Automated rigsite lithology data extraction

Abstract

A computerized method for evaluating drilling cuttings lithology includes acquiring digital images of drilling cuttings, and segmenting the acquired images at a rig site to delineate individual grains in the images. Features are extracted from selected ones of the delineated grains in the segmented images, and the delineated grains are further assigned to at least one group based on the extracted features at the rig site. A rig site user is prompted to input measured data, labels, or descriptions for the at least one group. The segmented images, extracted features, assigned groupings, and user inputs are electronically transmitted to a cloud computing resource that performs lithology classification of selected ones of the delineated grains and generates an integrated summary result of the lithology classification based on the extracted features and the user inputs.

Description

Cross-reference to related applications

[0001] none. Background Technology

[0002] In underground drilling operations such as those used for oil and gas exploration and production, formation cuttings are generated by drilling activities. These cuttings have long been assessed and characterized to better understand the properties and structure of subsurface rocks. In recent years, there has been a desire to perform such assessments and characterization automatically (or semi-automatically) to reduce labor costs and shorten interpretation turnaround times. However, achieving such automation is far from routine.

[0003] The automated evaluation and characterization of digital drilling cuttings images is particularly challenging because individual cuttings encompass a wide variety of sizes, shapes, textures, and colors. A large number of drilling cuttings images can be acquired during a single drilling operation. Furthermore, each image includes numerous cuttings particles mixed together on a tray, with individual particles sometimes in contact with and partially overlapping each other. Significant electronic storage and computing power are required to extract the particle features needed for subsequent AI-based processing. Managing these processing power requirements and providing reliable and timely results for each of many drilling operations globally is an extremely challenging problem. Improved automated methods for evaluating and classifying drilling cuttings images are needed within the industry. Attached Figure Description

[0004] To gain a more complete understanding of the disclosed subject matter and its advantages, reference is now made to the following description taken in conjunction with the accompanying drawings: Figure 1 An example drilling rig, including an example system for evaluating rock cuttings, is depicted.

[0005] Figure 2 A flowchart is depicted for an example method for generating cuttings descriptions during drilling operations.

[0006] Figure 3A and Figure 3B (Collectively referred to as Figure 3) depicts a flowchart of another example method for generating cuttings descriptions during drilling operations.

[0007] Figure 4 Example digital images of rock fragments, including a first group of rock fragments and a second group of rock fragments, are depicted. Detailed Implementation

[0008] A computerized method and system for evaluating the lithology of drilling cuttings are disclosed. An example implementation includes: acquiring a digital image of the drilling cuttings at a drilling site; segmenting the acquired digital image at the drilling site to delineate individual particles within the digital image; extracting features from selected particles from the delineated particles in the segmented image at the drilling site; further assigning the delineated particles to at least one group at the drilling site based on the extracted features; prompting a user at the drilling site to input measurement data, labels, or descriptions for the at least one group; and electronically transmitting the segmented image, extracted features, assigned groups, and user input to a cloud computing resource, which performs lithological classification of the selected particles from the delineated particles based on the extracted features and the user input, and generates an integrated summary result of the lithological classification.

[0009] As is known to those skilled in the art, drilling underground wells (e.g., in oil and gas exploration and / or production operations) produces a large quantity of rock cuttings that are available at the drilling rig site. In current drilling operations, these cuttings are typically evaluated by geologists, rock physicists, and / or other experts at the drilling rig site, for example, to assess the well's local lithology and oil-bearing potential. Cuttings description and evaluation are generally manual and repetitive tasks performed at the drilling rig site. The typical workflow requires rig personnel to identify each lithology type by performing chemical and / or physical measurements on cuttings samples. In identifying the lithology type in the cuttings samples, operators describe rock characteristics (color, hardness, grain size, etc.) on paper, for example.

[0010] Recently, there has been a strong desire to implement automated assessment of cuttings (e.g., by evaluating digital images of cuttings using artificial intelligence and neural networks). One incentive for implementing automated methods is to remove costly experts from the drilling site. While such automated processing and assessment is promising, it also presents many difficulties as described above in the background section of this disclosure. The disclosed implementation advantageously distributes the required computer processing between the drilling site and cloud processors. By distributing some of the heavy processing to the drilling site, the disclosed implementation tends to reduce the data accumulation at a single remote location (the cloud), and thereby tends to reduce latency and promote efficient and reliable service (in terms of service availability). As described in more detail below, the disclosed implementation can also adapt to (even prompt) user input, which tends to improve the overall accuracy of lithological classification.

[0011] Figure 1An example drilling rig 20 including system 100 is depicted for evaluating rock cuttings particles removed from circulating drilling fluid on the rig. The drilling rig 20 may be positioned above subsurface formations (not shown). The drilling rig 20 may include, for example, a derrick and lifting equipment (also not shown) for raising and lowering a drill string 30, which, as shown, extends into a wellbore 40 and includes, for example, a drill bit 32 and one or more downhole measurement tools 38 (e.g., logging-while-drilling tools or measurement-while-drilling tools) in a bottom hole assembly (BHA) above the drill bit 32. Suitable drilling systems, including drilling, steerable, logging, and other downhole tools, are well known in the art.

[0012] Although Figure 1 An onshore drilling rig 20 is depicted, but it should be understood that the disclosed embodiments are equally well applicable to onshore or offshore drilling rigs. As known to those skilled in the art, offshore drilling rigs typically comprise a platform deployed atop a riser extending from the seabed to the water surface. The drill string extends downward from the platform, through the riser, and into the wellbore via a blowout preventer (BOP) located on the seabed. The disclosed embodiments are not limited to these methods.

[0013] The drilling rig 20 also includes a surface system 50 for controlling the flow of drilling fluid used on the rig (e.g., when drilling the wellbore 40). In the illustrated example rig, drilling fluid 35 is pumped downhole, for example, via a conventional mud pump 57 (as depicted at 62). Drilling fluid 35 may be pumped to the drill string 30, for example, en route via riser 58 and mud hose 59. Drilling fluid 35 typically emerges from the drill string 30 at or near the drill bit 32, creating an upwelling flow 64 of mud through the wellbore annulus 42 (the annular space between the drill string and the wellbore wall). Drilling fluid 35 then flows through return pipe 52 to a mud pit system 56, where it can be recirculated. It should be understood that the terms drilling fluid and mud are used synonymously herein.

[0014] The circulating drilling fluid 35 is designed to perform a number of functions during drilling operations, one of which is to carry drilling cuttings 45 to the surface (in the upwelling 64). Drilling cuttings 45 are typically removed from the returning mud via a shale vibrating screen 55 (or other similar solids control equipment) in a return duct (e.g., immediately upstream of the mud pit 56). Formation gases released during drilling can also be carried to the surface in the circulating drilling fluid. These gases are typically removed from the drilling fluid, for example, via a degasser or gas trap 54 located in or near a top tank 53, which in the example depiction is immediately upstream of the shale vibrating screen 55. The drilling cuttings 45 can be evaluated to characterize the subsurface formation and / or estimate its various properties, as described in more detail below.

[0015] Drilling rig 20 may include system 100 configured to capture and evaluate digital images of drilling cuttings, as described in more detail below. System 100 advantageously employs an embedded computing unit 110 deployed at the drilling rig site (e.g., in field laboratory 80). As depicted, embedded computing unit 110 may be configured to communicate electronically with cloud computing resource 120. The system workflow executed in embedded computing unit 110 may be configured to combine artificial intelligence (AI) processing of digital images of drilling cuttings particles with measured cuttings properties input by rig personnel. By merging user input (e.g., as labels) with AI data output by computing unit 110, the intelligent workflow can pre-compute image data and feature extraction, thereby reducing the computational demands on cloud resource 120.

[0016] It should be understood that, in order to perform the functions described herein, system 100 (including embedded unit 110 and cloud resource 120) may include computer hardware and software configured to automatically or semi-automatically evaluate cuttings images. To perform these functions, the hardware may include one or more processors (e.g., microprocessors) that may be connected to one or more data storage devices (e.g., hard disk drives or solid-state drives). As known to those skilled in the art, the processor may also be connected to a network interface, for example, to receive images from a networked camera system (not shown) or another computer system, and to enable communication with various surface and / or downhole sensors, meters, etc., and with actuators and other surface and / or downhole equipment. Furthermore, the network interface may provide communication between embedded unit 110 and cloud resource 120. Of course, it should be understood that the disclosed embodiments are not limited to any particular network configuration, computer hardware, and / or software usage or configuration.

[0017] Continue to refer to Figure 1The drilling rig 20 (e.g., laboratory 80) can be configured to acquire digital images of selected rock cutting particles and measure a wide range of physical properties of the particles, and therefore can include a variety of measuring tools and instruments. For example, the drilling rig may include a calibrated image acquisition system comprising a digital camera, lenses, a light source, a rock cutting sample tray, and a computer system configured to capture and store digital images. Furthermore, the drilling rig 20 may include hardness or microhardness tools and a porosity meter configured to measure the hardness of individual rock cutting particles and to measure the porosity of individual rock cutting particles or groups of similar rock cutting particles. The drilling rig 20 may also include ultraviolet (UV) fluorescence measurement equipment and imaging equipment for measuring and classifying the particle size of rock cutting particles. The drilling rig 20 may also include other instruments, such as the SLB PetroFecta®, configured to perform, for example, X-ray fluorescence (XRF) elemental analysis, diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) mineralogical measurements, and to capture high-resolution visible and ultraviolet (UV) digital images of the cuttings. These measurements and / or images can be evaluated by trained rig personnel to determine the lithology, size, and grain size of the cuttings. Furthermore, the drilling rig 20 may include a chemical laboratory suitable for performing acid testing, in which the cuttings are exposed to strong acids such as hydrochloric acid or sulfuric acid.

[0018] Now go to Figure 2 This document describes a flowchart of an example method 200 for generating a description of drilling cuttings during drilling operations. Method 200 includes obtaining or photographing digital images of drilling cuttings particles at point 202. These cuttings particles can be generated, for example, during wellbore drilling and can be transported to the surface in circulating drilling fluid, as described above regarding... Figure 1 Described. Digital images acquired at the drilling rig (as shown jointly at 210) can be processed to group individual particles into one or more groups based on particle characteristics, and the drilling crew can be prompted to input measured physical property data and / or descriptions or classifications for one or more particle groups within a particle group (e.g., each particle group within a particle group). The digital images, groupings, and corresponding measured property data can be uploaded to the cloud, where further ML-based processing can be performed to identify particle lithology (or the lithology of each identified group) to determine grain size grades and distributions, extract texture features, and / or generate integrated lists or summaries of particle and group properties (as shown jointly at 220).

[0019] Continue to refer to Figure 2At 212, a segmentation algorithm can be used to process the acquired digital image to obtain a segmented image in which individual image rock cuttings are identified (e.g., highlighted, delineated, or otherwise delineated) within the image rock cuttings. At 214, features and / or properties of the identified particles in the segmented image can be extracted. Feature and / or property extraction can include both AI-based extraction of AI-generated features and non-AI-based extraction of particle features and / or properties (e.g., measurements related to color and shape). At 216, the segmented particles can be grouped based on the extracted features to form one or more particle groups (e.g., groups of particles with common color features and / or other features indicating the possibility of common lithology). Grouping can include both machine learning (ML)-based and non-ML-based processing. At 218, additional properties and measured characteristics of one or more groups identified at 216 can be received. For example, drilling personnel (i.e., users) can be prompted (or given the opportunity) to input additional information, such as measured data or written descriptions of one or more groups. At location 219, intermediate cuttings descriptions (including segmented images, groupings, and any corresponding measured property data and / or user comments) can be uploaded to cloud resource 120. Figure 1 At point 222, an ML-based algorithm can be used to perform lithological classification, and at point 224, the integrated results can be prepared and output (e.g., saved and / or downloaded to the drilling rig).

[0020] Further reference Figure 2 It should be understood that the disclosed embodiments are not limited to any particular image acquisition method at 202. In example embodiments, retrieved rock cuttings (e.g., from a shale vibrating screen or other solid control equipment) can be prepared for imaging, for example, by washing and then drying in an oven and / or sieving or screening the cuttings to remove oversized and / or undersized particles. The rock cuttings can also be placed in a tray, for example, with a high-contrast (bright) background color to enhance particle identification and segmentation in the acquired images. The particle tray can be placed in front of a digital camera (e.g., located in laboratory 80), and high-resolution digital photographs can be taken (e.g., using visible and / or UV light). The image acquisition process can utilize normalized and / or calibrated lighting, color enhancement, magnification, and / or focus / resolution settings. Images can be taken while the rock cuttings are wet or dry, with humidity generally controlled to obtain images of dry rock cuttings. Nevertheless, the disclosed embodiments are not limited to any image acquisition method.

[0021] The acquired images may optionally be labeled with a late arrival depth based on a known or calculated late arrival depth of the cuttings sample. It should be understood that the late arrival depth can be determined based on the time it takes for the drilling cuttings sample to emerge from the well or for the drilling cuttings to be collected from a shale vibrating screen, and the time delay (late arrival time) required to transport the cuttings upwards along the wellbore annulus, to generate the time at which cuttings are generated in the well (e.g., by subtracting the late arrival time from the time the cuttings reach the surface). This time can then be correlated with a measured depth log to determine the late arrival depth.

[0022] At 212, a segmentation algorithm can be used to process the acquired digital image to obtain a segmented image, in which multiple individual rock debris particles are identified (e.g., highlighted or delineated) in the segmented image. In an example embodiment, the segmentation algorithm may employ a mask-based convolutional neural network (Mask R-CNN) such as that disclosed in commonly assigned U.S. Patent 12,131,526. Mask R-CNN can be configured to identify individual rock debris particles in the digital image at 212 and thereby generate the segmented image. For example, Mask R-CNN may produce, for example, bounding boxes and mask images. A bounding box can be defined as a set of xy coordinates in the image indicating an image region containing an object of interest. The bounding box may include a confidence score for each object of interest ranging from 0 to 1 (e.g., where larger values ​​indicate higher relevance confidence). The mask image may indicate (e.g., highlight or otherwise demarcate) regions of interest with confidence scores exceeding a threshold.

[0023] It should be understood that Mask R-CNN is a model architecture belonging to the supervised learning category, meaning it typically requires a training dataset that includes images and corresponding labels. For example, images of rock cuttings of various sizes, shapes, colors, and lithological types can be used to train the model. It should also be understood that the R-CNN model can be continuously retrained during drilling operations or between drilling operations in an oil field. For example, segmentation errors can be identified and corrected, and then used to generate labeled training images that can be used to retrain (or further train) the R-CNN.

[0024] In an example implementation, the segmented image can depict multiple rock cuttings on a tray. Individual particles can be identified, for example, via particle outlines or other delineations. Furthermore, each identified particle can be identified by a corresponding set of pixels in the image. In other words, the segmented image can include pixel-by-pixel segmentation, where each pixel in the image is assigned to either the background or a single individual particle (instance).

[0025] At position 214, the extracted features may include, for example, color-related features, size and shape-related features, and AI-generated features extracted using ML-based algorithms. Color-related features may include, for example, the average (such as mean, median, or mode) red, green, and blue intensities per particle, or the distribution or standard deviation of red, green, and blue intensities and / or average brightness. Color-related features may also include histograms, variances, skewness, and / or kurtosis of red, green, and blue intensities. Furthermore, for infrared and / or ultraviolet images, color-related features may include the average (such as mean, median, or mode) infrared or ultraviolet intensity per particle, or the distribution or standard deviation of infrared or ultraviolet intensity and / or average infrared or ultraviolet brightness. Color-related features may also include histograms, variances, skewness, and / or kurtosis of infrared or ultraviolet intensities.

[0026] Size and shape-related features may include, for example, particle diameter, area, perimeter, maximum axis, minimum axis, particle aspect ratio, and interior angle measurements. Furthermore, geometrically related features may include the spatial relationships of grouped pixels within each particle to extract particle roundness, solidity, elongation, roundness, and / or convex hull area. Individual particles may also be classified as, for example, plate-like, fibrous, round or elliptical particles, acute-angled particles, etc., and within a size range across various size intervals (e.g., based on particle diameter, cross-sectional area, and / or perimeter).

[0027] The extracted features may also include texture-related features, which quantify various spatial relationships and / or directional variations in pixel color and / or brightness within each grain. The extracted texture-related features may include, for example, edge detection, pixel-to-pixel contrast, correlation, and / or entropy. Additionally, in some embodiments, techniques such as image texture filters (e.g., Gabor filters), autoencoders, or other deep learning-based techniques may be used to extract texture-related features. In example embodiments, AI-generated features may include AI-generated texture-related features extracted using deep learning (AI or ML) techniques.

[0028] At position 216, particles can be grouped, for example, based on the “position” of individual particles in a multidimensional space according to extracted color features, size features, shape features, and / or AI-generated features. For example, as described above, a set of particle features can be extracted and / or computed for each of the selected cutting particles. This set of particle features can include substantially any number of suitable features, and sometimes can include a large number of features, for example, up to 16 features (e.g., up to 32, 48, 64, or more features). In an example implementation, particles can be clustered in a multidimensional hyperspace of the extracted and / or computed features. While the disclosed implementations are not limited in this respect, clustering or grouping can include nearest neighbor regression based on the distance between individual cutting particles in the multidimensional hyperspace (e.g., Euclidean distance). It should be understood that when using a nearest neighbor regression method, there is no training step. In such an implementation, a local region or group can be defined by the number of K nearest neighbor reference cutting particles, where K can be substantially any number of nearest neighbors (such as 5, 10, 20, or 50). In other implementations, the grouping at 216 may include (or further include) using an ML algorithm or a retrained ML (or AI) algorithm.

[0029] Continue to refer to Figure 2 The system may prompt (or provide opportunities for) drilling personnel to input particle properties and / or measurement data, as well as other comments or descriptions. It should be understood that input is optional. For example, embedded unit 110 may provide the user with a list of particle groups and corresponding images of representative particles in each group, along with prompts for entering property and / or measurement data. Property and / or measurement data may include virtually any suitable information, such as hardness measurements, porosity measurements, UV fluorescence measurements, particle size measurements, XRF elemental analysis (or a summary thereof), DRIFTS mineralogical measurements, and / or results of acid tests in which the particles may have undergone acid (e.g., hydrochloric acid or sulfuric acid) and observed particle reactivity. It should be understood that, depending on the nature of the measurement, the measured properties may be quantitative and / or qualitative. In an example embodiment, inputting particle properties and / or measurement data may also include user-input lithological tags, such as sandstone, limestone, etc., such that each particle group or particle cluster receives a corresponding lithological tag.

[0030] At 222, the cloud resource is configured to use an ML-based algorithm to provide lithological classification for each particle (or each group of particles) within a particle group. In an example implementation, an ML-based algorithm can be used to process segmented images and extracted features to classify identified cuttings particles and / or groups of particles with corresponding lithological types, for example, as described in co-assigned U.S. Patent Application Publication 20230220770A1. In an example implementation, a convolutional neural network (CNN) or transformer model can be trained on a dataset of drilling cuttings sample images collected from lithologies around the world or on images obtained from sample images collected from lithologies obtained locally (e.g., in a local basin or oil field). Training images can be manually annotated at both the pixel level (assigning a rock type label to each image pixel) and / or at the rock or particle level (assigning a rock type label to each rock or particle in the image).

[0031] It should be understood that the lithology of a rock or rock fragment is a description of its physical properties as visible at outcrops, in hand or core samples, or under a low-power microscope. Lithology can refer to a detailed description of these physical properties or a summary of the overall physical properties of the rock. In a second sense, the lithology of a rock (or rock fragment) refers to the type of rock or the overall (or macroscopic) identification or classification of the rock. The term lithology is generally used in this second sense. Examples of lithologies or lithology types in this second sense include sandstone, limestone, slate, shale, basalt, coal, anhydrite, dolomite, gypsum, clay, flint, granite, etc. Therefore, the extracted features (such as color and / or texture features) represent the physical properties that can be measured, quantified, and evaluated from digital images to determine the lithology of a formation (in the second sense). For example, trained models (such as trained CNNs or transformer models) can be used to evaluate the color and / or texture features extracted from selected grains in a segmented image to classify the lithology type of those selected grains.

[0032] Figure 3A and Figure 3B (Collectively referred to as Figure 3) depicts a flowchart of another example method 300 for generating a description of drilling cuttings during drilling operations. Method 300 is similar to method 200 in that it includes: acquiring or photographing digital images of drilling cuttings particles at 302, and processing the acquired images at the drilling rig site (such as in...). Figure 3A(As shown in 310 common locations in the image), to identify grain groups and prompt well crew personnel to acquire and / or input measured physical property data for one or more grain groups (e.g., each grain group within a grain group). Digital images, groupings, and corresponding measured property data can be uploaded to the cloud, where further ML-based processing can be performed to identify grain lithology (or the lithology of each identified group), determine grain size class and distribution, extract texture features, and / or generate an integrated list or summary of grain and group properties (as shown in the image). Figure 3B (As shown in 350 places in the text).

[0033] exist Figure 3A In the process, digital images of drilling cuttings particles were acquired at location 302, for example, as mentioned above. Figure 2 As described, an image can be segmented at 304 to obtain a segmented image that includes the identified (e.g., highlighted, delineated, or otherwise delineated) rock fragments, as also described above. Figure 2 As described. In the example implementation, a segmentation quality control routine can be performed at 306 before outputting the segmented image. The segmentation quality control routine may include, for example, comparing the segmented particle geometry (e.g., size and shape) with predefined physical constraints and statistical representations of a typical rock cutting instance. In other words, the segmented instances may be compared with various upper and lower geometric thresholds that enable the identification of poor segmentation results (e.g., a single instance comprising multiple particles or a single particle being divided into multiple instances). In the example implementation, the quality control routine may also prompt the user for input and / or verification if necessary. Such an approach can improve the overall segmentation quality and also ensure that the grouping is both robust and user-verified.

[0034] The segmented image (including identified rock fragments) can be further processed at points 312, 314, and 316 to extract and / or measure various features of the identified fragments, and in doing so, output a set of alphanumeric data as intermediate processing results for image analysis. For example, as mentioned above regarding... Figure 2 As described, an ML algorithm can be used at position 312 to extract AI-generated features, size and shape-related features can be extracted and / or measured at position 314, and color-related features can be extracted and / or measured at position 316. It should be understood that color, size, and shape-related features generally do not require ML or AI algorithms.

[0035] The extracted features can also be evaluated at 320 to group particles into one or more groups with common or similar features. As described above, particles can be grouped at 320 based on the “position” of individual particles in a multidimensional space of extracted color features, size features, shape features, and AI-generated features. It should be understood that in some implementations, certain features among the extracted features may be given priority or heavier weighting compared to other features. By way of example only, in some implementations, color features may be given priority.

[0036] Figure 4 Example digital images depicting first and second groups of lithic fragments corresponding to the first and second lithologies are provided. In the depicted example images, the fragments are grouped using features extracted at 312, 314, and 316. The example first and second groups are indicated by labels 402 and 404. The tray background is indicated by label 406. Although... Figure 4 Black and white (grayscale) images are depicted, but it should be understood that the disclosed embodiments typically employ high-resolution color images of rock fragments.

[0037] Continue to refer to Figure 3A At point 330, particle properties and / or measurement data can be entered. In an example implementation, software built on top of the AI ​​layer can interact with the user to obtain additional information related to cuttings particles, to label data and particle groups, and thereby improve the quality of the analysis. For example, it can prompt well crew personnel to optionally enter data along with other comments or written particle descriptions. As described above, embedded unit 110 can provide the user with a list of particle groups, along with corresponding images of representative particles in each group, along with prompts for entering data and / or descriptions. (The above refers to...) Figure 2 The example input is described.

[0038] At point 340, the segmented images, extracted features, groupings, and user input information can be combined as intermediate results. These intermediate results can then be uploaded to the cloud or other central processing facility for further processing, as depicted at point 345. It should be understood that the intermediate results described above can be transmitted to an off-site location at virtually any suitable time or measured depth interval. In example embodiments, intermediate results can be uploaded one segmented image at a time (e.g., segmented images one by one). In other embodiments, intermediate results can be uploaded less frequently, for example, in batches of multiple segmented images, or at predetermined time or depth intervals including multiple segmented images (and corresponding data). The disclosed embodiments are explicitly not limited in this respect.

[0039] It should also be understood that intermediate results do not necessarily include user input information for every segmented image. Nor do they necessarily include user input information for every group identified in the segmented images. Furthermore, in some implementations, it may be advantageous to add more extensive input measurements when identifying new or previously unidentified groups of rock debris (e.g., representing new strata or subsurface layers).

[0040] exist Figure 3B At point 355, the uploaded intermediate results are received. User input data 330 can be added to the reference database 358. Therefore, user input (input data and comments) can be considered as labels and can be merged with AI-generated data. At point 360, AI-based lithology classification is performed to obtain labeled images, where selected lithic particles in the segmented lithic grains are labeled with their corresponding lithology type 362. For example, a trained CNN or transformer model can be used to perform the classification, as described above. Figure 2 As described, the segmented image (along with lithological classification) can also be evaluated at 370 to perform a grain size assessment of selected lithic particles (e.g., particles with a specific lithological label such as sandstone), and this assessment can be based on extracted color features, size features, shape features, and AI-generated features. The grain size assessment at 370 can output, for example, a grain size level (such as fine-grained, medium-grained, or coarse-grained) at 372 and a grain size distribution level (such as narrow distribution, medium distribution, or wide distribution) at 374.

[0041] Continue to refer to Figure 3B At 380, the received intermediate results can be further evaluated to perform texture measurements on the particles (or texture-related features extracted from the particles) in the segmented image. Texture measurements can be designed to quantify various spatial relationships and / or directional variations in pixel color and / or brightness within each particle. The extracted texture-related features can include, for example, edge detection, pixel-to-pixel contrast, correlation, and / or entropy. Additionally, in some implementations, techniques such as image texture filters (e.g., Gabor filters), autoencoders, or other deep learning-based techniques can be utilized to extract texture-related features. Furthermore, at 382, ​​directional variations can be evaluated, for example, with respect to symmetry and used to generate a spectrum, which can be compared to a reference spectrum to classify textures and assign them to each particle, such as homogeneous, heterogeneous, granular, laminated particles, etc. While in Figure 3B Not depicted in the text, grain size analysis at 370 can also consider texture features to characterize the grain size and / or grain size distribution within the rock fragments. The extracted features and user input can optionally be further evaluated in the cloud to perform additional feature extraction and / or classification related to the geometry, mechanical, and geological properties of the rock fragments, such as fragment shape, hardness, and porosity.

[0042] An integrated summary of the results can be generated at position 390, and may include, for example, segmented images, features extracted at positions 312, 314, and 316, and texture-related features extracted at position 390, lithological classification, and grain size and grain size distribution classification. The integrated summary of results can be saved to memory and / or electronically transmitted back to the drilling site where the digital images were acquired.

[0043] It should be understood that the disclosed systems and computerized methods for evaluating drilling cuttings images offer several advantages. For example, each embedded unit (and its corresponding drilling site) performs its own analysis, where the heavy (processor-intensive) computations are not all performed at the same location. This tends to reduce data accumulation at a single remote location (the cloud), and thus tends to reduce latency and promote efficient and reliable service (in terms of service availability). Furthermore, each embedded unit is autonomous and delivers pre-analyzed data in addition to segmented images. Such autonomy advantageously promotes adaptability, which in turn allows local processing to be tailored to local lithology and cuttings properties (e.g., by retraining or supplementing local AI algorithms using local knowledge and corresponding labels).

[0044] It should be understood that this disclosure includes numerous implementations. These implementations include, but are not limited to, the following implementations.

[0045] In a first embodiment, a computerized method for assessing the lithology of drilling cuttings includes: acquiring a digital image of drilling cuttings at a drilling site; segmenting the acquired digital image at the drilling site to delineate individual particles in the digital image; extracting features from selected particles from the delineated particles in the segmented image at the drilling site; assigning the delineated particles to at least one group at the drilling site based on the extracted features; prompting a user at the drilling site to input measurement data, labels, or descriptions for the at least one group; electronically transmitting the segmented image, extracted features, assigned groups, and user input to a cloud computing resource; using the cloud computing resource to perform a lithological classification of the selected particles in the delineated particles, the lithological classification being based on the extracted features and the user input; and generating an integrated summary result of the lithological classification of the segmented image.

[0046] The second implementation may include the first implementation, wherein acquiring the digital image includes: drilling a subsurface wellbore; collecting drilling cuttings from circulating drilling fluid; preparing the drilling cuttings for imaging; and taking the digital image of the prepared drilling cuttings at the drilling rig site.

[0047] The third implementation may include any one of the first to second implementations, wherein the segmented image includes pixel-by-pixel segmentation, wherein each pixel in the image is assigned to a single particle in the background or defined particles.

[0048] The fourth implementation scheme may include any one of the first to third implementation schemes, wherein the extracted properties include color-related features, size and shape-related features, and AI-generated features.

[0049] The fifth implementation may include any one of the first to fourth implementations, wherein assigning the delineated particles to at least one group further includes: clustering the delineated particles in a multidimensional hyperspace of selected features among the extracted features.

[0050] The sixth implementation may include any one of the first to fifth implementations, wherein the user input includes at least one of the following: hardness measurement, fluorescence measurement, chemical reactivity with acid measurement, and lithological description.

[0051] The seventh implementation scheme may include any one of the first to sixth implementation schemes, wherein the user input is stored in a reference database in the cloud computing resource, and the lithological classification is based on the extracted features and the user input in the reference database.

[0052] The eighth implementation may include any one of the first to seventh implementations, wherein the cloud resources are used to further process the segmented image to extract texture-related features that quantify the spatial relationships or directional variations of pixel color or brightness in each of the delineated particles.

[0053] The ninth implementation may include any of the first to eighth implementations, and also includes using cloud resources to perform a grain size assessment of selected grains in the defined grains, the grain size assessment being based on the lithological classification and extracted features, the grain size assessment outputting a grain size classification and a grain size distribution classification.

[0054] The tenth implementation may include any one of the first to ninth implementations, wherein the integrated summary result also includes texture-related features extracted from the cloud resource and granularity classification and granularity distribution classification determined by the cloud resource, wherein the texture-related features quantify the spatial relationship or directional variation of pixel color or brightness in each of the defined granules.

[0055] In an eleventh embodiment, a system for assessing the lithology of drilling cuttings includes: a digital camera located at a drilling rig site, the digital camera being configured to acquire digital images of drilling cuttings; a first processor located at the drilling rig site, the first processor being configured to: segment the acquired digital images to delineate individual particles and thereby generate a segmented image; extract features from selected particles among the delineated particles in the segmented image; assign the delineated particles to at least one group based on the extracted features; prompt a drilling rig site user to input measurement data, labels, or descriptions of the delineated particles in the at least one group; and electronically transmit the segmented image, the extracted features, the assigned groupings, and the user input to a cloud location; and a second processor located at the cloud location, the second processor being configured to: receive the electronic transmission from the first processor; perform a lithological classification of the selected particles among the delineated particles, the lithological classification being based on the extracted features and the user input; and generate an integrated summary result of the lithological classification of the segmented image.

[0056] The twelfth implementation may include the eleventh implementation, wherein the first processor is configured to: automatically segment the acquired digital image, extract the features, assign the delineated particles to at least one group, notify the drilling site user, and electronically transmit the segmented image, the extracted features, the assigned groups, and the user input to the cloud location; and the second processor is configured to: automatically receive the electronic transmission, perform the lithological classification, and generate the integrated summary result.

[0057] The thirteenth embodiment may include any one of the eleventh to twelfth embodiments, wherein the first processor is configured to segment the acquired digital image by pixel-by-pixel segmentation, wherein each pixel in the image is assigned to a single particle in the background or defined grains.

[0058] The fourteenth implementation may include any one of the eleventh to thirteenth implementations, wherein the first processor is configured such that the extracted features include color-related features, size and shape-related features, and AI-generated features.

[0059] The fifteenth implementation may include any one of the eleventh to fourteenth implementations, wherein the first processor is configured to assign the delineated particles by clustering the delineated particles in a multidimensional hyperspace of selected features among the extracted features.

[0060] The sixteenth embodiment may include any one of the eleventh to fifteenth embodiments, wherein the first processor is configured to prompt the user with at least one of the following: hardness measurement, fluorescence measurement, chemical reactivity with acid measurement, and lithological description.

[0061] The seventeenth implementation may include any one of the eleventh to sixteenth implementations, wherein the second processor is further configured to: save the user input to a reference database in the cloud location, and perform the lithological classification based on the extracted features and the user input in the reference database.

[0062] The eighteenth embodiment may include any one of the eleventh to seventeenth embodiments, wherein the second processor is further configured to process the segmented image to extract texture-related features that quantify the spatial relationships or directional variations of pixel color or brightness in each of the delineated particles.

[0063] The nineteenth implementation may include any one of the eleventh to eighteenth implementations, wherein the second processor is further configured to perform a grain size assessment of selected grains in the delineated grains, the grain size assessment being based on the lithological classification and extracted features, the grain size assessment outputting a grain size classification and a grain size distribution classification.

[0064] The twentieth implementation may include any one of the eleventh to nineteenth implementations, wherein the integrated summary result also includes texture-related features extracted from the cloud resource and granularity classification and granularity distribution classification determined by the cloud resource, wherein the texture-related features quantify the spatial relationship or directional variation of pixel color or brightness in each of the defined granules.

[0065] Although the extraction of lithological data from an automated drilling rig has been described in detail, it should be understood that various changes, substitutions and modifications may be made herein without departing from the spirit and scope of this disclosure as defined by the appended claims.

Claims

1. A computerized method for evaluating the lithology of drilling cuttings, the method comprising: Digital images of drilling cuttings were acquired at the drilling rig site; The acquired digital images are segmented at the drilling site to delineate individual particles within the digital images; Features are extracted from selected particles in the segmented image at the drilling site; At the drilling site, the defined particles are assigned to at least one group based on the extracted features; The drilling rig site user is prompted to input measurement data, labels, or descriptions for at least one group; The segmented images, extracted features, assigned groups, and user input are transmitted electronically to cloud computing resources. The cloud computing resources are used to perform lithological classification of selected particles within the defined particles, the lithological classification being based on extracted features and the user input; as well as Generate an integrated summary of the lithological classification of the segmented image.

2. The method according to claim 1, wherein acquiring the digital image comprises: drilling underground wellbore; Collect drilling cuttings particles from circulating drilling fluid; Prepare the drill cuttings particles for imaging; as well as The digital image of the prepared drilling cuttings particles was taken at the drilling rig site.

3. The method of claim 1, wherein the segmented image comprises pixel-by-pixel segmentation, wherein each pixel in the image is assigned to a single particle in the background or defined grains.

4. The method according to claim 1, wherein the extracted properties include color-related features, size and shape-related features, and AI-generated features.

5. The method of claim 1, wherein assigning the designated particles to at least one group further comprises: Cluster the defined particles in the multidimensional hyperspace of the selected features among the extracted features.

6. A system for evaluating the lithology of drilling cuttings, the system comprising: A digital camera, located at the drilling rig site, is configured to acquire digital images of drill cuttings. A first processor, located at the drilling site, is configured to: segment the acquired digital image to delineate individual particles and thereby generate a segmented image; and extract features from selected particles among the delineated particles in the segmented image. The defined particles are assigned to at least one group based on the extracted features. The system prompts the drilling rig site user to input measurement data, labels, or descriptions of the delineated particles in at least one group; and electronically transmits the segmented image, extracted features, assigned groups, and user input to the cloud location. as well as A second processor, located at the cloud location, is configured to: receive electronic transmissions from the first processor; perform lithological classification of selected particles from the delineated particles, the lithological classification being based on extracted features and the user input; and generate an integrated summary result of the lithological classification of the segmented image.

7. The system according to claim 6, wherein: The first processor is configured to: automatically segment the acquired digital image, extract the features, assign the defined particles to at least one group, notify the drilling rig site user, and electronically transmit the segmented image, the extracted features, the assigned groups, and the user input to the cloud location; and The second processor is configured to automatically receive the electronic transmission, perform the lithological classification, and generate the integrated summary results.

8. The system of claim 6, wherein the first processor is configured to prompt the user with at least one of the following: hardness measurement, fluorescence measurement, chemical reactivity with acid measurement, and lithological description.

9. The system of claim 6, wherein the second processor is further configured to: The user input is saved to a reference database in the cloud location, and the lithology classification is performed based on the extracted features and the user input in the reference database. The segmented image is processed to extract texture-related features, which quantify the spatial relationships or directional changes in pixel color or brightness in each of the defined particles; as well as Perform a grain size assessment on selected grains within the defined grains, the grain size assessment being based on the lithological classification and the extracted features, the grain size assessment outputting a grain size classification and a grain size distribution classification.

10. The system of claim 6, wherein the integrated summary result further includes texture-related features extracted from cloud resources and granularity classification and granularity distribution classification determined by the cloud resources, wherein the texture-related features quantify the spatial relationship or directional variation of pixel color or brightness in each of the defined granules.

Images