JP2024016883A - Ground state determination system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a ground state determination system capable of determining the ground state in real-time while boring without depending on human determination, which is to solve problems in a conventional art.

SOLUTION: A ground state determination system of the present invention is a system to determine a state of the ground to be bored or excavated, being provided with learned model creation means and ground state determination means. The learned model creation means is means to create a learned model through machine learning of teacher data on which a geological label and a soil property label are attached, and the ground state determination means is means to determine a geological category and a soil property category of the objective ground by inputting construction data acquired during boring to the learned model.

SELECTED DRAWING: Figure 2

COPYRIGHT: (C)2024,JPO&INPIT

Description

The present invention relates to a technology for determining the properties of the target ground for boring, and more specifically, to determine the properties of the target ground in real time during construction using a learned model generated in advance. This relates to a ground property determination system that can perform

Boring is the process of forming a relatively long cylindrical underground hole with a relatively small diameter, and is usually carried out by drilling the hole using a rod and a bit attached to its tip. This boring is sometimes carried out for the purpose of ground investigation, such as core collection and standard penetration testing, but it is also sometimes carried out as part of so-called permanent construction work. For example, in tunnel excavation using blast excavation, boring is performed using a drill jumbo in order to load gunpowder (dynamite), and in tunnel construction using NATM (New Australian Tunneling Method), boring is also performed using a drill jumbo in order to insert rock bolts. . Furthermore, when performing mechanical agitation or high-pressure injection for the purpose of soil improvement through deep mixing, boring is performed vertically downward into the ground.

In recent years, heavy rains associated with abnormal weather have caused deep failures of slopes and slopes, as well as large-scale landslides, making slope countermeasures more important than ever. Anchor systems have now become indispensable as countermeasures against deep collapses and landslides on slopes. This anchor system was originally called an "earth anchor" because one fixed end of the tendon is installed in the ground, and is now called a "ground anchor." Ground anchors have been used for nearly 60 years since they were first introduced in Japan in 1957 as a measure to stabilize dams, and because they are economical and have excellent construction properties, they have become increasingly popular with society as society continues to grow rapidly. Due in part to the acceleration of capital development, it has been used so frequently that an estimated 24,000 construction projects have been completed. This ground anchor is also bored using a boring machine such as a rotary percussion type.

When carrying out boring for permanent construction (hereinafter referred to as "boring work"), a ground survey is generally conducted in advance, and holes are drilled according to the results of the survey. For example, during boring work to install a ground anchor, a ground survey is conducted to determine the location of the anchoring layer and the physical properties of the anchoring layer (such as circumferential frictional resistance). Although the holes were drilled in accordance with the design plan based on the results of the ground investigation, it was of course impossible to visually see the inside of the ground where the holes were being drilled, and it was not possible to carry out the work according to the design plan. However, it is extremely difficult to accurately judge whether there is a problem.

On the other hand, it is known that it is extremely useful to instantly (in real time) grasp the properties of the ground during boring work. For example, if you can grasp the position of the anchoring layer and its physical properties in real time while drilling a hole for a ground anchor, you may be able to use a shorter drilling length (free length or anchoring length) than planned. In some cases, it may turn out that the ground anchor will not work unless the hole is drilled longer than planned, and as a result, an efficient and effective ground anchor can be installed.

Therefore, various efforts have been made to grasp the properties of the ground in real time during boring work. Typical methods include a technology that estimates the hardness and softness of the ground from electrical signals (torque, thrust, impact pressure, etc.) from construction machines, as disclosed in Patent Document 1, and Techniques include determining the geological classification and soil type by visual judgment by specialized engineers or by chemical tests using chemicals, etc., based on the appearance of the slime.

JP2021-127655A

However, all conventional methods for understanding the properties of the ground in real time have had problems. For example, a method of estimating the hardness of the ground from electrical signals from construction machines (hereinafter referred to as the "machine information estimation method") can determine the hardness and softness of the ground and the presence or absence of cavities, but Not suitable for judgment. On the other hand, although the method of determining the appearance of slime (hereinafter referred to as "slime appearance determination method") can determine the geological classification and soil quality classification, it is difficult to determine the hardness and softness of the ground and the presence or absence of cavities.

In addition, in machine information estimation methods, information may be obtained from various measuring instruments and devices installed on construction machines, but each is independent information and has different formats and dimensions (1-3 dimensions). Even if the information is uploaded and aggregated into one central device, it is not easy to comprehensively judge the information. In other words, determining properties from disparate information that is not organically linked relies on human knowledge, experience, and technical ability. As a result, the results vary depending on the ability of the judge, making it impossible to obtain stable judgments. Furthermore, because people deliberate and deliberate, it was not possible to determine the ground properties (geological classification, soil classification, etc.) in real time, and the obtained information could not be fed back to construction in a timely manner.

As described above, the slime appearance determination method is a method in which an engineer with specialized knowledge visually determines the geological classification or soil quality classification based on the slime appearance. Therefore, when carrying out boring work, it is necessary to secure specialized engineers and have them permanently stationed, and a budget must be set aside for their labor costs. In particular, in cases where multiple construction teams are deployed to drill holes, multiple specialized engineers must be stationed at all times, which not only costs money but also makes it difficult to secure the required number in the first place. Furthermore, it is difficult to quickly aggregate the judgment results of multiple specialized engineers, and as a result, it is not possible to judge the ground properties in real time, and the obtained information cannot be fed back to construction in a timely manner. . Furthermore, it is unavoidable that the judgment results differ depending on the skill of the expert engineer, and it is not possible to obtain stable judgments.

An object of the present invention is to solve the problems faced by the prior art, namely, to provide a ground property determination system that can grasp the ground properties during drilling in real time without relying on human judgment. It is.

The ground property determination system of the present invention generates a trained model by learning construction data with type labels, and calculates geological classification and soil type classification in real time by inputting the construction data into the trained model during construction. This invention was made with a focus on the fact that a judgment can be made based on an unprecedented idea.

The ground property determination system of the present invention is a system for determining the properties of the ground to be drilled or excavated, and includes a trained model generation means and a ground property determination means. Among these, the trained model generation means is a means for generating a trained model by machine learning training data to which geological classification labels and soil classification labels are attached, and the ground property determination means is a means for generating trained models during drilling (or This is a means of determining the geological classification and soil quality classification of the target ground by inputting construction data acquired during excavation (during excavation) into a trained model. The training data includes construction machine data related to construction obtained from a drilling machine (or excavation machine), camera image data obtained by photographing ground slime obtained by drilling (or excavation), and ground slime captured by a spectral camera. One or more data selected from spectral image data obtained by spectral image data obtained by drilling a hole (or after drilling) and borehole image data obtained by photographing with a borehole camera after drilling (or after excavation), geological classification labels and soil quality. It is generated by adding a category label. Furthermore, the construction data input to the trained model is one or more data selected from construction machine data, camera image data, spectral image data, and borehole image data, and is the same as the teaching data used for machine learning. This is the same type of data.

The ground property determination system of the present invention can also generate a learned model by performing machine learning on teacher data labeled with ground hardness/softness labels. In this case, the ground property determining means also determines the hardness and softness of the target ground when construction data is input to the learned model.

The ground property determination system of the present invention can also generate a learned model by performing machine learning on teacher data labeled with ground cavities. In this case, the ground property determining means also determines the condition of cavities in the target ground when construction data is input to the learned model.

The ground property determination system of the present invention may further include one or more terminal devices and a central device capable of communicating with the terminal devices. In this case, construction data input using the terminal device is transmitted to the central device. Then, the central device receives the construction data and determines the geological classification and soil type of the target ground by inputting the construction data into the ground property determining means. Further, the terminal device can output the results determined by the ground property determining means in real time.

The ground property determination system of the present invention is a three-dimensional (or two-dimensional) display that superimposes the actual result of drilling (or excavation) and the planned drilling shape (or excavation shape) as construction progresses. It can also be made possible.

The ground property determination system of the present invention has the following effects.
(1) Information about the ground that is invisible (geological classification, soil type, hardness and softness of the ground, presence or absence of cavities) can be determined in real time, and as a result, necessary information can be quickly fed back to the construction team. , more appropriate boring work can be performed.
(2) Since it does not depend on human knowledge, human error and human errors are eliminated, and stable and consistent judgment results can be obtained.
(3) Even in cases where multiple construction teams are deployed to drill holes, the construction data from each team can be consolidated into a central device, allowing various parties to check and share it. This makes it possible to provide appropriate advice from various sources, such as predicting uncertainties in the construction environment and appropriate construction management.
(4) The actual results of drilling (or excavation) as construction progresses, the planned drilling shape (or excavation shape), and the determined ground information (geological classification, soil type, hardness and softness of the ground) By superimposing and displaying 3D (or 2D) information such as the presence or absence of cavities, it is possible to understand the properties of the ground more clearly, and for example, it is also possible to confirm the planned anchorage layer and support layer. .
(5) By using the information obtained during construction as training data and continuously performing deep learning, it is possible to obtain a trained model with further improved estimation accuracy.

A model diagram schematically showing ground properties determined based on construction data. FIG. 1 is a block diagram showing the main configuration of the ground property determination system of the present invention. FIG. 1 is a block diagram showing a ground property determination system in which a central device and terminal devices are distributed. An upper list diagram showing an example of construction machine data obtained from a drilling machine. (a) is a model image diagram schematically showing a ground surface and a ground anchor head displayed in 3D, and (b) is a model image diagram schematically showing a ground cross section and a ground anchor displayed in 3D. (a) is a model image diagram schematically showing a planned borehole displayed by the output control means, and (b) is a model image diagram schematically showing the overlapping display of the planned borehole and finished borehole displayed by the output control means. Model image diagram. A flow diagram showing the main process flow from performing test construction to generating a learned model using the ground property determination system. 1 is a flowchart showing the flow of main processes up to determining the ground properties using the ground properties determination system during this construction.

An example of the ground property determination system of the present invention will be explained based on the drawings.

1. Overall Overview The present invention is a technology that can determine the properties of the target ground in real time during boring work or ground excavation. More specifically, as shown in FIG. 1, the ground properties are determined based on data obtained during construction (hereinafter referred to as "construction data"). Note that the present invention can determine the ground properties for each drilling depth during the boring work, and can also determine the ground properties for each excavation depth during the ground excavation work, but for convenience, this will not be described here. This will be explained using the example of boring work.

Examples of construction data include "construction machine data," "slime photographs," "spectrum data," and "borehole photographs," as shown in FIG. 1, for example. Among these, the construction machine data is various information obtained from drilling machines (boring machines, drill jumbo, etc.) used for boring work, as will be described later. In addition, slime photography is image data obtained by photographing slime raised on the ground during drilling with a digital camera, etc., and spectral data is data obtained by photographing the slime with a hyperspectral camera or multispectral camera. be. However, it is preferable to take slime photos and spectrum data together with a color sample before acquiring them. A borehole photograph is image data obtained by photographing the inner wall of a borehole with a borehole camera. Construction data is acquired periodically (or intermittently) during construction; for example, construction machine data is acquired every time the excavation is approximately 15 cm, slime photographs and spectral data are acquired every time the excavation is approximately 1 to 2 m, and borehole photographs are acquired every time the excavation is approximately 1 to 2 m. Obtained when drilling is completed.

On the other hand, the ground properties include, for example, "geological classification", "soil classification", "hardness/softness of the ground", and "cavity condition" as shown in FIG. 1. Among these, geological classifications are, for example, classifications such as sandstone, granite, and slate, and soil classifications are classifications such as sandy soil, gravelly soil, and clayey soil. In addition, the hardness and softness of the ground is an indicator of its strength, such as the so-called rock class classification, and the condition of cavities literally refers to the conditions related to cavities, such as the presence or absence of cavities, their size, and their scale.

A ``trained model'' generated by machine learning is used to determine ground properties based on construction data. In other words, when construction data is input, the learned model outputs ground properties. Note that in order to generate the learned model, various conventionally used machine learning techniques, including deep learning, can be employed.

Teacher data is used to generate a trained model using machine learning. This training data is obtained by adding a correct label to the construction data, that is, a known ground property as a label (hereinafter simply referred to as a "ground property label"). In order to obtain training data, it is necessary to conduct a trial construction (hereinafter simply referred to as "test construction") in advance.In other words, after carrying out the test construction, actual construction (hereinafter simply referred to as " The ground properties will be determined during the actual construction. In the test construction, it is desirable to use a machine similar to that used in the main construction (particularly a drilling machine), and at least obtain construction data similar to that obtained in the main construction. Therefore, in order to distinguish between the two types of construction data, the data obtained in the test construction will be referred to as "test construction data" and the data obtained in the actual construction will be referred to as "main construction data." Alternatively, actual construction carried out in the past can be treated as test construction. In other words, training data is created by attaching ground property labels to the construction data based on various data obtained from actual construction carried out in the past, and a trained model is generated by learning the training data. That's why. Of course, training data can also be created by using information obtained from test construction and past actual construction.

2. Ground Property Determination System The ground property determination system of the present invention will be explained in detail. FIG. 2 is a block diagram showing the main configuration of the ground property determination system 100 of the present invention. As shown in this figure, the ground property determination system 100 of the present invention includes a learned model generation means 101 and a ground property determination means 102, and also includes a construction machine data acquisition means 103, an output control means 104, and an output means 105. , the learned model storage means 108, the test boring data storage means 109, the test construction data storage means 110, and the actual construction data storage means 111. 107, terminal equipment, and a central device.

Among the main elements constituting the ground property determination system 100, the learned model generation means 101, the ground property determination means 102, and the output control means 104 can be manufactured as dedicated devices, or can be manufactured using general-purpose computer equipment. You can also. This computer device is equipped with a processor such as a CPU, memory such as ROM and RAM, input means such as a mouse and keyboard, and a display, and includes a personal computer (PC), a server, a tablet PC such as an iPad (registered trademark), and a smartphone. It can be configured by a mobile terminal, etc. If a computer device equipped with a display is used, the display may be used as the output means 105.

The learned model storage means 108, the test boring data storage means 109, the test construction data storage means 110, and the main construction data storage means 111 can use a storage device of a general-purpose computer, or can be constructed in a database server. can. When building a database server, it can be placed on a local network (LAN: Local Area Network), or it can be a cloud server that stores it via the Internet. Note that these storage means can be constructed separately, or two or more storage means can be constructed together.

Although it has been explained that some of the elements constituting the ground property determination system 100 can be made up of general-purpose computer devices, they can also be made up of terminal equipment and a central device as shown in FIG. 3. The terminal equipment may be placed near the construction team that actually performs the boring work, and may be a personal computer, tablet PC, smartphone, or the like. On the other hand, the central device may be placed in a management team different from the construction team, such as a management office or a supervisor's office, and may utilize a server, personal computer, or the like. Further, by providing the terminal side transmitting/receiving means 106 in the terminal device and the central side transmitting/receiving means 107 in the central device, the terminal device and the central device can communicate, and thereby various information (e.g. , actual construction data), and various information (for example, determination results) can be transmitted from the central device to the terminal device. Note that although Figure 3 shows an example in which terminal devices placed in two construction teams are each connected to the central device, this is not the only option. It can also be connected and operated.

Hereinafter, each main element constituting the ground property determination system 100 of the present invention will be explained in detail.

(Learned model generation means)
The trained model generating means 101 is a means for generating a trained model. Specifically, this trained model is generated by learning teacher data (data in which ground property labels are attached to construction data) using machine learning technology such as deep learning. The training data can be obtained by performing test construction as described above. An example of the procedure for obtaining teacher data will be described below.

First, a test borehole (hereinafter simply referred to as "test borehole") is formed. For example, if the main construction is ground anchor installation work and the plan is to drill a hole using a rotary percussion type boring machine, the test boring hole will be made using the planned (or the same standard) rotary percussion type boring machine. It is best to drill the holes. At this time, the holes are drilled while acquiring the construction data that is planned to be acquired during the main construction. For example, if you are planning to acquire construction machine data, camera image data, spectral image data, and borehole image data during main construction, you will also acquire construction machine data, camera image data, spectral image data, and borehole image data during test construction. Drill the hole while

As mentioned above, the construction machine data here refers to various information obtained from the drilling machines used for boring work (boring machines, drill jumbo, etc.) It is acquired by the machine data acquisition means 103. For example, construction machine data includes raw data such as drilling speed, feeding force, torque, water supply pressure, impact pressure, and water flow rate of a boring machine, as shown in Figure 4, and the raw data that can be obtained from these raw data. Examples include impact energy, number of impact blows, breaking energy value, and drilling energy value. Of course, the information is not limited to these, and various information regarding the drilling machine can also be acquired as construction machine data, including measured values such as the position of the bit during drilling and the insertion angle of the rod. During the test construction, construction data (that is, test construction data) such as construction machine data, camera image data, spectral image data, and borehole image data acquired during the test construction is stored in the test construction data storage means 110 (FIG. 2).

Once the test borehole is formed, the obtained boring core is observed. In addition, we conduct uniaxial (or triaxial) compression tests on the obtained boring cores, perform in-hole horizontal loading tests, perform standard penetration tests during drilling, and conduct desired physical tests to determine the ground conditions. Obtain physical test data. Then, based on the observation results of the borehole core and the physical test data, a person (for example, a specialized engineer) evaluates the ground properties of the test borehole. Specifically, the test borehole is divided into blocks according to depth, and the geological classification and soil type (or one or the other) are evaluated for each block. At this time, in addition to geological classification and soil type classification, it is also possible to evaluate the hardness and softness of the ground and the condition of cavities. The physical test data obtained here and the ground properties for each block (geological classification, soil classification, etc.) are stored in the test boring data storage means 109 (FIG. 2).

On the other hand, while drilling test boreholes, test construction data such as construction machine data, camera image data, spectral image data, and borehole image data are obtained, especially construction machine data, camera image data, and spectral image data. Multiple data are obtained for each excavation depth. Based on the depth at which these test construction data were acquired, ground properties (geological classification, soil classification, etc.) are associated with each test construction data, and each test construction data is associated with a geological classification label and soil classification. Teacher data is created by adding soil property labels such as labels, hardness/softness labels, and cavity labels.

The training data may be created according to the type of ground properties evaluated. For example, when a specialized engineer evaluates the geological classification and soil type, the training data is created by adding geological classification labels and soil type labels to the test construction data, and the specialized engineer evaluates the geological classification, soil type, and soil type. When hardness/softness and cavity conditions are evaluated, training data is created by adding geological classification labels, soil type classification labels, hardness/softness labels, and cavity labels to the test construction data.

Further, the teacher data may be created according to the type of the acquired test construction data. For example, in a case where only construction machine data is acquired as test construction data, training data is created by adding a ground property label to the construction machine data. In cases where image data and borehole image data have been acquired, training data is created by adding ground property labels to each of the construction machine data, camera image data, spectral image data, and borehole image data. Alternatively, in a case where construction machine data, camera image data, and spectral image data are acquired as test construction data, a dataset consisting of construction machine data, camera image data, and spectral image data (data set according to drilling depth) In other words, it is also possible to create training data by assigning a ground property label to each data set.

As mentioned above, training data does not necessarily need to be created after carrying out test construction (in this case, meaning not actual construction), but training data can be created based on various information obtained from past actual construction. Alternatively, training data can be created based on various information obtained from test construction and various information obtained from past actual construction.

The learned model generation means 101 generates a learned model by learning a large amount of teacher data obtained through such a procedure. Note that the teacher data can be created from one test borehole, or can be created after forming two or more test boreholes. The trained model generated by the trained model generation means 101 is stored in the trained model storage means 108 (FIG. 2).

(Means for determining ground properties)
During the actual construction, construction data (that is, actual construction data) such as acquired construction machine data, camera image data, spectral image data, and borehole image data is stored in the actual construction data storage means 111 (FIG. 2). Then, the ground property determining means 102 reads out the actual construction data from the actual construction data storage means 111, and outputs the ground properties (geological classification, soil classification, etc.) by inputting the actual construction data into the learned model. As mentioned above, construction data is acquired periodically (or intermittently) during construction. Therefore, the ground property determining means 102 can also output the ground property each time the main construction data is acquired, in other words, for each drilling depth. For example, when drilling a hole for a ground anchor, it is possible to grasp the position of the anchoring layer and its physical properties in real time, and as a result, it is possible to find out that a shorter hole length than planned is sufficient On the other hand, it is possible to realize efficient and effective construction by discovering that it will not work unless the hole is drilled longer than planned.

(Output control means)
The ground properties (geological classification, soil classification, etc.) determined by the ground property determining means 102 are output to an output means 105 such as a display or a printer. At this time, the ground properties for each drilling depth can be output in list format or in drawing format. In addition, the output control means 104 controls the shape of the borehole formed by the main construction (hereinafter simply referred to as "the main borehole") as well as the ground properties for each drilling depth in two-dimensional (2D) or three-dimensional (3D). It can also be displayed.

In cases where measurement values such as the position of the bit during drilling and the insertion angle of the rod are acquired as construction machine data, it is possible to understand the shape of the drilling hole (in other words, the actual finished shape) according to the progress of construction. can. Therefore, the output control means 104 can display the main borehole in 3D (or 2D) on the output means 105 as shown in FIG. In addition, the output control means 104 controls the shape (drilling shape) of a planned borehole (hereinafter referred to as "planned borehole DP") as shown in FIG. It is also possible to overlap the shape of the "as-built borehole DA" and display it in 3D (or 2D) on the output means 105. In addition, in FIG. 6(a), only the planned borehole DP is shown by a broken line, and FIG. 6(b) shows an overlapping display of the planned borehole DP and the completed borehole DA.

Further, in FIG. 6(b), the ground properties for each drilling depth are displayed on the completed borehole DA. At this time, different ground properties may be displayed by changing colors, shading, and patterns. Furthermore, when the system is configured with terminal equipment and a central device as shown in FIG. 3, even the construction team performing boring work can check the information displayed on the output means 105 of the terminal equipment in real time. For example, the main construction data is transmitted to the central device using the terminal-side transmitting/receiving means 106 of the terminal device, the central transmitting/receiving means 107 receives this, and the ground property determining means 102 determines the ground property for each drilling depth. At the same time, by transmitting the information to the terminal device, the construction team can sequentially confirm the ground properties for each drilling depth along with the planned borehole DP and finished borehole DA.

(Example of use)
An example of using the ground property determination system 100 of the present invention will be described with reference to FIGS. 7 and 8. FIG. 7 is a flowchart illustrating the main process flow from performing test construction to generating a learned model using the ground property determination system 100. Moreover, FIG. 8 is a flowchart showing the flow of the main process of determining the ground properties using the ground properties determination system 100 and displaying the results while carrying out the main construction. In FIGS. 7 and 8, the center column shows the process to be performed, the left column shows information necessary for the process, and the right column shows information generated from the process.

To generate a trained model, first, as shown in FIG. 7, drilling is performed to form a test borehole (Step 211 in FIG. 7). At this time, it is advisable to drill holes using the same excavating machine as that planned for this construction. During drilling of the test borehole, a boring core is collected (Step 212 in Figure 7), construction data (test construction data) similar to that planned in the main construction is obtained (Step 213 in Figure 7), and Perform standard penetration tests as necessary.

Once the test borehole is formed, physical test data of the ground is obtained by conducting physical tests such as a horizontal loading test in the hole and a compression test of the boring core (Step 214 in Figure 7), and the obtained boring core is observed. . Then, for example, a specialized engineer evaluates the ground properties for each depth block of the test borehole based on the physical test data and the observation results of the boring core. Here, as ground properties, it is possible to evaluate the geological classification and soil type (or either one), or in addition to the geological classification and soil type, it is also possible to evaluate the hardness and softness of the ground and the condition of cavities (Fig. 7, Step 215).

Once the ground properties are evaluated, training data is created by associating the ground properties with each test construction data based on the acquired depth and further adding a ground property label to each test construction data (Step 216 in Figure 7). ). It is advisable to create training data according to the type of ground properties evaluated. For example, when a specialized engineer evaluates the geological classification and soil type, the training data is created by adding geological classification labels and soil type labels to the test construction data, and the specialized engineer evaluates the geological classification, soil type, and soil type. When hardness/softness and cavity conditions are evaluated, training data is created by adding geological classification labels, soil type classification labels, hardness/softness labels, and cavity labels to the test construction data.

Further, the teacher data may be created according to the type of the acquired test construction data. For example, in a case where only construction machine data is acquired as test construction data, training data is created by adding a ground property label to the construction machine data, and the construction machine data, camera image data, and spectral image are used as test construction data. In the case where data and borehole image data have been acquired, training data is created by adding ground property labels to each of the construction machine data, camera image data, spectral image data, and borehole image data. Alternatively, in cases where construction machine data, camera image data, and spectral image data are acquired as test construction data, a ground property label can be assigned to each data set consisting of construction machine data, camera image data, and spectral image data. You can also create data.

Once the training data is created, the trained model generation means 101 generates a trained model by learning the training data using machine learning technology such as deep learning (Step 217 in FIG. 7).

After carrying out the test construction and generating a trained model, the main construction is performed. When carrying out this construction, it is preferable to arrange a central device in the management team and to arrange terminal equipment near the construction team, as shown in FIG. After the central equipment and terminal equipment are arranged and other necessary temporary construction work and other preparatory work is completed, drilling is performed to form the main borehole as shown in FIG. 8 (Step 221 in FIG. 8). At this time, it is advisable to drill the hole using the same excavating machine as that used in the test construction.

During drilling of the main borehole, at least construction data similar to that obtained in the test construction (main construction data), that is, construction data of the same type as the teacher data is acquired. For example, in FIG. 8, construction machine data is acquired by the construction machine data acquisition means 103 each time drilling progresses (Step 222 in FIG. 8), and camera image data and spectral image data are acquired each time slime is obtained ( Step 223 in FIG. 8). Additionally, in this example, in addition to these actual construction data, measured values such as the position of the bit during drilling and the insertion angle of the rod are acquired each time drilling progresses. The actual shape of the hole (drilled shape) is known (Step 224 in FIG. 8).

When the actual construction data is acquired, the actual construction data is input to the ground property determining means 102, and the soil property determining means 102 outputs the ground property related to the actual construction data (Step 225 in FIG. 8). At this time, since the same type of construction data as the teacher data has been acquired, the same type of construction data as the teacher data is input to the ground property determining means 102. For example, in a case where machine learning is performed using training data in which ground property labels are attached to construction machine data, camera image data, and spectral image data, the construction machine data, camera image data, and spectral image data are input to the ground property determination means 102. That's why.

Also, the ground property determining means 102 naturally outputs the same type of ground property as the teacher data used to generate the learned model. For example, in a case where training data to which a geological classification label and a soil classification label are assigned is subjected to machine learning, the ground property determining means 102 outputs the geological classification and soil classification as the ground properties, and outputs the geological classification label, soil classification label, hardness and softness. In the case where training data to which labels and cavity labels have been assigned is subjected to machine learning, the ground property determining means 102 outputs the geological classification, soil classification, hardness and softness of the ground, and cavity status as the ground properties. For example, if training data is created by assigning soil property labels to each of the construction machine data, camera image data, and spectral image data, the ground properties will be assigned to each of the input construction machine data, camera image data, and spectral image data. Output. On the other hand, if the teacher data is created by adding a ground property label to a data set consisting of construction machine data, camera image data, and spectral image data, for example, the ground property is output for each data set.

As shown in FIG. 3, when the system is configured by a terminal device and a central device, the main construction data can be input using the terminal device, and the ground property determining means 102 of the central device can output the ground property. That is, when the actual construction data is input to a terminal device placed near the construction team, the actual construction data is transmitted by the terminal-side transmitting/receiving means 106 to the central-side transmitting/receiving means 107, and the ground property determining means 102 of the central device reads the actual construction data. The actual construction data is taken in and the ground properties are output. In this case, the ground properties output by the ground property determining means 102 may be transmitted to the terminal device and displayed on the output means 105 thereof.

When the ground properties are outputted by the ground properties determination means 102, the output control means 104 displays the main borehole in 3D (or 2D) on the output means 105 as shown in FIG. 5 (Step 226 in FIG. 8). Alternatively, as shown in Figure 6, the shape of the planned borehole DP (drilled hole shape) and the shape of the finished borehole DA may be displayed overlappingly, and the type of ground properties may be indicated using colors, shading, and patterns. can.

The ground property determination system of the present invention can be used in various construction works that involve boring, such as ground anchor construction as a measure against slopes, tunnel construction, and ground improvement work by deep mixing. According to the present invention, highly accurate information-based construction is possible, and as a result, high-quality social infrastructure such as tunnel structures and foundations can be constructed, and appropriate construction on slopes, etc. Considering that this invention can be used as a countermeasure against disasters, it can be said that this invention is not only useful in industry, but can also be expected to make a significant contribution to society.

100 Ground property determination system of the present invention 101 Learned model generation means (of the ground property determination system) 102 Ground property determination means (of the soil property determination system) 103 Construction machine data acquisition means (of the soil property determination system) 104 (Soil property Output control means (of the judgment system) 105 Output means (of the ground property judgment system) 106 Terminal-side transmission/reception means (of the ground property judgment system) 107 Central-side transmission/reception means (of the ground property judgment system) 108 Learning (of the ground property judgment system) Completed model storage means 109 Test boring data storage means (of the soil property determination system) 110 Test construction data storage means (of the soil property determination system) 111 Main construction data storage means (of the soil property determination system) DA As-built borehole DP Plan borehole

Claims (5)

A system for determining the properties of target ground for drilling or excavation,
a trained model generating means for generating a trained model by machine learning training data attached with geological classification labels and/or soil classification labels;
A ground property determining means for determining the geological classification and/or soil classification of the target ground by inputting construction data acquired during drilling or excavation into the learned model,
The teacher data includes construction machine data related to construction obtained from a drilling machine or an excavation machine, camera image data obtained by photographing ground slime obtained by drilling or excavation, and data obtained by capturing the ground slime with a spectral camera. The geological classification label and/or the soil quality classification label is attached to one or more data selected from spectral image data obtained by photographing with a borehole camera after drilling or excavation. is generated with
The construction data input to the trained model is one or more data selected from the construction machine data, the camera image data, the spectral image data, and the borehole image data, and is used for machine learning. is the same type of data as the teacher data obtained.
A ground property determination system characterized by: The learned model generation means generates the learned model by machine learning the teacher data labeled with the hardness/softness of the ground,
The ground property determining means determines the hardness and softness of the target ground when the construction data is input to the learned model.
The ground property determination system according to claim 1, characterized in that: The learned model generation means generates the learned model by machine learning the teacher data labeled with a ground cavity label,
When the construction data is input to the learned model, the ground property determining means determines the condition of the cavity in the target ground.
The ground property determination system according to claim 1, characterized in that: one or more terminal devices;
further comprising a central device capable of communicating with the terminal device,
The construction data input using the terminal device is transmitted to the central device,
The central device receives the construction data and determines the geological classification and/or soil classification of the target ground by inputting the construction data into the ground property determining means,
The terminal device can output the results determined by the ground property determining means in real time.
The ground property determination system according to any one of claims 1 to 3, characterized in that: The terminal device can display the actual result of drilling or excavation according to the progress of construction and the planned drilling shape or excavation shape in a two-dimensional or three-dimensional manner.
5. The ground property determination system according to claim 4.

Images