JP2024058780A - Prediction device, prediction method, and computer program - Google Patents

Abstract

[Problem] To provide a technology that can output information that is easy to use in formulating plans to renew buried pipes such as water pipes and that makes it easy to explain the basis for formulating the plan. [Solution] An acquisition unit of a prediction device acquires attribute information that represents the attributes of the buried pipe to be predicted. The prediction unit calculates the predicted years of use of the buried pipe to be predicted using a years prediction model and the attribute information of the buried pipe to be predicted. The years prediction model is generated by machine learning relationship data between actual years of use information that represents the years of use of the buried pipe from installation to breakage and the attribute information of the buried pipe, and inputs the attribute information of the buried pipe to be predicted and outputs the predicted years of use of the buried pipe. The prediction unit further calculates the time when the buried pipe to be predicted is broken as the predicted breakage time using the predicted years of use and installation time information that represents the installation time included in the attribute information of the buried pipe to be predicted. [Selected Figure] Figure 9

Description

This invention is a technology that provides information on when to replace (update) buried pipes such as water pipes.

As one of our lifelines, waterworks are expected to provide a continuous supply of tap water. However, water pipes deteriorate over time, and damage to the pipes caused by aging can lead to accidents such as water leaks and bursting. Such accidents hinder the continuous and stable supply of tap water. For this reason, in order to prevent such accidents from occurring, construction work is being carried out to gradually replace (upgrade) old water pipes with new ones.

Patent Document 1 shows how to calculate the burial period from when a buried pipe is buried until the corrosion depth of the buried pipe is estimated to reach the allowable corrosion depth. Furthermore, Patent Document 1 shows how to calculate the remaining period until the allowable corrosion depth is reached by subtracting the actual burial period of the buried pipe from the calculated burial period. Patent Document 2 shows a technology for evaluating the risk of a water leakage accident in a pipeline in a water facility based on pipeline attributes. Patent Document 3 shows a technology for estimating the amount of leakage for each area used when planning a leakage investigation plan for multiple areas divided into a water pipe network.

JP 2021-56224 A JP 2021-140332 A JP 2015-94665 A

In Patent Documents 1 to 3, the depth of corrosion of buried pipes, the risk of water leakage accidents in the pipeline, and the amount of water leakage in an area are used as information (indicators) that indicate the deterioration state of buried water pipes. When planning water pipe renewal work, the timing of water pipe renewal is determined using such information that indicates the deterioration state of the water pipes. Even if numerical values that indicate the depth of corrosion of buried pipes, the risk of water leakage accidents, and the amount of water leakage as described above are presented as the basis for determining the renewal timing of water pipes determined in this way, there is a problem that, for example, it is difficult for a person formulating a renewal plan to understand the basis for when renewal should be performed.

The present invention was conceived to solve the above problems. That is, the main objective of the present invention is to provide a technology that can be easily used to formulate plans for updating buried pipes such as water pipes, and that can output information that makes it easy to explain the basis for formulating the plans.

In order to achieve the above object, a prediction device according to the present invention includes, as one aspect thereof,
An acquisition unit that acquires attribute information representing attributes of a buried pipe that is a prediction target;
a years prediction model that is generated by machine learning relationship data between actual years of use information, which indicates the years of use of a buried pipe from its installation until its failure, and attribute information of the buried pipe, and that receives attribute information of the buried pipe to be predicted as an input and outputs the years of use predicted from its installation until its failure as an predicted years of use, and a prediction unit that calculates the predicted years of use of the buried pipe to be predicted using the attribute information of the buried pipe to be predicted, and further calculates the time when the buried pipe to be predicted will be broken as an predicted failure time using the predicted years of use and installation time information, which indicates the time of installation, included in the attribute information of the buried pipe to be predicted;
and an output unit that outputs predicted breakage time information that expresses the predicted breakage time using a calendar year or an annual year.

In addition, one aspect of the prediction method according to the present invention is to
By computer,
Acquire attribute information representing attributes of the buried pipe to be predicted;
a predicted service life of the buried pipe to be predicted is calculated using a service life prediction model generated by machine learning relationship data between actual service life information, which indicates the service life of a buried pipe from its installation until its failure, and attribute information of the buried pipe, the model taking the attribute information of the buried pipe to be predicted as an input and outputting the service life predicted from the installation of the buried pipe to its failure as a predicted service life, and the attribute information of the buried pipe to be predicted; and further using the predicted service life and installation time information, which indicates the installation time included in the attribute information of the buried pipe to be predicted, the time when the buried pipe to be predicted is predicted to fail is calculated as a predicted failure time;
The predicted breakage time information is output, which indicates the predicted breakage time using a calendar year or an annual year.

Furthermore, one aspect of the computer program according to the present invention is
A process of acquiring attribute information representing attributes of a buried pipe to be predicted;
a process of calculating a predicted service life of the buried pipe to be predicted using the attribute information of the buried pipe to be predicted and a service life prediction model generated by machine learning relationship data between actual service life information indicating the service life of the buried pipe from its installation until its failure and attribute information of the buried pipe, the model taking the attribute information of the buried pipe to be predicted as an input and outputting the service life predicted from the installation of the buried pipe to its failure as a predicted service life, and further calculating the time when the buried pipe to be predicted is to be failed as a predicted failure time using the predicted service life and installation time information indicating the installation time included in the attribute information of the buried pipe to be predicted;
and outputting predicted breakage time information that expresses the predicted breakage time using a calendar year or an annual year.

The present invention makes it possible to output information that is easy to use in formulating plans to renew buried pipes such as water pipes, and that makes it easy to explain the basis for formulating the plans.

FIG. 1 is a diagram illustrating a configuration of a prediction device according to a first embodiment of the present invention. 11 is a diagram illustrating an example of a manner in which predicted breakage time information is displayed on a display device. FIG. 13 is a diagram illustrating an example of a mode in which an explanation of the basis for the calculated predicted years of use is displayed on a display device. FIG. 11 is a diagram illustrating an example of a manner in which information regarding the calculated predicted years of use is displayed on a display device. FIG. 11 is a diagram illustrating an example of an aspect in which predicted failure time information is superimposed on a map and displayed on a display device. FIG. 4 is a flowchart illustrating an example of an operation of the prediction device according to the first embodiment. FIG. 11 is a diagram illustrating a second embodiment. FIG. 13 is a diagram illustrating another embodiment. FIG. 13 is a block diagram illustrating a configuration of a prediction device according to another embodiment of the present invention. 13 is a flowchart illustrating an example of an operation of a prediction device according to another embodiment.

Below, an embodiment of the present invention will be described with reference to the drawings.

First Embodiment
FIG. 1 is a diagram for explaining the configuration of a prediction device according to a first embodiment of the present invention. This prediction device 1 is a device for predicting when a buried pipe will break. A buried pipe is a pipe buried underground. There are multiple types of buried pipes, such as water pipes, sewer pipes, and gas pipes. Here, the configuration of the prediction device will be explained using a water pipe, which is one type of buried pipe, as an example. In addition, "a buried pipe (water pipe) breaks" here means that the water pipe has been damaged, such as cracked or perforated, due to deterioration of the water pipe.

The prediction device 1 is a computer device, and is connected to an information source 3, a database 4, an input device 5, and a display device 7. The prediction device 1 can also be connected to a printer 8 as necessary. The input device 5 is composed of a keyboard, a mouse, etc., and is a device that inputs information to the prediction device 1 in response to user operations. The display device 7 is a device that displays information on a screen using characters and images. The printer 8 is a device that receives data such as characters, images, and figures from a computer and prints it on paper, etc.

The information source 3 is an information source that can provide information used in the processing executed by the prediction device 1. Specific examples of the information source 3 include a computer device or a database of a system that manages water pipes. Information related to the water pipe to be predicted is provided to the prediction device 1 from such information source 3. The database 4 is a storage device, and stores information provided to the prediction device 1 from the information source 3, information input to the prediction device 1 by the input device 5, and the like.

The prediction device 1 includes a calculation device 10 and a storage device 40. The storage device 40 includes a storage medium for storing data and a computer program (hereinafter also referred to as a program) 41. There are multiple types of storage devices, such as magnetic disk devices and semiconductor memory elements, and there are multiple types of semiconductor memory elements, such as RAM (Random Access Memory) and ROM (Read Only Memory). The type of storage device 40 included in the prediction device 1 is not limited to one. A computer device is often equipped with multiple types of storage devices. Here, the type and number of storage devices 40 included in the prediction device 1 are not limited, and a description thereof will be omitted. Furthermore, when the prediction device 1 is equipped with multiple types of storage devices 40, they will be collectively referred to as storage devices 40.

The arithmetic device 10 is composed of a processor such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit). The arithmetic device 10 can have various functions based on a program 41 stored in a storage device 40 by reading and executing the program 41. Here, the arithmetic device 10 has an acquisition unit 11, a prediction unit 12, and an output unit 13 as functional units related to a breakage time prediction process that predicts when a water pipe will break.

The acquisition unit 11 acquires attribute information that represents the attributes of the water pipe to be predicted. The attribute information of the water pipe acquired here is used to predict the number of years of use of the water pipe. Here, the number of years of use of the water pipe is the number of years from when the water pipe is buried until the water pipe breaks. The main cause of water pipe breakage is deterioration. In other words, deterioration causes damage such as cracks and holes in the water pipe. The number of years of use of the water pipe is related to the rate at which deterioration progresses in the water pipe. For this reason, the attribute information of the water pipe acquired by the acquisition unit 11 includes information that is assumed to be related to the rate at which deterioration progresses in the water pipe (hereinafter, also referred to as deterioration-related information). In other words, deterioration-related information is information that represents factors related to the deterioration of the water pipe.

For example, the attribute information of the water pipe acquired by the acquisition unit 11 includes information on the installation location and information on the installation time, and further includes information on the size of the pipe, information on the type of pipe, information on the use, and information on the soil in which the pipe is buried as deterioration-related information. The installation location information includes the address and the length of the pipe, which indicate the location where the water pipe is buried. For example, the address information included in the installation location information is address information such as 1-3-1, XXX-cho, △△-ku. The pipe length information included in the installation location information is pipe length information such as 300 meters of pipe distance. Note that the address of the water pipe to be predicted is an address based on a predetermined rule, for example, an address corresponding to a position half the length of the water pipe to be predicted. The installation time information is information indicating the time when the water pipe was buried, and is, for example, information expressed by a calendar year (for example, the Gregorian calendar) or a fiscal year.

Pipe size information is information that indicates the size of a water pipe, and includes, for example, information that indirectly or directly indicates the diameter and pipe thickness (wall thickness). Pipe type information is type information that classifies water pipes according to the material of which they are made. Specific examples of water pipe types include ductile cast iron pipes, stainless steel pipes, and polyethylene pipes. Use information is information that indicates the use of a water pipe, and is information related to the water pressure, water volume, and water flow of the tap water flowing inside the pipe. Information on the buried soil is, for example, information such as sand, silt, and clay. In addition, soil information may include information that indicates soil quality, such as moisture content and acidity.

Furthermore, the attribute information may include information representing the quality of tap water flowing through the water pipe (for example, the content of components such as chlorine) as deterioration-related information. Furthermore, the attribute information may include information on factors that may affect the water pipe (for example, information on the traffic volume of the road if the water pipe is buried in a road, and information on the water pressure, water volume, and water flow of the tap water flowing through the water pipe) as deterioration-related information. Since information representing the state of tap water, such as the water quality, water pressure, and water volume of tap water flowing through the water pipe, may fluctuate due to some cause, for example, the water quality, water pressure, and water volume may be measured continuously or intermittently (for example, once a day), and the measurement data may be stored. Then, for example, an average value for the most recent predetermined period (for example, three months) is calculated from the measurement data. This average value may be included in the attribute information as information representing the state of the fluid (tap water) flowing inside the water pipe during a predetermined period. In other words, the attribute information of the water pipe to be predicted may include information that describes the state of the tap water flowing inside the water pipe during a predetermined period of time.

The attribute information of water pipes as described above is stored as electronic data, for example, in a database of a water pipe management system established in a water utility or in a database connected to a personal computer used by staff of the water utility. The prediction device 1 reads out water pipe attribute information from such a database (i.e., information source 3) used by the water utility in response to, for example, a user's operation of the input device 5 or at predetermined time intervals, and stores the water pipe attribute information in the database 4. In the database 4, the water pipe attribute information is associated with pipe identification information that identifies the water pipe.

When a user operates the input device 5 to request the start of a breakage time prediction process and pipe identification information corresponding to the water pipe to be predicted is input to the prediction device 1, the acquisition unit 11 acquires attribute information of the water pipe to be predicted from the database 4 using the pipe identification information. The attribute information acquired by the acquisition unit 11 is attribute information that is predetermined as the information to be acquired from among the multiple types of attribute information as described above, and here includes at least one piece of deterioration-related information and installation time information.

For example, it is conceivable that the attribute information of water pipes held by each water utility may differ. In such a case, the type of attribute information of water pipes that the acquisition unit 11 acquires is determined for each water utility. Then, the acquisition unit 11 acquires the type of attribute information according to the water utility that manages the water pipe to be predicted. In such a case, for example, the pipe identification information that identifies the water pipe includes information that represents the managing water utility or information that represents the city, town, or village that operates the water utility. Furthermore, when there are multiple water pipes to be predicted and multiple pieces of pipe identification information corresponding to each of the water pipes to be predicted are input, the acquisition unit 11 acquires the attribute information of the water pipes to be predicted together from the database 4.

The prediction unit 12 predicts when the water pipe to be predicted will break. That is, the prediction unit 12 first calculates the predicted years of use of the water pipe to be predicted. The predicted years of use is the predicted number of years from when the water pipe is buried until the water pipe breaks. The prediction unit 12 uses AI (Artificial Intelligence) technology to calculate the predicted years of use. That is, the prediction unit 12 calculates the predicted years of use of the water pipe to be predicted using a years prediction model generated by AI technology. The years prediction model is a model that inputs attribute information of the water pipe to be predicted and outputs the predicted years of use of the water pipe to be predicted, and is generated by machine learning the following teacher data. The teacher data used to generate the years prediction model is data in which the actual years of use of the water pipe and the attribute information of the water pipe are associated (relationship data between the actual years of use and the attribute information of the water pipe). The actual years of use of the water pipe is the number of years from when the water pipe is laid until the water pipe actually breaks. The attribute information associated with this actual years of use is attribute information of the actually damaged water pipe, and includes at least one piece of deterioration-related information. The attribute information of the water pipe to be predicted, which is the input of the age prediction model, is attribute information of the type included in the attribute information of the training data used in the machine learning of the age prediction model, and includes installation location information, installation time information, and at least one piece of deterioration-related information.

For example, the type of attribute information of water pipes held by different water utilities may differ. In this case, the type of attribute information that can be used to generate (learn) the age prediction model may differ depending on the water utility. In such a case, for example, multiple types of age prediction models are generated for each water utility (in other words, for each area). In other words, multiple types of age prediction models are prepared according to the difference in the type of attribute information of water pipes used to generate the age prediction model. In this way, multiple types of age prediction models that the prediction unit 12 uses to calculate the predicted years of use may be prepared. In this case, the prediction unit 12 selects, from the multiple types of age prediction models, a age prediction model according to the type of information included in the attribute information of the water pipe to be predicted as the age prediction model to be used to calculate the predicted years of use. Then, the prediction unit 12 calculates the predicted years of use of the water pipe to be predicted using the selected age prediction model and the attribute information of the water pipe to be predicted.

There are several methods for generating a year prediction model. As described above, when a year prediction model is prepared for each water utility, a year prediction model generated by the same generation method may be prepared for each water utility. However, instead of preparing a year prediction model generated by the same generation method for each water utility, a year prediction model generated by a generation method according to the water utility may be prepared as follows. That is, first, multiple year prediction models using different generation methods are generated for each water utility, and the predicted years of use are actually calculated using each of the multiple generated year prediction models. In this calculation process, attribute information of the water pipe for which information on the actual years of use is obtained is input to the year prediction model. Then, the output result output from the year prediction model is compared with the actual years of use, and the year prediction model generated by the generation method determined to be suitable for the water utility based on the comparison result is used as the year prediction model corresponding to the water utility.

One type of AI technology is explainable AI (Artificial Intelligence). Explainable AI is an AI technology that can output not only the results of calculations from a model, but also an explanation of the basis for calculating the results. Here, the years prediction model is an explainable model generated using the explainable AI technology. As a result, the years prediction model not only outputs the predicted years of use by inputting attribute information of the water pipe to be predicted, but also outputs information explaining the basis for outputting the predicted years of use. Note that there are multiple types of explainable AI technology, and the type of explainable AI technology adopted here is not limited, so its explanation is omitted.

Furthermore, the prediction unit 12 calculates the time when the water pipe to be predicted is broken (herein also referred to as the predicted breakage time) using the predicted years of use output from the years prediction model and the installation time information included in the attribute information for the water pipe to be predicted. That is, the prediction unit 12 calculates the predicted breakage time by adding the predicted years of use to the installation time (laying year) included in the installation time information. Specifically, for example, if the installation time of the water pipe to be predicted is 1975 and the predicted years of use is 52 years, the prediction unit 12 calculates that the predicted breakage time is 2027 by adding 52 years to 1975. Note that, if the installation time of the installation time information included in the attribute information is expressed by fiscal year instead of calendar year, the predicted breakage time calculated by the prediction unit 12 will also be expressed by fiscal year.

In this way, the prediction unit 12 calculates the predicted time of failure for the water pipe to be predicted, in a manner that expresses the predicted time of failure in terms of a calendar year or fiscal year. The calculated information on the predicted time of failure is stored in the database 4 in a state in which the pipe identification information, attribute information, predicted years of use information, and information explaining the basis for calculating the predicted years of use are associated with the corresponding water pipe.

The output unit 13 outputs predicted breakage time information representing the predicted breakage time calculated by the prediction unit 12. For example, when a display request for predicted breakage time information and information specifying the water pipe for which the user wants to display the predicted breakage time information (i.e., the water pipe to be displayed) are input to the prediction device 1 by the user's operation of the input device 5, the output unit 13 causes the display device 7 to display the predicted breakage time information of the water pipe to be displayed. That is, the output unit 13 has a function of controlling the display operation of the display device 7. For example, a display template used to display the predicted breakage time information is given to the prediction device 1, and the output unit 13 uses the display template to display the predicted breakage time information of the water pipe to be displayed on the screen of the display device 7. The form of the display template is not limited, but an example is shown in FIG. 2. In the example of FIG. 2, the display template is in the form of a table, and in the table, the predicted breakage time information of the water pipe to be displayed is associated with the attribute information of the water pipe and the information of the predicted service life. In the first embodiment, as described above, the years prediction model used to calculate the predicted service life is an explainable model. For this reason, an explanation of the basis for the calculated predicted years of use is output from the years prediction model. This explanation of the basis is stored in database 4 in association with the information on the predicted years of use. When the output unit 13 displays the information on the predicted years of use on the display device 7, the explanation of the basis may also be output to the display device 7 and displayed on the display device 7. For example, the explanation of the basis may be incorporated into a table as shown in FIG. 2 and displayed as a pop-up as shown in FIG. 3.

When the output unit 13 displays information including predicted breakage time information for multiple water pipes in a table such as that shown in FIG. 2, the output unit 13 may have a function of sorting the information in response to a request input by a user operating the input device 5. In other words, the output unit 13 may display information including predicted breakage time information for multiple water pipes to be predicted in order of predicted breakage time, in order of oldest installed year, or grouped together by predefined region in response to a user request.

The output unit 13 may also display on the display device 7 an explanation of the basis of the predicted breakage time information for the water pipe in a display format as shown in FIG. 4. In the example of FIG. 4, the explanation of the basis output from the age prediction model for the water pipe for which an explanation was requested (in the example of FIG. 4, the water pipe with pipe identification information: Ab-fmw) is displayed using a table or graph. In other words, the display device 7 displays information on the water pipe's feature values used by the age prediction model to calculate the predicted years of use of the water pipe in the form of a table, and also displays the degree of involvement (importance) of the feature values that affect the predicted years of use of the target water pipe (for example, the water pipe with pipe identification information: Ab-fmw). Furthermore, in the example of FIG. 4, the display device 7 displays a sentence explaining the basis based on the items shown in the table or graph.

As another example of the display mode in which the output unit 13 displays the predicted breakage time information on the display device 7, there is a display mode in which the predicted breakage time information is superimposed on a map 30 as shown in FIG. 5. In the example of FIG. 5, the map 30 has color or pattern information representing the predicted breakage time of the water pipe superimposed on the road portion in which the water pipe to be displayed is buried. In other words, if the predicted breakage time of the water pipe is within the period up to 2025, the information representing the predicted breakage time is determined to be the color or hatching of pattern A, and if the predicted breakage time of the water pipe is within the period from 2026 to 2030, the information representing the predicted breakage time is determined to be the color or hatching of pattern B. In the map 30, the information of the predicted breakage time of the water pipe represented by a predetermined color or hatching is superimposed on the road portion in which the water pipe to be displayed is buried. As shown in FIG. 5, when information on the predicted breakage time of a water pipe is superimposed on the map 30, information indicating the area in which the water pipe to be displayed (the water pipe to be displayed) is buried is input to the output unit 13 by the user operating the input device 5. This allows the output unit 13 to display on the display device 7 the map 30 including the area in which the water pipe to be displayed is buried.

When a user operates the input device 5 to input an instruction to the prediction device 1 to print predicted damage time information, the output unit 13 causes the printer 8 to print the predicted damage time information for the requested item to be printed.

The prediction device 1 of the first embodiment is configured as described above. Next, an example of the operation of the damage time prediction process in the prediction device 1 will be described with reference to FIG. 6. FIG. 6 is a flowchart explaining an example of the operation of the damage time prediction process in the prediction device 1.

For example, when a request to start a breakage time prediction process is input to the prediction device 1 by a user operating the input device 5, the acquisition unit 11 receives the request to start the process (step 101 in FIG. 6). This request to start the process is associated with pipe identification information corresponding to the water pipe to be predicted. The acquisition unit 11 uses the pipe identification information to acquire attribute information of the water pipe to be predicted from the database 4 (step 102).

Then, the prediction unit 12 calculates the predicted years of use of the water pipe to be predicted (step 103). In the first embodiment, a years prediction model using AI technology is used to calculate the predicted years of use. In other words, the prediction unit 12 inputs attribute information of the water pipe to be predicted into the years prediction model, and calculates the predicted years of use by a calculation operation using the years prediction model.

Then, the prediction unit 12 calculates the predicted failure time expressed in terms of a calendar year or fiscal year by adding the installation date information included in the attribute information and the predicted years of use for the water pipe to be predicted (step 104). The prediction unit 12 stores the information on the predicted failure time calculated in this way in the database 4. The information on the predicted failure time stored in the database 4 is associated with the pipe identification information, attribute information, information on the predicted years of use, and information explaining the basis for calculating the predicted years of use for the corresponding water pipe.

The failure time prediction process is executed in this manner. When a request to display the predicted failure time information calculated by this process and information identifying the water pipe to be displayed are input to the prediction device 1 by the user operating the input device 5, the output unit 13 causes the display device 7 to display the predicted failure time information for the water pipe to be displayed.

As described above, the prediction device 1 of the first embodiment calculates the predicted years of use using a years prediction model based on AI technology. The years prediction model is generated by machine learning relationship data between information on the actual years of use of the water pipe and attribute information of the water pipe (attribute information including information on the installation location, information on the installation time, and at least one piece of deterioration-related information). In other words, in the first embodiment, in order to calculate the predicted years of use, attention is paid to the relationship between the actual years of use of the water pipe and the attribute information, rather than the damage to the water pipe itself. As a result, in the first embodiment, the years prediction model can output the predicted years of use (the number of years from when the water pipe is buried until it is predicted to break) in response to input of the pipe identification information and attribute information of the water pipe to be predicted.

Furthermore, the prediction unit 12 calculates information on the predicted breakage time (the time when the predicted water pipe is predicted to break) expressed by the calendar year or fiscal year by adding the predicted years of use to the laying year (calendar year) or laying year that indicates the laying time of the water pipe to be predicted. This calculated information on the predicted breakage time is output by the output unit 13 as predicted breakage time information. Since this predicted breakage time information is information expressed by the calendar year or fiscal year, it can be used as is, for example, when determining the timing of water pipe renewal work. In other words, the predicted breakage time information is information that is easy to use in formulating a water pipe renewal plan. Furthermore, since the predicted breakage time information is information expressed by the calendar year or fiscal year, by presenting the predicted breakage time information, it is possible to clearly communicate the time when the water pipe must be renewed. In other words, the predicted breakage time information is information that makes it easy to explain the renewal time of the water pipe in the water pipe renewal plan formulated using the information.

Furthermore, in the first embodiment, the age prediction model that calculates the predicted years of use used to calculate the predicted breakage time is an explainable model, and therefore an explanation of the basis for the calculated predicted years of use is output from the age prediction model. When a water pipe renewal plan is formulated using the predicted breakage time information that is the output of the prediction device 1, the explanation output from the age prediction model is used to explain the renewal plan, thereby making the renewal plan more persuasive.

When generating the age prediction model by machine learning, information on the causal relationship between the actual age information of the water pipe obtained by causal analysis technology and the attribute information of the water pipe may also be machine learned. In this way, the explanation of the basis for the predicted age output from the age prediction model, which is an explainable model, includes an explanation of the causal relationship between the predicted age and the attribute information of the buried pipe, which can further increase the persuasiveness (reliability) of the calculated predicted age and predicted time of failure.

The prediction device 1 may also be connected to a terminal device 6 as shown by the dotted line in FIG. 1. The terminal device 6 is an information device equipped with a communication function and a display function, and here, as a specific example, a portable information device such as a smartphone or a tablet terminal may be mentioned. When the prediction device 1 is connected to the terminal device 6, the output unit 13 may have the following functions. That is, when the output unit 13 receives a request for predicted breakage time information and information specifying the water pipe that is the subject of the request from the terminal device 6, it returns (outputs) predicted breakage time information of the requested water pipe in response to the request to the terminal device 6. In the first embodiment, the terminal device 6 (or user) that is permitted to receive predicted breakage time information of the water pipe is registered in advance in the storage device 40 or the like. The prediction device 1 performs authentication processing of the terminal device 6 (or user) using this registration information and a predetermined authentication method. The output unit 13 outputs predicted breakage time information to the terminal device 6 (or user) that has been authenticated by the authentication processing.

Furthermore, the prediction device 1 may be connected to another computer device (e.g., a server) 9 as shown by the dotted line in FIG. 1. When the prediction device 1 is connected to the computer device 9, the output unit 13 may further have the following functions. That is, when the output unit 13 receives a request for predicted breakage time information and information specifying the requested water pipe from the computer device 9, it returns (outputs) the predicted breakage time information of the requested water pipe in response to the request to the computer device 9. For example, the computer device 9 uses the received predicted breakage time information to generate a water pipe renewal plan including an estimate of the future pipe renewal work volume and work costs. Note that the computer device 9, which is the output destination to which the predicted breakage time information is output, is, for example, a computer device authenticated by authentication processing, similar to the terminal device 6.

Second Embodiment
A second embodiment of the present invention will be described below. In the description of the second embodiment, components having the same functions as those in the description of the first embodiment are denoted by the same reference numerals, and descriptions of the overlapping parts will be omitted.

In the second embodiment, the acquisition unit 11, the prediction unit 12, and the output unit 13 in the prediction device 1 described in the first embodiment are incorporated into the plan development device 20 as shown in FIG. 7. That is, the plan development device 20 is a computer device equipped with a function of developing a water pipe renewal plan, and includes a calculation device 25 composed of a processor and a storage device 26. The storage device 26 includes a storage medium for storing data and a program 27. The calculation device 25 can have various functions based on the program 27 by executing the program 27. As part of the function of the calculation device 25, the plan development device 20 includes the acquisition unit 11, the prediction unit 12, and the output unit 13 related to the breakage time prediction process. That is, the plan development device 20 is a device that also functions as a prediction device that predicts the time when a buried pipe will break (predicted breakage time). The functions (configurations) of the acquisition unit 11, the prediction unit 12, and the output unit 13 in the second embodiment are similar to the functions (configurations) of the acquisition unit 11, the prediction unit 12, and the output unit 13 in the first embodiment.

The computing device 25 further includes a planning unit 28 as a functional unit. The planning unit 28 generates a plan (renewal plan) for water pipe renewal work using the predicted breakage time calculated by the prediction unit 12. One example of a method in which the planning unit 28 generates a renewal plan is to use AI technology. There are various types of AI technology for generating renewal plans, and the type of AI technology employed by the planning unit 28 is not limited, and a description thereof will be omitted.

The planning unit 28 stores the generated water pipe renewal plan in the database 4. When a user operates the input device 5 to request display of the display device 7, the water pipe renewal plan stored in the database 4 is displayed on the display device 7 by the plan development device 20. The display mode on the display device 7 is not limited, and a description thereof will be omitted.

The planning device 20 (in other words, the prediction device) in the second embodiment has the same functions as the prediction device 1 in the first embodiment, and can therefore achieve the same effects as the first embodiment. In addition, in the planning device 20, the planning unit 28 generates a water pipe renewal plan using the predicted breakage time calculated by the prediction unit 12. The predicted breakage time is information expressed in terms of a calendar year or fiscal year. As a result, for example, the planning unit 28 of the planning device 20 can reduce the load of converting the water pipe renewal time into a calendar year or fiscal year in the process of determining the water pipe renewal time, compared to when a plan is generated using information on the corrosion level of the water pipe.

The planning unit 28 may generate a plan for water pipe renewal work by considering not only the predicted breakage time by the prediction unit 12, but also information on the impact and earthquake resistance of the water pipe. The impact is information that indicates the degree of impact in the event of a major accident such as a large amount of water leakage from the water pipe or a water pipe bursting. This impact is calculated by the planning unit 28 using, for example, use information and flow rate information included in the attribute information of the water pipe. The earthquake resistance is information on the earthquake resistance of the water pipe, and is determined, for example, by the material of the water pipe and the type of ground in which it is buried. The earthquake resistance of the water pipe is obtained, for example, from an information source 3 (for example, a computer device or database of a system that manages the water pipe) as one of the attribute information of the water pipe.

<Other embodiments>
The present invention is not limited to the first and second embodiments, and may be embodied in various ways. For example, in the first and second embodiments, an example is shown in which a plurality of year prediction models are prepared for each water utility as the year prediction model used by the prediction unit 12. Alternatively, for example, a common year prediction model independent of the water utility may be generated by using a type of common attribute information that can be acquired from any water utility to generate the year prediction model. In this case, the prediction unit 12 inputs a common type of attribute information into the common year prediction model, and calculates the predicted years of use using the year prediction model.

In the first and second embodiments, the years prediction model used by the prediction unit 12 is a model that is generated using explainable AI technology and outputs an explanation of the basis for the predicted years of use. Alternatively, the years prediction model does not have to be an explainable model. In this case, an explanation of the basis for the calculated predicted years of use is not output from the years prediction model, but the following information may be presented as reference information. The reference information includes, for example, an explanation of matters that are known in advance about the impact that elements represented in the attribute information of the water pipe (for example, the soil quality of the soil in which it is buried) have on the years of use of the water pipe.

Furthermore, in the first and second embodiments, the configuration of the prediction device is described using a buried water pipe as an example, but a prediction device similar to the first and second embodiments can also be applied to the calculation of the predicted breakage time for other types of buried pipes, such as sewer pipes and gas pipes.

Furthermore, the prediction device 1 of the first embodiment may include an analysis unit 15 as shown in FIG. 8. Note that the prediction device 1 shown in FIG. 8 also includes a storage device 40, and the calculation device 10 includes an acquisition unit 11, a prediction unit 12, and an output unit 13, but these are omitted from FIG. 8.

The analysis unit 15 has a function of analyzing the deterioration of a plurality of water pipes, for example, by region, using the calculation results calculated by the prediction unit 12 and attribute information of the water pipes. The analysis results are stored in the database 4 in association with the information used in the analysis. Furthermore, the analysis results by the analysis unit 15 can be displayed on the display device 7 by the function of the output unit 13. An example of this display is shown in FIG. 8. Furthermore, the analysis unit 15 may be provided in the planning device 20.

Figure 9 is a block diagram showing the configuration of a prediction device according to another embodiment of the present invention. This prediction device 50 is, for example, a computer device, and can have the following functional units by executing a computer program. That is, the prediction device 50 includes an acquisition unit 51, a prediction unit 52, and an output unit 53.

The acquisition unit 51 acquires attribute information representing the attributes of the buried pipe to be predicted. The prediction unit 52 calculates the predicted years of use of the buried pipe to be predicted using the years prediction model and the attribute information of the buried pipe to be predicted. The predicted years of use is the predicted years of use of the buried pipe from its installation to its failure. The years prediction model is generated by machine learning the relationship data between the actual years of use information representing the years of use of the buried pipe from its installation to its failure and the attribute information of the buried pipe, and inputs the attribute information of the buried pipe to be predicted and outputs the predicted years of use of the buried pipe. The prediction unit 52 further calculates the time when the buried pipe to be predicted is to be broken as the predicted failure time using the predicted years of use and the laying time information representing the laying time included in the attribute information of the buried pipe to be predicted. The output unit 53 outputs predicted failure time information representing the predicted failure time using a calendar year or a fiscal year.

Next, an example of the operation of the prediction device 50 will be described with reference to FIG. 10. FIG. 10 is a flowchart illustrating an example of the operation of the prediction device 50.

For example, when the acquisition unit 51 acquires attribute information that represents the attributes of the buried pipe to be predicted (step 201), the prediction unit 52 calculates the predicted years of use using the attribute information and the years prediction model (step 202). Furthermore, the prediction unit 52 calculates the predicted time of failure using the calculated predicted years of use and installation time information that represents the installation time included in the attribute information of the buried pipe to be predicted (step 203). Thereafter, for example, when output of the predicted time of failure is requested, the output unit 53 outputs the predicted time of failure information (step 204). The predicted time of failure information is information that represents the predicted time of failure using a calendar year or fiscal year.

As described above, the prediction device 50 calculates the predicted years of use using a years prediction model generated by machine learning the relationship data between actual years of use information and attribute information of the buried pipe. Furthermore, the prediction device 50 outputs the predicted time of failure using the calculated predicted years of use and installation time information, and further outputs predicted failure time information that expresses the predicted time of failure using a calendar year or fiscal year. Since this predicted failure time information is expressed using a calendar year or fiscal year, it is easy to use in formulating a buried pipe renewal plan, and the basis for formulating the plan is easy to understand. In other words, the prediction device 50 can output information that is easy to use in formulating a buried pipe renewal plan, and that makes it easy to explain the basis for formulating the plan.

1, 50 Prediction device 11, 51 Acquisition unit 12, 52 Prediction unit 13, 53 Output unit

Claims (8)

An acquisition unit that acquires attribute information representing attributes of a buried pipe that is a prediction target;
a years prediction model that is generated by machine learning relationship data between actual years of use information, which indicates the years of use of a buried pipe from its installation until its failure, and attribute information of the buried pipe, and that receives attribute information of the buried pipe to be predicted as an input and outputs the years of use predicted from its installation until its failure as an predicted years of use, and a prediction unit that calculates the predicted years of use of the buried pipe to be predicted using the attribute information of the buried pipe to be predicted, and further calculates the time when the buried pipe to be predicted will be damaged as an predicted failure time using the predicted years of use and installation time information, which indicates the time of installation, included in the attribute information of the buried pipe to be predicted;
and an output unit that outputs predicted failure time information that expresses the predicted failure time using a calendar year or fiscal year. The age prediction model is an explainable model that also outputs an explanation of the basis for the calculated predicted age of use,
The prediction device according to claim 1 , wherein the output unit outputs information including an explanation of a basis for the output from the age prediction model, in addition to the predicted failure time information. There are multiple types of information that the acquisition unit can acquire as attribute information of the buried pipe,
A plurality of types of age prediction models are prepared based on different types of buried pipe attribute information used to generate the age prediction models;
The prediction device of claim 1, wherein the prediction unit selects from the multiple types of age prediction models a age prediction model that corresponds to the type of attribute information of the buried pipe to be predicted acquired by the acquisition unit, and calculates a predicted number of years of use of the buried pipe to be predicted using the selected age prediction model and the attribute information of the buried pipe to be predicted acquired by the acquisition unit. The attribute information of the buried pipe to be predicted acquired by the acquisition unit includes at least deterioration-related information representing factors related to deterioration of the buried pipe,
The age prediction model is generated by machine learning relationship data between actual service age information of the buried pipe and attribute information including deterioration-related information of the buried pipe,
The prediction device according to claim 1 , wherein the prediction unit calculates a predicted number of years of use of the buried pipe to be predicted using attribute information including deterioration-related information acquired by the acquisition unit and the number of years prediction model. The prediction device according to claim 1 , wherein the attribute information of the buried pipe to be predicted acquired by the acquisition unit includes information representing a state of the fluid flowing inside the buried pipe during a predetermined period of time. The prediction device of claim 2, wherein the age prediction model, which is an explainable model, is generated by machine learning information about the causal relationship obtained by causal analysis technology between the actual age information of buried pipes and the attribute information of the buried pipes, and the explanation of the basis for the predicted age of use output from the age prediction model includes an explanation of the causal relationship between the predicted age of use and the attribute information of the buried pipes. By computer,
Acquire attribute information representing attributes of the buried pipe to be predicted;
a predicted service life of the buried pipe to be predicted is calculated using a service life prediction model generated by machine learning relationship data between actual service life information, which indicates the service life of a buried pipe from its installation until its failure, and attribute information of the buried pipe, the model taking the attribute information of the buried pipe to be predicted as an input and outputting the service life predicted for the buried pipe from its installation until its failure as a predicted service life, and the attribute information of the buried pipe to be predicted; and further using the predicted service life and installation time information, which indicates the installation time included in the attribute information of the buried pipe to be predicted, the time when the buried pipe to be predicted is predicted to fail is calculated as a predicted failure time;
A prediction method that outputs predicted failure time information that expresses the predicted failure time using a calendar year or fiscal year. A process of acquiring attribute information representing attributes of a buried pipe to be predicted;
a process of calculating a predicted service life of the buried pipe to be predicted using the attribute information of the buried pipe to be predicted and a service life prediction model which is generated by machine learning relationship data between actual service life information indicating the service life of the buried pipe from its installation until its failure and attribute information of the buried pipe, the model taking the attribute information of the buried pipe to be predicted as an input and outputting the service life predicted for the buried pipe from its installation until its failure as a predicted service life, and further calculating the time when the buried pipe to be predicted is to be broken as a predicted failure time using the predicted service life and installation time information indicating the installation time included in the attribute information of the buried pipe to be predicted;
and outputting predicted breakage time information that expresses the predicted breakage time using a calendar year or fiscal year.

Images