubing, tubing string or downhole tractor, communication can be supported using, for example, wireless protocols (e.g., EM, acoustic, etc.) and / or measurements and logging data can be stored in local memory for later retrieval.

[0025] The downhole tool 34 may include one or more hydrophones, a microphone, a hydrophone array, or a microphone array. Such arrays may include one or more hydrophones and / or hydrophones that collect data from a drill well at various stages of its lifecycle, from the initial stages when drill wells are drilled and fabricated, to when the drill well is used during a production process (e.g., hydrocarbon extraction or carbon sequestration process) and / or after a drill well is taken out of service.

[0026] Although Figures 1A and 1B show specific wellbore configurations, it is understood that this disclosure is equally suitable for use in drilling wells with other orientations, including vertical, horizontal, inclined, multilateral and similar drilling wells. Although Figures 1A and 1B depict an onshore operation, it should also be understood that this disclosure is equally suitable for use in offshore operations. Furthermore, this disclosure is not limited to the environments represented in Figures 1A and 1B, and may also be used, for example, in other well operations such as production piping operations, articulated piping operations, operations of Petition 870250077104, dated 08 / 29 / 2025, page 32 / 69 12 / 35 spiral piping, combinations of these operations and similar.

[0027] The scope of this disclosure is not limited to the environment shown in Figures 1A and 1B, as the methods of this disclosure can be applied in other environments. The methods and apparatus of this disclosure can process acoustic data received from one or more microphones, hydrophones, piezoelectric sensors, or other equipment capable of detecting acoustic signals, such as subsonic, sonic, or ultrasonic signals. This processing may include performing evaluations that allow parts of the received acoustic data to be identified based on characteristics known to be representative of specific types of sound sources. Characteristics that can be associated with a sound source include, but are not limited to, one or more frequencies emitted by the sound source and / or information that can be used to identify the location of the sound source.Furthermore, or alternatively, the acoustic noise characteristic of a sound source can be associated with a sound amplitude, a power, or a power spectral density of the noise emitted by the sound source.

[0028] Figure 2 illustrates the actions that can be performed by a machine learning process that can be used to refine the machine learning process. In block 210, measurement data can be accessed. This measurement data may have been collected from a borehole using any type of sensing equipment known in the art. Thus, the measurement data accessed in block 210 can be acquired using a tool Petition 870250077104, dated 08 / 29 / 2025, page 33 / 69 13 / 35 acoustics, an electromagnetic imaging tool, a nuclear magnetic resonance tool, or another type of tool. In block 220, a machine learning (ML) computational model may be invoked according to a set of constraints. This may allow the ML computer model to generate ML data that may match measured or detected data, so that materials included in the borehole can be identified.

[0029] The operation of an ML computer model may be restricted to generate ML data that have a specific frequency response. This may cause the ML computer model to filter out or suppress frequency components that do not match a frequency response known to be associated with a specific type of borehole detection equipment or materials located in a borehole.For example, a filtering function can remove frequencies from a computer-generated dataset that are below a first frequency and above a second frequency based on knowledge of frequencies that are not consistent with a particular type of detection equipment.

[0030] Alternatively or additionally, the constraints of the ML computer model may cause the ML computer model to make assessments based on a type or class of material present in the borehole. This may mean that assessments may be limited to a class or type of material that is presumed to be in the borehole or has been shown to exist there. For example, the well reflection characteristics of Petition 870250077104, dated 08 / 29 / 2025, p. 34 / 69 14 / 35 drilling can be consistent with sandstone of a given porosity, and this characteristic can allow the AM computer model to limit the evaluations to a set of constraints based on a rule associated with the sandstone and that given porosity. These constraints may include identifying the first and second frequencies of a bandpass filter, such as the filter discussed above, or they may include identifying other constraints known to be associated with the sandstone.

[0031] Determination block 230 can then identify whether the ML data in block 220 matches the measurement data at a threshold level. This threshold level may require that a measured parameter or value be within a certain percentage of a value of the same parameter or value generated by the ML computer model operation. For example, when an attenuation prediction by the ML computer model identifies that an acoustic or electromagnetic signal attenuation should have a value of 50 decibels, a measured value of that attenuation should be within 50 decibels plus or minus ten percent. When determination block 220 identifies that the ML data does not match the measurement data within the threshold level, the program flow can proceed to block 240, where the ML computer model is updated. The ML computer model can then be invoked again in block 220 using the updated ML computer model.At least some of the actions performed in Figure 2 can be executed iteratively until the ML data generated by the ML computer model matches the measurement data up to the threshold level (i.e., match). Petition 870250077104, dated 08 / 29 / 2025, p. 35 / 69 15 / 35 to a limiting degree). This can allow for multiple evaluations with different constraints or formula coefficients associated with material identification. For example, sets of constraints can be associated with a frequency bandpass function or specific material types, characteristics, or material responses. These constraints can also correspond to the material properties modeled by the ML computer model. Thus, blocks 220, 230, and 240 can be used to generate or choose one or more ML algorithms based on what can be called a learning process.

[0032] When determination block 230 identifies that the ML data matches the measurement data, the program flow can proceed to block 250, where the ML computer model is put into service. This means that the determinations made by the ML computer model can be considered when drilling well decisions are made in a real ML application. For example, the ML computer model may identify that the material properties of a location within an underground formation are suitable for hydrocarbon extraction or carbon dioxide (CO2) sequestration. Based on this determination, an automated process or a drilling well operator can authorize hydrocarbon extraction or CO2 sequestration for the underground formation.The methods of this disclosure, therefore, allow ML learning to replace conventional inversion techniques and can use measured or detected physical responses in such a way that the properties... Petition 870250077104, dated 08 / 29 / 2025, pp. 36 / 69 16 / 35 of the material can be identified more efficiently.

[0033] Figure 3 illustrates other actions that can be performed by a machine learning process when the machine learning process is refined. In block 310, measurement data can be accessed as discussed in relation to block 210 of Figure 2. One or more processors can then execute instructions from an ML computer model to identify a calculated dataset in block 320. This may include applying the constraints discussed in relation to the actions in Figure 2. This calculated dataset may have its bandwidth limited similarly to the bandpass filter discussed above, or this calculated data may be used to identify the material property or type of material located in the borehole.

[0034] In block 330, the mismatch gradient data can be calculated. This may involve performing calculations according to formulas 1 and 2 below. In formulas 1 to 5 below, Σ represents a summation function, F(m) and F(mi-1) represent model functions m associated with different values, d represents data collected by sensors, S is a fitness estimate between a current model and previously known information (prior knowledge), λ is a weighting factor, mi is a model value m at a specific update, mi-1 is the model value m immediately before the specific update, σ is an update step along a direction inverse to the gradient, V represents a gradient function, and F'(mi-1) is a derivative called the modeling gradient function. Petition 870250077104, dated 08 / 29 / 2025, page 37 / 69 17 / 35 direct. Error = Value of the cost function = Σ (dF(m))2 + Prior knowledge Formula 1: Error = Value of the Cost Function = Σ (dF(m))2 + AS (m) Formula 2: mi = mi -1 - σ (V [Σ (dF (mi-1))2+AS (mi-1)]) Formula 3: mi = mi - 1 - σ (V [Σ (dF (mi-1))2+ VAS (mi-1)]) Formula 4: (V [Σ (dF (mi-1))2]) = -2 [F ' (mi-1) (dF (mi-1))] Formula 5:

[0035] Formulas 1 and 2 can be called machine learning-assisted or artificial intelligence (AI)-assisted inversion and can be used to map a function model space F(m) to a data space. In this way, an error or cost function can be identified using the measured data from the ML model data of the function F(m). This error or cost function can be calculated based on prior knowledge, which can be expressed as AS(m).

[0036] This prior knowledge can be general knowledge about a particular material. For example, in seismic inversion, it can be assumed that the reflectivity is Petition 870250077104, dated 08 / 29 / 2025, pp. 38 / 69 18 / 35 sparse or limited to certain dimensions. The reason such an assumption can be made is that recorded seismic waves may have a narrow bandwidth (limited frequency response); however, the Earth's reflectivity may have a larger or total bandwidth. To fill a high- and low-frequency loss gap associated with the measured data, the sparsity assumption can be used to make an inversion stable by limiting the computer-modeled data to the high- and low-frequency loss characteristics assumed to be associated with the measured data. A formula λ| | rMx1| |1 can be used, where λ is a previously known coefficient, r is the Earth's reflectivity, its dimension can be Mx1; || rMx1||1 is the normal level of Earth's reflectivity.

[0037] In the physics of wellbore sensors, such as wellbore detection, we also have many sets of prior knowledge. For example, when acoustic measurements are used, we know that the impedance between the wellbore and a pipe should correspond to a step function. But perhaps we don't know how strong the impedance contrast is. The methods of the present disclosure may have to take this into account when prior knowledge is used to perform computer-modeled inversions.

[0038] In another example, an electromagnetic wave can be used to detect erosion in pipelines. In this case, we can know approximately the permeability and resistivity of the pipe, but what may be unknown is the thickness of the pipe due to the loss or gain of metal. Therefore, during inversion, we can define the range of Petition 870250077104, dated 08 / 29 / 2025, pp. 39 / 69 19 / 35 permeability and resistivity as a guide to stabilizing the inversion and achieving a reasonable pipe thickness. In summary, the prior knowledge used is a case-by-case estimate, which may need to be customized. This may require altering a mathematical formula to accommodate prior knowledge so that the inversion workflow can be expected to produce reliable results.

[0039] The model values ​​m (e.g., mi and mi-1) can be calculated according to formula 3 or 4 above. The equation V[Σ(dF(mi-1))2] can be called the gradient mismatch equation, and the equation VAS(mi-1) can be called the prior knowledge constraint gradient equation. Classical methods can be used to calculate the values ​​of the function F(mi-1), so the gradient mismatch equation can be used to calculate the values. Machine learning can be used to replace the techniques associated with traditional inversion. Machine learning or AI can therefore be used to replace the numerical solutions identified using classical advanced numerical modeling and / or classical solutions that identify data gradient mismatches. Formula 5 is a simplified data gradient mismatch equation. Formula 5 can be used to simplify the calculations performed by a machine learning process.

[0040] In block 340, a mapping can be generated of the measurement data accessed in block 310, the data calculated in block 320, and the mismatch gradient data identified in block 330. The block of Petition 870250077104, dated 08 / 29 / 2025, pp. 40 / 69 20 / 35 determination 350 can then identify whether the data calculated by the ML computer model corresponds to the measurement data; if not, the program flow can proceed to block 360, where an adjustment can be made to the ML computer model. Blocks 320, 330, 340, and 350 can be executed iteratively until determination block 350 identifies that the data calculated by the ML computer model corresponds to the measurement data, so that the ML computer model can be put into service in block 370. As discussed in Figure 2, putting the ML computer model into service can allow the determinations made by the ML computer model to be considered when decisions are made about the drilling well.Again, this could mean that the ML computational model can be used to identify locations where the material properties of an underground formation are suitable for hydrocarbon extraction, a CO2 sequestration process, or some other well drilling process.

[0041] Regarding the latent space discussed below, when working with especially complex and high-dimensional data. For example, data that has the complexity of images or audio files may not be easy to work with directly, as they may contain a lot of unnecessary information. In addition, or alternatively, it may be difficult to find patterns in this data that are useful for a specific assessment.

[0042] The term latent essentially means “hidden” or “underlying.” Therefore, latent space represents the underlying or hidden resources that are most important in Petition 870250077104, dated 08 / 29 / 2025, page 41 / 69 21 / 35 of our data. Imagine looking at a bunch of photos of people's faces. To a computer, these photos are just a large array of numbers (pixels). But, for people, there are certain important features that distinguish one face from another, such as eye shape, nose size, hair color, etc. Latent space is a way to transform raw data (the pixels, in this case) into a simpler, more compact representation that captures these important features. In this transformed space, similar faces will be close together and different faces will be further apart. This makes working with data much easier, as it's possible to operate directly on the important features, instead of having to deal with each pixel individually.

[0043] In machine learning, the model can learn to automatically extract these important features and represent the data in latent space. The learning process of this transformation is usually done through a method called training, in which the model learns from a large amount of example data. After the model is trained, it can map new data into latent space, even if it has never seen that exact data before.

[0044] Figure 4 illustrates a flow of actions that can be performed during a machine learning training stage and during a machine learning application stage. A first flow of actions 400 shown in Figure 4 includes blocks 410T, 420T, 430T, and 450T that can be implemented with the training of an ML computer model. In block 410T, the measured data d Petition 870250077104, dated 08 / 29 / 2025, page 42 / 69 22 / 35 data acquired from a drilling well can be reduced in dimensions. This dimensional reduction can include mapping three-dimensional (3D) data into two-dimensional (2D) data, for example. By reducing the number of dimensions associated with a measured dataset d, calculations or other evaluations performed based on this calculated data can be simplified. In block 420T, the measured data d mapped in block 410 can be represented in the latent space of measured data Hd, for example, in a 2D mapping. In block 430T, a dataset generated based on the operation of a computer model ML m can be reduced in dimensions. This can include mapping data identified by the operation of a computer model ML in a 3D space to a 2D space. In block 440T, the data from the computer model ML m mapped in block 430T can be represented in the latent space of modeled data Hm. For example, here again in 2D space.In block 450, a machine learning (ML) process can be used to map the measured data from latent space Hd to the modeled data from latent space Hm. This process can be performed iteratively and after executing any of the actions discussed in Figures 2 and 3. Once the errors associated with the differences between the measured data and the computer-modeled ML data are reduced to the threshold level, training stage 400 can be considered complete, and the operation of the computer-ML model can proceed to the application stage.

[0045] A second action flow 460 shown in Figure 4 shows that blocks 410A, 420A, 430A, 440A and 450A can Petition 870250077104, dated 08 / 29 / 2025, page 43 / 69 23 / 35 can be implemented when the ML computer model is used to identify the material properties of a borehole from measured data. Here again, a number of dimensions from a measurement dataset can be reduced in block 410A (e.g., from a 3D representation to a 2D representation). This reduced set of measured data can be represented in latent space (e.g., 2D space), and the measured data from latent space Hd can be mapped to ML computer model operations in block 450. In block 440A, the ML computer model can be configured to represent the data modeled by the ML computer in latent space Hm, and the dimension reduction performed in block 430A can reconstruct the m data associated with the ML computer model in latent space Hm.

[0046] Figure 5 illustrates a series of actions that can be performed when a machine learning computational model performs evaluations that compare measured data with data calculated or otherwise determined by the machine learning computational model. In block 510, measurement data detected in a borehole can be accessed. Determination block 520 can identify a process flow that should be used when making determinations of this disclosure. This means that actions in any of blocks 530, 540, 550, or 560 can be performed. A process flow could include an ML computer model performing calculations. A calculated dataset can then be filtered using a filter (e.g., a bandpass filter) in block 530. Alternatively or additionally, this process may include Petition 870250077104, dated 08 / 29 / 2025, pp. 44 / 69 24 / 35 performing gradient calculations in block 540 and / or identifying a property classification in block 550. In block 560, the ML computational model data can be mapped to a latent space when the ML computational model is trained, as discussed in Figure 4. Determination block 570 can then identify whether the calculated data matches the measurement data to a degree or threshold level; if not, the program flow can return to determination block 520. When determination block 570 identifies that the calculated data matches the measurement data, the ML computational model can be put into service in block 580.

[0047] Figure 6 illustrates an example architecture 600 of a computing device that can implement the various technologies and techniques described herein. The various implementations will be apparent to those skilled in the art when practicing the present technology. Those skilled in the art will also readily realize that other system implementations or examples are possible. The components of the computing device architecture 600 are shown in electrical communication with each other using a connection 605, such as a bus. The example computing device architecture 600 includes a processing unit (CPU or processor) 610 and a computing device connection 605 that couples various computing device components, including computing device memory 615, such as read-only memory (ROM) 620 and random-access memory (RAM) 625, for the processor 610. Petition 870250077104, dated 08 / 29 / 2025, pp. 45 / 69 25 / 35

[0048] The computing device architecture 600 may include a high-speed memory cache directly connected to, in immediate proximity to, or integrated as part of the processor 610. The computing system architecture 600 may copy data from memory 615 and / or storage device 630 to the cache 612 for fast access by the processor 610. In this way, the cache may provide a performance boost that prevents the processor 610 from lagging while waiting for data. These and other modules may control or be configured to control the processor 610 to perform various actions. Other computing device memory 615 may also be available for use. Memory 615 may include multiple different types of memory with different performance characteristics.The 610 processor may include any general-purpose processor and a hardware or software service, such as service 1 632, service 2 634, and service 3 636 stored in storage device 630, configured to control the 610 processor, as well as a special-purpose processor where software instructions are incorporated into the processor design. The 610 processor may be a standalone system containing multiple cores or processors, a bus, memory controller, cache, etc. A multi-core processor may be symmetric or asymmetric.

[0049] To enable user interaction with the computing device 600, an input device 645 can represent any number of input mechanisms, such as a microphone for speech, a touchscreen, Petition 870250077104, dated 08 / 29 / 2025, pp. 46 / 69 26 / 35 touch for gesture input, keyboard, mouse, motion input, speech, and so on. An output device 635 may also be one or more of a number of output mechanisms known to those skilled in the art, such as a monitor, projector, television, speaker device, etc. In some cases, multimodal computing devices may allow a user to provide multiple types of input to communicate with the computing device architecture 600. The communication interface 640 may generally govern and manage user input and computing device output. There is no restriction regarding operation in any particular hardware arrangement, and therefore the basic features described in this document may be easily superseded by improved hardware or firmware arrangements as they are developed.

[0050] The storage device 630 is non-volatile memory and may be a hard disk or other types of computer-readable media that can store data accessible by a computer, such as magnetic cassettes, flash memory cards, solid-state memory devices, digital versatile disks, cartridges, random access memories (RAMs) 625, read-only memory (ROM) 620, and hybrids thereof. The storage device 630 may include software modules 632, 634, 636 to control the processor 610. Other hardware or software modules are contemplated. The storage device 630 may be connected to the computing device connection 605. In one aspect, a hardware module that performs a specific function may include the stored software component. Petition 870250077104, dated 08 / 29 / 2025, pp. 47 / 69 27 / 35 in a computer-readable medium in connection with the necessary hardware components, such as processor 610, connector 605, output device 635, and so on, to perform the function.

[0051] For clarity of explanation, in some cases, the present technology may be presented as including individual functional blocks, including functional blocks comprising devices, device components, steps or routines, in a method embodied in software or combinations of hardware and software.

[0052] In some cases, computer-readable storage devices, media, and memories may include a cable or wireless signal containing a bit stream and the like. However, when mentioned, computer-readable non-transient storage media expressly exclude media such as energy, carrier signals, electromagnetic waves, and signals per se.

[0053] Methods according to the examples described above can be implemented using computer-executable instructions that are stored or otherwise available from computer-readable media. Such instructions may include, for example, instructions and data that cause or otherwise configure a general-purpose computer, special-purpose computer, or processing device to perform a particular function or group of functions. Portions of the computer resources used may be accessible over a network. Computer-executable instructions may be, for example, binary, intermediate-format instructions such as ensemble language, firmware, source code, etc. Examples of computer-readable media Petition 870250077104, dated 08 / 29 / 2025, pp. 48 / 69 28 / 35 Computers that can be used to store instructions, information used and / or information created during the methods according to the examples described include magnetic or optical disks, flash memory, USB devices supplied with non-volatile memory, network storage devices, and so on.

[0054] Devices implementing methods according to these disclosures may include hardware, firmware, and / or software and may assume any of a variety of form factors. Typical examples of such form factors include laptops, smartphones, small form factor personal computers, personal digital assistants, rack-mounted devices, standalone devices, and so forth. The functionality described in this document may also be incorporated into peripheral or add-on cards. Such functionality may also be implemented on a circuit board between different chips or different processes running on a single device, by way of further example.

[0055] The instructions, means for transmitting such instructions, computing resources for executing them, and other structures for supporting such computing resources are example means for providing the functions described in this disclosure.

[0056] In the preceding description, aspects of the application are described with reference to specific embodiments thereof, but those skilled in the art will recognize that the application is not limited to them. Thus, although illustrative embodiments of the application have been described in detail in this document, it should be understood that the concepts Petition 870250077104, dated 08 / 29 / 2025, pp. 49 / 69 29 / 35 disclosed may be incorporated and employed in other ways, and the appended claims are intended to be interpreted to include such variations, except as limited by the state of the art. Several features and aspects of the subject described above may be used individually or in combination. Furthermore, the embodiments may be used in any number of environments and applications beyond those described in this document, without departing from the broader spirit and scope of the descriptive report. The descriptive report and drawings should therefore be considered illustrative and not restrictive. For illustrative purposes, the methods have been described in a particular order. It should be appreciated that in alternative embodiments, the methods may be carried out in a different order than that described.

[0057] Where components are described as being “configured to perform certain operations, such configuration may be achieved, for example, by designing electronic circuits or other hardware to perform the operation, by programming programmable electronic circuits (e.g., microprocessors or other suitable electronic circuits) to perform the operation, or any combination thereof.

[0058] The various logic blocks, modules, circuits, and illustrative algorithm steps described in connection with the examples disclosed in this document can be implemented as electronic hardware, computer software, firmware, or combinations thereof. To clearly illustrate this interchangeability of hardware and software, various components, blocks, modules, circuits, and steps are shown. Petition 870250077104, dated 08 / 29 / 2025, pages 50 / 69 30 / 35 illustrative examples have been described above, generally in terms of their functionality. Whether this functionality is implemented as hardware or software depends on the particular application and the design constraints imposed on the overall system. Persons skilled in the art may implement the described functionality in various ways for each specific application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

[0059] The techniques described herein may also be implemented in electronic hardware, computer software, firmware, or any combination thereof. These techniques may be implemented in any of a variety of devices, such as general-purpose computers, wireless communication devices, or multi-purpose integrated circuit devices, including application in wireless communication devices and other devices. Any features described as modules or components may be implemented together in an integrated logic device or separately as distinct but interoperable logic devices. If implemented in software, the techniques may be performed, at least in part, by a computer-readable data storage medium comprising program code including instructions that, when executed, perform one or more of the methods, algorithms, and / or operations described above.Computer-readable data storage media can be part of a computer program product, which may include packaging materials. Petition 870250077104, dated 08 / 29 / 2025, pp. 51 / 69 31 / 35

[0060] Computer-readable media may include memory or data storage media, such as random access memory (RAM), such as synchronous dynamic random access memory (SDRAM), read-only memory (ROM), non-volatile random access memory (NVRAM), electrically erasable programmable read-only memory (EEPROM), FLASH memory, magnetic or optical data storage media and the like. The techniques may additionally, or alternatively, be implemented, at least in part, by a computer-readable communication medium that carries or communicates the program code in the form of instructions or data structures and that can be accessed, read and / or executed by a computer as propagated signals or waves.

[0061] Other forms of dissemination can be practiced in network computing environments with many types of computer system configurations, including personal computers, portable devices, multi-processor systems, microprocessor-based or programmable consumer electronics, network PCs, minicomputers, mainframe computers, and the like. Forms can also be practiced in distributed computing environments where tasks are performed by local and remote processing devices that are connected (either by wired links, wireless links, or a combination thereof) through a communications network. In a distributed computing environment, program modules may be located on both local and remote memory storage devices.

[0062] The various modalities described above are provided Petition 870250077104, dated 08 / 29 / 2025, pp. 52 / 69 32 / 35 are for illustrative purposes only and should not be interpreted to limit the scope of the disclosure. For example, the principles in this document apply equally to optimization as well as general improvements. Various modifications and changes may be made to the principles described in this document without following the example modalities and applications illustrated and described in this document, and without departing from the spirit and scope of the disclosure. Claim language stating “at least one of a set” indicates that one member of the set or multiple members of the set satisfy the claim. ASPECTS OF DISSEMINATION:

[0063] Aspect 1: A method of the present disclosure may include accessing wellbore measurement data; identifying, through a machine learning process, data that corresponds to wellbore measurement data; implementing a function that restricts the machine learning process data by mapping the restricted machine learning process data to wellbore measurement data; comparing the restricted machine learning process data with wellbore measurement data; and identifying, based on the comparison, that wellbore measurement data corresponds to the restricted machine learning process data based on an error smaller than an error threshold.

[0064] Aspect 2: The method of Aspect 1, in which the function that restricts the data of the machine learning process is implemented based on an assumption regarding the measurement data of the drilling well. Petition 870250077104, dated 08 / 29 / 2025, pp. 53 / 69 33 / 35

[0065] Aspect 3: The method of Aspect 1 or 2, where the function that restricts the machine learning process data extracts features from the machine learning process data.

[0066] Aspect 4: The method of any of Aspects 1 to 3, further comprising the classification of the material properties of the machine learning process data, narrowing an output range associated with the machine learning process data.

[0067] Aspect 5: The method of Aspect 4, further comprising the generation of a mapping of a space that associates the data from the restricted machine learning process with known properties of the borehole.

[0068] Aspect 6: The method of Aspect 4, further comprising performing calculations to identify an error value by: subtracting predicted values ​​from data identified by operating a computer model on portions of wellbore measurement data to create a set of difference values; squaring each of the set of difference values; and generating a sum of the set of squared difference values.

[0069] Aspect 7: The method of Aspect 6, where the generated sum also includes a set of weighted fitness estimate values.

[0070] Aspect 8: The method of Aspect 6, further comprising the iterative application of a data mismatch gradient equation and a prior knowledge constraint equation to identify the predicted data values. Petition 870250077104, dated 08 / 29 / 2025, pp. 54 / 69 34 / 35

[0071] Aspect 9: The method of Aspect 6, further comprising the generation of a mapping of a space that associates the data from the restricted machine learning process with known properties of the borehole. [007 2] Aspect 10: The method of Aspect 4, further comprising performing calculations to identify an error value by: subtracting predicted values ​​from data identified by operating a computer model on portions of wellbore measurement data to create a set of difference values; squaring each of the set of difference values; and generating a sum of the set of squared difference values.

[0073] Aspect 11: The method of Aspect 10, where the generated sum also includes a set of weighted fitness estimate values.

[0074] Aspect 12: The method of Aspect 10, further comprising the iterative application of a data mismatch gradient equation and a prior knowledge constraint equation to identify the predicted data values.

[0075] Aspect 13: The method of any of Aspects 1 to 12, where the function that restricts the machine learning process data limits a bandwidth associated with a portion of the borehole measurement data.

[0076] Aspect 14: The method of Aspect 13, where the bandwidth is limited by applying a bandpass filter to the measurement data portion of the borehole.

[0077] The device of this disclosure may include one or Petition 870250077104, dated 08 / 29 / 2025, pp. 55 / 69 35 / 35 plus sensors, memory, and one or more processors executing instructions from memory to perform any of the Disclosure Aspects discussed above. Furthermore, any of the Aspects discussed above can be implemented by a computer-readable non-transient storage medium where one or more processors execute instructions from memory. Petition 870250077104, dated 08 / 29 / 2025, pp. 56 / 69

Claims

1 / 5 CLAIMS 1. A method, characterized in that it comprises: accessing wellbore measurement data; identifying, through a machine learning process, the data that corresponds to the wellbore measurement data; implementing a function that restricts the machine learning process data by mapping the restricted machine learning process data to the wellbore measurement data; comparing the restricted machine learning process data with the wellbore measurement data; and identifying, based on the comparison, that the wellbore measurement data corresponds to the restricted machine learning process data based on an error smaller than an error threshold.

2. A method according to claim 1, characterized in that the function that restricts the data of the machine learning process is implemented based on an assumption regarding the measurement data of the drilling well.

3. A method according to claim 1, characterized in that the function that restricts the machine learning process data extracts features from the machine learning process data.

4. Method according to claim 3, characterized in that it further comprises classifying the material properties of the machine learning process data by narrowing an output range associated with the machine learning process data. Petition 870250077104, dated 08 / 29 / 2025, pp. 57 / 69 2 / 5 5. Method, according to claim 4, characterized in that it further comprises: generating a mapping of a space that associates the data from the restricted machine learning process with known properties of the borehole.

6. Method according to claim 4, characterized in that it further comprises: performing calculations to identify an error value; subtracting predicted data values ​​identified by operating a computer model from portions of the borehole measurement data to create a set of difference values; squaring each of the difference values ​​in the set; and generating a sum of the set of squared difference values.

7. Method according to claim 6, characterized in that the generated sum also includes a set of weighted fitness estimate values.

8. Method according to claim 6, characterized in that it further comprises the iterative application of a data mismatch gradient equation and a prior knowledge constraint equation to identify the predicted data values.

9. Method, according to claim 6, characterized in that it further comprises: generating a mapping of a space that associates data from the restricted machine learning process with known properties of the borehole.

10. Method, according to claim 4, characterized in Petition 870250077104, dated 08 / 29 / 2025, page 58 / 69 3 / 5 by the fact that it further comprises: performing calculations to identify an error value; subtracting predicted data values ​​identified by operating a computer model from portions of the borehole measurement data to create a set of difference values; squaring each of the difference values ​​in the set; and generating a sum of the set of squared difference values.

11. Method, according to claim 10, characterized in that the generated sum also includes a set of weighted fitness estimate values.

12. Method according to claim 10, characterized in that it further comprises the iterative application of a data mismatch gradient equation and a prior knowledge constraint equation to identify the predicted data values.

13. Method, according to claim 1, characterized in that the function that restricts the machine learning process data limits a bandwidth associated with a portion of the drilling well measurement data.

14. Method, according to claim 13, characterized in that the bandwidth is limited by applying a bandpass filter to the measurement data portion of the borehole.

15. Non-transient, computer-readable storage medium, characterized in that it has embedded instructions that, when executed by one or more processors, result in one or more processors: accessing wellbore measurement data; identifying, through a machine learning process, the data that corresponds to the wellbore measurement data; implementing a function that restricts the machine learning process data by mapping the restricted machine learning process data to the wellbore measurement data; comparing the restricted machine learning process data with the wellbore measurement data; and identifying, based on the comparison, that the wellbore measurement data corresponds to the restricted machine learning process data based on an error smaller than an error threshold.

16. A non-transient, computer-readable storage medium according to claim 15, characterized in that the function that restricts the data of the machine learning process is implemented based on an assumption regarding the measurement data of the drilling well.

17. A non-transient, computer-readable storage medium according to claim 15, characterized in that the function that restricts the machine learning process data extracts features from the machine learning process data.

18. Non-transient, computer-readable storage medium according to claim 17, characterized in that it further comprises the classification of the material properties of the machine learning process data, narrowing an output range associated with the machine learning process data.

19. A non-transient, computer-readable storage medium according to claim 18, characterized in that one or more processors execute instructions to: generate a mapping of a space that associates the data from the restricted machine learning process with known properties of the borehole.

20. Device, characterized in that it comprises: a memory; and one or more processors that execute instructions from the memory to: access the drilling well measurement data; identify, through a machine learning process, the data that corresponds to the drilling well measurement data; implement a function that restricts the machine learning process data by mapping the restricted machine learning process data to the drilling well measurement data; compare the restricted machine learning process data with the drilling well measurement data; and identify, based on the comparison, that the drilling well measurement data corresponds to the restricted machine learning process data based on an error smaller than an error threshold. Petition 870250077104, dated 08 / 29 / 2025, pp. 61 / 69

Images