JP7756064B2 - Non-destructive diagnostic method, non-destructive diagnostic device, and non-destructive diagnostic program for porous concrete layers - Google Patents

Description

The present invention relates to a non-destructive diagnostic method, a non-destructive diagnostic device, and a non-destructive diagnostic program for porous concrete layers.

Activated carbon adsorption basins and sand filtration basins are installed at water purification plants in designated areas depending on the condition of the raw water. Activated carbon adsorption basins and sand filtration basins are equipped with underdrainage devices. Among these, perforated plate-type underdrainage devices have a porous concrete layer as one of their components. The porous concrete layer is installed on the top layer of the underdrainage device and has the functions of supporting filter media such as activated carbon and sand, and of evenly passing treated water filtered by the filter media and collecting it in a pressure chamber. The porous concrete layer is formed using a different method than the regular concrete used in bridges and tunnels, and is porous and permeable. In other words, the porous concrete layer is formed in a coarser state than regular concrete.

If a defect such as a crack or fissure occurs in the porous concrete layer, there is a risk that filter media such as activated carbon or sand will pass through the porous concrete layer and become mixed into the treated water in the pressure chamber. Furthermore, if the defect such as a crack is serious, the filter media may fall into the pressure chamber, causing problems in the operation of the equipment. In other words, if a defect such as a crack or fissure occurs in the porous concrete layer, it may not be able to perform its function. Therefore, there is a need for a non-destructive method for diagnosing the integrity of the porous concrete layer.

Patent Document 1 discloses a non-destructive evaluation method and non-destructive diagnostic device for the soundness of concrete bodies such as bridge decks, girders, piers, abutments, tunnel linings, retaining walls, revetments, and buildings. Concrete bodies such as those described in Patent Document 1 are formed in a denser state than porous concrete layers. Therefore, images based on the exploration data obtained by exploring concrete bodies such as those described in Patent Document 1 with an electromagnetic radar probe have a relatively clear background.

In contrast, as mentioned above, porous concrete layers are formed in a coarser state than regular concrete. Therefore, in images based on the exploration data obtained by exploring the porous concrete layer with an electromagnetic radar probe, the background is less clear than in the case of concrete bodies such as those described in Patent Document 1. Therefore, while the non-destructive evaluation method and non-destructive diagnostic device described in Patent Document 1 can non-destructively diagnose the soundness of regular concrete bodies used in bridges, tunnels, etc., they may not be able to non-destructively diagnose the soundness of the porous concrete layer in under-drainage systems.

JP 2010-107259 A

The present invention was made in consideration of the above circumstances, and aims to provide a non-destructive diagnostic method, device, and program for non-destructively diagnosing a porous concrete layer that can non-destructively diagnose the soundness of the porous concrete layer in an under-drainage device.

A first aspect of the present invention is a method for non-destructively diagnosing the soundness of a porous concrete layer in an underdrainage device installed in at least one of an activated carbon adsorption basin and a sand filtration basin, comprising the steps of: exploring the interior of the healthy porous concrete layer with an electromagnetic radar probe and storing the acquired first exploration data; performing machine learning using the stored first exploration data as training data to generate a learning model for the healthy porous concrete layer; storing the generated learning model; exploring the interior of the porous concrete layer after it has been installed and used as part of the underdrainage device with the electromagnetic radar probe and acquiring second exploration data; and diagnosing the soundness of the porous concrete layer installed as part of the underdrainage device using the acquired second exploration data as input values and the stored learning model.

According to a first aspect of the present invention, the interior of a healthy porous concrete layer is probed using an electromagnetic radar probe and first probe data obtained is stored in advance. Machine learning is performed using the stored first probe data as training data, and a learning model for the healthy porous concrete layer is generated and stored. Then, after the porous concrete layer has been installed and used as part of an underdrainage device, the interior of the porous concrete layer is probed using the electromagnetic radar probe and second probe data obtained is used as input values, and the soundness of the porous concrete layer installed as part of an underdrainage device is diagnosed using the stored learning model. As a result, the non-destructive diagnosis method for porous concrete layers according to the first aspect of the present invention can non-destructively diagnose the soundness of a porous concrete layer that has been formed in a coarser state than normal concrete.

A second aspect of the present invention is a non-destructive diagnosis method for a porous concrete layer, wherein the first exploration data is exploration data obtained by exploring the interior of a specimen of the sound porous concrete layer using the electromagnetic wave radar exploration device.

According to the second aspect of the present invention, the exploration data acquired from a sound porous concrete layer is exploration data acquired by exploring the interior of a specimen of a sound porous concrete layer using an electromagnetic radar probe. Therefore, machine learning can be performed using the exploration data for a simple porous concrete layer as the specimen as training data to generate a learning model, and the soundness of the porous concrete layer to be diagnosed can be diagnosed. As a result, the non-destructive diagnosis method for a porous concrete layer according to the second aspect of the present invention can diagnose the soundness of a porous concrete layer in a simpler manner.

A third aspect of the present invention is a non-destructive diagnosis method for a porous concrete layer, characterized in that, in the first aspect of the present invention, the first exploration data is exploration data obtained by exploring the interior of the porous concrete layer with the electromagnetic radar exploration device before it is installed and used as part of the under-drainage device.

According to the third aspect of the present invention, the exploration data obtained from a healthy porous concrete layer is exploration data obtained by using an electromagnetic radar probe to explore the interior of the porous concrete layer before it is installed and used as part of an underdrainage device. Therefore, machine learning can be performed using the exploration data from a porous concrete layer that has actually been installed as part of an underdrainage device as training data to generate a learning model and diagnose the soundness of the porous concrete layer. As a result, the non-destructive diagnosis method for a porous concrete layer according to the third aspect of the present invention can diagnose the soundness of the porous concrete layer with greater accuracy.

A fourth aspect of the present invention is a non-destructive diagnosis method for a porous concrete layer, characterized in that, in the first aspect of the present invention, the first exploration data is exploration data obtained by using the electromagnetic radar probe to explore the interior of the porous concrete layer before it is installed and used as part of the under-drainage device installed in at least one of an activated carbon adsorption basin different from the activated carbon adsorption basin and a sand filtration basin different from the sand filtration basin.

According to a fourth aspect of the present invention, the exploration data acquired from a healthy porous concrete layer is exploration data acquired by using an electromagnetic radar probe to explore the interior of a porous concrete layer installed in at least one of an activated carbon adsorption basin and a sand filtration basin other than the activated carbon adsorption basin or sand filtration basin in which the porous concrete layer to be diagnosed is installed and before it is used. Therefore, machine learning can be performed using the exploration data for a porous concrete layer installed in at least one of an activated carbon adsorption basin and a sand filtration basin other than the activated carbon adsorption basin or sand filtration basin in which the porous concrete layer to be diagnosed is installed as training data to generate a learning model, and the soundness of the porous concrete layer to be diagnosed can be diagnosed. As a result, the non-destructive porous concrete layer diagnosis method according to the fourth aspect of the present invention can diagnose the soundness of a porous concrete layer to be diagnosed that is installed as part of an underdrainage system, even in an activated carbon adsorption basin or sand filtration basin after the porous concrete layer has already been used.

A fifth aspect of the present invention is a non-destructive diagnosis method for a porous concrete layer, characterized in that, in any one of aspects 1 to 4 of the present invention, it further comprises the steps of: searching the interior of the porous concrete layer having a defect using the electromagnetic radar search device; storing the acquired third search data; and performing machine learning using the stored first search data and third search data as training data to generate a learning model regarding the presence or absence of the defect.

According to a fifth aspect of the present invention, machine learning is performed using, as training data, exploration data relating to a sound porous concrete layer and exploration data relating to a porous concrete layer having defects, to generate a learning model regarding the presence or absence of defects within the porous concrete layer. As a result, the non-destructive diagnosis method for a porous concrete layer according to the fifth aspect of the present invention can diagnose the soundness of the porous concrete layer with greater accuracy.

A sixth aspect of the present invention is a non-destructive diagnosis method for a porous concrete layer, characterized in that, in the fifth aspect of the present invention, the third exploration data is exploration data obtained by exploring the interior of a specimen of the porous concrete layer having the defect using the electromagnetic wave radar exploration device.

According to the sixth aspect of the present invention, the probe data acquired from a porous concrete layer having a defect is probe data acquired by using an electromagnetic radar probe to probe the interior of a specimen of the porous concrete layer having a defect. Therefore, machine learning can be performed using the probe data for a simple porous concrete layer as the specimen as training data to generate a learning model, and the soundness of the porous concrete layer to be diagnosed can be diagnosed. As a result, the non-destructive diagnosis method for a porous concrete layer according to the sixth aspect of the present invention can diagnose the soundness of the porous concrete layer in a simpler manner.

A seventh aspect of the present invention is a non-destructive diagnosis method for a porous concrete layer according to the fifth aspect of the present invention, characterized in that the third exploration data is exploration data obtained by using the electromagnetic radar probe to explore the interior of the porous concrete layer having the defect after it has been installed and used as part of the under-drainage device installed in at least one of an activated carbon adsorption basin different from the activated carbon adsorption basin and a sand filtration basin different from the sand filtration basin.

According to a seventh aspect of the present invention, the probe data acquired from the defective porous concrete layer is probe data acquired by using an electromagnetic radar probe to probe the interior of a defective porous concrete layer that has been installed and used in at least one activated carbon adsorption basin and/or sand filtration basin other than the activated carbon adsorption basin and/or sand filtration basin in which the porous concrete layer to be diagnosed is installed, and that has exceeded its design life. Therefore, machine learning can be performed using the probe data for the defective porous concrete layer installed in at least one activated carbon adsorption basin and/or sand filtration basin other than the activated carbon adsorption basin and/or sand filtration basin in which the porous concrete layer to be diagnosed is installed as training data to generate a learning model, and the soundness of the porous concrete layer to be diagnosed can be diagnosed. Thus, the non-destructive porous concrete layer diagnosis method according to the seventh aspect of the present invention can diagnose the soundness of a porous concrete layer to be diagnosed that is installed as part of an underdrainage system by using the probe data for the porous concrete layer installed in at least one activated carbon adsorption basin and/or sand filtration basin other than the activated carbon adsorption basin and/or sand filtration basin in which the porous concrete layer to be diagnosed is installed.

An eighth aspect of the present invention is a device for non-destructively diagnosing the soundness of a porous concrete layer in an underdrainage device installed in at least one of an activated carbon adsorption basin and a sand filtration basin. The device comprises: a whole-body healthy data storage unit that stores first exploration data acquired by exploration with an electromagnetic radar probe, the first exploration data relating to the interior of the healthy porous concrete layer; a machine learning unit that performs machine learning using the first exploration data stored in the whole-body healthy data storage unit as training data to generate a learning model for the healthy porous concrete layer; a learning model storage unit that stores the learning model generated by the machine learning unit; a post-operation exploration data storage unit that stores second exploration data acquired by exploration with the electromagnetic radar probe, the second exploration data relating to the interior of the porous concrete layer after it has been installed and used as part of the underdrainage device; and a soundness diagnosis unit that uses the exploration data stored in the post-operation exploration data storage unit as input values and the learning model stored in the learning model storage unit to diagnose the soundness of the porous concrete layer installed as part of the underdrainage device.

According to an eighth aspect of the present invention, the overall healthy inspection data storage unit pre-stores first inspection data acquired by inspecting the interior of a healthy porous concrete layer using an electromagnetic radar probe. The machine learning unit then performs machine learning using the first inspection data stored in the overall healthy inspection data storage unit as training data to generate a learning model for a healthy porous concrete layer. The learning model storage unit stores the learning model generated by the machine learning unit. The post-operation inspection data storage unit also stores second inspection data acquired by inspecting the interior of a porous concrete layer after it has been installed and used as part of an underdrainage device using the electromagnetic radar probe. The soundness assessment unit then uses the second inspection data stored in the post-operation inspection data storage unit as input values and assesses the soundness of the porous concrete layer installed as part of an underdrainage device using the learning model stored in the learning model storage unit. This allows the non-destructive diagnosis device for porous concrete layers according to the eighth aspect of the present invention to non-destructively assess the soundness of a porous concrete layer formed in a coarser state than normal concrete.

A ninth aspect of the present invention is a program executed by a computer of an apparatus for non-destructively diagnosing the soundness of a porous concrete layer in an underdrainage device installed in at least one of an activated carbon adsorption basin and a sand filtration basin, the program causing the computer to execute the following steps: storing first exploration data obtained by exploring the interior of the healthy porous concrete layer with an electromagnetic radar probe; performing machine learning using the stored first exploration data as training data to generate a learning model for the healthy porous concrete layer; storing the generated learning model; exploring the interior of the porous concrete layer after it has been installed and used as part of the underdrainage device with the electromagnetic radar probe to obtain second exploration data; and diagnosing the soundness of the porous concrete layer installed as part of the underdrainage device using the acquired second exploration data as input values and the stored learning model.

According to a ninth aspect of the present invention, a computer is caused to execute an operation of pre-storing first exploration data obtained by exploring the interior of a healthy porous concrete layer using an electromagnetic radar exploration device, performing machine learning using the stored first exploration data as training data, and generating and storing a learning model for the healthy porous concrete layer. Then, second exploration data obtained by exploring the interior of the porous concrete layer after it has been installed and used as part of an underdrainage device is used as input values, and diagnosing the soundness of the porous concrete layer installed as part of an underdrainage device based on the stored learning model.

As a result, the non-destructive diagnostic program for porous concrete layers according to the ninth aspect of the present invention can non-destructively diagnose the soundness of porous concrete layers that are formed in a coarser state than normal concrete.

The present invention provides a non-destructive porous concrete layer diagnostic method, non-destructive diagnostic device, and non-destructive diagnostic program that can non-destructively diagnose the soundness of the porous concrete layer in an under-drainage device.

1 is a schematic perspective view showing an underdrain device having a porous concrete layer to which a non-destructive diagnostic method according to an embodiment of the present invention is applied; FIG. 2 is an enlarged schematic perspective view of the underdrain device of the present embodiment. FIG. 2 is a cross-sectional view showing the underdrain device of the present embodiment. 1 is a block diagram showing a configuration of a main part of a nondestructive diagnostic device according to an embodiment of the present invention. 1 is a block diagram showing a specific configuration of a main part of a nondestructive diagnostic device according to an embodiment of the present invention. 1 is an example of an image relating to the exploration data obtained from a specimen of a porous concrete layer. 1 is a flowchart illustrating a nondestructive diagnostic method according to the present embodiment.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
The embodiments described below are preferred examples of the present invention, and therefore various technically preferable limitations are applied thereto, but the scope of the present invention is not limited to these aspects unless otherwise specified in the following description to the effect that the present invention is particularly limited. Furthermore, in each drawing, similar components are designated by the same reference numerals, and detailed descriptions thereof will be omitted as appropriate.

FIG. 1 is a schematic perspective view showing an underdrainage device having a porous concrete layer to which a non-destructive diagnostic method according to an embodiment of the present invention is applied.
FIG. 2 is a schematic perspective view showing an enlarged underdrain device of this embodiment.
FIG. 3 is a cross-sectional view showing the underdrain device of this embodiment.
1 and 2 show a state in which some members are cut away for the sake of convenience of explanation.

The non-destructive diagnostic method according to this embodiment is a method for non-destructively diagnosing the soundness of the porous concrete layer 51 of the underdrain device 5 installed in at least one of an activated carbon adsorption basin and a sand filtration basin. The following explanation uses as an example a case in which the underdrain device 5 is installed in an activated carbon adsorption basin.

As shown in Figure 1, an activated carbon adsorption basin 2 is installed in a water purification plant depending on the condition of the water 3 to be treated, for example, when the water 3 to be treated is contaminated with organic matter. An underdrain device 5 is installed in the activated carbon adsorption basin 2. The underdrain device 5 has a porous concrete layer 51, a dispersed gravel layer 52, a slit plate 53, and an air dispersion beam 54. As described above, the non-destructive diagnostic method according to this embodiment is applied to the porous concrete layer 51 of the underdrain device 5. Therefore, the underdrain device 5 will first be described with reference to Figures 1 to 3.

The porous concrete layer 51 is located on the top layer of the under-drainage device 5. The porous concrete layer 51 is formed by cementing gravel with a particle size of approximately 6 mm to 8 mm. Sand is not mixed with the gravel or cement. In other words, the method of forming the porous concrete layer 51 differs from the method of forming ordinary concrete used in buildings such as bridges and tunnels. The porous concrete layer 51 is formed in a coarser state than ordinary concrete. The gravel used in the porous concrete layer 51 is uniform or homogeneous. Therefore, the porous concrete layer 51 is formed with a uniform or homogeneous distribution of voids, which are the areas other than the solid gravel solidified with cement. The uniform or homogeneous distribution of voids allows treated water to pass through the porous concrete layer 51 evenly. Specifically, the porosity of the porous concrete layer 51 is approximately 30±5%. Here, "void ratio" refers to the proportion of voids per unit volume of the porous concrete layer 51, expressed as a percentage.

In the case of an activated carbon adsorption basin, activated carbon is installed on top of the porous concrete layer 51 as the filter medium 4. In the case of a sand filter basin, filter sand is installed on top of the porous concrete layer 51 as the filter medium 4. The activated carbon used as the filter medium 4 adsorbs and removes suspended solids in the treated water 3 supplied to the activated carbon adsorption basin 2, as well as odors, color, and organic matter dissolved in the treated water 3 supplied to the activated carbon adsorption basin 2. In other words, the filter medium 4 functions to adsorb and filter the treated water 3 supplied to the activated carbon adsorption basin 2. The porous concrete layer 51 supports the filter medium 4 from below and evenly passes the treated water filtered by the filter medium 4 and collects it in the pressure chamber 21. The function of the porous concrete layer 51 will be described in detail below. The thickness t1 of the porous concrete layer 51 (see Figure 3) is approximately 70 mm. Note that the thickness t1 is not limited to 70 mm.

The dispersed gravel layer 52 is provided between the porous concrete layer 51 and the slit plate 53. For example, the layer of dispersed gravel layer 52 provided above the slit plate 53 is formed of gravel with a particle size of approximately 6 mm or more and 8 mm or less. The thickness t2 (see Figure 3) of the layer of dispersed gravel layer 52 provided above the slit plate 53 is approximately 20 mm. However, thickness t2 is not limited to 20 mm. For example, the layer of dispersed gravel layer 52 provided inside the slit plate 53 is formed of gravel with a particle size of approximately 8 mm or more and 12 mm or less.

The slit plate 53 is provided between the dispersed gravel layer 52 and the air dispersion beam 54. The slit plate 53 has a main body and a mesh portion. The main body of the slit plate 53 is made of reinforced concrete such as stainless steel bars. The mesh portion of the slit plate 53 is installed on top of the main body and is made of, for example, stainless steel. The slit plate 53 prevents the dispersed gravel layer 52 from falling into the pressure chamber 21. The thickness t3 of the slit plate 53 (see Figure 3) is approximately 60 mm. However, the thickness t3 is not limited to 60 mm.

The air dispersion beam 54 is located on the lowest level of the underwater collection device 5. The air dispersion beam 54 is made of reinforced concrete and supports the slit plate 53 from below. The thickness t4 of the air dispersion beam 54 (see Figure 3) is approximately 360 mm. However, the thickness t4 is not limited to 360 mm.

The treated water filtered by the filter media 4 passes through the under-water collection device 5 and is collected in the pressure chamber 21 located below the under-water collection device 5. The treated water collected in the pressure chamber 21 passes through the pressure chamber 21 and the collection culvert 22 before being supplied to a distribution reservoir or the like.

As mentioned above, the activated carbon used as the filter medium 4 captures suspended solids in the water to be treated 3 supplied to the activated carbon adsorption basin 2, as well as odors, colors, and organic matter dissolved in the water to be treated 3 supplied to the activated carbon adsorption basin 2. Therefore, after the activated carbon adsorption basin 2 starts operating, cleaning is periodically performed to remove suspended solids adhering to the filter medium 4. Cleaning the filter medium 4 is known as backwashing, and is performed, for example, by supplying water to the filter medium 4 while blowing air into it. The "backwashing" of this embodiment will be further explained with reference to Figures 2 and 3.

When backwashing is performed, for example, as indicated by arrow A1 in Figures 2 and 3, backwash water passes from the pressure chamber 21 through the under-drainage device 5 and is supplied to the filter medium 4. The flow rate of the backwash water is, for example, about 0.3 ^m3 /min or more and 0.5 ^m3 /min or less per 1 ^m2 . However, the flow rate of the backwash water is not limited to this range.

Furthermore, when backwashing is performed, backwash air is supplied to the filter medium 4 simultaneously with the supply of backwash water. That is, for example, as indicated by arrow A2 in FIG. 2, backwash air supplied from the cleaning air conduit 23 (see FIG. 1) is guided into the air dispersion beam 54. Next, as indicated by arrow A3 in FIGS. 2 and 3, the backwash air guided into the air dispersion beam 54 is blown out from a plurality of holes formed on the side of the air dispersion beam 54, passes through the underwater collection device 5, and is supplied to the filter medium 4. The flow rate of the backwash air is, for example, approximately 0.5 m ³ /min or more and 1.0 m ³ /min or less per 1 m ^2. However, the flow rate of the backwash air is not limited to this range.

During the backwashing process, the backwash water and backwash air vigorously agitate the activated carbon filter media 4, causing the suspended solids adhering to the filter media 4 to peel off. Furthermore, as the backwash water is supplied to the filter media 4, the water level in the activated carbon adsorption basin 2 rises. Just before the water in the activated carbon adsorption basin 2 flows into the backwash water outflow conduit 24 (see Figure 1), the supply of backwash air is stopped. Subsequently, only the backwash water is supplied to the filter media 4, and the captured suspended solids (i.e., suspended solids) suspended within the layer of the filter media 4 due to the simultaneous supply of backwash water and backwash air are discharged through the backwash water outflow conduit 24.

Cleaning of the filter media 4 (i.e., backwashing) must be carried out uniformly throughout the activated carbon adsorption basin 2. Therefore, the porous concrete layer 51 must have the ability to evenly distribute the backwash water and backwash air.

As described above with reference to Figures 1 to 3, the functions required of the porous concrete layer 51 include supporting the filter media 4, allowing the treated water filtered by the filter media 4 to pass evenly and collect in the pressure chamber 21, and evenly distributing the backwash water and backwash air.

If a defect such as a crack or fissure occurs in the porous concrete layer 51, the filter material 4, such as activated carbon, may fall through the porous concrete layer 51 and become mixed with the treated water stored in the pressure chamber 21. Furthermore, if the defect, such as a crack, is serious, the filter material 4 may fall into the pressure chamber 21, causing disruption to the operation of the equipment. In other words, if a defect such as a crack or fissure occurs in the porous concrete layer 51, the porous concrete layer 51 may not be able to function properly. Therefore, it is desirable to non-destructively diagnose the integrity of the porous concrete layer 51. However, as mentioned above, the porous concrete layer 51 is formed in a coarse state compared to ordinary concrete. Therefore, it is difficult to non-destructively diagnose the integrity of the porous concrete layer 51 using methods for non-destructively diagnosing the integrity of ordinary concrete bodies. Furthermore, knowledge and skill are required to identify the defective areas to be diagnosed and diagnose the soundness of the porous concrete layer 51 using only images based on the acquired data obtained by scanning the porous concrete layer 51 with an electromagnetic radar probe.

In contrast, the non-destructive diagnostic method, non-destructive diagnostic device, and non-destructive diagnostic program of this embodiment perform machine learning using exploration data obtained from a healthy porous concrete layer as training data, generate and store a learning model for a healthy porous concrete layer, and diagnose the soundness of the porous concrete layer 51 based on the learning model.
The nondestructive diagnostic method, the nondestructive diagnostic device, and the nondestructive diagnostic program according to this embodiment will be further described below with reference to the drawings.

FIG. 4 is a block diagram showing the configuration of the main parts of the nondestructive diagnostic device according to this embodiment.
The nondestructive diagnostic device 6 according to this embodiment includes a computer 61 and a storage unit 62. The computer 61 includes a control unit 63 (see FIG. 5 ) and reads out a program 621 stored in the storage unit 62 to perform various calculations and processes. The term "computer" as used herein is not limited to a personal computer, but also includes an arithmetic processing unit, a microcomputer, etc. included in information processing equipment, and is a general term for equipment and devices that can realize the functions of the present invention by a program.

The storage unit 62 stores a program 621 executed by the computer 61. Examples of the storage unit 62 include a semiconductor memory or hard disk drive (HDD) built into the non-destructive diagnostic device 6. Alternatively, the storage unit 62 may be an external storage device connected to the computer 61.

Program 621 in this embodiment is an example of the "non-destructive diagnostic program" of the present invention. Program 621 also includes a sequence program for measurement, an image processing program for image processing, and a calculation program. Program 621 does not necessarily have to be stored in storage unit 62; it may be pre-stored on a computer-readable storage medium and distributed, or downloaded to non-destructive diagnostic device 6 via a network.

FIG. 5 is a block diagram showing a specific configuration of the main parts of the nondestructive diagnostic device according to this embodiment.
FIG. 6 is an example of an image relating to the exploration data acquired from a specimen of a porous concrete layer.

The non-destructive diagnostic device 6 according to this embodiment includes a control unit 63, a memory unit 62, a touch panel 64, and a communication unit 65. The control unit 63 is, for example, a CPU (central processing unit) and reads a program 621 (see FIG. 4) stored in the memory unit 62 to perform various calculations and processes. The control unit 63 includes an image processing unit 631, a machine learning unit 632, and a soundness diagnosis unit 633. The image processing unit 631, the machine learning unit 632, and the soundness diagnosis unit 633 are realized by the computer 61 executing the program 621 stored in the memory unit 62. The image processing unit 631, the machine learning unit 632, and the soundness diagnosis unit 633 may be realized by hardware or a combination of hardware and software. The memory unit 62 includes a healthy whole body exploration data storage unit 622, a defective body exploration data storage unit 623, a learning model storage unit 624, and a post-operation exploration data storage unit 625.

The image processing unit 631 performs image processing on the exploration data acquired by the electromagnetic radar exploration device 7 during exploration of the interior of the porous concrete layer 51. The electromagnetic radar exploration device 7 is a well-known electromagnetic radar exploration device that emits electromagnetic waves from a transmitting antenna and receives, via a receiving antenna, the reflected waves reflected at the boundary surface between materials with different dielectric constants within the porous concrete layer 51, thereby exploring the interior of the porous concrete layer 51 and acquiring exploration data. For example, the image processing unit 631 performs at least one of background processing, processing to identify the surface position of the porous concrete layer 51, noise removal processing, and gain processing on the exploration data received from the electromagnetic radar exploration device 7 via the communication unit 65. The image processing unit 631 also performs processing to display an image of the interior of the porous concrete layer 51 as a result of the image processing on the touch panel 64 or external monitor 8, thereby visualizing the interior of the porous concrete layer 51.

Note that the display unit on which the image processing unit 631 displays the image of the interior of the porous concrete layer 51 is not limited to the touch panel 64 and the external monitor 8. The touch panel 64 and the external monitor 8 are examples of display units.

The machine learning unit 632 performs machine learning using the exploration data stored in the healthy whole body exploration data storage unit 622 as training data, and generates a learning model for the healthy porous concrete layer 51. The machine learning unit 632 also performs machine learning using the exploration data stored in the healthy whole body exploration data storage unit 622 and the exploration data stored in the defective body exploration data storage unit 623 as training data, and generates a learning model for the presence or absence of defects inside the porous concrete layer 51.

The healthy whole inspection data storage unit 622 stores inspection data related to the interior of a sound porous concrete layer 51 that is free of defects. The inspection data stored in the healthy whole inspection data storage unit 622 is an example of the "first inspection data" of the present invention. The inspection data stored in the healthy whole inspection data storage unit 622 is, for example, inspection data acquired by an electromagnetic radar probe 7 inspecting a specimen of a sound porous concrete layer 51 that is not installed as part of an underdrainage device 5. In other words, in this example, the inspection data acquired by the electromagnetic radar probe 7 inspecting a specimen of a sound porous concrete layer 51 is pre-stored in the healthy whole inspection data storage unit 622. The image shown in Figure 6(a) is an example of an image obtained by image processing the inspection data stored in the healthy whole inspection data storage unit 622 by the image processing unit 631. Note that the specimen used to acquire the first inspection data may be a sound specimen of a single layer of porous concrete, or may be a sound specimen with the same component structure as an underdrainage device actually installed in an activated carbon adsorption basin, etc.

Alternatively, the exploration data stored in the healthy overall exploration data storage unit 622 is, for example, exploration data acquired by an electromagnetic wave radar probe 7 exploring the interior of a porous concrete layer 51 that has been installed as part of the under-drainage device 5 and before use. In other words, in this example, the exploration data acquired by the electromagnetic wave radar probe 7 exploring a healthy porous concrete layer 51 before use is pre-stored in the healthy overall exploration data storage unit 622.

Alternatively, the exploration data stored in the healthy overall exploration data storage unit 622 is, for example, exploration data obtained by an electromagnetic wave radar probe 7 exploring the interior of a porous concrete layer 51 installed in at least one of an activated carbon adsorption basin and a sand filtration basin other than the activated carbon adsorption basin and sand filtration basin in which the porous concrete layer 51 to be diagnosed is installed and before use. In other words, in this example, the exploration data obtained by the electromagnetic wave radar probe 7 exploring a healthy porous concrete layer 51 other than the porous concrete layer 51 to be diagnosed and before use is stored in advance in the healthy overall exploration data storage unit 622.

The defect body exploration data storage unit 623 stores exploration data related to the interior of the porous concrete layer 51 having a defect. The exploration data stored in the defect body exploration data storage unit 623 is an example of the "third exploration data" of the present invention. The exploration data stored in the defect body exploration data storage unit 623 is, for example, exploration data acquired by an electromagnetic wave radar probe 7 exploring a specimen of the porous concrete layer 51 having a defect that is not installed as part of the underdrainage device 5. In other words, the exploration data acquired by the electromagnetic wave radar probe 7 exploring the specimen of the porous concrete layer 51 having a defect is pre-stored in the defect body exploration data storage unit 623. The specimen for acquiring the third exploration data may be a specimen having a simulated defect in a single layer of porous concrete, or may be a specimen having a simulated defect with the same component structure as an underdrainage device actually installed in an activated carbon adsorption basin, etc.

Alternatively, the exploration data stored in the defective body exploration data storage unit 623 is, for example, exploration data acquired by an electromagnetic wave radar probe 7 exploring the interior of a defective porous concrete layer 51 that has been installed and used in at least one of an activated carbon adsorption basin and a sand filtration basin other than the activated carbon adsorption basin and sand filtration basin in which the porous concrete layer 51 to be diagnosed is installed, and has exceeded its design life. In other words, in this example, the exploration data acquired by the electromagnetic wave radar probe 7 exploring the defective porous concrete layer 51 other than the porous concrete layer 51 to be diagnosed after the porous concrete layer 51 has been used is pre-stored in the defective body exploration data storage unit 623.

The images shown in Figures 6(b) to 6(d) are examples of images obtained by image processing by the image processing unit 631 of the exploration data stored in the defect exploration data storage unit 623. That is, Figure 6(b) is an example of an image obtained by image processing by the image processing unit 631 of exploration data related to a specimen of a porous concrete layer 51 having a 5 mm crack within range A11. Figure 6(c) is an example of an image obtained by image processing by the image processing unit 631 of exploration data related to a specimen of a porous concrete layer 51 having a 3 mm crack within range A12. Figure 6(d) is an example of an image obtained by image processing by the image processing unit 631 of exploration data related to a specimen of a porous concrete layer 51 having a 2 mm crack within range A13. As in the example images shown in Figures 6(a) to 6(d), the background in images obtained by image processing of exploration data related to the interior of the porous concrete layer 51 is relatively unclear, so extensive knowledge is required to visually assess the soundness of the porous concrete layer 51.

As described above, the machine learning unit 632 of this embodiment performs machine learning using the exploration data stored in the healthy whole exploration data storage unit 622 as training data, and generates a learning model for the healthy porous concrete layer 51. Furthermore, the machine learning unit 632 of this embodiment performs machine learning using the exploration data stored in the healthy whole exploration data storage unit 622 and the exploration data stored in the defective body exploration data storage unit 623 as training data, and generates a learning model for the presence or absence of defects inside the porous concrete layer 51.

As described above with reference to Figures 1 to 3, the gravel used in the porous concrete layer 51 is uniform or homogeneous. Therefore, the voids formed in the porous concrete layer 51 are distributed uniformly or homogeneously within the porous concrete layer 51. Therefore, for example, the machine learning unit 632 performs machine learning using as training data the uniformity, evenness, or continuity of at least one of the gravel and voids in images obtained by image processing the exploration data stored in the healthy overall exploration data storage unit 622 by the image processing unit 631, and generates a learning model regarding the presence or absence of defects within the healthy porous concrete layer 51. Furthermore, for example, the machine learning unit 632 performs machine learning using as training data the uniformity, evenness, and continuity of at least one of the gravel and voids in images obtained by image processing by the image processing unit 631 from the exploration data stored in the healthy whole exploration data storage unit 622, and the non-uniformity, non-uniformity, and discontinuity of at least one of the gravel and voids in images obtained by image processing by the image processing unit 631 from the exploration data stored in the defective body exploration data storage unit 623, and generates a learning model regarding the presence or absence of defects inside the porous concrete layer 51.

The learning model storage unit 624 stores the learning model generated by the machine learning unit 632.

The soundness diagnosis unit 633 uses the exploration data stored in the post-operation exploration data memory unit 625 as input values, and uses the learning model stored in the learning model memory unit 624 to diagnose the soundness of the porous concrete layer 51 installed as part of the under-drainage device 5, and outputs the presence or absence of defects as a diagnosis result. The learning model used by the soundness diagnosis unit 633 may be a learning model related to a sound porous concrete layer 51 stored in the learning model memory unit 624, or a learning model related to the presence or absence of defects inside the porous concrete layer 51 stored in the learning model memory unit 624.

The post-operation exploration data storage unit 625 stores exploration data acquired by the electromagnetic radar probe 7 after the porous concrete layer 51 installed as part of the under-drainage device 5 has been used. The exploration data stored in the post-operation exploration data storage unit 625 is an example of the "second exploration data" of the present invention. The exploration data stored in the post-operation exploration data storage unit 625 is exploration data acquired by the electromagnetic radar probe 7 exploring the usage state of the porous concrete layer 51 after it has been installed and used as part of the under-drainage device 5. In other words, the exploration data acquired by the electromagnetic radar probe 7 exploring the usage state of the porous concrete layer 51 after the filter media 4 has been placed on the porous concrete layer 51 is stored in the post-operation exploration data storage unit 625.

After the under-drainage device 5 begins operation, the filter media 4 is replaced, for example, once every two to three years. When the filter media 4 is replaced, the water to be treated 3 (see Figure 1) and filter media 4 are removed from the activated carbon adsorption basin 2, exposing the surface (specifically, the top surface) of the porous concrete layer 51. At this time, for example, the electromagnetic wave radar probe 7 can probe the usage status of the porous concrete layer 51 and obtain probe data, which can then be stored in the post-operation probe data storage unit 625.

FIG. 7 is a flowchart illustrating the nondestructive diagnostic method according to this embodiment.
First, in step S11, the interior of the healthy porous concrete layer 51 is probed by the electromagnetic wave radar probe 7, and the obtained probe data is stored in the healthy overall probe data storage unit 622. As described above with reference to FIGS. 5 and 6 , the probe data stored in the healthy overall probe data storage unit 622 is, for example, probe data related to a specimen of the healthy porous concrete layer 51 that is not installed as part of the underdrainage device 5. Alternatively, the probe data stored in the healthy overall probe data storage unit 622 is, for example, probe data related to the porous concrete layer 51 before it is installed as part of the underdrainage device 5 and used. Alternatively, the probe data stored in the healthy overall probe data storage unit 622 is, for example, probe data related to the porous concrete layer 51 before it is installed and used in at least one of an activated carbon adsorption basin and a sand filtration basin other than the activated carbon adsorption basin and the sand filtration basin in which the porous concrete layer 51 to be diagnosed is installed.

Next, in step S12, the interior of the defective porous concrete layer 51 is probed using the electromagnetic radar probe 7, and the obtained probe data is stored in the defective body probe data storage unit 623. As described above with reference to Figures 5 and 6, the probe data stored in the defective body probe data storage unit 623 is, for example, probe data related to a specimen of the defective porous concrete layer 51 that is not installed as part of the under-drainage device 5. Alternatively, the probe data stored in the defective body probe data storage unit 623 is, for example, probe data related to a defective porous concrete layer 51 that has been installed and used in at least one of an activated carbon adsorption basin and a sand filtration basin other than the activated carbon adsorption basin and sand filtration basin in which the porous concrete layer 51 to be diagnosed is installed, and that has exceeded its design life.

Next, in step S13, the machine learning unit 632 uses the exploration data stored in the healthy whole body exploration data storage unit 622 as training data, performs machine learning using a known machine learning algorithm such as a neural network, and generates a learning model for the healthy porous concrete layer 51. The machine learning unit 632 also uses the exploration data stored in the healthy whole body exploration data storage unit 622 and the exploration data stored in the defective body exploration data storage unit 623 as training data, performs machine learning using a known machine learning algorithm such as a neural network, and generates a learning model for the presence or absence of defects inside the porous concrete layer 51.

Next, in step S14, the learning model generated by the machine learning unit 632 is stored in the learning model storage unit 624.

Next, in step S15, the interior of the porous concrete layer 51 after it has been installed and used as part of the under-drainage device 5 is scanned by the electromagnetic radar scanner 7, and the acquired scanning data is stored in the post-operation scanning data storage unit 625. As described above with reference to Figure 5, the scanning data stored in the post-operation scanning data storage unit 625 is scanning data relating to the usage state of the porous concrete layer 51 after it has been installed and used as part of the under-drainage device 5.

Next, in step S16, the soundness diagnosis unit 633 uses the exploration data stored in the post-operation exploration data memory unit 625 as input values, diagnoses the soundness of the porous concrete layer 51 installed as part of the under-drainage device 5 using the learning model stored in the learning model memory unit 624, and outputs the presence or absence of defects as the diagnosis result. The learning model used by the soundness diagnosis unit 633 is, for example, a learning model related to a sound porous concrete layer 51 stored in the learning model memory unit 624. Alternatively, the learning model used by the soundness diagnosis unit 633 is, for example, a learning model related to the presence or absence of defects inside the porous concrete layer 51 stored in the learning model memory unit 624.

According to the non-destructive diagnostic method, non-destructive diagnostic device, and non-destructive diagnostic program of this embodiment, the machine learning unit 632 performs machine learning using the exploration data acquired from a healthy porous concrete layer as training data, and generates a learning model for the healthy porous concrete layer 51. The learning model memory unit 624 then stores the learning model generated by the machine learning unit 632. Furthermore, the soundness diagnosis unit 633 uses the exploration data stored in the post-operation exploration data memory unit 625 as input values and diagnoses the soundness of the porous concrete layer 51 installed as part of the underdrain device 5 using the learning model stored in the learning model memory unit 624, i.e., the learning model for the healthy porous concrete layer 51. This allows the soundness of the porous concrete layer 51, which is formed in a coarser state than normal concrete, to be non-destructively diagnosed.

If the exploration data stored in the overall healthy exploration data memory unit 622 is exploration data acquired by exploring the interior of a test specimen of a healthy porous concrete layer 51 using an electromagnetic radar probe 7, the machine learning unit 632 performs machine learning using the exploration data for a simple porous concrete layer 51 as the test specimen as training data to generate a learning model, and the soundness diagnosis unit 633 can diagnose the soundness of the porous concrete layer 51 being diagnosed. This allows the soundness diagnosis unit 633 to diagnose the soundness of the porous concrete layer 51 in a simpler manner.

If the exploration data stored in the overall soundness exploration data memory unit 622 is exploration data acquired by using an electromagnetic radar probe 7 to explore the interior of the porous concrete layer 51 before it was installed and used as part of the under-drainage device 5, the machine learning unit 632 performs machine learning using the exploration data related to the porous concrete layer 51 actually installed as part of the under-drainage device 5 as training data to generate a learning model, and the soundness diagnosis unit 633 can diagnose the soundness of the porous concrete layer 51. This allows the soundness diagnosis unit 633 to diagnose the soundness of the porous concrete layer with greater accuracy.

If the exploration data stored in the overall health exploration data memory unit 622 is exploration data obtained by using an electromagnetic radar probe 7 to explore the interior of a porous concrete layer 51 installed in at least one activated carbon adsorption basin and sand filtration basin other than the activated carbon adsorption basin or sand filtration basin in which the porous concrete layer 51 to be diagnosed is installed and before use, the integrity diagnosis unit 633 can diagnose the integrity of the porous concrete layer 51 installed as part of the lower drainage device 5 even in an activated carbon adsorption basin or sand filtration basin after the porous concrete layer 51 has already been used.

The machine learning unit 632 performs machine learning using the exploration data acquired from the sound porous concrete layer and the porous concrete layer 51 having defects as training data, and generates a learning model regarding the presence or absence of defects in the porous concrete layer 51. The learning model memory unit 624 stores the learning model generated by the machine learning unit 632. The soundness diagnosis unit 633 uses the exploration data stored in the post-operation exploration data memory unit 625 as input values, and diagnoses the soundness of the porous concrete layer 51 installed as part of the underdrain device 5 using the learning model stored in the learning model memory unit 624, i.e., the learning model regarding the presence or absence of defects in the porous concrete layer 51. This enables the soundness of the porous concrete layer 51 to be diagnosed with greater accuracy.

If the exploration data stored in the defect exploration data memory unit 623 is exploration data for a specimen of a porous concrete layer 51 having a defect that is not installed as part of the under-drainage device 5, the machine learning unit 632 performs machine learning using the exploration data for the simple porous concrete layer 51 as the specimen as training data to generate a learning model, and the soundness diagnosis unit 633 can diagnose the soundness of the porous concrete layer 51 being diagnosed. This allows the soundness diagnosis unit 633 to diagnose the soundness of the porous concrete layer 51 in a simpler manner.

If the exploration data stored in the defect exploration data memory unit 623 is exploration data relating to a porous concrete layer 51 having a defect after it was installed and used in at least one activated carbon adsorption basin and/or sand filtration basin other than the activated carbon adsorption basin and/or sand filtration basin in which the porous concrete layer 51 to be diagnosed is installed, the soundness diagnosis unit 633 can diagnose the soundness of the porous concrete layer 51 to be diagnosed that is installed as part of the undercatchment device 5 by using the exploration data relating to the porous concrete layer 51 installed in at least one activated carbon adsorption basin and/or sand filtration basin other than the activated carbon adsorption basin and/or sand filtration basin in which the porous concrete layer 51 to be diagnosed is installed.

The above describes an embodiment of the present invention. However, the present invention is not limited to the above embodiment, and various modifications can be made without departing from the scope of the claims. The configurations of the above embodiment can be partially omitted or arbitrarily combined in a different manner than described above.

2: Activated carbon adsorption basin, 3: Water to be treated, 4: Filter material, 5: Under-water collection device, 6: Non-destructive diagnostic device, 7: Electromagnetic wave radar probe, 8: External monitor, 21: Pressure chamber, 22: Water collection culvert, 23: Cleaning air culvert, 24: Backwash water outflow culvert, 51: Porous concrete layer, 52: Dispersed gravel layer, 53: Slit plate, 54: Air dispersion beam, 61: Computer, 62: Memory unit, 63: Control unit, 64: Touch panel, 65: Communication unit, 621: Program, 622: Overall healthy body inspection data memory unit, 623: Defective body inspection data memory unit, 624: Learning model memory unit, 625: Post-operation inspection data memory unit, 631: Image processing unit, 632: Machine learning unit, 633: Soundness diagnosis unit

Claims (9)

A method for non-destructively diagnosing the soundness of a porous concrete layer of an under-drainage device installed in at least one of an activated carbon adsorption basin and a sand filtration basin, comprising:
a step of detecting the inside of the sound porous concrete layer using an electromagnetic wave radar probe and storing first detection data obtained;
performing machine learning using the stored first exploration data as training data to generate a learning model for the sound porous concrete layer;
storing the generated learning model;
a step of acquiring second exploration data by using the electromagnetic wave radar probe to explore the interior of the porous concrete layer after it has been installed and used as part of the underdrainage device;
a step of diagnosing the soundness of the porous concrete layer installed as part of the underdrain device using the acquired second exploration data as an input value and the stored learning model;
A non-destructive diagnostic method for a porous concrete layer, comprising: The non-destructive diagnosis method for porous concrete layers described in claim 1, characterized in that the first exploration data is exploration data obtained by exploring the interior of the specimen of the sound porous concrete layer using the electromagnetic wave radar exploration device. The non-destructive diagnosis method for a porous concrete layer described in claim 1, characterized in that the first exploration data is exploration data obtained by exploring the interior of the porous concrete layer using the electromagnetic radar exploration device before it is installed and used as part of the under-drainage device. The non-destructive diagnosis method for a porous concrete layer described in claim 1, characterized in that the first exploration data is exploration data obtained by using the electromagnetic radar probe to explore the interior of the porous concrete layer before it is installed and used as part of the under-drainage device installed in at least one of an activated carbon adsorption basin different from the activated carbon adsorption basin and a sand filtration basin different from the sand filtration basin. a step of storing third exploration data obtained by exploring the interior of the porous concrete layer having defects using the electromagnetic wave radar exploration device;
performing the machine learning using the stored first inspection data and the stored third inspection data as training data to generate a learning model regarding the presence or absence of the defect;
The non-destructive diagnosis method for a porous concrete layer according to claim 1, further comprising: The non-destructive diagnosis method for porous concrete layers described in claim 5, characterized in that the third inspection data is inspection data obtained by inspecting the interior of a specimen of the porous concrete layer having the defect using the electromagnetic wave radar inspection device. The non-destructive diagnosis method for porous concrete layers described in claim 5, characterized in that the third exploration data is exploration data obtained by using the electromagnetic radar probe to explore the interior of the porous concrete layer having the defect after it has been installed and used as part of the under-drainage device installed in at least one of an activated carbon adsorption basin different from the activated carbon adsorption basin and a sand filtration basin different from the sand filtration basin. A device for non-destructively diagnosing the soundness of a porous concrete layer of a lower drainage device installed in at least one of an activated carbon adsorption basin and a sand filtration basin,
a sound whole inspection data storage unit that stores first inspection data acquired by inspection using an electromagnetic wave radar inspection device, the first inspection data relating to the inside of the sound porous concrete layer;
a machine learning unit that performs machine learning using the first exploration data stored in the healthy whole exploration data storage unit as training data to generate a learning model for the healthy porous concrete layer;
a learning model storage unit that stores the learning model generated by the machine learning unit;
a post-operation exploration data storage unit that stores second exploration data acquired by exploration by the electromagnetic wave radar exploration device, the second exploration data relating to the inside of the porous concrete layer after it has been installed and used as part of the under-drainage device;
a soundness diagnosis unit that uses the second exploration data stored in the post-operation exploration data storage unit as an input value and diagnoses the soundness of the porous concrete layer installed as part of the underdrain device using the learning model stored in the learning model storage unit;
A non-destructive diagnostic device for porous concrete layers, comprising: A program executed by a computer of a device for non-destructively diagnosing the soundness of a porous concrete layer of an under-drainage device installed in at least one of an activated carbon adsorption basin and a sand filtration basin,
The computer,
a step of detecting the inside of the sound porous concrete layer using an electromagnetic wave radar probe and storing first detection data obtained;
performing machine learning using the stored first exploration data as training data to generate a learning model for the sound porous concrete layer;
storing the generated learning model;
a step of acquiring second exploration data by using the electromagnetic wave radar probe to explore the interior of the porous concrete layer after it has been installed and used as part of the underdrainage device;
a step of diagnosing the soundness of the porous concrete layer installed as part of the underdrain device using the acquired second exploration data as an input value and the stored learning model;
A non-destructive diagnostic program for porous concrete layers, characterized by executing the above.