Machine learning model-based borehole breakout detection system

Abstract

The present invention relates to a machine learning model-based borehole breakout detection system comprising: a second input material generation unit (800) configured to receive borehole ultrasonic imaging log data for learning and borehole ultrasonic imaging log data for inspection, in which an X-axis represents an angle of 0 to 360 degrees, a Y-axis represents depth, and an amplitude is represented by color, and generate input layer data for learning, input layer data for inspection, and output layer data for learning; a borehole breakout depth unit detection model generation unit (900) configured to generate a machine learning model trained by a dataset of the input layer data for learning and the output layer data for learning; and a borehole breakout depth unit detection unit (1000) configured to obtain data on whether a depth unit borehole breakout is present by inputting the input layer data for inspection to the machine learning model generated from the borehole breakout depth unit detection model generation unit.

Description

Machine learning model-based compression and crushing zone detection system

[0001] The present invention relates to a machine learning model-based compression fracture zone detection system, and more specifically, to a machine learning model-based compression fracture zone detection system configured to receive training borehole ultrasonic image logging data and inspection borehole ultrasonic image logging data, and to obtain depth-unit data on the existence of a compression fracture zone, coordinate and size information (x, y, w, h) of a bounding box surrounding the compression fracture zone, compression fracture zone attribute information per bounding box, and compression fracture zone attribute information per pixel.

[0002] The characteristics of stress acting on the Earth's crust (magnitude, direction, anisotropy, etc.) are important factors in understanding the stability of underground space development and the mechanisms of seismic and fault activity. Recently, as shallow crust less than 1 km deep is being developed for various purposes such as underground carbon dioxide storage and high-level nuclear waste disposal facilities, identifying stress in underground spaces is directly linked to stability. Stress fields in the shallow crust are measured using various geotechnical methods, such as hydraulic fracturing, overcoring, fault line inversion, and borehole stress indicators. Among these, the borehole stress indicator method is widely used because it allows for the relatively rapid and continuous acquisition of stress field information through compressional or tensile fracture zones observable in borehole walls. Stress indicators appearing in borehole walls include drilling-induced tensile fractures and borehole breakouts. These fracture zones occur where stress around the borehole relaxes or concentrates during the drilling process, providing information on the direction and magnitude of on-site stress. Among these, compressional fracture zones occur when stress concentrates in a direction orthogonal to the direction of maximum horizontal principal stress, and the stress magnitude at that point exceeds the rock strength. The presence of compressional fracture zones can be determined by analyzing amplitude and travel time data acquired through ultrasound imaging logging. Due to the characteristics of principal stress, compressional fracture zones occur at 180° intervals; in this case, the amplitude decreases and the travel time is measured to be large. Therefore, the direction of the maximum horizontal principal stress at different depths can be inferred from the logging data.

[0003] Generally, compression fracture zones are identified manually by experts based on image logs. This process is subject to variations in analysis results depending on the expert's proficiency, and it is difficult to exclude subjectivity. Furthermore, the precision of the analysis can vary depending on the given time and budget, and there was a problem in that analysis required a significant amount of time, particularly for high-density drilling or deep-depth data with large volumes.

[0004] To address these issues, Dias et al. (2020) developed a model using a deep learning-based Fast-RCNN (fast-region convolution neural network) technique to improve the efficiency of manual image log analysis. This model automatically detects compression fracture zones and fractures in image logs.

[0005] However, models using the deep learning-based Fast-RCNN technique had a problem in that the application of deep learning models was limited when available data was scarce.

[0006] [Prior Art Literature]

[0007] [Non-patent literature]

[0008] (Non-patent Document 1) Dias, LO, Bom, CR, Faria, EL, Valentin, MB, Correia, MD, de Albuquerque, MP, de Albuquerque, MP, and Coelho, JM, 2020. Automatic detection of fractures and breakouts patterns in acoustic borehole image logs using fast-region convolutional neural networks, Journal of Petroleum Science and Engineering, 191, 107099.

[0009] Accordingly, the present invention has been made in consideration of the above-mentioned situation, and the objective of the present invention is to provide a machine learning model-based compression fracture zone detection system that can improve false detections occurring in object detection models (problems where a compression fracture zone is predicted to exist when it does not), generate a machine learning model for compression fracture zone detection even with limited available training data, and improve the detection precision of compression fracture zones even when compression fracture zones exist at both ends of inspection image data.

[0010] To achieve the above objective, a machine learning model-based compression fracture zone detection system according to an embodiment of the present invention comprises: a second input data generation unit configured to generate training input layer data, inspection input layer data, and training output layer data by receiving training borehole ultrasonic image logging data and inspection borehole ultrasonic image logging data, wherein the X-axis represents an angle of 0 to 360 degrees, the Y-axis represents depth, and the color represents amplitude; and a compression fracture zone depth unit detection model generation unit configured to generate a machine learning model learned by the data set of the training input layer data and the training output layer data. The invention includes a compression fracture zone depth unit detection unit configured to acquire depth unit compression fracture zone existence data by inputting the inspection input layer data into a machine learning model generated from the compression fracture zone depth unit detection model generation unit; wherein the learning input layer data and the inspection input layer data are amplitude value data for all orientations at depths (mn) to (m+n), and the learning output layer data is data on the presence of a compression fracture zone at a depth of (m), wherein m represents the detection depth and n represents an arbitrary number of depths.

[0011] A machine learning model-based compression fracture zone detection system according to the above embodiment comprises: a first input data generation unit that receives the training borehole ultrasonic image logging data and the inspection borehole ultrasonic image logging data and generates a plurality of training image data having a set interval from the top and bottom of the compression fracture zone and a plurality of inspection image data separated by a set depth interval; an image selection unit configured to receive depth-unit compression fracture zone existence data obtained from the compression fracture zone depth-unit detection unit and to select a plurality of inspection image data in which a compression fracture zone exists among the plurality of inspection image data; a preprocessing unit configured to preprocess the plurality of training image data and the selected inspection image data to generate grayscale training image data and inspection image data having a set resolution; a data augmentation unit configured to receive the grayscale training image data having a set resolution and to generate augmented grayscale training image data; and a rolling unit configured to receive the preprocessed inspection image data, translate the angle column with the lowest average amplitude value to the first column, and then translate it to the right. It may further include: a compression fracture zone object unit detection model generation unit configured to generate a machine learning model learned from a dataset of grayscale training image data augmented by the data augmentation unit and coordinate and size information (x, y, w, h) of a bounding box surrounding the compression fracture zone of the augmented grayscale training image data; and a compression fracture zone object unit detection unit configured to acquire coordinate and size information (x, y, w, h) of a bounding box surrounding the compression fracture zone by inputting inspection image data translated by the rolling unit into the machine learning model generated from the compression fracture zone object unit detection model generation unit.

[0012] The machine learning model-based compression crushing zone detection system according to the above embodiment may further include a post-processing unit configured to receive coordinate and size information (x, y, w, h) of a bounding box from the compression crushing zone object unit detection unit and receive translated inspection image data and translation information from the rolling unit, and to post-process to generate compression crushing zone attribute information per bounding box and compression crushing zone attribute information per pixel unit.

[0013] The machine learning model-based compression fracture zone detection system according to the above embodiment may further include a result display unit configured to display one or more of the following: depth-unit compression fracture zone existence data obtained from the compression fracture zone depth-unit detection unit, bounding box coordinate and size information (x, y, w, h) obtained from the compression fracture zone object-unit detection unit, compression fracture zone attribute information per bounding box and compression fracture zone attribute information per pixel unit generated by the post-processing unit.

[0014] In a machine learning model-based compression crushing zone detection system according to the above embodiment, the preprocessing unit may include: a resolution adjustment unit that receives the plurality of training image data and the selected inspection image data and adjusts them to have a set resolution; and a grayscale conversion unit that converts the training image data and inspection image data having a set resolution into a plurality of grayscale training image data and inspection image data.

[0015] In a machine learning model-based compression fracture zone detection system according to the above embodiment, the post-processing unit receives bounding box coordinate and size information (x, y, w, h) from the compression fracture zone object unit detection unit and receives translated inspection image data and translation information from the rolling unit, and calculates the center depth, azimuth angle, opening angle, and length of the compression fracture zone for each bounding box using the input bounding box coordinate and size information, translated inspection image data, and translation information; and a clustering model that classifies compression fracture zone image data classified based on bounding boxes in the translated inspection image data into two types by using a clustering algorithm to assign a cluster with a relatively small average pixel value in the image data as 1 and a cluster with a relatively large average value as 0. and may include a second compression fracture zone attribute information calculation unit that calculates the median value by depth of the compression fracture zone clusters in the image data clustered into the two types above, and calculates the center depth, azimuth angle, opening angle, and length of the compression fracture zones on a pixel-by-pixel basis in the clustered image data in which the median value was calculated.

[0016] In the machine learning model-based compression crushing zone detection system according to the above embodiment, the center depth, azimuth angle, opening angle, and length of the compression crushing zone for each bounding box calculated by the first compression crushing zone attribute information calculation unit can be calculated by the following [Equation 1].

[0017] [Mathematical Formula 1]

[0018] Center depth (depth, m) = Start depth (start depth) + (y × tm)

[0019] Azimuth (°) = x × 360° + roll

[0020] Opening angle (°) = ω × 360°

[0021] Length (m) = h × tm

[0022] [Here, the starting depth represents the top depth of the image data output from the rolling unit, t represents the vertical length of the image data output from the rolling unit, x and y represent the center coordinates of the bounding box, w and h represent the horizontal and vertical lengths of the bounding box, and roll represents information performed by the rolling unit, indicating that the column with the lowest average amplitude value among columns from 0 to 360 degrees in the image data from the rolling unit is translated to the first column and then translated again to the right.]

[0023] In the machine learning model-based compression fracture zone detection system according to the above embodiment, the center depth, azimuth angle, opening angle, and length of the compression fracture zone per pixel unit calculated by the second compression fracture zone attribute information calculation unit can be calculated by the following [Equation 2].

[0024] [Mathematical Formula 2]

[0025] Center depth (depth, m) = (depth min +depth max ) / 2

[0026] Azimuth (°) = Average of the median values ​​by depth of the compression fracture zone cluster

[0027] Opening angle (°) = Average value of the opening angle at different depths of the compression fracture zone cluster

[0028] length(length, m) = depth max - depth min

[0029] [Here, depth max and depth min [ indicates the maximum and minimum depths of the compression fracture zone in pixel units]

[0030] According to the machine learning model-based compression fracture zone detection system of the embodiment of the present invention, by utilizing depth-unit compression fracture zone existence data obtained from a compression fracture zone depth-unit detection unit to select multiple inspection image data containing a compression fracture zone among multiple inspection image data, and by configuring the selected multiple inspection image data to be used for compression fracture zone object-unit detection, there is an excellent effect of improving false detection (a problem where a compression fracture zone is predicted to exist when it does not exist) that occurred in object detection models.

[0031] Furthermore, according to the machine learning model-based compression fracture zone detection system of the embodiment of the present invention, a plurality of training image data and inspection image data generated by receiving training borehole ultrasonic image logging data and inspection borehole ultrasonic image logging data are preprocessed to generate grayscale training image data and inspection image data having a set resolution, the generated grayscale training image data is augmented, a machine learning model is generated by the augmented grayscale training image data and a data set of coordinate and size information of a bounding box surrounding the compression fracture zone of the training image data, and inspection image data translated by a rolling part is input to the machine learning model to obtain coordinate and size information (x, y, w, h) of a bounding box surrounding the compression fracture zone, thereby having another excellent effect of being able to generate a machine learning model for compression fracture zone detection even if there is little available training data.

[0032] Furthermore, according to the machine learning model-based compression fracture zone detection system of the embodiment of the present invention, grayscale inspection image data is input, and the column with the lowest average amplitude value among columns from 0 degrees to 360 degrees in the image data is configured to be translated to the first column and then translated again to the right, thereby having another excellent effect of improving the detection precision of the compression fracture zone even if a compression fracture zone exists at both ends of the inspection image data.

[0033] FIG. 1 is a block diagram of a machine learning model-based compression crushing zone detection system according to an embodiment of the present invention.

[0034] Figure 2 is a detailed block diagram of the preprocessing unit of Figure 1.

[0035] Figure 3 is a detailed block diagram of the post-processing section of Figure 1.

[0036] Figure 4 is a diagram showing the learning and inspection borehole ultrasonic image logging data input to the first input data generation unit and the second input data generation unit of Figure 1.

[0037] Figure 5 is a diagram showing the image data for learning and testing generated in the first input data generation unit of Figure 1.

[0038] Figure 6 is a diagram showing grayscale learning and inspection image data converted by the grayscale conversion unit of Figure 2.

[0039] Figure 7 is a diagram showing original image data, which is grayscale learning image data input to the data augmentation unit of Figure 2, translated image data, horizontally flipped image data, and 180-degree rotated image data augmented by the data augmentation unit.

[0040] Figure 8 is a flowchart showing the process performed by the rolling part of Figure 3.

[0041] FIG. 9 is a diagram for explaining the coordinate and size information (x, y, w, h) of the bounding box input to the first compression crushing zone attribute information calculation unit of FIG. 3, the image data in which the compression crushing zone is detected, and the attribute information of the compression crushing zone for each bounding box calculated by the first compression crushing zone attribute information calculation unit.

[0042] Figure 10 is a diagram illustrating the process of clustering by the clustering model of Figure 3, the classification of 1 and 0 values ​​of clustered image data, and the calculation of the median of the compressed fracture zone cluster.

[0043] FIG. 11 is a diagram illustrating the process of calculating attribute information of a compressed fracture zone on a pixel-by-pixel basis from clustered image data in which the median value is calculated by the second compressed fracture zone attribute information calculation unit of FIG. 3.

[0044] FIG. 12 is a flowchart illustrating the process of obtaining data on the presence of a depth-unit compression fracture zone according to an embodiment of the present invention.

[0045] In describing the embodiments of the present invention, if it is determined that a detailed description of known technology related to the present invention may unnecessarily obscure the essence of the present invention, such detailed description will be omitted. Furthermore, the terms described below are defined considering their functions in the present invention, and these may vary depending on the intentions or practices of the user or operator. Therefore, such definitions should be based on the content throughout this specification. Terms used in the detailed description are intended merely to describe the embodiments of the present invention and should not be interpreted restrictively. Unless explicitly stated otherwise, expressions in the singular form include the meaning of the plural form. In this description, expressions such as "include" or "comprise" are intended to refer to certain characteristics, numbers, steps, actions, elements, parts thereof, or combinations thereof, and should not be interpreted as excluding the existence or possibility of one or more other characteristics, numbers, steps, actions, elements, parts thereof, or combinations thereof other than those described.

[0046] In each system illustrated in the drawings, elements in some cases may have the same or different reference numbers, suggesting that the represented elements may be different or similar. However, elements may have different implementations and may operate with some or all of the systems shown or described herein. The various elements illustrated in the drawings may be the same or different. It is optional which is referred to as the first element and which is referred to as the second element.

[0047] In this specification, the phrase “transmits,” “delives,” or “provides” data or signals from one component to another component includes not only the direct transmission of data or signals from one component to another component, but also the transmission of data or signals to another component through at least one other component.

[0048] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[0049] FIG. 1 is a block diagram of a machine learning model-based compression crushing zone detection system according to an embodiment of the present invention.

[0050] A machine learning model-based compression / crushing zone detection system according to an embodiment of the present invention includes, as illustrated in FIG. 1, a second input data generation unit (800), a compression / crushing zone depth unit detection model generation unit (900), a compression / crushing zone depth unit detection unit (1000), a first input data generation unit (100), an image selection unit (150), a preprocessing unit (200), a data augmentation unit (300), a rolling unit (350), a compression / crushing zone object unit detection model generation unit (400), a compression / crushing zone object unit detection unit (500), a postprocessing unit (600), and a result display unit (700). The second input data generation unit (800), the compression crushing zone depth unit detection model generation unit (900), the compression crushing zone depth unit detection unit (1000), the first input data generation unit (100), the image selection unit (150), the preprocessing unit (200), the data augmentation unit (300), the rolling unit (350), the compression crushing zone object unit detection model generation unit (400), the compression crushing zone object unit detection unit (500), the postprocessing unit (600), and the result display unit (700) may be composed of a single terminal device (e.g., a laptop, personal computer, PDA, PMP, smartphone, etc.).

[0051] The second input data generation unit (800) receives learning borehole ultrasonic image logging data and inspection borehole ultrasonic image logging data (where the X-axis represents an angle from 0 to 360 degrees, the Y-axis represents depth, and the color represents amplitude) and generates learning input layer data, inspection input layer data, and learning output layer data. The learning input layer data and inspection input layer data are amplitude value data for all orientations at depths (mn) to (m+n), and the learning output layer data is data on whether there is a compression fracture zone at a depth of (m). Here, m represents the detection depth, and n represents an arbitrary number of depths.

[0052] The compression fracture zone depth unit detection model generation unit (900) plays the role of generating a machine learning model trained by the data set of learning input layer data and learning output layer data generated by the second input data generation unit (800).

[0053] The compression fracture zone depth unit detection unit (1000) plays the role of acquiring depth unit compression fracture zone existence data by inputting inspection input layer data generated by the second input data generation unit (800) into the machine learning model generated from the compression fracture zone depth unit detection model generation unit (900).

[0054] The first input data generation unit (100) receives learning borehole ultrasonic image logging data and inspection borehole ultrasonic image logging data (where the X-axis represents an angle of 0 to 360 degrees, the Y-axis represents depth, and the color represents amplitude) and generates multiple learning image data and multiple inspection image data that have a fixed size and are separated by a set depth interval from the top and bottom of the compression fracture zone.

[0055] FIG. 4 is a diagram showing training borehole ultrasonic image logging data and inspection borehole ultrasonic image logging data input to the first input data generation unit and the second input data generation unit of FIG. 1, where the X-axis represents an angle from 0 to 360 degrees, the Y-axis represents depth, and the color represents amplitude. Here, the darkest color indicates the lowest amplitude and is highly likely to be a compressional fracture zone.

[0056] FIG. 5 is a diagram showing training image data and inspection image data generated in the first input data generation unit of FIG. 1, wherein a plurality of training image data each have a set interval (e.g., 0.5 m) from the top and bottom of the compression crushing platform and have a constant size.

[0057] The image selection unit (150) receives depth-unit compression fracture zone existence data obtained from the compression fracture zone depth-unit detection unit (1000) and selects multiple inspection image data in which a compression fracture zone exists among multiple inspection image data generated by the first input data generation unit (100) based on the depth-unit compression fracture zone existence data.

[0058] The preprocessing unit (200) preprocesses a plurality of training image data generated by the first input data generation unit (100) and a plurality of inspection image data selected by the image selection unit (150) to generate grayscale training image data and inspection image data having a set resolution. More specifically, as shown in FIG. 2, the preprocessing unit (200) includes a resolution adjustment unit (210) and a grayscale conversion unit (220).

[0059] The resolution adjustment unit (210) receives a plurality of training image data generated by the first input data generation unit (100) and a plurality of inspection image data selected by the image selection unit (150), and adjusts them to have a set resolution (e.g., 640×640).

[0060] The grayscale conversion unit (220) serves to convert the training image data and inspection image data having a set resolution adjusted by the resolution adjustment unit (210) into grayscale training image data and inspection image data. FIG. 6 is a diagram showing the grayscale training image data and inspection image data converted by the grayscale conversion unit of FIG. 2.

[0061] The data augmentation unit (300) uses grayscale training image data with a set resolution preprocessed by the preprocessing unit (200) as original image data, generates translated image data by translating the original image data by a set angle, generates horizontally flipped image data by flipping the original image data horizontally, generates vertically flipped image data by flipping the original image data vertically, and generates 180-degree rotated image data by rotating the original image data 180 degrees. That is, 20 image data can be augmented for one original image data. FIG. 7 is a diagram showing original image data, which is grayscale training image data input to the data augmentation unit of FIG. 1, translated image data, horizontally flipped image data, vertically flipped image data, and 180-degree rotated image data augmented by the data augmentation unit. In this way, by augmenting a single original image data, which is training image data, with translation image data, horizontal flip image data, vertical flip image data, and 180-degree rotation image data, a machine learning model for detecting compression fracture zones can be generated even if there is little available training image data.

[0062] The rolling unit (350) receives grayscale inspection image data with a set resolution preprocessed by the preprocessing unit (200) as shown in FIG. 8 (S351), translates the angle column with the lowest average amplitude value to the first column (S352), and then translates it to the right (S353). Furthermore, the rolling unit (350) can provide translation information to the postprocessing unit (600).

[0063] The compressed fragmentation zone object unit detection model generation unit (400) plays the role of generating a machine learning model learned from a dataset of grayscale training image data augmented by the data augmentation unit (300), and coordinate and size information (x, y, w, h) of a bounding box surrounding the compressed fragmentation zone of the augmented grayscale training image data.

[0064] The compression crushing zone detection unit (500) inputs inspection image data, which has been translated by the rolling unit (350), into a machine learning model generated by the compression crushing zone object unit detection model generation unit (400), and obtains coordinate and size information (x, y, w, h) of the bounding box surrounding the compression crushing zone.

[0065] The post-processing unit (600) receives the coordinates and size information (x, y, w, h) of the bounding box from the compressed crushing unit object unit detection unit (500), and receives the translated inspection image data and translation information from the rolling unit (350) to perform post-processing and generate compressed crushing unit attribute information per bounding box and compressed crushing unit attribute information per pixel.

[0066] More specifically, the post-processing unit (600) includes a first compression crushing zone attribute information calculation unit (620), a clustering model (610), and a second compression crushing zone attribute information calculation unit (630), as shown in FIG. 3.

[0067] The first compression crushing zone attribute information calculation unit (620) receives the coordinates and size information (x, y, w, h) of the bounding box from the compression crushing zone object unit detection unit (500) and receives the translated inspection image data and translation information from the rolling unit (350), and calculates the center depth, azimuth angle, opening angle, and length of the compression crushing zone for each bounding box using the input coordinates and size information of the bounding box, the translated inspection image data, and translation information. The center depth, azimuth angle, opening angle, and length of the compression crushing zone for each bounding box can be calculated by the following [Equation 1].

[0068] [Mathematical Formula 1]

[0069] Center depth (depth, m) = Start depth (start depth) + (y × tm)

[0070] Azimuth (°) = x × 360° + roll

[0071] Opening angle (°) = ω × 360°

[0072] Length (m) = h × tm

[0073] [Here, the starting depth represents the top depth of the image data output from the rolling unit, t represents the vertical length of the image data output from the rolling unit, x and y represent the center coordinates of the bounding box, w and h represent the horizontal and vertical lengths of the bounding box, and roll represents information performed by the rolling unit, indicating that the column with the lowest average amplitude value among the columns from 0 degrees to 360 degrees in the image data output from the input to the rolling unit is translated to the first column and then translated again to the right.]

[0074] FIG. 9 is a diagram for explaining the coordinate and size information (x, y, w, h) of the bounding box input to the first compression crushing zone attribute information calculation unit of FIG. 3, the image data in which the compression crushing zone is detected, and the attribute information of the compression crushing zone for each bounding box calculated by the first compression crushing zone attribute information calculation unit.

[0075] The clustering model (610) performs the role of classifying compressed crushed zone image data classified based on bounding boxes in the parallel-translated inspection image data obtained by the rolling unit (350) into two types by using a clustering algorithm to classify clusters with relatively small pixel average values ​​in the image data as 1 and clusters with relatively large average values ​​as 0.

[0076]

[0077] Figure 10 illustrates the process of clustering by the clustering model of Figure 3, the classification of 1 and 0 values ​​of clustered image data, and the calculation of the median of the compressed fracture zone cluster.

[0078] The clustering process by the clustering model (610) is explained.

[0079] First, the parallel-translated inspection image data obtained by the rolling unit (350) is classified into compression crush zones based on the bounding box (classified into left and right images based on the middle compression crush zone detection image data in FIG. 10).

[0080] Next, the compressed fragmentation zone image data classified based on bounding boxes in the translated inspection image data is clustered into two types using a clustering algorithm, with clusters having a relatively small average pixel value in the image data assigned 1 and clusters having a relatively large average value assigned 0 (clustered into white clusters and black clusters in Fig. 10).

[0081] The second compression fracture zone attribute information calculation unit (630) calculates the median of the compression fracture zone clusters (median of the black clusters in the left and right images of FIG. 10) from the image data clustered into two types of 1 and 0 by the clustering model (610), and calculates the center depth, azimuth angle, opening angle, and length of the compression fracture zones on a pixel-by-pixel basis using the following [Equation 2] from the clustered image data for which the median has been calculated (see FIG. 11).

[0082] [Mathematical Formula 2]

[0083] Center depth (depth, m) = (depth min +depth max ) / 2

[0084] Azimuth (°) = Average of the median values ​​by depth of the compression fracture zone cluster

[0085] Opening angle (°) = Average value of the opening angle by depth of the compression fracture zone cluster

[0086] length(length, m) = depth max - depth min

[0087] [Here, depth max and depth min [ indicates the maximum and minimum depths of the compression fracture zone in pixel units]

[0088] The result display unit (600) serves to display depth unit compression fracture zone existence data obtained from the compression fracture zone depth unit detection unit (1000), bounding box coordinate and size information (x, y, w, h) obtained from the compression fracture zone object unit detection unit (500), or bounding box-specific compression fracture zone attribute information (center depth, azimuth, opening angle and length of the compression fracture zone per bounding box) and pixel-unit compression fracture zone attribute information (center depth, azimuth, opening angle and length of the compression fracture zone per pixel) generated by the post-processing unit (600).

[0089] The operation of the machine learning model-based compression and crushing zone detection system according to the embodiment of the present invention configured as described above will be explained.

[0090] First, the second input data generation unit (800) receives learning borehole ultrasonic image logging data and inspection borehole ultrasonic image logging data, in which the X-axis represents an angle of 0 to 360 degrees, the Y-axis represents depth, and the color represents amplitude, and generates learning input layer data, inspection input layer data, and learning output layer data (S100 of FIG. 12).

[0091] Next, the compression fracture zone depth unit detection model generation unit (900) generates a machine learning model trained by the data set of learning input layer data and learning output layer data generated by the second input data generation unit (800) (S200 of FIG. 12).

[0092] Next, the compression fracture zone depth unit detection unit (1000) inputs the inspection input layer data generated by the second input data generation unit (800) into the machine learning model generated by the compression fracture zone depth unit detection model generation unit (900) to obtain depth unit compression fracture zone existence data (S300 of FIG. 12).

[0093] Next, the depth unit compression fracture zone existence data obtained from the compression fracture zone depth unit detection unit (1000) is displayed through the result display unit (700) (S400 of FIG. 12).

[0094] Next, the first input data generation unit (100) receives learning borehole ultrasonic image logging data and inspection borehole ultrasonic image logging data, in which the X-axis represents an angle of 0 to 360 degrees, the Y-axis represents depth, and the color represents amplitude, and generates multiple learning image data and multiple inspection image data separated by set depth intervals from the top and bottom of the compression fracture zone.

[0095] Next, the image selection unit (150) receives depth-unit compression crushing zone existence data obtained from the compression crushing zone depth-unit detection unit (1000) and selects multiple inspection image data in which a compression crushing zone exists among multiple inspection image data generated by the first input data generation unit (100).

[0096]

[0097] Next, the preprocessing unit (200) preprocesses a plurality of training image data generated by the first input data generation unit (100) and inspection image data selected by the image selection unit (150) to generate grayscale training image data and inspection image data having a set resolution.

[0098] Next, the data augmentation unit (300) receives grayscale training image data having a set resolution generated by the preprocessing unit (200) and generates augmented grayscale training image data.

[0099] Next, the rolling unit (350) receives grayscale inspection image data with a set resolution generated by the preprocessing unit (200), translates the angle column with the lowest average amplitude value to the first column, and then translates it to the right.

[0100]

[0101] Next, the compression crushing zone object unit detection model generation unit (400) generates a machine learning model trained on a dataset of grayscale training image data augmented by the data augmentation unit (300), and coordinate and size information (x, y, w, h) of a bounding box surrounding the compression crushing zone of the augmented grayscale training image data.

[0102] Next, the compression crushing unit object unit detection unit (500) loads the machine learning model generated from the compression crushing unit object unit detection model generation unit (400), and inputs the inspection image data translated by the rolling unit (350) into the machine learning model to obtain the coordinates and size information (x, y, w, h) of the bounding box surrounding the compression crushing unit.

[0103] Next, the post-processing unit (600) receives the coordinates and size information (x, y, w, h) of the bounding box from the compressed crushing zone object unit detection unit (500), and receives the translated inspection image data and translation information from the rolling unit (350) to post-process and generate compressed crushing zone attribute information per bounding box and compressed crushing zone attribute information per pixel.

[0104] Next, the coordinate and size information (x, y, w, h) of the bounding box obtained from the object-unit detection unit (500) of the compressed crushing zone, or the attribute information of the compressed crushing zone per bounding box and the attribute information of the compressed crushing zone per pixel unit generated from the post-processing unit (600) are displayed through the result display unit (700).

[0105] According to a machine learning model-based compression fracture zone detection system of an embodiment of the present invention, by utilizing depth-unit compression fracture zone existence data obtained from a compression fracture zone depth-unit detection unit to select multiple inspection image data in which a compression fracture zone exists among multiple inspection image data, and by configuring the selected multiple inspection image data to be used for compression fracture zone object-unit detection, it is possible to improve the false detection (problem of predicting that a compression fracture zone exists when it does not) that occurred in the object detection model.

[0106] Furthermore, according to the machine learning model-based compression fracture zone detection system of the present invention, a plurality of training image data and inspection image data generated by receiving training borehole ultrasonic image logging data and inspection borehole ultrasonic image logging data are preprocessed to generate grayscale training image data and inspection image data having a set resolution, the generated grayscale training image data is augmented, a machine learning model is generated by the augmented grayscale training image data and a data set of coordinate and size information of a bounding box surrounding the compression fracture zone of the training image data, and inspection image data translated by a rolling part is input to the machine learning model to obtain coordinate and size information (x, y, w, h) of a bounding box surrounding the compression fracture zone, thereby enabling the generation of a machine learning model for compression fracture zone detection even if there is little available training data.

[0107] Furthermore, according to the machine learning model-based compression fracture zone detection system of the embodiment of the present invention, grayscale inspection image data is input, and the column with the lowest average amplitude value among columns from 0 degrees to 360 degrees in the image data is configured to be translated to the first column and then translated again to the right, thereby improving the detection precision of the compression fracture zone even if a compression fracture zone exists at both ends of the inspection image data.

[0108] Optimal embodiments have been disclosed in the drawings and specification, and specific terms have been used, but these are used only for the purpose of describing embodiments of the invention and are not intended to limit the meaning or the scope of the invention as described in the claims. Therefore, those skilled in the art will understand that various modifications and equivalent alternative embodiments are possible therefrom. Accordingly, the true technical scope of protection of the invention should be determined by the technical spirit of the appended claims.

Claims

1. A second input data generation unit (800) configured to receive learning borehole ultrasonic image logging data and inspection borehole ultrasonic image logging data, wherein the X-axis represents an angle of 0 to 360 degrees, the Y-axis represents depth, and the color represents amplitude, and to generate learning input layer data, inspection input layer data, and learning output layer data; A compression fracture zone depth unit detection model generation unit (900) configured to generate a machine learning model learned by the data set of the above-mentioned learning input layer data and learning output layer data; and A compression fracture zone depth unit detection unit (1000) configured to acquire depth unit compression fracture zone existence data by inputting the inspection input layer data into a machine learning model generated from the compression fracture zone depth unit detection model generation unit; The above-mentioned learning input layer data and inspection input layer data are amplitude value data for all orientations at depths (mn) to (m+n), and The above training output layer data is data on whether there is a compression fracture zone at a depth of (m), and Here, m represents the detection depth, and n represents an arbitrary number of depths. Machine learning model-based compression and crushing zone detection system.

2. In Paragraph 1, A first input data generation unit (100) that receives the above-mentioned learning borehole ultrasonic image logging data and inspection borehole ultrasonic image logging data and generates a plurality of learning image data and a plurality of inspection image data separated by a set depth interval from the upper and lower ends of the compression fracture zone; An image selection unit (150) configured to receive depth-unit compression fracture zone existence data obtained from the compression fracture zone depth-unit detection unit (1000) and to select multiple inspection image data in which a compression fracture zone exists among the multiple inspection image data; A preprocessing unit (200) configured to preprocess the plurality of training image data and the selected inspection image data to generate grayscale training image data and inspection image data having a set resolution; A data augmentation unit (300) configured to receive grayscale training image data having the above-set resolution and generate augmented grayscale training image data; A rolling unit (350) that receives the above-mentioned inspection image data, translates the angle column with the lowest average amplitude value to the first column, and then translates it to the right; A compression fracture zone object unit detection model generation unit (400) configured to generate a machine learning model trained by a dataset of grayscale training image data augmented by the data augmentation unit, and coordinate and size information (x, y, w, h) of a bounding box surrounding the compression fracture zone of the augmented grayscale training image data; and A machine learning model-based compression / crushing zone detection system further comprising: a compression / crushing zone object unit detection unit (500) configured to acquire coordinate and size information (x, y, w, h) of a bounding box surrounding a compression / crushing zone by inputting inspection image data translated by the rolling unit into a machine learning model generated from the above compression / crushing zone object unit detection model generation unit.

3. In Paragraph 2, A machine learning model-based compression / crushing zone detection system further comprising a post-processing unit (600) configured to receive coordinate and size information (x, y, w, h) of a bounding box from the above compression / crushing zone object unit detection unit (500), receive parallel-translated inspection image data and parallel-translation information from the above rolling unit (350), and perform post-processing to generate compression / crushing zone attribute information per bounding box and compression / crushing zone attribute information per pixel unit.

4. In any one of paragraphs 1 to 3, A machine learning model-based compression fracture detection system further comprising a result display unit (700) configured to display one or more of the following: depth-unit compression fracture presence data obtained from the compression fracture depth unit detection unit (1000), bounding box coordinate and size information (x, y, w, h) obtained from the compression fracture object unit detection unit (500), and compression fracture attribute information per bounding box and compression fracture attribute information per pixel unit generated by the post-processing unit (600).

5. In Paragraph 2, The above preprocessing unit (200) A resolution adjustment unit (210) that receives the plurality of training image data and the selected inspection image data and adjusts them to have a set resolution; and A machine learning model-based compression and crushing zone detection system comprising: a grayscale conversion unit (220) that converts the training image data and inspection image data having a set resolution into a plurality of grayscale training image data and inspection image data.

6. In Paragraph 3, The above post-processing unit (600) A first compression crushing zone attribute information calculation unit (620) that receives coordinate and size information (x, y, w, h) of a bounding box from the above compression crushing zone object unit detection unit (500), receives translated inspection image data and translation information from the above rolling unit (350), and calculates the center depth, azimuth angle, opening angle, and length of the compression crushing zone for each bounding box using the input coordinate and size information of the bounding box, translated inspection image data, and translation information; A clustering model (610) that classifies compressed fragmentation zone image data classified based on bounding boxes in the above-mentioned parallel-translated inspection image data into two types by using a clustering algorithm to assign a cluster with a relatively small average pixel value in the image data to 1 and a cluster with a relatively large average value to 0; and A machine learning model-based compression fracture zone detection system comprising: a second compression fracture zone attribute information calculation unit (630) that calculates the median value for each depth of the compression fracture zone clusters in the image data clustered into the two types above, and calculates the center depth, azimuth angle, opening angle, and length of the compression fracture zone in pixel units in the clustered image data in which the median value was calculated.

7. In Paragraph 6, A machine learning model-based compression zone detection system in which the center depth, azimuth angle, opening angle, and length of the compression zone for each bounding box, calculated by the first compression zone attribute information calculation unit (620) above, are calculated by the following [Equation 1]. [Mathematical Formula 1] Center depth (depth, m) = Start depth (start depth) + (y × tm) Azimuth (°) = x × 360° + roll Opening angle (°) = ω × 360° Length (m) = h × tm [Here, the starting depth represents the top depth of the image data output from the rolling unit, t represents the vertical length of the image data output from the rolling unit, x and y represent the center coordinates of the bounding box, w and h represent the horizontal and vertical lengths of the bounding box, and roll represents information performed by the rolling unit, indicating that the image data output from the rolling unit is translated to the first column by the column with the lowest average amplitude value among columns from 0 degrees to 360 degrees, and then translated again to the right.] 8. In Paragraph 6, A machine learning model-based compression fracture detection system in which the center depth, azimuth angle, opening angle, and length of the compression fracture zone per pixel unit, calculated by the second compression fracture zone attribute information calculation unit (630) above, are calculated by the following [Equation 2]. [Mathematical Formula 2] Center depth (depth, m) = (depth min +depth max ) / 2 Azimuth (°) = Average of the median values ​​by depth of the compression fracture zone cluster Opening angle (°) = Average value of the opening angle at different depths of the compression fracture zone cluster length(length, m) = depth max - depth min [Here, depth max and depth min [ indicates the maximum and minimum depths of the compression fracture zone in pixel units]

Images