Face evaluation system and face evaluation method

The face evaluation system integrates drilling, RGB, and spectral image data with geology-based weighting to accurately assess face properties in tunnel excavation, addressing inaccuracies in existing methods.

JP2024157655AActive Publication Date: 2024-11-08PENTA OCEAN CONSTRUCTION CO LTD +1
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
JP2023072128
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-04-26
Publication Date
2024-11-08
Estimated Expiration
2043-04-26

AI Technical Summary

Technical Problem

Existing face property evaluation methods in tunnel excavation, whether through drilling data analysis or machine learning-based image evaluation, face properties cannot be accurately assessed due to variations in ground geology and mismatch between training data and actual conditions.

Method used

A face evaluation system and method that integrates drilling data, RGB image data, and spectral image data using machine learning models to calculate a comprehensive face evaluation value, with data weighting based on ground geology, incorporating measurements like rotational pressure, impact pressure, drilling speed, and spectral image analysis.

Benefits of technology

Enables accurate and appropriate face property evaluation by reducing errors and variations, ensuring precise determination of face properties through multi-data integration and geology-specific weighting.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024157655000001_ABST
    Figure 2024157655000001_ABST
Patent Text Reader

Abstract

To provide a face evaluation system capable of appropriately and accurately evaluating face properties.SOLUTION: A face evaluation system comprises an input unit to which drilling data, RGB image data of a face, and spectral image data of the face are input, a first calculation unit 30 that outputs a first calculated value using a drilling learning model generated by machine learning using teacher data having the drilling data as an input variable and the face evaluation by a skilled engineer as an output variable, a second calculation unit 40 that outputs a second calculated value using an RGB learning model generated by machine learning using teacher data having the RGB image data as an input variable and the face evaluation by a skilled engineer as an output variable, a third calculation unit 50 that outputs a third calculated value using a spectral learning model generated by machine learning using teacher data having the spectral image data as an input variable and the face evaluation by a skilled engineer as an output variable, and a face evaluation value calculation unit 60 that integrates the first calculated value, the second calculated value, and the third calculated value to calculate a face evaluation value.SELECTED DRAWING: Figure 1
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a face evaluation system and a face evaluation method. [Background technology]

[0002] In tunnel excavation work, it is important to properly understand the geology of the ground in order to proceed with construction safely and efficiently. Constructors grasp the geology of the ground by evaluating the face characteristics (e.g. hardness or softness of the rock, presence or absence of cracks, spring water conditions, etc.). Then, based on the results of that evaluation, constructors take appropriate safety measures and determine the appropriate amount of explosives to be used when blasting.

[0003] Such evaluation of face properties is performed by an engineer by visual inspection of the face, but since it requires skill and is based on human knowledge and experience, there is a risk of variation in the accuracy of the evaluation. Therefore, a method of automatically judging the evaluation of face properties from data measured during drilling and a method of automatically judging using artificial intelligence (AI) have been proposed. For example, the following Patent Document 1 discloses a method of acquiring data (e.g., impact pressure, feed pressure, drilling speed, etc.) measured during drilling of a charge hole and evaluating the face properties based on the data. In addition, the following Patent Document 2 discloses a method of evaluating the face from images using a face judgment model that has been machine-learned from multiple learning face images photographed and obtained during tunnel construction. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent Publication No. 2022-37786 [Patent Document 2] Patent Publication No. 2021-92031 Summary of the Invention [Problem to be solved by the invention]

[0005] However, in the method of evaluating the face properties by acquiring measurement data during drilling, although it is possible to judge the hardness or softness of the geological condition of the face, it is not possible to evaluate using weathering of the face, the state of cracks, the spacing between cracks, etc., and therefore there is a risk that the face properties cannot be evaluated appropriately. In addition, since the geology of the ground varies depending on the formation of the ground and the region, the face property evaluation method using the above-mentioned face judgment model may not be able to properly evaluate the face properties if the ground of the face image used in the machine learning differs from the ground actually evaluated.

[0006] The present invention has been made in consideration of the above-mentioned circumstances, and aims to provide a face evaluation system and a face evaluation method that can appropriately and accurately evaluate the face. [Means for solving the problem]

[0007] In order to achieve the above object, the present invention provides an input unit for inputting drilling data measured by a measuring device when drilling, RGB image data of the face acquired by an RGB (red (R), green (G), blue (B)) image acquisition device, and spectral image data of the face acquired by a spectral image acquisition device, a first calculation unit for using the drilling data as input data and outputting a first calculated value as an output variable by using a drilling learning model generated by machine learning using teacher data in which the drilling data is an input variable and the face evaluation by a skilled engineer corresponding to the drilling data is an output variable, and a second calculation unit for using the RGB image data as an input variable and the face evaluation by a skilled engineer corresponding to the RGB image data is an output variable by using a drilling learning model generated by machine learning using teacher data in which the RGB image data is an input variable and the face evaluation by a skilled engineer corresponding to the RGB image data is an output variable by using a first calculation unit for using the RGB image data as input data and outputting a first calculated value as an output variable by using a drilling learning model generated by machine learning using teacher data in which the RGB image data is an input variable and the face evaluation by a skilled engineer corresponding to the RGB image data is an output variable by using a first calculation unit for using the RGB image data as input variable and the face evaluation by a skilled engineer corresponding to the RGB image data is an output variable by using a first calculation unit for using the RGB image data as input variable and the face evaluation by a skilled engineer corresponding to the RGB image data as ... The present invention provides a face evaluation system comprising: a second calculation unit that uses an RGB learning model generated by performing machine learning using teacher data in which an operator's face evaluation is an output variable, to take RGB image data as an input variable and output a second calculated value as an output variable; a third calculation unit that uses a spectral learning model generated by performing machine learning using teacher data in which the spectral image data is an input variable and an operator's face evaluation by a skilled technician corresponding to the spectral image is an output variable, to take spectral image data as input data and output a third calculated value as an output variable; and a face evaluation value calculation unit that integrates the first calculated value, the second calculated value, and the third calculated value to calculate a face evaluation value.

[0008] In addition, in the above-mentioned face evaluation system, when generating a hole-drilling learning model, the first calculation unit may weight the hole-drilling data, which is an explanatory variable of the hole-drilling learning model, according to the geology of the natural ground.

[0009] In addition, the drilling data to be weighted may include at least two or more pieces of data: position information data of the drilling location, data on the feed pressure, rotational pressure, impact pressure, and drilling speed of the drilling machine at the drilling location corresponding to the position information data, and spring water pressure data at the drilling location.

[0010] Furthermore, when integrating the first calculated value, the second calculated value, and the third calculated value, the face evaluation value calculation unit may weight each calculated value according to the geology of the natural ground.

[0011] In addition, the face evaluation value calculation unit may use the first calculated value, the second calculated value, and the third calculated value as input variables, and a face evaluation learning model generated by machine learning using teacher data in which the first calculated value, the second calculated value, and the third calculated value are input variables, and a face evaluation by a skilled technician corresponding to the input variables is an output variable, to calculate the face evaluation value as an output variable.

[0012] The present invention also provides a face evaluation method that includes acquiring drilling data using a measuring device when drilling a hole, acquiring RGB image data of the face using an RGB image capturing device, acquiring spectral image data of the face using a spectral image capturing device, inputting the drilling data as an input variable into a drilling learning model generated by performing machine learning using teacher data having the drilling data as an input variable and the face evaluation by a skilled engineer corresponding to the drilling data as an output variable, inputting the drilling data as an input variable into a drilling learning model generated by performing machine learning using teacher data having the RGB image data as an input variable and the face evaluation by a skilled engineer corresponding to the RGB image data as an output variable, and calculating a first calculated value. Inputting the RGB image data as an input variable into an RGB learning model generated by performing machine learning using teacher data having the spectral image data as an input variable and the face evaluation by a skilled engineer corresponding to the spectral image data as an output variable, and calculating a third calculated value. The first calculated value, the second calculated value, and the third calculated value are integrated to calculate a face evaluation value.

[0013] In addition, in the above-mentioned face evaluation method, when generating a drilling learning model, the drilling data, which is an explanatory variable of the drilling learning model, may be weighted according to the geology of the natural ground.

[0014] In addition, the drilling data to be weighted may include at least two or more pieces of data: position information data of the drilling location, data on the feed pressure, rotational pressure, impact pressure, and drilling speed of the drilling machine at the drilling location corresponding to the position information data, and spring water pressure data at the drilling location.

[0015] Furthermore, when the first calculated value, the second calculated value, and the third calculated value are integrated, each calculated value may be weighted in accordance with the geology of the natural ground.

[0016] In addition, to calculate the face value, the first calculated value, the second calculated value, and the third calculated value are input variables, and a face evaluation learning model generated by machine learning using teacher data in which the first calculated value, the second calculated value, and the third calculated value are used as input variables and the face evaluation value by a skilled technician corresponding to the input variables is used as an output variable, and the first calculated value, the second calculated value, and the third calculated value are used as input data and the face evaluation value is calculated as an output variable. Effect of the Invention

[0017] According to the present invention, the tunnel face can be appropriately and accurately evaluated. [Brief description of the drawings]

[0018] [Figure 1] FIG. 1 is a block diagram showing an example of the configuration of a face evaluation system according to a first embodiment. [Diagram 2] FIG. 2 is a diagram showing a data flow in the face evaluation system according to the first embodiment. [Diagram 3] 11 is a flowchart showing an example of a method for generating a hole-drilling learning model. [Figure 4] (a) is a diagram to explain face evaluation by a skilled technician, and (b) is a diagram to explain position information of drilling data. [Diagram 5] FIG. 1A is a diagram showing an example of engineer evaluation data, and FIG. 1B is a diagram showing an example of drilling data. [Figure 6] FIG. 11 is a diagram illustrating an example of weighting data. [Figure 7] 1 is a flowchart showing an example of the operation of the face evaluation system when evaluating a face. [Figure 8] FIG. 11 is a diagram illustrating an example of weighting data. [Figure 9] FIG. 11 is a diagram for explaining a second embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0019] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. However, the present invention is not limited thereto. In addition, in the drawings, in order to explain the embodiment, some parts may be enlarged or emphasized, and the scale may be appropriately changed.

[0020] Fig. 1 is a block diagram showing an example of the configuration of a tunnel face evaluation system 100 according to the first embodiment. In addition to the tunnel face evaluation system 100, Fig. 1 also shows a measurement device 70, an RGB image capturing device 80, and a spectral image capturing device 90. The tunnel face evaluation system 100 is a system for evaluating a tunnel face and judging tunnel face properties.

[0021] The measuring device 70 is a device that measures drilling data. The drilling data is data indicating the operating status of a drilling machine (for example, a face drilling machine such as a drill jumbo) when drilling a hole or the spring water status at the drilling location. The data indicating the operating status of the drilling machine is, for example, data on the rotation pressure, impact pressure, and drilling speed of the drilling bit when the drilling machine drills a hole while pressing the rotating drilling bit. The data indicating the spring water status is, for example, data on the pressure of spring water (spring water pressure) and the amount of spring water at the drilling location. The measuring device 70 is configured to include various sensors, and is, for example, a sensor mounted on the drilling machine. In this embodiment, the measuring device 70 acquires the above-mentioned data on the rotation pressure, impact pressure, drilling speed, and spring water pressure together with position information data indicating the drilling location of the face. In this embodiment, the drilling data includes position information data of the drilling location, data of the feed pressure, rotation pressure, impact pressure, and drilling speed of the drilling machine at the drilling location corresponding to the position information data, and water pressure data of the drilling location, but may be configured to include at least two or more of these data. The position information data is, for example, planar or spatial coordinate data.

[0022] The RGB image capturing device 80 is a device that captures RGB image data of the tunnel face before the tunnel face is drilled (before the measuring device 70 captures the drilling data). The RGB image capturing device 80 is, for example, a color camera.

[0023] The spectral image capturing device 90 is a device that acquires spectral image data (hereinafter, may be simply referred to as "SP image data") of the tunnel face before drilling the tunnel face. The spectral image capturing device 90 is, for example, a multispectral camera. The spectral image data acquired by the spectral image capturing device 90 is, for example, image data that records electromagnetic waves in a plurality of wavelength bands, and electromagnetic waves in wavelength bands such as visible light, ultraviolet light, infrared light, and far infrared light are recorded.

[0024] The measuring device 70 , the RGB image capturing device 80 , and the spectral image capturing device 90 are each capable of data communication with the face evaluation system 100 .

[0025] 1, the face evaluation system 100 has an input unit 10, a storage unit 20, a first calculation unit 30, a second calculation unit 40, a third calculation unit 50, and a face evaluation value calculation unit 60. The face evaluation system 100 is configured by a computer.

[0026] The input unit 10 is a processing unit that inputs the drilling data measured by the measuring device 70, the RGB image data of the face acquired by the RGB image capturing device 80, and the spectral image data of the face acquired by the spectral image capturing device 90. The input unit 10 also associates the position information (e.g., coordinate data) of the drilling location with each of the drilling data input from the measuring device 70, the RGB image data input from the RGB image capturing device 80, and the spectral image data input from the spectral image capturing device 90.

[0027] The storage unit 20 stores various data input or output during the operation of the rock face evaluation system 100. For example, the storage unit 20 stores drilling data, RGB image data, and spectral image data input from the input unit 10, drilling teacher data used to generate the drilling learning model 31, weighting data, and engineer evaluation data, all of which will be described later. The drilling teacher data used to generate the drilling learning model 31 is data in which drilling data (see FIG. 5(b)) of a drilling location at the rock face is associated with data including a rock face evaluation score by a skilled engineer for the drilling location (see FIG. 5(a)) for a given evaluation item (e.g., rock strength, weathering alteration).

[0028] The storage unit 20 also stores a program for implementing the processing described below by the face evaluation system 100. When the calculation means of the computer constituting the face evaluation system 100 executes the program, each function such as a first calculation unit 30, a second calculation unit 40, a third calculation unit 50, and a face evaluation value calculation unit 60 is implemented.

[0029] 2 is a diagram showing the flow of data in the cut face evaluation system 100 according to the first embodiment. Next, the configuration of the cut face evaluation system 100 will be described with reference to FIG.

[0030] The first calculation unit 30 is a processing unit that generates a drilling learning model 31 by machine learning using teacher data (hereinafter, referred to as drilling teacher data to distinguish it from other teacher data) in which drilling data is an input variable and a face evaluation by a skilled engineer corresponding to the drilling data is an output variable. The first calculation unit 30 is also a processing unit that calculates a first calculated value 32, which is a face evaluation, using the drilling learning model 31 with the drilling data 71 input to the input unit 10 as input data.

[0031] The second calculation unit 40 is a processing unit that calculates a second calculated value 42, which is a face evaluation, using an RGB learning model 41 generated by machine learning using teacher data (hereinafter referred to as RGB teacher data to distinguish it from other teacher data) in which RGB image data is an input variable and a face evaluation by a skilled technician corresponding to the RGB image data is an output variable, using the RGB image data 81 input to the input unit 10 as input data.

[0032] The third calculation unit 50 is a processing unit that calculates a third calculated value 52, which is a face evaluation, using the spectral image data 91 input to the input unit 10 as input data, by using a spectral learning model 51 generated by machine learning using teacher data (hereinafter, referred to as spectral teacher data to distinguish it from other teacher data) in which the spectral image data is an input variable and the face evaluation by a skilled engineer corresponding to the spectral image is an output variable. Note that hereinafter, the spectral learning model may be simply referred to as the "SP learning model."

[0033] The face evaluation value calculation unit 60 is a processing unit that calculates a face evaluation value 62, which is a face evaluation calculated comprehensively by integrating the face evaluation first calculation value 32, the second calculation value 42, and the third calculation value 52 calculated by the first calculation unit 30, the second calculation unit 40, and the third calculation unit 50. The face evaluation value calculation unit 60 calculates the face evaluation value 62 using a face evaluation learning model 61 generated by machine learning using teacher data.

[0034] Here, a method for generating the hole-drilling learning model 31 will be described. The hole-drilling learning model 31 is generated by machine learning using drilling teacher data. A specific generation method is, for example, as follows. FIG. 3 is a flowchart showing an example of a method for generating the hole-drilling learning model 31. FIG. 4(a) is a diagram for explaining face evaluation by a skilled engineer. FIG. 5(a) is a diagram showing an example of engineer evaluation data.

[0035] In the process shown in FIG. 3, the first calculation unit 30 acquires face evaluation point data (engineer evaluation data) by a skilled engineer (step S01). The engineer evaluation data acquired in step S01 is created, for example, as follows. First, in the image data IM of the face C, the image of the face C is divided into a plurality of sections. For example, as shown in FIG. 4(a), the image of the face C is divided into meshes of an arbitrary size. The file data of the divided images D, D, etc. are associated with file numbers (for example, "101", "102", "103"), respectively. Then, for each divided image D, a skilled engineer actually visually evaluates the face and determines an evaluation point for a predetermined evaluation item of the face (for example, rock strength, cracks), thereby creating engineer evaluation data. In the example shown in FIG. 5(a), the evaluation point of the rock strength of the area corresponding to the file number "101" in the face C is "10", the evaluation point of the cracks is "20", and the evaluation point of the rock strength of the area corresponding to the file number "102" is "15", etc. The engineer evaluation data is output variable data constituting the above-mentioned drilling teaching data.

[0036] The first calculation unit 30 also acquires drilling data (step S02). FIG. 4(b) is a diagram for explaining the position information of the drilling data. In step S02, a drilling number (e.g., "1", "2", etc.) is associated with each of the multiple drilling holes H1, H2, H3, etc. of the face C, and the position information and drilling data of each are acquired. The acquired position information and drilling data are associated with each of the drilling holes H. For example, as shown in FIG. 4(b), data of local coordinates (X, Y) when the horizontal position on the face is set to X and the vertical position is set to Y is used as the position information of the drilling hole H. For example, the position information of the drilling hole H1 is "(X1, Y1)", and the position information of the drilling hole H2 is "(X2, Y2)". FIG. 5(b) is a diagram showing an example of the drilling data. In the example shown in FIG. 5(b), the coordinates (e.g. (X1, Y1)) of the drilling number "1" (e.g. drilling H1) are (1, 1), the feed pressure is "40 MPa", and the coordinates of the drilling number "2" of hole H are "(2, 2)". The drilling data acquired in step S02 constitutes the drilling teacher data together with the engineer evaluation data. The drilling teacher data is teacher data used when machine learning the drilling learning model 31, and is data in which the drilling data is an input variable and the face evaluation by a skilled engineer corresponding to the drilling data is an output variable. The drilling teacher data is data in which the evaluation score by a skilled engineer and the drilling data are associated with each evaluation item, and therefore can be easily acquired (collected).

[0037] Following step S02, weighting is performed on the drilling data (step S03). In step S03, weighting is performed on the drilling data input to the drilling learning model 31 when the drilling learning model 31 is machine-learned. Since the natural ground has different geology depending on its origin and region, the drilling data is weighted according to the natural ground of the drilled face. In step S03, first, a weighting value is read from the data in the storage unit 20. Specifically, a weighting value is read from both data: geological level data (not shown) indicating the geological level of each natural ground, and weighting data including a weighting value set according to the level of the geology, both of which are data stored in advance in the storage unit 20. FIG. 6 is a diagram showing an example of this weighting data. In the example shown in FIG. 6, the natural ground geology is divided into levels such as "1", "2", and "3", and a weighting value is set for each level. That is, a weighting value is set according to the natural ground geology. For example, when the geological level of the drilled natural ground is "1", the weighting coefficient for the "feed pressure" value is "2.1", and when the geological level is "2", the weighting coefficient for the "coordinates (X, Y)" is "1.0". In step S03, the geological level of the drilled natural ground (e.g., "1", "2", "3", ...) is read from the geological level data, and a value related to the weighting of the read geological level is read from the weighting data. Then, the drilling data is weighted using the value read from the weighting data as a weighting coefficient. In this way, when the drilling learning model 31 is machine-learned, the input variables to the drilling learning model 31 are weighted according to the geology of the natural ground.

[0038] In the above-mentioned step S03, weighting is applied to all data constituting the drilling data. That is, weighting is applied to data including the position information data of the drilling location, the feed pressure, rotation pressure, impact pressure, and drilling speed data of the drilling machine at the drilling location corresponding to the position information data, and the well water pressure data of the drilling location. In step S03, weighting may be applied to only a part of the data constituting the drilling data, rather than to all the data constituting the drilling data in this manner. For example, in step S03, weighting may be applied to at least two or more data among the position information data, the feed pressure, rotation pressure, impact pressure, and drilling speed data, and the well water pressure data.

[0039] Next, machine learning is performed (step S04). In step S04, machine learning is performed using drilling teacher data, and when the weighted drilling data values ​​are input, machine learning is performed so that the face evaluation point corresponding to the weighted drilling data is output. At this time, the face evaluation point to be output is determined based on the position information of the input drilling data (i.e., the position of the drilling hole H), and the evaluation value of the area including the position indicated by the position information on the face is used. For example, when the drilling data of the drilling hole H1 at the position (X1, Y1) is input, machine learning is performed so that the face evaluation point of the image D1 (area R) (see FIG. 4(b)) including the position (X1, Y1) of the drilling hole H1 is output. By such machine learning, the drilling learning model 31 is generated. The drilling learning model 31 is a learning model with the drilling data as an explanatory variable and the first calculated value 32 as an objective variable, and when each numerical value of the drilling data is input, the first calculated value 32 is output.

[0040] The RGB learning model 41 outputs the second calculated value 42 when the RGB image data is input. The spectrum learning model 51 outputs the third calculated value 52 when the spectrum image data is input. The RGB learning model 41 and the spectrum learning model 51 are generated by machine learning using each teacher data in the same manner as the generation method of the above-mentioned hole drilling learning model 31, except that weighting setting (step S03) is not required. The RGB teacher data used when generating the RGB learning model 41 includes data in which the evaluation score of a skilled engineer in a predetermined area of ​​the face is associated with an RGB image of the area, and the RGB image data is an input variable and the face evaluation by the skilled engineer corresponding to the RGB image data is an output variable. The spectrum teacher data used when generating the spectrum learning model 51 includes data in which the evaluation score of a skilled engineer in a predetermined area of ​​the face is associated with a spectral image of the area, and the spectral image data is an input variable and the face evaluation by the skilled engineer corresponding to the spectral image is an output variable. A generation method of the face evaluation learning model 61 will be described later.

[0041] The tunnel face evaluation system 100 evaluates a tunnel face, for example, at a tunnel construction site, using the above-mentioned learning model.

[0042] Next, the operation of the cut face evaluation system 100 will be described. Fig. 7 is a flowchart showing an example of the operation of the cut face evaluation system 100 when evaluating the cut face. In Fig. 7, steps S11 to S13 are executed by the first calculation unit 30, steps S21 to S24 are executed by the second calculation unit 40, and steps S31 to S34 are executed by the third calculation unit 50. Furthermore, steps S14, S25, S35, S41, and S42 are executed by the cut face evaluation value calculation unit 60. The following will be described with reference to the flowchart in Fig. 7.

[0043] As shown in FIG. 7, first, drilling data of the tunnel face is acquired (step S11). The drilling data acquired here is the drilling data of the tunnel face that is the subject of evaluation and judgment. In tunnel construction and the like, a part of the tunnel face is drilled with a drilling machine to provide search holes, blast holes, and the like. In step S11, drilling data measured by the measuring device 70 during such drilling is acquired via the input unit 10. The drilling data input here includes position information data of the drilling location (for example, the above-mentioned coordinate (X, Y) data).

[0044] Also, RGB image data of the face is acquired (step S21). In step S21, RGB image data of the face captured by the RGB image capturing device 80 before drilling is acquired via the input unit 10. The acquired RGB image is then cut out according to the drilling positions (step S22). In step S22, from the input RGB image data, images corresponding to each of the drilling positions are cut out based on the position information data input in step S11. For example, for each drilling location, an image is cut out within a 1m radius centered on the drilling position. 2 A square RGB image is then cut out (see Fig. 9(b)). In this way, an RGB image is generated for each drilling location.

[0045] Also, SP image data of the face is acquired (step S31). In step S31, the SP image data of the face captured by the spectral image capturing device 90 before drilling is acquired via the input unit 10. The acquired spectral images are then cut out according to the drilling positions (step S32). In step S32, for the input SP image data, images corresponding to each of the drilling positions are cut out based on the position information data acquired in step S11. For example, for each drilling location, an image is cut out within a 1m radius centered on the drilling position. 2 In this way, an RGB image is generated for each drilling location.

[0046] The RGB image data and SP image data generated for each drilling position in step S22 and step S32 are linked to the drilling data of the corresponding drilling position. Therefore, data linking each of the data of the drilling position information data (e.g., coordinate data), the drilling data, the RGB image data, and the SP image data is generated for each drilling H of the face C.

[0047] Following step S11, the drilling data is input to the drilling learning model 31 (step S12), and the first calculated value 32 is calculated (step S13). Following step S22, the RGB image data is input to the RGB learning model 41 (step S23), and the second calculated value 42 is output (calculated) (step S24). The second calculated value 42 is data that quantifies the face properties based on the RGB image data. Following step S32, the SP image data is input to the SP learning model 51 (step S33), and the third calculated value 52 is output (calculated) (step S34). The third calculated value 52 is data that quantifies the face properties based on the SP image data. The output (calculated) first to third calculated values ​​32, 42, 52 are explanatory variables for determining the face evaluation value 62.

[0048] Next, weighting is applied to the first calculated value 32 calculated in step S13 (step S14). Also, weighting is applied to the second calculated value 42 calculated in step S24 (step S25). Also, weighting is applied to the third calculated value 52 calculated in step S34 (step S35). That is, weighting is applied to the numerical values ​​of the explanatory variables for determining the face evaluation value 62.

[0049] In steps S14, S25, and S35, the weighting coefficient is determined by reading both the geological level data indicating the geological level of each natural ground, which are stored in advance in the storage unit 20, and the weighting data including the weighting value determined according to the geological level. At this time, the geological level of the natural ground to be evaluated (e.g., "1", "2", "3", etc.) is read from the geological level data, and the weighting value related to the read geological level is read from the weighting data. FIG. 8 is a diagram showing an example of weighting data. In the example shown in FIG. 8, when the geological level of the drilled natural ground is "1", the weighting value set for the first calculated value 32 is "1.1", and when the geological level is "2", the weighting value set for the first calculated value 32 is "1.0", and the weighting coefficient value for each calculated value is set according to the geological level of the natural ground.

[0050] Following steps S14, S25, and S35, the first calculated value 32, the second calculated value 42, and the third calculated value 52 after weighting are input to the face evaluation learning model 61 (step S41). As a result, the weighted first to third calculated values ​​32, etc. are integrated, and the face evaluation value 62 calculated by the face evaluation learning model 61 is output. In this manner, the face evaluation value 62 related to the selected evaluation item is calculated. Then, the face properties are judged based on the calculated face evaluation value 62.

[0051] The face evaluation learning model 61 used in step S41 is generated by machine learning using teacher data. The teacher data used here includes data in which a comprehensive face evaluation by a skilled engineer is associated with the first calculated value 32, the second calculated value 42, and the third calculated value 52. The model is generated by machine learning so that when the numerical values ​​of the first calculated value 32, the second calculated value 42, and the third calculated value 52 are input, the corresponding engineer evaluation data of the face is output.

[0052] According to the face evaluation system 100 and face evaluation method described above, the following effects can be achieved. That is, in the above system and method, a plurality of data (hole drilling data, RGB image data, and SP image data) having different properties that affect the judgment of the face are acquired, and the acquired data is used to calculate the face evaluation value 62 using a plurality of calculation means (first to third calculation units 30, 40, 50 and face evaluation value calculation unit 60), and the face evaluation value 62 is calculated by integrating the calculated first calculation value 32, second calculation value 42, and third calculation value 52, so that the face evaluation value 62 can be calculated with high accuracy. Therefore, the face properties can be appropriately judged based on the face evaluation value 62 calculated with high accuracy in this way.

[0053] In addition, when drilling teacher data is input into the drilling learning model 31, and when the first to third calculated values ​​32, 42, 52 are integrated, the input variables and each calculated value 32, 42, 52 are weighted according to the geology of the natural ground, so that an appropriate face evaluation value 62 according to the geology of the natural ground can be obtained.

[0054] In addition, the drilling data includes position information data of the drilling location, data on the feed pressure, rotational pressure, impact pressure, and drilling speed of the drilling machine at the drilling location corresponding to the position information data, and wellhead pressure data at the drilling location.By using various measurement data in this manner, errors and variations in the face evaluation can be reduced, and a more accurate face evaluation value can be obtained.

[0055] In addition, the face evaluation value calculation unit 60 calculates the face evaluation value 62 using a face evaluation learning model 61 having the first calculated value 32, the second calculated value 42, and the third calculated value 52 as explanatory variables, so that a more appropriate face evaluation value 62 can be obtained.

[0056] Next, a face evaluation system according to a second embodiment will be described. Fig. 9 is a diagram for explaining the second embodiment. In the description of this embodiment, the description of the same or similar content as the first embodiment will be omitted.

[0057] In the first embodiment described above, the explanatory variables for calculating the face evaluation value 62 of the face evaluation learning model 61 were the first calculated value 32, the second calculated value 42, and the third calculated value 52, but in the second embodiment, in addition to these, as shown in Fig. 9(a), distance, angle (direction), maximum value, minimum value, average value, and variance value (hereinafter, these are referred to as "distance, etc.") are added to the explanatory variables for calculating the face evaluation value 62. That is, in the second embodiment, the numerical values ​​such as these distances and maximum values ​​are also input into the face evaluation learning model 61 to calculate the face evaluation value 62. In addition, the face evaluation learning model 61 used in this case is generated by machine learning using training data. The training data used here includes data in which the face evaluation points by skilled technicians are associated with the first calculated value 32, the second calculated value 42, the third calculated value 52, and distance, etc., and the first calculated value 32, the second calculated value 42, the third calculated value 52, and distance, etc. are input variables, and the face evaluations by skilled technicians corresponding to the first calculated value 32, the second calculated value 42, the third calculated value 52, and distance, etc. are output variables.

[0058] In the explanatory variables for calculating the face evaluation value 62 of the face evaluation learning model 61, the distance refers to the distance between the drilled holes H, and for example, refers to the Euclidean distance between adjacent drilled holes H on the face as shown in Figures 9(b) and (c). Also, the angle (direction) refers to, for example, the angle (horizontal angle) between the horizontal line and the line segment connecting the drilled holes H on the face. Note that Figure 9(c) illustrates the Euclidean distance and angle between drilled holes H4 and H6. In addition, the maximum, minimum, average, and variance values ​​of the explanatory variables mentioned above refer to the maximum, minimum, average, and variance values ​​of the data relating to a plurality of holes H including a given hole H when the face evaluation value 62 of an area including a hole H of the face C is calculated using data relating to the plurality of holes H including the given hole H, and are, for example, the maximum, minimum, average, and variance values ​​of the Euclidean distances between a given hole H and the plurality of holes H surrounding it, as well as the maximum, minimum, average, and variance values ​​of the first calculated value 32, the second calculated value 42, and the third calculated value 52 for the plurality of holes H.

[0059] As described above, in the process of calculating the first calculated value 32, the second calculated value 42, and the third calculated value 52, data is generated for each drilled hole H of the face C, linking each of the data of its position information data (e.g., local coordinate data), drilling data, RGB image data, and SP image data with each other. When focusing on the drilled hole H1, in the above-mentioned first embodiment, the face evaluation value 62 is calculated using the position information data of this drilled hole H1, the drilling data of this hole H1, the RGB image data of this hole H1, and the SP image data of this hole H1. In contrast, in the second embodiment, the face evaluation value 62 is further calculated using various data related to the drilled holes H2 to H7 adjacent to this hole H1.

[0060] As shown in Figure 9(b), when there are six drilling holes H5 to H10 adjacent to and surrounding drilling hole H4 at the face C, if focus is placed on drilling hole H4, the face evaluation value 62 of the area including drilling hole H4 is calculated, for example, as follows.

[0061] First, the first calculation value 32, the second calculation value 42, and the third calculation value 52 corresponding to each of the drilled holes H4 to H10 are calculated using the drilling data of H4 to H10, the RGB image data of H4 to H10 (images A, a to f), and the SP image data of the drilled holes H4 to H10. Each of the first calculation value 32, the second calculation value 42, and the third calculation value 52 includes a plurality of data calculated separately for each evaluation item (for example, rock strength, cracks, etc.). In addition, the average value and variance value of each of the calculated first calculation value 32, the second calculation value 42, and the third calculation value 52 are calculated. In addition, the Euclidean distance between the drilled hole H4 and each of the drilled holes H5 to H10 surrounding it are calculated, and the average value and variance value of these Euclidean distances are further calculated. In addition, the above-mentioned angles (directions) determined by the positions of the drilled holes H4 and the drilled holes H5 to H10 are calculated.

[0062] Then, the calculated data, i.e., the first calculated value 32 and its maximum value and minimum value corresponding to each of the drilled holes H4 to H10, the second calculated value 42 and its maximum value and minimum value corresponding to each of the positions of the drilled holes H4 to H10, the third calculated value 52 and its maximum value and minimum value corresponding to each of the positions of the drilled holes H4 to H10, the average value and variance value of the first calculated value 32, the second calculated value 42, and the third calculated value 52 related to the drilled holes H4 to H10, the calculated Euclidean distance and its maximum value, minimum value, average value, and variance value, and the angle (direction) are input to the face evaluation learning model 61, whereby the face evaluation value 62 is calculated. In this way, in the second embodiment, the first calculated value 32, the second calculated value 42, and the third calculated value 52 are integrated, and further, the face evaluation value 62 is calculated using data such as distance. According to such a second embodiment, the face evaluation value 62 can be calculated with even greater accuracy.

[0063] Although the embodiment of the present invention has been described above, the technical scope of the present invention is not limited to the scope described in the above embodiment. Various modifications or improvements can be made to the above embodiment without departing from the spirit of the present invention. In addition, the first calculated value, the second calculated value, and the third calculated value are used as input variables to the face evaluation value learning model to calculate the face value 62, but only the first calculated value and the second calculated value, or only the first calculated value and the third calculated value, may be used as input variables, and one or more of the requirements described in the above embodiment may be omitted. Such modifications, improvements, and omitted forms are also included in the technical scope of the present invention. [Explanation of symbols]

[0064] 10 Input section 30 First Calculation Section 31 Drilling Learning Model 32 First calculated value 40 Second Calculation Section 42 Second calculated value 50 Third Calculation Section 52 3rd calculated value 60 Face evaluation value calculation section 70 Measuring Equipment 80 RGB image capture device 90 Spectral Imaging Device 100 Face Evaluation System

Claims

1. an input unit to which drilling data measured by a measuring device when drilling a hole, RGB image data of the face acquired by an RGB image acquisition device, and spectral image data of the face acquired by a spectral image acquisition device are input; a first calculation unit that uses the drilling data as an input variable, and a drilling learning model generated by machine learning using teacher data in which a face evaluation by a skilled engineer corresponding to the drilling data is an output variable, and outputs a first calculated value as an output variable; A second calculation unit that uses the RGB image data as an input variable and an RGB learning model generated by machine learning using teacher data having a face evaluation by a skilled engineer corresponding to the RGB image data as an output variable, and outputs a second calculated value as an output variable; a third calculation unit that uses the spectral image data as input data and outputs a third calculated value as an output variable using a spectral learning model generated by machine learning using teacher data having the spectral image data as an input variable and a face evaluation by a skilled engineer corresponding to the spectral image as an output variable; A face evaluation system comprising: a face evaluation value calculation unit that calculates a face evaluation value by integrating the first calculated value, the second calculated value, and the third calculated value.

2. The face evaluation system of claim 1, wherein when generating the drilling learning model, the first calculation unit weights the drilling data, which is an explanatory variable of the drilling learning model, according to the geology of the natural ground.

3. The face evaluation system described in claim 2, wherein the drilling data to be weighted includes at least two or more pieces of data: position information data of the drilling location, data on the feed pressure, rotation pressure, impact pressure, and drilling speed of the drilling machine at the drilling location corresponding to the position information data, and spring water pressure data at the drilling location.

4. The face evaluation system according to any one of claims 1 to 3, wherein when the face evaluation value calculation unit integrates the first calculated value, the second calculated value, and the third calculated value, it weights each calculated value according to the geology of the ground.

5. The face evaluation value calculation unit uses the first calculated value, the second calculated value, and the third calculated value as input variables, and calculates the face evaluation value as an output variable using a face evaluation learning model generated by machine learning using teacher data in which the first calculated value, the second calculated value, and the third calculated value are input variables and the face evaluation by a skilled technician corresponding to the input variables is an output variable. The face evaluation system of claim 4,

6. acquiring drilling data by a measuring device when drilling a hole; Acquiring RGB image data of a face by an RGB image capturing device; Acquiring spectral image data of the face by a spectral image capture device; Inputting the drilling data as an input variable into a drilling learning model generated by machine learning using teacher data in which the drilling data is an input variable and the face evaluation by a skilled engineer corresponding to the drilling data is an output variable, and calculating a first calculated value; Calculating a second calculated value by inputting the RGB image data as an input variable into an RGB learning model generated by performing machine learning using teacher data in which the RGB image data is used as an input variable and the face evaluation by a skilled engineer corresponding to the RGB image data is used as an output variable; inputting the spectral image data as an input variable into a spectral learning model generated by machine learning using teacher data having the spectral image data as an input variable and a face evaluation by a skilled engineer corresponding to the spectral image data as an output variable, and calculating a third calculated value; Calculating a face evaluation value by integrating the first calculated value, the second calculated value, and the third calculated value; A face evaluation method including:

7. The face evaluation method according to claim 6, wherein when generating the drilling learning model, the drilling data which serves as an explanatory variable of the drilling learning model is weighted according to the geology of the natural ground.

8. The face evaluation method described in claim 7, wherein the weighted drilling data includes at least two or more pieces of data: position information data of the drilling location, data on the feed pressure, rotation pressure, impact pressure, and drilling speed of the drilling machine at the drilling location corresponding to the position information data, and spring water pressure data at the drilling location.

9. The face evaluation method according to any one of claims 6 to 8, wherein when integrating the first calculated value, the second calculated value, and the third calculated value, each calculated value is weighted according to the geology of the natural ground.

10. The face evaluation method according to claim 9, further comprising the steps of: calculating the face value using a face evaluation learning model generated by machine learning using teacher data in which the first calculated value, the second calculated value, and the third calculated value are input variables; and a face evaluation by a skilled technician corresponding to the input variables is used as an output variable; and the first calculated value, the second calculated value, and the third calculated value are used as input data to calculate the face evaluation value as an output variable.

Citation Information

Patent Citations

  • Face plane evaluation system and face plane evaluation method

    JP2018071165A

  • Tunnel face state display system, tunnel face state display method, and moving measuring object

    JP2021046663A

  • Face property estimation device and estimation method

    JP2021107650A

  • Working face evaluation system, working face evaluation learning system, working face evaluation method, working face evaluation learning method, and computer program

    JP2021183774A

  • Rock fall prediction device, machine learning device, rock fall prediction method, and machine learning method

    JP2023034745A