Model generation method and cavity detection method

Abstract

This invention provides a model generation method that can shorten the time required for cavity investigation, improve quality (reduce variability), and reduce manpower. [Solution] For each of the multiple measurement points along the length of the tunnel, waveform data of reflected waves obtained by irradiating with electromagnetic waves is acquired. From the waveform data of the reflected waves, positive polarity extrema exceeding a positive threshold and negative polarity extrema exceeding a negative threshold are extracted. Based on the extracted positive polarity extrema and negative polarity extrema, feature data is generated that includes a multi-wave feature that shows that positive polarity extrema and negative polarity extrema appear alternately in the depth direction, a continuity feature that shows that the multi-wave feature appears continuously across multiple adjacent measurement points along the length of the tunnel, and an extremity occurrence count feature that shows the number of times the positive polarity extrema and negative polarity extrema appear. A trained model for determining the presence or absence of cavities is generated using the feature data and training data that shows the presence or absence of cavities at multiple measurement points.

Description

Technical Field

[0006] , , ,

[0005] , , , , , ,

[0001] The present invention relates to a model generation method and a cavity determination method.

Background Art

[0002] If there is a cavity behind the lining of a tunnel, sudden collapses or the like may occur. Therefore, tunnel maintenance management operators regularly or irregularly conduct cavity surveys on the back of the tunnel lining. In recent years, many of these cavity surveys are performed using an electromagnetic wave radar detector.

[0003] A cavity survey using an electromagnetic wave radar detector is performed by irradiating electromagnetic waves at a measurement point, receiving the reflected waves obtained, generating an electromagnetic wave radar reflected wave image (radar image), and having a technician refer to the radar image. The following Patent Documents 1 and 2 disclose techniques for performing cavity surveys of lining concrete using such an electromagnetic wave radar detector or the like.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0005] By the way, in a cavity survey using an electromagnetic wave radar detector, a veteran technician determines the presence or absence of a cavity from the acquired radar image. However, this work requires know - how based on many years of experience and also requires a great deal of time for data confirmation. Here, when performing a cavity survey, shortening the survey period by improving work efficiency is desired.

[0006] This invention has been made in view of the above circumstances, and aims to provide a model generation method and a void detection method that can shorten the time required for void investigation of lining concrete, improve quality (reduce variability), and reduce manpower. [Means for solving the problem]

[0007] To solve the above problems, a cavity determination device (1) according to a first aspect of the present invention includes: an input unit (31) that inputs measurement data (D1), which is reflected wave data obtained by irradiating a measurement point with electromagnetic waves; a preprocessing unit (32) that performs preprocessing to obtain extreme value data relating to the extreme values ​​of the reflected waves from the measurement data input from the input unit; a determination unit (33) that uses the extreme value data relating to the extreme values ​​of the reflected waves and training data (D12) indicating the presence or absence of a cavity at the measurement point where the reflected wave was obtained, to determine the presence or absence of a cavity at the measurement point where the measurement data input to the input unit was obtained, a trained model (MD) that shows the relationship between the extreme values ​​of the reflected waves and the presence or absence of a cavity, and the extreme value data obtained by the preprocessing unit.

[0008] Furthermore, in the cavity determination device according to the second aspect of the present invention, the preprocessing unit acquires the extreme value data relating to the extreme values ​​of the reflected wave that have a magnitude exceeding a predetermined threshold, in the cavity determination device according to the first aspect.

[0009] Furthermore, in the cavity determination device according to the third aspect of the present invention, the learned model has learned the relationship between a first feature quantity in which positive polarity extrema and negative polarity extrema alternately appear for the reflected wave obtained at least at one measurement point, a second feature quantity in which the first feature quantity for each of the reflected waves obtained at a plurality of adjacent measurement points is continuous in the direction of the measurement point, and the presence or absence of the cavity, and the determination unit determines the presence or absence of the first feature quantity and the second feature quantity in the extremity data acquired by the preprocessing unit, thereby determining the presence or absence of the cavity at the measurement point.

[0010] Furthermore, in the cavity detection device according to the fourth aspect of the present invention, the cavity detection device according to any one of the first to third aspects includes, in the extreme value data, data indicating the number of times the extreme value of the reflected wave appears.

[0011] Furthermore, in a cavity determination device according to a fifth aspect of the present invention, in a cavity determination device according to any of the first to fourth aspects, if the determination unit determines whether or not there is a cavity at a first measurement point located in the middle of three adjacent measurement points, and the determination result of whether or not there is a cavity at a second measurement point, which is one of the other two measurement points, the determination unit changes the determination result of whether or not there is a cavity at the first measurement point to the determination result of whether or not there is a cavity at the second measurement point.

[0012] Furthermore, a cavity detection device according to a sixth aspect of the present invention includes a generation unit (23) that generates the trained model based on extreme value data relating to the extreme values ​​of the reflected wave and training data (D12) indicating the presence or absence of a cavity at the measurement point where the reflected wave was obtained, in a cavity detection device according to any of the first to fifth aspects of the present invention.

[0013] Furthermore, in the cavity detection device according to the seventh aspect of the present invention, the training data is set to extreme value data relating to the extreme values ​​of the reflected waves obtained at a plurality of adjacent measurement points, in the cavity detection device according to any of the first to sixth aspects.

[0014] A cavity determination method according to one aspect of the present invention includes an input step (S21) in which an input unit (31) inputs measurement data (D1), which is data of reflected waves obtained by irradiating a measurement point with electromagnetic waves; a preprocessing step (S22) in which a preprocessing unit (32) performs preprocessing to obtain extreme value data relating to the extreme values ​​of the reflected waves from the measurement data input from the input unit; and a determination step (S23) in which a determination unit (33) uses the extreme value data relating to the extreme values ​​of the reflected waves and training data (D12) indicating the presence or absence of a cavity at the measurement point where the reflected waves were obtained to determine the presence or absence of a cavity at the measurement point where the measurement data input to the input unit was obtained, a trained model (MD) showing the relationship between the extreme values ​​of the reflected waves and the presence or absence of a cavity, which was learned using the extreme value data relating to the extreme values ​​of the reflected waves and training data (D12) indicating the presence or absence of a cavity at the measurement point where the reflected waves were obtained, and the extreme value data obtained by the preprocessing unit to determine the presence or absence of a cavity at the measurement point where the measurement data input to the input unit was obtained.

[0015] A cavity determination program according to one aspect of the present invention causes a computer to perform the following steps: an input step (S21) in which measurement data (D1) is data of reflected waves obtained by irradiating a measurement point with electromagnetic waves; a preprocessing step (S22) in which preprocessing is performed to obtain extremum data relating to the extremum of the reflected wave from the measurement data input in the input step; a determination step (S23) in which a trained model (MD) showing the relationship between the extremum of the reflected wave and the presence or absence of the cavity is trained using the extremum data relating to the extremum of the reflected wave and training data (D12) showing the presence or absence of a cavity at the measurement point where the reflected wave was obtained, and the extremum data obtained in the preprocessing step to determine the presence or absence of a cavity at the measurement point where the measurement data input in the input step was obtained. [Effects of the Invention]

[0016] According to the present invention, it is possible to shorten the time required for investigating voids in lining concrete, improve quality (reduce variability), and reduce manpower. [Brief explanation of the drawing]

[0017] [Figure 1] This is a block diagram showing an example of the hardware configuration of a cavity detection device according to one embodiment of the present invention. [Figure 2] It is a diagram showing an example of display of measurement data in a cavity determination device according to an embodiment of the present invention. [Figure 3] It is a block diagram showing an example of a functional configuration related to generation of a learned model of a cavity determination device according to an embodiment of the present invention. [Figure 4] It is an explanatory diagram showing an example of a learning dataset used in an embodiment of the present invention. [Figure 5] It is a diagram for explaining preprocessing performed in an embodiment of the present invention. [Figure 6] It is a flowchart showing an example of model generation processing performed by a cavity determination device according to an embodiment of the present invention. [Figure 7] It is a block diagram showing an example of a functional configuration related to cavity determination of a cavity determination device according to an embodiment of the present invention. [Figure 8] It is a flowchart showing an example of cavity determination processing performed by a cavity determination device according to an embodiment of the present invention.

Embodiments for Carrying Out the Invention

[0018] Hereinafter, a cavity determination device, a cavity determination method, and a cavity determination program according to embodiments of the present invention will be described in detail with reference to the drawings. The configurations of the following embodiments are examples, and the present invention is not limited to the configurations of the embodiments.

[0019] 〈Cavity determination device〉 FIG. 1 is a block diagram showing an example of the hardware configuration of a cavity determination device according to an embodiment of the present invention. As shown in FIG. 1, the cavity determination device 1 includes an operation unit 11, a display unit 12, a communication unit 13, an input / output unit 14, a storage unit 15, and a processing unit 16. Such a cavity determination device 1 is realized by, for example, a desktop-type, notebook-type, or tablet-type computer, or a workstation or the like.

[0020] The operation unit 11 is equipped with an input device such as a keyboard or a pointing device, and outputs instructions (instructions for the cavity detection device 1) to the processing unit 16 in accordance with the operation of the worker using the cavity detection device 1. The display unit 12 is equipped with a display device such as a liquid crystal display device, and displays various information output from the processing unit 16. The operation unit 11 and the display unit 12 may be physically separate, or they may be physically integrated, such as a touch panel type liquid crystal display device that combines display and operation functions.

[0021] The communication unit 13 communicates with external devices (not shown) under the control of the processing unit 16. The communication unit 13 may communicate directly with the external devices or via a network such as the Internet (not shown). The communication unit 13 may also communicate via wired or wireless connections. The input / output unit 14 performs data input and output to an external recording medium (not shown) under the control of the processing unit 16. The external recording medium may be, for example, a magnetic disk, optical disk, or memory card.

[0022] The storage unit 15 includes, for example, an auxiliary storage device such as an HDD (hard disk drive) or an SSD (solid state drive), and stores various data used by the cavity detection device 1. The storage unit 15 stores, for example, measurement data D1 and a trained model MD. The storage unit 15 may also store a program that implements the functions of the cavity detection device 1.

[0023] Measurement data D1 is data of reflected waves obtained by irradiating measurement points in the tunnel under investigation with electromagnetic waves. Measurement data D1 is formed in the RADAN format, which is the data format of Geophysical Survey Systems, Inc. (hereinafter referred to as GSSI). The trained model MD is a model that shows the relationship between the extreme values ​​of reflected waves obtained by irradiating the tunnel with electromagnetic waves and the presence or absence of cavities, and is generated by machine learning using past survey results. Details of the trained model MD will be described later.

[0024] The processing unit 16 is implemented, for example, by a CPU (Central Processing Unit), RAM (Random Access Memory), etc., and comprehensively controls the operation of the cavity detection device 1 based on operation instructions input from the operation unit 11. The processing unit 16 performs, for example, preprocessing of the measurement data D1 necessary for determining the presence or absence of a cavity, processing to determine the presence or absence of a cavity using the trained model MD and the preprocessed measurement data D1, and other processing.

[0025] Furthermore, the processing unit 16 can also display the measurement data D1 on the display unit 12 based on the operation instructions input from the operation unit 11. For example, the processing unit 16 can display an image (A mode) showing the waveform of the reflected wave obtained at a single measurement point (1 trace) from the measurement data D1 obtained at that point. In addition, the processing unit 16 can display a radar image (B mode) in which the reflected waves obtained at each measurement point are color-coded according to their magnitude from the measurement data D1 obtained at multiple measurement points (multiple traces).

[0026] Figure 2 shows an example of display of measurement data in a cavity detection device according to one embodiment of the present invention. Figure 2(a) is an example of display when measurement data D1 is displayed in A mode, and Figure 2(b) is an example of display when measurement data D1 is displayed in B mode. As shown in Figure 2(a), A mode is a graph with the measured value of measurement data D1 on the horizontal axis and depth on the vertical axis. B mode is a two-dimensional image with distance (distance in the length direction of the tunnel) on the horizontal axis and depth on the vertical axis. The engineer operates the operation unit 11 to display A mode or B mode on the display unit 12 and determines the presence or absence of a cavity.

[0027] The functions of the cavity detection device 1 (functions necessary for detecting cavities) may be realized, for example, by a program (including a program recorded on a recording medium not shown) that implements those functions being executed by hardware such as a CPU. In other words, the functions of the cavity detection device 1 may be realized through the cooperation of software and hardware resources. Of course, the functions of the cavity detection device 1 may also be realized using hardware such as FPGAs, LSIs, and ASICs.

[0028] <Generating a pre-trained model> Figure 3 is a block diagram showing an example of a functional configuration for generating a trained model of a cavity detection device according to one embodiment of the present invention. As shown in Figure 3, the cavity detection device 1 comprises an input unit 21, a pre-processing unit 22, a generation unit 23, a model generation storage unit 24, and a model storage unit 25. The input unit 21, the pre-processing unit 22, and the generation unit 23 are realized by the processing unit 16 shown in Figure 1. That is, the processing unit 16 executes a model generation program stored in the storage unit 15, thereby realizing the functions of the input unit 21, the pre-processing unit 22, and the generation unit 23. The model generation storage unit 24 and the model storage unit 25 are realized by the storage unit 15 shown in Figure 1.

[0029] The input unit 21 receives the training dataset DS. Methods of inputting the training dataset DS to the input unit 21 include, for example, input from the operation unit 11 shown in Figure 1, input from an external device via the communication unit 13 shown in Figure 1, and input from a recording medium via the input / output unit 14 shown in Figure 1. The training dataset DS input to the input / output unit 14 may be temporarily stored in the storage unit 15 shown in Figure 1.

[0030] Figure 4 is an explanatory diagram showing an example of a training dataset used in one embodiment of the present invention. As shown in Figure 4, the training dataset DS includes training samples SP1 to SPn (where n is an integer of 2 or more). Each training sample SP1 to SPn is obtained from the results of a tunnel survey conducted in the past and includes measurement data D11 and training data D12.

[0031] Measurement data D11 is data of reflected waves obtained by irradiating measurement points in tunnels that have been surveyed in the past with electromagnetic waves. Measurement data D11 is formed in the RADAN format, which is the data format of GSSI, for example, similar to measurement data D1 shown in Figure 1. Training data D12 is data indicating the presence or absence of cavities at the measurement points where reflected waves were obtained. Training data D12 is generated, for example, by an engineer displaying measurement data D11 in mode A as shown in Figure 2(a) or mode B as shown in Figure 2(b), and determining the presence or absence of cavities by referring to the displayed content.

[0032] Here, the measurement data obtained from the cavity survey (measurement data D1, D11) is divided into files, with each file containing data in 25-meter sections along the length of the tunnel. In the cavity survey, 70 traces are measured per meter. In other words, there are 70 measurement points per meter where measurements are taken by irradiating with electromagnetic waves. Therefore, the estimated number of traces per 25 meters along the length of the tunnel is 1750 traces.

[0033] However, due to distance measurement errors in the tunnel's longitudinal direction or errors caused by radar contact with uneven lining surfaces, the actual number of traces per 25 meters in the tunnel's longitudinal direction may differ from 1750 traces. Therefore, it is desirable to enlarge the training data D12 every meter if the actual number of traces per 25 meters is less than 1750, and to shrink it if it is more, setting the training data D12 every meter.

[0034] If the above-mentioned scaling is performed, the actual corresponding measurement points (the traces observed to set the training data D12) become unknown. Therefore, it is desirable to set the training data D12 for multiple traces near the grid points set every meter. For example, it is desirable to set the training data D12 for a total of 5 traces: the trace at the grid point and the two traces before and after that grid point. The feature quantity of the maximum value of the extrema of these 5 traces can be considered a feature of the traces near the training data D12, regardless of which value of the 5 traces was selected.

[0035] Alternatively, training data D12 may be set for a radar image formed by the measurement values ​​of consecutive traces (measurement data obtained at consecutive measurement points). When setting such training data D12, it is desirable to set the training data D12 by making a comprehensive judgment that includes the situation of adjacent traces, rather than looking only at individual traces.

[0036] The preprocessing unit 22 performs preprocessing on the training dataset DS input from the input unit 21, which is necessary for creating the trained model MD. Specifically, the preprocessing unit 22 performs preprocessing to obtain extremum data related to the extremum of reflected waves from the measurement data D11 included in the training dataset DS.

[0037] Figure 5 is a diagram illustrating the preprocessing performed in one embodiment of the present invention. The graph shown in Figure 5 is an example of measurement data D11 included in the training dataset DS, with the measurement value on the horizontal axis and depth on the vertical axis. The depth on the vertical axis is the depth from the position of the surface of the lining concrete to which the electromagnetic waves are irradiated, and the position of depth "0" is the position of the surface of the lining concrete.

[0038] The preprocessor 22 acquires extremum data for reflected waves whose magnitude exceeds predetermined thresholds TH1 and TH2. The purpose of acquiring such extremum data is to exclude small reflected waves from sources other than the interface, such as multiple waves. In the example shown in Figure 5, the preprocessor 22 acquires extremum data for nine extremums PK1 to PK9. Extremums PK1, PK3, PK6, and PK8 have a magnitude exceeding threshold TH1, while extremums PK2, PK4, PK5, PK7, and PK9 have an absolute magnitude exceeding the absolute value of threshold TH2.

[0039] The preprocessor 22 acquires data indicating the depth of each of the extreme values ​​PK1 to PK9 and data indicating the magnitude (measured value) of each of the extreme values ​​PK1 to PK9 as extreme value data. The preprocessor 22 also acquires data indicating the number of times an extreme value (extreme values ​​PK1 to PK9) with a magnitude exceeding the thresholds TH1 and TH2 appears (in the example shown in Figure 5, it is "9") as extreme value data.

[0040] Here, reflected waves obtained in areas without cavities tend to show little change in magnitude (change in measured value), while reflected waves obtained in areas with cavities tend to show significant change in magnitude (change in measured value). Therefore, for example, by using the number of extreme value occurrences, it is possible to exclude measurement data obtained in areas without cavities and use measurement data obtained in areas with cavities as the target for training. In other words, it is possible to screen the measurement data to be used for training.

[0041] The generation unit 23 trains the learning model using the learning dataset DS that has been preprocessed by the preprocessing unit 22, and generates the trained model MD. Here, since the learning dataset DS that has been preprocessed by the preprocessing unit 22 includes the extreme value data and training data D12 acquired by the preprocessing unit 22, the generated trained model MD can be said to be a model that shows the relationship between the extreme values ​​of reflected waves obtained by irradiating a tunnel with electromagnetic waves and the presence or absence of voids.

[0042] Here, it is desirable that the trained MD model has learned the relationship between at least the first and second features shown below and the presence or absence of cavities. Alternatively, the trained MD model may have learned the relationship between the first to third features shown below and the presence or absence of cavities. • First feature (multi-wave feature): A feature in which positive and negative extremes alternate for the reflected wave obtained at a single measurement point. • Second feature (continuity feature): A feature that is continuous in the direction of the measurement point (the length of the tunnel) for each of the first feature obtained from multiple adjacent measurement points. • Third feature (polarity feature): A feature that indicates the polarity of the first extreme value that appears in the reflected wave obtained at the measurement point.

[0043] Furthermore, the trained MD model may also have learned the relationship between the fourth and fifth features shown below and the presence or absence of cavities, in addition to the first to third features described above. • Fourth feature (multi-waveform feature): A feature in which a strong second wave appears curved in the opposite direction to the first wave, resulting in an eyeball-like pattern in the radar image. • Fifth feature (delayed waveform feature): A feature that indicates that large extreme values ​​appear in the deeper parts due to the influence of waterlogged cavities, etc.

[0044] The trained model MD generated by the generation unit 23 is, for example, a decision tree model. However, the trained model MD is not limited to a decision tree model; it can also be a model represented using a neural network, or a model represented using other algorithms such as random forests, support vector machines, or Gaussian process regression.

[0045] The model generation storage unit 24 stores the trained model and parameters being trained by the generation unit 23. The model storage unit 25 stores the trained model MD generated by the generation unit 23.

[0046] Figure 6 is a flowchart showing an example of the model generation process performed by a cavity detection device according to one embodiment of the present invention. The process shown in the flowchart in Figure 6 is started, for example, when a user of the cavity detection device 1 operates the operation unit 11 of the cavity detection device 1 to issue a learning start instruction. For the sake of simplicity, it is assumed that the model generation storage unit 24 shown in Figure 3 stores the learning model and parameters.

[0047] When the processing shown in the flowchart in Figure 6 begins, first, the input unit 21 of the cavity detection device 1 acquires the training dataset DS (step S11). Next, the preprocessing unit 22 of the cavity detection device 1 performs preprocessing on the acquired training dataset DS (step S12). For example, it extracts extreme values ​​PK1 to PK9 that have a magnitude exceeding the thresholds TH1 and TH2 explained using Figure 5, and acquires threshold data for these extreme values ​​PK1 to PK9. The preprocessing unit 22 may also perform standardization or normalization processing on the measurement data D11 included in the training dataset DS.

[0048] Next, the generation unit 23 of the cavity detection device 1 performs machine learning using the pre-processed training dataset DS (step S13). Here, since the training dataset DS includes the training data D12, the generation unit 23 performs supervised learning. The generation unit 23 may also perform machine learning by excluding training samples SP1 to SPn included in the training dataset DS that have fewer occurrences of extreme values ​​in the threshold data acquired by the pre-processing unit 22 than a predetermined value. Subsequently, the generation unit 23 determines whether or not the machine learning is complete (step S14).

[0049] If it is determined that machine learning is not yet complete (if the result of the determination in step S14 is "NO"), the generation unit 23 continues machine learning (step S13). For example, it performs machine learning using training samples SP1 to SPn included in the training dataset DS that have not been used in machine learning. On the other hand, if it is determined that machine learning is complete (if the result of the determination in step S14 is "YES"), the generation unit 23 stores the generated trained model MD in the model storage unit 25 (step S15). The series of processes shown in Figure 6 is completed by the above processing.

[0050] In this embodiment, an example is described in which the trained model MD is generated by the cavity detection device 1. However, the trained model MD may be generated by a device other than the cavity detection device 1. If the trained model MD is generated by a device other than the cavity detection device 1, the trained model MD may be supplied from that device to the model storage unit 25.

[0051] <Determining the presence or absence of a cavity> Figure 7 is a block diagram showing an example of the functional configuration related to cavity determination in a cavity determination device according to one embodiment of the present invention. As shown in Figure 7, the cavity determination device 1 includes an input unit 31, a pre-processing unit 32, a determination unit 33, and a determination result display unit 34, in addition to a model storage unit 25. The input unit 31 to the determination unit 33 are realized by the processing unit 16 shown in Figure 1. That is, the processing unit 16 realizes the functions of the input unit 31 to the determination unit 33 by executing the cavity determination program stored in the storage unit 15. The model storage unit 25 is realized by the storage unit 15 shown in Figure 1, and the determination result display unit 34 is realized by the display unit 12 shown in Figure 1.

[0052] The input unit 31 receives the measurement data D1 stored in the storage unit 15 of the cavity detection device 1 shown in Figure 1. In other words, the input unit 31 receives the data of reflected waves obtained by irradiating the measurement points of the tunnel being investigated for the presence or absence of cavities with electromagnetic waves. If the measurement data D1 is not stored in the storage unit 15, the input of the measurement data D1 may be from an external device via the communication unit 13 shown in Figure 1, or from a recording medium via the input / output unit 14 shown in Figure 1.

[0053] The preprocessing unit 32 performs preprocessing necessary for determining the presence or absence of cavities on the measurement data D1 input from the input unit 31. Specifically, the preprocessing unit 32 performs preprocessing to acquire extreme value data related to the extreme values ​​of reflected waves from the measurement data D1, similar to the preprocessing unit 22 shown in Figure 3. For example, as explained using Figure 5, the preprocessing unit 32 acquires extreme value data for reflected wave extrema that have a magnitude exceeding predetermined thresholds TH1 and TH2, in order to exclude small reflected waves from sources other than the interface that originate from multiple waves, etc.

[0054] Similar to the preprocessor 22, the preprocessor 32 acquires data indicating the depth of each extremum and data indicating the magnitude (measured value) of each extremum as extremum data. The preprocessor 32 also acquires data indicating the number of times an extremum with a magnitude exceeding the thresholds TH1 and TH2 appears as extremum data.

[0055] The determination unit 33 uses the trained model MD stored in the model storage unit 25 and the extreme value data acquired by the preprocessing unit 32 to determine whether or not there is a cavity at the measurement point where the measurement data D1 input to the input unit 31 was obtained. Specifically, the determination unit 33 determines whether or not there is a cavity at the measurement point by determining the presence or absence of the first feature (multi-wave feature) and the second feature (continuity feature) described above in the extreme value data acquired by the preprocessing unit 32.

[0056] Furthermore, the determination unit 33 modifies the determination result for the presence or absence of a cavity by considering the determination results for the presence or absence of cavities at adjacent measurement points. Specifically, if the determination result for the presence or absence of a cavity at the middle measurement point of three adjacent measurement points differs from the determination results for the presence or absence of cavities at the other two measurement points, the determination result for the presence or absence of a cavity at the middle measurement point is modified. This modification is made because sections with cavities and sections without cavities tend to be continuous, and the presence or absence of cavities rarely changes from trace to trace.

[0057] For example, if the measurement result at the center point is "cavity present" and the results at the other two measurement points are "no cavity present," the result at the center point will be changed to "no cavity present." Conversely, if the measurement result at the center point is "no cavity present" and the results at the other two measurement points are "cavity present," the result at the center point will be changed to "cavity present."

[0058] The determination result display unit 34 displays the determination result of the determination unit 33 (determination result of whether or not there is a cavity). In addition to the determination result of the determination unit 33, the determination unit 33 may also display the extreme value data acquired by the preprocessing unit 32, the determination result of whether or not the first feature (multi-wave feature) and second feature (continuity feature) mentioned above are present, and other information.

[0059] Figure 8 is a flowchart showing an example of a cavity detection process performed by a cavity detection device according to one embodiment of the present invention. The process shown in the flowchart in Figure 8 is started, for example, when a user of the cavity detection device 1 operates the operation unit 11 of the cavity detection device 1 to issue a command to start the cavity detection process.

[0060] When the flowchart shown in Figure 8 starts, the input unit 31 of the cavity detection device 1 first acquires measurement data D1, which is data of reflected waves obtained by irradiating measurement points in the tunnel being investigated with electromagnetic waves. For example, the input unit 31 acquires the measurement data D1 stored in the storage unit 15.

[0061] Next, the preprocessing unit 32 of the cavity detection device 1 performs preprocessing on the acquired measurement data D1. For example, the preprocessing unit 32 extracts extreme values ​​PK1 to PK9 that have a magnitude exceeding the thresholds TH1 and TH2 explained using Figure 5, and acquires threshold data for these extreme values ​​PK1 to PK9.

[0062] Next, the determination unit 33 of the cavity detection device 1 uses the trained model MD stored in the model storage unit 25 and the extreme value data acquired by the preprocessing unit 32 to determine whether or not there is a cavity at the measurement point where the measurement data D1 input to the input unit 31 was obtained (step S23). Specifically, the determination unit 33 determines whether or not there is a cavity at the measurement point by determining the presence or absence of the first feature (multi-wave feature) and the second feature (continuity feature) described above in the extreme value data acquired by the preprocessing unit 32.

[0063] Furthermore, the determination unit 33 modifies the determination result for the presence or absence of a cavity by taking into consideration the determination results for the presence or absence of a cavity at adjacent measurement points. For example, if the determination result at the measurement point located in the middle is "cavity present" and the determination results at the other two measurement points are "no cavity", the determination result at the measurement point located in the middle is changed to "no cavity".

[0064] The determination unit 33 outputs the determination result of whether or not there is a cavity to the determination result display unit 34 (step S24). As a result, the determination result of whether or not there is a cavity by the determination unit 33 is displayed on the determination result display unit 34.

[0065] As described above, the cavity detection device 1 according to this embodiment has a preprocessing unit 32 that performs preprocessing to acquire extreme value data related to the extreme values ​​of reflected waves from the measurement data D1 input from the input unit 31, and a determination unit 33 that uses a trained model MD and the extreme value data acquired by the preprocessing unit 32 to determine whether or not there is a cavity at the measurement point from which the measurement data D1 input to the input unit 31 was obtained. This makes it possible to shorten the time required for cavity investigation of lining concrete, improve quality (reduce variation), and reduce manpower.

[0066] Furthermore, the cavity detection device 1 according to this embodiment includes a generation unit 23 that generates a trained model MD using training samples SP1 to SPn (training dataset DS) which include measurement data D11 and training data D12. This makes it possible to generate a trained model MD that reflects accumulated know-how and past survey results, thus enabling effective utilization of accumulated know-how and past survey results.

[0067] Furthermore, the functions of the input unit 21, pre-processing unit 22, and generation unit 23 shown in Figure 3, and the functions of the input unit 31, pre-processing unit 32, and determination unit 33 shown in Figure 7, were explained as being realized by the processing unit 16 executing a program. In other words, the functions of the input unit 21, pre-processing unit 22, generation unit 23, input unit 31, pre-processing unit 32, and determination unit 33 were described as examples realized through the cooperation of software and hardware. However, the functions of the input unit 21, pre-processing unit 22, generation unit 23, input unit 31, pre-processing unit 32, and determination unit 33 can also be realized using hardware (including the circuitry) such as LSI (Large Scale Integration), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), and GPU (Graphics Processing Unit).

[0068] Furthermore, while several embodiments of the present invention have been described herein, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These embodiments can be carried out in a variety of other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims and their equivalents.

[0069] Furthermore, the program (information processing program) for realizing the cavity detection device 1 described above may be stored in a computer-readable storage medium, and the program may be loaded into a computer system and executed. Here, "computer system" includes hardware such as the OS and peripheral devices. "Computer-readable storage medium" refers to portable media such as flexible disks, magneto-optical disks, ROMs, CD-ROMs, and storage devices such as hard disks built into a computer system. Moreover, "computer-readable storage medium" also includes volatile memory (RAM) inside a computer system that acts as a server or client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line, which holds the program for a certain period of time. Furthermore, the above program may be transmitted from the computer system that stores the program in a storage device, etc., to another computer system via a transmission medium or by transmission waves in the transmission medium. Here, the "transmission medium" for transmitting the program refers to a medium that has the function of transmitting information, such as a network such as the Internet or a communication line such as a telephone line. Furthermore, the above program may be for realizing only a part of the functions described above. Furthermore, the aforementioned functions may be implemented in combination with programs already stored in the computer system, such as so-called differential files (differential programs). [Explanation of Symbols]

[0070] 1 Cavity determination device 23 Generation part 31 Input section 32 Pre-processing section 33 Judgment section D1 Measurement Data D12 Training Data MD pre-trained model

Claims

1. A model generation method for generating a trained model for determining the presence or absence of voids behind the concrete lining of a tunnel, The steps include acquiring measurement data for each of the multiple measurement points along the length of the tunnel, including waveform data of reflected waves obtained by irradiating them with electromagnetic waves, For each of the plurality of measurement points, the steps include extracting positive polarity extrema exceeding a positive threshold and negative polarity extrema exceeding a negative threshold from the waveform data of the reflected wave, Based on the extracted positive polarity extrema and negative polarity extrema, A multiwave feature that shows that the positive polarity extrema and the negative polarity extrema appear alternately in the depth direction, A continuity feature that indicates the multi-wave feature appears continuously across the plurality of adjacent measurement points in the length direction of the tunnel, An extreme value occurrence count feature quantity indicating the number of occurrences of the positive polarity extreme value and the negative polarity extreme value, The steps include generating feature data that includes, A step of generating a trained model for determining the presence or absence of a cavity using the feature data and training data indicating the presence or absence of cavities at the plurality of measurement points, A model generation method having the following characteristics.

2. The model generation method according to claim 1, wherein the positive threshold and the negative threshold are defined to be values ​​that can exclude small reflected waves originating from multiple waves.

3. The model generation method according to claim 1, wherein the feature data further includes, for each of the plurality of measurement points, a polarity feature indicating the polarity of the first extreme value appearing in the depth direction.

4. The model generation method according to claim 1, wherein the feature data further includes at least one of a multi-wave waveform feature that indicates that a strong second wave appears curved in the opposite direction to the first wave, and a delayed waveform feature that indicates that a large extremum appears in the depths.

5. The model generation method according to any one of claims 1 to 4, wherein the training data is set for a plurality of measurement points including a measurement point corresponding to one of a plurality of grid points set at predetermined distances along the length of the tunnel, and a plurality of measurement points located before and after the said measurement point.

6. A method for determining the presence or absence of voids behind the concrete lining of a tunnel, The steps include acquiring measurement data for each of the multiple measurement points along the length of the tunnel, including waveform data of reflected waves obtained by irradiating them with electromagnetic waves, For each of the plurality of measurement points, the steps include extracting positive polarity extrema exceeding a positive threshold and negative polarity extrema exceeding a negative threshold from the waveform data of the reflected wave, Based on the extracted positive polarity extrema and negative polarity extrema, A multiwave feature that shows that the positive polarity extrema and the negative polarity extrema appear alternately in the depth direction, A continuity feature that indicates the multi-wave feature appears continuously across the plurality of adjacent measurement points in the length direction of the tunnel, An extreme value occurrence count feature quantity indicating the number of occurrences of the positive polarity extreme value and the negative polarity extreme value, The steps include generating feature data that includes, The steps include: inputting the feature data into a trained model generated using the feature data and training data indicating the presence or absence of cavities to determine the presence or absence of cavities at the multiple measurement points; A method for determining a cavity.

7. After the step of determining whether or not there is a cavity, The cavity detection method according to claim 6, further comprising the step of changing the determination result at the central measurement point to the determination result at the two measurement points on either side if the determination result at the central measurement point differs from the determination results at the two measurement points on either side of the plurality of measurement points.

Images