Ai pipeline deterioration prediction system, ai pipeline deterioration prediction method, and ai pipeline deterioration prediction program
The AI pipe deterioration prediction system uses a mathematical model trained with event and attribute data to accurately forecast pipeline deterioration, addressing the limitations of existing systems by predicting specific causes and handling varied data availability.
Patent Information
- Application Number
- JP2023184087
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-10-26
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2043-04-28
AI Technical Summary
Existing systems struggle to accurately predict the deterioration of pipelines, particularly in identifying specific causes of damage and handling pipelines with limited data, without providing comprehensive models for individual pipe deterioration prediction.
An AI pipe deterioration prediction system that utilizes a mathematical model trained with event information and pipeline attribute data, including classification and integration of data from multiple sources, to predict deterioration based on event types and regional characteristics, accommodating pipelines with and without event history.
The system enhances the accuracy of pipeline deterioration prediction by considering event types, regional characteristics, and integrating data from various sources, enabling precise forecasting even for pipelines with limited data and identifying abnormal deterioration.
Smart Images

Figure 2024159427000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to an AI pipeline deterioration prediction system, an AI pipeline deterioration prediction method, and an AI pipeline deterioration prediction program. [Background technology]
[0002] Companies that manage pipelines that supply fluids, such as water and gas pipes, need to replace the pipelines when they reach the end of their service life or when they are damaged. However, there are a huge number of pipelines that each company must manage, making it very difficult to identify which pipelines need to be replaced.
[0003] Patent Document 1 discloses a system that predicts the likelihood of future pipe damage by using historical data on damage history and a damage model created based on machine learning. [Prior art documents] [Patent documents]
[0004] [Patent Document 1] Special Publication No. 2021-519433
[0005] Here, the technology described in Patent Document 1 makes predictions using a model that highly emphasizes damage (leakage model) and a model that does not include damage history (general-purpose model), but does not disclose a technology for making predictions for each cause of damage to a pipeline. Summary of the Invention [Problem to be solved by the invention]
[0006] An object of the present invention is to provide a novel technique for predicting deterioration of individual pipelines using a mathematical model. [Means for solving the problem]
[0007] In order to solve the above problems, the present invention provides an AI pipeline deterioration prediction system for predicting deterioration of a pipeline that supplies a fluid, comprising: a storage unit that stores event information including an event type and timing information of an event that has occurred in a pipeline, pipeline attribute information indicating attributes of the pipeline, and a mathematical model or parameters that have been trained using the event information and the pipeline attribute information; a prediction unit that uses the mathematical model corresponding to an event type to predict pipeline deterioration based on the pipeline attribute information and determines a prediction result; Equipped with.
[0008] With this configuration, prediction of pipeline deterioration can be performed according to the type of event that occurs, such as the cause of leakage.
[0009] In a preferred embodiment of the present invention, the prediction unit utilizes the mathematical model to predict pipeline deterioration based on the pipeline attribute information of pipelines belonging to areas where event investigations have been conducted and / or pipelines in which an event has occurred, and determines the prediction result.
[0010] With this configuration, it is possible to predict pipeline deterioration using pipeline attribute information for pipelines for which data on the occurrence of events has been acquired, such as areas that have already been surveyed for water leakage.
[0011] In a preferred embodiment of the present invention, the AI pipeline deterioration prediction system further includes a classification unit that classifies pipelines based on the pipeline attribute information, The prediction unit predicts deterioration of a pipeline based on a classification result and the pipeline attribute information, and determines a prediction result.
[0012] With this configuration, it is possible to predict pipeline deterioration by utilizing the classification results even for pipelines with little data, such as those that do not have event information.
[0013] In a preferred embodiment of the present invention, the pipeline attribute information used for learning the mathematical model includes data indicating regional characteristics.
[0014] With this configuration, it is possible to predict pipeline deterioration taking into account the characteristics of each region.
[0015] In a preferred embodiment of the present invention, the AI pipeline deterioration prediction system further includes a learning unit that learns the mathematical model using the newly added event information and the pipeline attribute information of the pipeline to which the event information has been added.
[0016] With this configuration, the mathematical model can be trained using data on the added pipelines, and the system can be updated.
[0017] In a preferred embodiment of the present invention, the AI pipeline deterioration prediction system further includes an acquisition unit that collects the event information and the pipeline attribute information linked to the event information from a plurality of base stations, and stores the pipeline attribute information and the event information in association with each other; The learning unit learns an integrated model based on the collected event information and the pipeline attribute information.
[0018] With this configuration, it is possible to generate an integrated model that can perform general-purpose prediction of pipeline deterioration by utilizing pipeline data acquired at each site.
[0019] In a preferred embodiment of the present invention, the learning unit learns the integrated model using the pipeline attribute information and the event information acquired at each base station to generate a base station model.
[0020] With this configuration, it is possible to generate site models corresponding to the prediction of pipeline deterioration at each operator's site.
[0021] In a preferred embodiment of the present invention, the pipeline attribute information used for learning the mathematical model includes abnormality information representing a pipeline exhibiting abnormal deterioration.
[0022] There may be cases where pipelines show abnormal deterioration due to reasons such as poor installation techniques or materials. With this configuration, normal pipelines and abnormal pipelines that become noise during learning can be distinguished and used for learning the mathematical model, thereby improving the accuracy of pipeline deterioration predictions using the mathematical model.
[0023] In a preferred embodiment of the present invention, the AI pipeline deterioration prediction system further includes an abnormality detection unit that detects installed pipelines showing abnormal deterioration based on the number of events per specified period and assigns abnormality information to the pipeline.
[0024] With this configuration, for example, abnormality information can be added to pipelines in the year of installation, when the number of events occurring is very high, thereby improving the accuracy of pipeline deterioration determined by the mathematical model.
[0025] In a preferred embodiment of the present invention, the learning section uses the pipeline attribute information of pipelines classified into specific event types to learn the mathematical model corresponding to each event type.
[0026] With this configuration, it is possible to perform pipeline deterioration prediction corresponding to various event types by using a mathematical model that has been trained for each event type.
[0027] In a preferred embodiment of the present invention, the pipeline that supplies the fluid is a water supply system, and the pipeline attribute information used for learning the mathematical model includes hydraulic calculation feature values related to hydraulic calculations.
[0028] With this configuration, it is possible to perform pipeline deterioration prediction that is particularly suited to pipelines that supply liquid, utilizing hydraulic calculation features.
[0029] In a preferred embodiment of the present invention, the AI pipeline deterioration prediction system further includes a combination processing unit that performs a combination process of identifying multiple pipelines that should be treated as a single pipeline and combining the pipeline attribute information of the identified multiple pipelines, The mathematical model is a model that is trained using the pipeline attribute information processed by the combination processing unit.
[0030] With this configuration, multiple pipelines that should be treated as a single pipeline can be linked together and used for learning as a single pipeline, enabling more accurate prediction of pipeline deterioration. Effect of the Invention
[0031] The present invention can provide a novel technology for predicting pipeline deterioration corresponding to an event type by utilizing a trained mathematical model. [Brief description of the drawings]
[0032] [Figure 1] Block diagram showing the overall system configuration [Diagram 2] System hardware configuration diagram [Diagram 3] Example of data structure of pipeline attribute information and event information [Figure 4] Example of anomaly information assignment and merging process [Diagram 5] Example of data structure for learning data [Figure 6] Example of generation of integrated model and prediction of pipeline deterioration [Figure 7] Learning process flowchart [Figure 8] Flowchart of pipeline deterioration prediction process DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0033] The present invention will be described in more detail below with reference to the accompanying drawings. Although the drawings show a preferred embodiment, the present invention can be implemented in different forms and is not limited to the embodiment described in this specification. In this embodiment, the configuration and operation of the integrated device and the base device are described, but the same effects can be achieved by a method executed by the device, a computer program, etc. The computer program may be provided as a non-transient recording medium readable by a computer.
[0034] In the embodiments described below, the AI pipeline deterioration prediction system 0 predicts deterioration of pipelines that supply fluids, and predicts deterioration of pipelines that supply liquids, such as water supply pipes. However, as long as the prediction is related to pipelines, it may also predict deterioration of pipelines such as gas pipes.
[0035] The AI pipeline deterioration prediction system 0 is a system that uses AI (artificial intelligence) such as a trained mathematical model. AI is a technology that performs various processes by imitating human intellectual abilities. In this embodiment, pipeline deterioration prediction is performed using a trained mathematical model that has been trained using supervised learning, but pipeline deterioration prediction may also be performed using a trained model generated by other machine learning techniques such as unsupervised learning, reinforcement learning, and deep learning.
[0036] In this embodiment, the AI pipeline deterioration prediction system 0 uses a mathematical model that has been trained by supervised learning. In this embodiment, pipeline deterioration prediction is performed using a trained mathematical model that has been trained using a method such as a decision tree or k-nearest neighbor method as a method of supervised learning, but a trained model that has been trained using a method such as a neural network or support vector machine may also be used.
[0037] <1.1. System configuration> Fig. 1 is a block diagram showing the configuration of a system according to an embodiment. As shown in Fig. 1, an AI pipeline deterioration prediction system 0 includes an integrated device 1 and a learning DB arranged at a central location, and a base device 2 and a pipeline information management device 4 arranged at one or more business operator bases.
[0038] <1.1.1. Data definition> In this embodiment, data related to pipelines in the base managed by the pipeline information management device 4 is regarded as raw data related to the pipeline. In the integrated device 1 or the base device 2, learning and prediction are performed using the processed data related to the pipeline. This raw data may be subjected to conversion processing such as replacing, adding, or deleting information as necessary so that it can be used within the system, and for example, the data that has been subjected to this conversion processing becomes the processed data related to the pipeline. Note that if processing is not required, the raw data is used as it is as the processed data. Also, in this embodiment, the processed data that has been subjected to various conversion processing for use in learning the mathematical model is used as the learning data. The processed data and raw data are data related to the pipeline, and include pipeline attribute information and event information.
[0039] The processed data and raw data include data in which past event occurrences are recorded, data in which no event has occurred in the past, and data in which it is unclear whether an event has occurred in the past. Data in which past event occurrences are recorded is, for example, data relating to an event-occurred pipeline in which an event has been detected in the past. Data in which it is recorded that no event has occurred in the past is, for example, data relating to pipelines in an event-investigated area in which an investigation of the pipeline has been conducted and it has been confirmed that no event has occurred. Data in which it is unclear whether an event has occurred in the past is, for example, data relating to pipelines in an event-uninvestigated area in which no event has occurred and no investigation has been conducted. The data in which past event occurrences are recorded and the data in which it is recorded that no event has occurred in the past are used to learn the mathematical model.
[0040] <1.1.2. Overview of processing in the AI pipeline deterioration prediction system> In the embodiment described below, pipeline deterioration prediction at multiple locations including a central location which is the location of the administrator of the system and an operator location which is the location of an operator that uses the system and manages the pipelines will be described, however, processing related to pipeline deterioration prediction and learning of the mathematical model may also be performed within a single location.
[0041] The central base is a central base managed by an administrator of the AI pipeline deterioration prediction system. The business operator base is a base of a business operator that manages the pipelines. In this embodiment, the business operator is a water business operator that manages a water supply system, etc.
[0042] At the central base, a learning DB that stores learning data and an integration device 1 that learns a mathematical model (integrated model) for predicting pipeline deterioration are disposed. The integration device 1 disposed at the central base is configured to be able to realize a functional configuration described below. In this embodiment, the integration device 1 disposed at the central base collects learning data including pipeline attribute information and event information from a plurality of business bases, and performs processing related to learning the mathematical model (integrated model) using the collected data. In this embodiment, the integration device 1 performs processing related to learning the mathematical model using the learning data, but may also collect raw data before processing and perform processing related to learning the mathematical model using the collected raw data. In addition, information about pipelines used for learning the mathematical model in this embodiment is information about pipelines in an area where an event has been investigated or pipelines where an event has occurred.
[0043] The learning DB located at the central base is a DB that stores learning data based on pipeline attribute information and event information collected from each operator base, and may be configured in the same information processing device as the integration device 1.
[0044] At the business site of the business operator, there are disposed a base device 2 that learns a mathematical model (base model) and predicts pipeline deterioration, a terminal device 3, and a pipeline information management device 4. In this embodiment, the pipeline information management device 4 functions as a GIS (Geographic Information System) that manages information related to pipelines within the business site.
[0045] The base device 2 receives information about pipelines managed by the pipeline information management device 4, and generates data for the base. Among the data for the base, learning data is then passed to the integrating device 1. The integrating device 1 then performs processing related to learning the integrated model using this learning data. The base device 2 further learns the integrated model that has been learned and returned to the base using pipeline data at the operator's base, thereby generating a base model, and pipeline deterioration prediction is performed using the generated base model.
[0046] In this embodiment, the base device 2 arranged at the business base is configured to be able to realize the functional configuration described below. The base device 2 may be configured with one or more information processing devices 10, and part of the functional configuration included in the base device 2 may be realized by another information processing device 10 communicably connected to the base device 2. In this embodiment, the prediction of pipeline deterioration at the business base is performed using a base model, which is a mathematical model that has been trained using data from the pipeline DB, but may also be performed using an integrated model. Also, in this embodiment, part or all of the functional configuration of the base device 2 may be configured in the same information processing device 10 as the pipeline information management device 4.
[0047] In this embodiment, the pipeline information management device 4 arranged at the business operator's base includes a pipeline information management unit 41 and a pipeline DB, which is a database of information related to pipelines managed by the business operator. As the pipeline information management device 4, for example, an information processing device such as a general-purpose server or a personal computer may be used.
[0048] The pipeline information management device 4 used at the base of the business operator is a device equipped with the functions of a pipeline information management unit 41 (GIS, etc.) for the business operator to manage information related to the pipelines within the base, and manages pipeline attribute information and event information. In this embodiment, the pipeline DB included in the pipeline information management device 4 is a database of pipelines managed by the business operator. Note that one or more pieces of location information indicating the location of the pipeline are added to the pipeline attribute information or the pipeline attribute information and event information stored in the pipeline DB by the pipeline information management unit 41, such as addresses, codes such as coordinates and EPSG codes, hashed location information such as Geohash and Quadkey, mesh codes such as standard area meshes, etc. (collectively referred to as location information).
[0049] The terminal device 3 may be used for inputting and outputting information, such as transferring data used for learning to the base device 2. As the terminal device 3, devices such as a personal computer, a smartphone, a tablet terminal, and a personal computer may be used. A plurality of terminal devices 3 may be used, or only one terminal device 3 may be used.
[0050] In addition, in this embodiment, the AI pipeline deterioration prediction system 0 further includes an external data distribution device that accesses external systems such as a weather information providing system and a map information providing system that provide weather data, acquires data to be used for pipeline deterioration prediction, and stores the data in an external DB, and an external DB in which the acquired data is stored.
[0051] In this embodiment, the external data distribution device converts data acquired from outside for use in pipeline deterioration into a format usable by the system, stores the data acquired from outside in an external DB, and passes the stored data to the integration device 1 and the base device 2, but the acquired data may be passed on as is. The conversion process in the external data distribution device refers to converting data acquired from outside into a predetermined number of digits or unit, or performing data replacement, and converting the data into a format that can be handled by the integration device 1 and the base device 2. The external data distribution device, for example, transmits location information to an API such as a web API (Application Programming Interface) provided by an external system, collects external data, and performs a predetermined conversion.
[0052] The external DB is a database that stores external data acquired by the external data distribution device. The external data distribution device distributes data stored in the external DB. The data stored in the external DB and used for predicting pipe deterioration is provided to the base device 2 using a portable storage medium such as a USB memory or a CD-ROM, but may also be provided via a network.
[0053] In this embodiment, the external data distribution device is a device managed by the administrator of the system, and performs a process of acquiring external information regarding pipelines from an external database, etc., and storing it in an external DB, but it may also perform a process of passing on the acquired information as is.
[0054] In addition, the information processing device 10, such as the integrated device 1 or the base device 2, may perform processing to obtain information necessary for predicting pipeline deterioration directly from an external system, such as a weather information providing system or a map information providing system, which is communicatively connected via a network, etc., by using various methods, such as using an API, without going through an external data distribution device.
[0055] In this embodiment, the base device 2 and the pipeline information management device 4 installed at the business operator's base exchange data between the bases using portable storage media such as USB memory and CD-ROM, from the viewpoint of the risk of the data being handled being leaked. On the other hand, for some or all of these business operators, the base device 2, the terminal device 3 connected to the base device 2, and the integrated device 1 may be configured to be able to communicate with each other via any network such as a VPN (Virtual Private Network) on an IP (Internet Protocol) network (not shown), and data may be exchanged via the network.
[0056] <1.2. Hardware configuration> Fig. 2 is a hardware configuration diagram. As shown in Fig. 2(a), the information processing device 10 (integrated device 1, base device 2) has a processing unit 101, a storage unit 102, and a communication unit 103, and is used to perform the functions of each unit and each process. Also, as shown in Fig. 2(b), the terminal device 3 has a processing unit 301, a storage unit 302, a communication unit 303, an input unit 304, and an output unit 305, and is used to perform the functions of each unit and each process.
[0057] One or more information processing devices 10, such as a general-purpose server or a personal computer, can be used as the integrating device 1 and the base device 2. The integrating device 1 and the base device 2 have functional configurations described below, but some of the functional configurations of the integrating device 1 and the base device 2 may be disposed in another device configured to be able to communicate with the integrating device 1 and the base device 2.
[0058] The processing unit 101 and the processing unit 301 have a processor such as a CPU (Central Processing Unit) capable of executing an instruction set, and execute an OS (Operating System), as well as a learning program (in the case of the integrated device 1 or the base device 2), an AI pipeline deterioration prediction program (in the case of the base device 2), or a pipeline management program (in the case of the pipeline information management device 4), and an integrated device utilization program or a base device utilization program (in the case of the terminal device 3).
[0059] The storage unit 102 and the storage unit 302 have a volatile memory such as a RAM (Random Access Memory) capable of storing an instruction set, and a non-volatile recording medium such as an HDD (Hard Disk Drive) or SSD (Solid State Drive) capable of recording an OS, etc. The storage unit 102 of the integrated device 1 stores a learning program. The storage unit 102 of the base device 2 stores an AI pipeline deterioration prediction program including the learning program, as well as a trained mathematical model for pipeline deterioration prediction or parameters of the trained mathematical model. The storage unit 302 of the terminal device 3 stores an integrated device utilization program or a base device utilization program, etc. The storage unit 302 of the pipeline information management device 4 stores a pipeline management program.
[0060] The communication unit 103 and the communication unit 303 each have an interface for physically connecting to a network, and execute communication control with the network to communicate with other information processing devices 10 and terminal devices 3 .
[0061] The input unit 304 has an operation input device capable of input processing such as a touch panel or a keyboard, an audio input device capable of audio input such as a microphone, etc. The output unit 305 has a display device capable of display processing such as a display, and an audio output device such as a speaker.
[0062] <1.3. Functional configuration> As shown in FIG. 1, the integration device 1 includes an acquisition unit 11 and a learning unit 12.
[0063] 1, the base device 2 includes an acquisition unit 21, a learning unit 12, an abnormality detection unit 13, a combination processing unit 14, a prediction unit 15, and a classification unit 16. The pipeline information management device 4 includes a pipeline information management unit 41. This is information processing by software (a program stored transiently or non-transiently in a storage unit or the like) is specifically realized by hardware (a processing unit or the like).
[0064] <1.3.1. Configuration of Acquisition Unit 11 and Acquisition Unit 21> The acquisition unit 11 and the acquisition unit 21 acquire data such as the passed-over pipeline attribute information and event information from databases that are communicatively connected, such as a pipeline DB and a learning DB, and perform processing to pass the data on to the integrated device 1 and the base device 2.
[0065] In this embodiment, the acquisition unit 11 and the acquisition unit 21 access databases that store information about pipelines including pipeline attribute information and event information, such as a learning DB and a pipeline DB, to acquire various necessary information and pass it on to the learning unit 12 and the prediction unit 15. However, the acquisition unit 11 and the acquisition unit 21 may also acquire data stored in an external DB from an external data distribution device that is communicatively connected via a network, or may access an external information providing system using an API or the like to acquire necessary data.
[0066] In addition, when the formats of data managed by the pipeline information management device 4 at each base are not uniform, such as when the names of pipelines are different for each business operator or when the data formats are different, the acquisition unit 11 or the acquisition unit 21 performs a process of converting the data into a format that can be used in the system, such as a process of unifying pipeline names and formats. In this embodiment, the learning data provided to the integration device 1 at the central base is data converted in the base device 2. In addition, the storage unit 102 of the integration device 1 or the base device 2 may store data used for conversion, such as dictionary data and a correspondence table used for converting the names of pipelines. In addition, the acquisition unit 11 or the acquisition unit 21 may acquire information from an external information providing system or an external DB, and may add data such as environmental features and regional features described later to the pipeline attribute information and / or event information stored in the pipeline DB by the pipeline information management unit 41, or the converted information. At this time, if necessary, the data acquired from the outside may be converted into a format handled by the integration device 1 or the base device 2, such as by converting the data into a predetermined number of digits or unit, or by replacing the data, and then added.
[0067] The acquisition unit 11 is disposed in the integration device 1 at the central base, and acquires information about pipelines, including pipeline attribute information and event information, from the operator's base, and stores the information in the learning DB. It also acquires learning data from the learning DB and transfers the data to the learning unit 12 of the integration device 1. The acquisition unit 11 of the integration device 1 may also acquire data to be used for learning from a device such as an external data distribution device connected to the integration device 1 so as to be able to communicate with the data, and may perform preprocessing such as complementing and aligning the format before storing the processed data in the learning DB.
[0068] In this embodiment, the acquisition unit 11 receives information about pipelines provided from the business site using a portable storage medium via a terminal device (not shown) at the central site and stores the information in a learning DB. The acquisition unit 11 may also perform processing to complement and standardize the format of information about pipelines used for pipeline deterioration prediction provided via a portable storage medium or a network.
[0069] The acquisition unit 21 is disposed in the base device 2 at the business operator's base, and performs a process of acquiring information on pipelines including pipeline attribute information and event information used for learning the integrated model from the pipeline DB, and transferring the information to the learning unit 12 or the prediction unit 15 of the base device 2. In this embodiment, the acquisition unit 21 at the business operator's base receives a mathematical model used for pipeline deterioration prediction acquired via a storage medium, and stores the mathematical model in the storage unit 102. In addition, at the business operator's base, the acquisition unit 21 may perform a process of accepting information on pipelines input via the terminal device 3. The acquisition unit 21 performs a process of converting the information on pipelines used for learning the mathematical model acquired from the pipeline DB and for pipeline deterioration prediction into an easy-to-use form, such as a process of complementing the information and unifying the format.
[0070] <1.3.2. Data structure> FIG. 3 shows an example of the data configuration that is transferred to the integrated device 1 or the base device 2 and used for learning processing or pipeline deterioration prediction.
[0071] FIG. 3(a) shows an example of the data configuration of pipeline attribute information related to the characteristics of each pipeline. In this embodiment, the pipeline attribute information is information related to the pipeline and is data acquired from a database of a company that manages the pipeline, but some of the pipeline attribute information may be data input by a company that manages the pipeline or an administrator of the system via a terminal device 3 that is communicably connected to the integrated device 1 and the base device 2. The pipeline attribute information includes feature amounts related to the pipeline used in the learning unit 12 or the prediction unit 15. In this embodiment, the learning unit 12 and the prediction unit 15 perform learning or pipeline deterioration prediction using about 60 to 70 feature amounts related to the pipeline included in the pipeline attribute information, but the number of feature amounts included in the pipeline attribute information is not limited to this range.
[0072] The pipeline attribute information shown in Figure 3(a) is information about the pipeline, and includes data linked to the features contained in the pipeline attribute information regarding the year in which the pipeline was installed, the pipe type, which is information about the material of the pipeline such as ductile cast iron pipe (DIP) or ordinary cast iron pipe (CIP), the pipeline diameter, and whether or not a leak investigation has been conducted, which is information about the presence or absence of event information such as whether a leak investigation has been conducted, using a unique ID that identifies the pipeline.
[0073] In this embodiment, the pipeline attribute information used for pipeline deterioration prediction includes, in addition to the information shown in Figure 3(a), hydraulic calculation features, which are features related to hydraulic calculations, environmental information related to the environment at the pipeline installation location, and information related to the pipeline, such as abnormality information indicating a pipeline showing abnormal deterioration and adjacent pipeline information related to adjacent pipelines including the pipe type of the adjacent pipeline.
[0074] In this embodiment, the environmental information on the environment of the location where the pipeline is installed includes information such as the average temperature of the installation location, the average precipitation, the distance from a major road, and the traffic volume of the road. In addition, other information such as the installation location of the pipeline may be included as the pipeline attribute information as long as it is a feature related to the deterioration of the pipeline. In addition, the acquisition unit 11 and the acquisition unit 21 may acquire data used for pipeline deterioration prediction or learning from an external information distribution system, an external DB, or the like via an external data distribution device, and store the acquired data in the storage unit 102 as pipeline attribute information.
[0075] The hydraulic calculation feature values are feature values related to hydraulic calculations including feature values obtained by performing hydraulic calculations based on feature values included in the pipeline attribute information and actual measured values obtained by field surveys, and include the flow rate and pressure of the water flow in the pipeline, the water supply population, flow rate, flow velocity, etc. In this embodiment, the field survey for acquiring the hydraulic calculation feature values includes water pressure measurement, flow rate measurement, and water supply usage amount.
[0076] The pipeline attribute information includes regional features that are data indicating regional characteristics. The regional characteristics data is data indicating characteristics of the region where the pipeline is laid, and may be data regarding soil characteristics or climate of the region where the pipeline is laid, for example.
[0077] 3(b) shows an example of the data configuration of the event information acquired by the acquisition unit 11 or the acquisition unit 21. The event information is information on events such as water leakage, breakage, leakage, etc. that have occurred in pipelines such as water and gas pipes, and in this embodiment, includes information on events such as the occurrence history of water leakage acquired by an event investigation by a business entity that manages the pipelines, such as a water utility, and information on events that have already occurred in the pipelines, acquired by a water utility or the like that has repaired or replaced a pipeline where a sudden water leakage accident has occurred.
[0078] An event investigation is an investigation carried out by a business operator that manages a pipeline as to whether an event such as a water leak has occurred in the pipeline, and in this embodiment, it is a pipeline water leak investigation. In this embodiment, the event information includes information on events such as a water leak accident or a gas pipe breakage that occurs suddenly in an uninvestigated area where an event investigation has not yet been carried out. In addition, such event information on an event that occurs suddenly in a pipeline does not need to be linked to pipeline attribute information.
[0079] The event information shown in FIG. 3(b) includes a water leakage data ID, which is a unique ID for identifying the event, the pipeline ID of the pipeline in which the event occurred, the cause of the water leakage, and the event occurrence year (year of water leakage) related to the year in which the event occurred as timing information. The timing information is information related to the timing of the event occurring in the pipeline. In the data shown in FIG. 3(b), data related to the year in which the event occurred is included as timing information, but the survival time of the pipeline until the event occurrence obtained from the year in which the pipeline was installed and the year in which the event occurred may also be included as timing information. In this embodiment, the event information and timing information are managed by linking them to the same water leakage data ID.
[0080] In this embodiment, the event type included in the event information is information indicating the cause of damage to the pipeline, such as the cause of leakage, and whether or not an event has occurred, and for example, information regarding the cause of leakage, such as corrosion deterioration of the water pipe, electrolytic leakage, and vibration leakage, is stored. Also, for a pipeline that does not have a leakage, information indicating "no leakage" may be stored as the event type. In this embodiment, the event information is information regarding leakage from a water pipe, but may also be information regarding other events occurring in the pipeline, such as a burst or breakage of a gas pipe.
[0081] It is also assumed that information related to pipelines, such as pipeline attribute information and event information, may have different formats for each business operator and may be missing information. The acquisition unit 11 or the acquisition unit 21 may perform preprocessing on the information related to pipelines, such as pipeline attribute information and event information, such as unifying the data format and complementing missing data. In addition, in this embodiment, when the acquired information related to pipelines includes data that does not have event information, such as data acquired in an area where an event has not been investigated, the acquisition unit 11 or the acquisition unit 21 generates learning data that excludes data of pipelines for which event information has not been acquired, such as data acquired in an area where an event has not been investigated.
[0082] <1.3.3. Configuration of the anomaly detection unit 13> The abnormality detection unit 13 assigns abnormality information to pipelines that show abnormal deterioration. In this embodiment, the abnormality detection unit 13 detects an abnormal year based on the number of events in which water leakage, gas leakage, etc. occurred for each number of laid pipelines in a year, and assigns abnormality information to pipelines that were laid in the detected abnormal year. The abnormality detection unit 13 calculates the number of events for each number of laid pipelines based on the number of events for each year and the number of laid pipelines for each year. If the number of events for each number of laid pipelines in a particular year is large and the laid pipelines in a particular year show abnormal deterioration, the abnormality detection unit 13 assigns abnormality information to the pipelines that were laid in that year.
[0083] In this embodiment, the abnormality information provided by the abnormality detection unit 13 is an abnormal year flag indicating whether or not the pipeline was installed in a specific installation year and shows abnormal deterioration due to reasons such as the installation work method used by a specific operator in that specific installation year, but the information may be any other information that indicates a pipeline that shows abnormal deterioration.
[0084] 4(a) shows the event occurrence rate for each year of installation of the pipeline managed by a specific utility UA. The following formula (1) shows a method for calculating the event occurrence rate in this embodiment.
[0085]
number
[0086] FIG. 4(a) shows an example of a graph of the event occurrence rate of pipelines managed by a specific utility UA. As shown in FIG. 4(a), pipelines laid in 1959 show a very high event occurrence rate compared to pipelines laid in other years. The abnormality detection unit 13 assigns abnormality information to pipelines laid in 1959 in which the event occurrence rate per number of laid pipelines calculated by formula (1) is very high compared to other years. In this embodiment, the abnormality detection unit 13 counts the number of occurrences and occurrence rate of events, and assigns abnormality information to pipelines laid in years in which the number of occurrences of events is abnormally high using an exponentially weighted moving average. In this embodiment, even if the abnormality detection unit 13 assigns abnormality information to pipelines laid in 1959 managed by a specific utility UA, it does not assign abnormality information to pipelines laid in 1959 managed by another utility UB.
[0087] <1.3.4. Configuration of the joining processing unit 14> The combining processing unit 14 calculates the similarity of the pipeline attribute information of two or more pipelines, and when the similarity is equal to or greater than a threshold, performs a combining process of linking the pipeline attribute information to the same pipeline ID. In addition, when the event information is not linked to the pipeline attribute information, the combining processing unit 14 may perform a process of linking the pipeline attribute information of the same type of pipeline or a pipeline having similar pipeline attribute information in the vicinity of where the event occurred to the event information. In this embodiment, the combining processing unit 14 performs a process of linking the pipeline attribute information of multiple pipelines having similar pipeline attribute information in order to link one event information to multiple pipelines that exist in the vicinity and can be considered to be almost the same.
[0088] 4(b), when multiple pipelines are connected by joints, the connection processing unit 14 performs a process of calculating the similarity of pipeline attribute information other than the pipe type, diameter, and installation year for multiple adjacent pipelines that have the same pipe type, diameter, and installation year. The connection processing unit 14 may also perform a process of calculating the similarity using all pipeline attribute information including the pipe type, diameter, and installation year.
[0089] 4(b), the similarity between pipes A and C is calculated using the following formula: When the similarity calculated using the following formula (2) is equal to or greater than a threshold, the combination processing unit 14 links the pipeline attribute information of pipes A and C to the same pipeline ID and stores the same pipeline ID in the storage unit 102.
[0090] In this embodiment, the combination processing unit 14 performs processing to calculate the similarity using the cosine similarity shown below, but may perform processing to obtain the similarity using other methods such as Euclidean distance, Jaccard coefficient, Dice coefficient, etc.
[0091]
number
[0092] In this embodiment, the abnormality detection unit 13 and the combination processing unit 14 are provided in the base device 2, but may be provided in the integrating device 1.
[0093] <1.3.5. Configuration of learning unit 12> The learning unit 12 learns a mathematical model for predicting deterioration of a pipeline using a combination of the pipeline attribute information and the event information as teacher data. The learning unit 12 learns a plurality of mathematical models corresponding to each event type.
[0094] Fig. 5 shows an example of the data structure of learning data including pipeline attribute information and event information collected by the acquisition unit 11 or the acquisition unit 21 and formatted for learning. The learning data shown in Fig. 5 is learning data generated based on the pipeline attribute information and event information collected by the acquisition unit 11 or the acquisition unit 21 and used for learning a mathematical model in the learning unit 12. In this embodiment, the learning unit 12 performs learning using the learning data formatted by the acquisition unit 11, but learning may also be performed using the pipeline attribute information and event information shown in Fig. 3.
[0095] The learning data is information about a pipeline, and includes a leakage data ID, a year of installation, a pipeline ID, a survival time, which is the time from the installation of the pipeline to the occurrence of an event or the time from the previous event to the next event, a teacher label, which is a label related to an event type, the number of events, such as the number of past leakage occurrences at that time in the pipeline where the event occurred, and abnormality information related to the abnormal year. In this embodiment, the teacher label included in the learning data is data related to the event type, and is a label assigned to each event type. In addition, the learning data may include any data included in the pipeline attribute information. The acquisition unit 11 or the acquisition unit 21 may perform a process of generating learning data to which a teacher label based on each event type is assigned from the acquired information about the pipeline.
[0096] The learning unit 12 learns a mathematical model corresponding to the event type by using the pipeline attribute information and the event information. In this embodiment, the learning unit 12 learns a plurality of mathematical models corresponding to each event type based on a plurality of event types including corrosion deterioration, no leakage, electrolytic corrosion leakage, vibration leakage, and other defective leakage.
[0097] In this embodiment, the learning unit 12 learns a mathematical model for predicting pipeline deterioration using learning data based on teacher labels related to event types. For example, when generating a corrosion deterioration model using a mathematical model that predicts pipeline deterioration due to corrosion deterioration, the learning unit 12 generates the corrosion deterioration model from a mathematical model learned using data from pipe A with a leakage data ID of 1 and data from pipe C with a leakage data ID of 2, to which labels related to corrosion deterioration are assigned as teacher labels. In addition, when generating an other defect model using a mathematical model that predicts pipeline deterioration due to defects other than corrosion deterioration, the learning unit 12 generates the other defect model from a mathematical model learned using data from pipe C with a leakage data ID of 3, to which labels related to other defects are assigned as teacher labels.
[0098] The generation of the integrated model and the generation of the base model will be described with reference to FIG.
[0099] The integration process for generating the integrated model will be described with reference to Fig. 6(a). The integration device 1 collects pipeline attribute information and event information acquired at the bases of each of the operators UA, UB, and UC at a central base of the administrator of the system, and generates the integrated model, which is a mathematical model that has been trained using the collected data.
[0100] The integration device 1, which is placed at a central location by the administrator of the system, performs integration processing to learn using data collected from each operator's location and integrated at the central location, and to generate an integrated model using the generated mathematical model. The learning unit 12 of the integration device 1 acquires information about pipelines, including pipeline attribute information and event information, collected at each operator's location of the operator UA, the operator UB, and the operator UC, and performs integration processing to generate an integrated model, which is a mathematical model that has been learned using the collected information about pipelines.
[0101] In this embodiment, the learning unit 12 learns the integrated model by performing fine tuning using new data on the integrated model that has been trained using pipeline attribute information and event information previously acquired by the system administrator. If the system administrator does not have a trained integrated model, the learning unit 12 may perform training using the acquired pipeline attribute information and event information to generate a new integrated model that is a trained mathematical model.
[0102] The learning unit 12 of the integration device 1 learns the integrated model using learning data based on the collected event information on pipelines managed by each operator and pipeline attribute information linked to the event information. In this embodiment, the integrated model is periodically updated, and the learning unit 12 re-learns the integrated model using learning data generated based on the event information collected at the operator's base between the previous update and the current update and the pipeline attribute information linked to the collected event information.
[0103] In this embodiment, the learning unit 12 performs learning using event information of a pipeline in which event information has been acquired in an area where an event has been investigated or in a pipeline where an event has occurred, and pipeline attribute information linked to the event information.
[0104] Furthermore, at the operator base of each operator, the learning unit 12 of the base device 2 generates a base model corresponding to the pipeline of each operator used at the operator base of each operator.
[0105] In this embodiment, the learning unit 12 of the base device 2 re-learns the integrated model using information about the pipelines stored in the pipeline DB at the business base, and generates a base model corresponding to each business base. For example, when generating a base model for performing corresponding pipeline deterioration prediction for pipelines managed by a business UA that uses the system, the learning unit 12 generates a base model MA corresponding to the pipeline deterioration prediction at the business base of the business UA by learning the integrated model using information about the pipelines including pipeline attribute information and event information acquired at the base BA.
[0106] In this embodiment, the learning unit 12 performs the above-mentioned integration process and the process related to the generation of the site model when sufficient data is collected at the business site or the central site of each business, and updates the mathematical model. In this embodiment, the learning unit 12 re-learns the trained integrated model using additional data by fine tuning, but the process may be performed using any method as long as the method trains the trained model using additional data.
[0107] In this embodiment, the learning unit 12 repeatedly learns the integrated model using newly acquired data and learns the site model by re-learning the trained integrated model. The learning unit 12 at the central site generates the integrated model for this year by further learning the integrated model for last year using data collected this year. The learning unit 12 at the operator site re-learns the site model for this year by further learning the generated integrated model for this year using data collected at the operator site this year.
[0108] <1.3.6. Configuration of prediction unit 15 and classification unit 16> The prediction unit 15 performs pipeline deterioration prediction based on the pipeline attribute information by using a mathematical model corresponding to each event type. In this embodiment, the prediction unit 15 performs pipeline deterioration prediction by using a mathematical model corresponding to a plurality of event types including corrosion deterioration, no leakage, electrolytic corrosion leakage, vibration leakage, and other defective leakage. The prediction unit 15 performs pipeline deterioration prediction based on information about the pipeline, and outputs a pipeline deterioration rate, which is the probability of an event occurring in a specific period for each event type, as a prediction result. The prediction unit 15 may output the probability of an event occurring within one year, the probability of an event occurring within three years, the probability of an event occurring within five years, and the probability of an event occurring within ten years for each event type.
[0109] Furthermore, the prediction unit 15 calculates the importance of the feature, and outputs the name of the pipeline attribute information having a high degree of contribution based on the acquired importance. In this embodiment, the prediction unit 15 outputs the names of the pipeline attribute information having the top three degrees of contribution. Furthermore, in this embodiment, the prediction unit 15 uses the Shapley value as an index indicating the importance of the feature, but the importance of other feature such as a correlation coefficient may also be used.
[0110] In addition, in this embodiment, the prediction unit 15 predicts pipeline deterioration by using the pipeline's lifetime calculated using the pipeline installation year included in the pipeline attribute information and the event occurrence year included in the event information as timing information.
[0111] The prediction unit 15 may perform pipeline deterioration prediction using a mathematical model configured by any method, such as a gradient descent method, boosting, a decision tree, a neural network, a logistic regression, a k-nearest neighbor method, or a method combining multiple methods, as long as the mathematical model is a machine learning-based mathematical model. The prediction unit 15 may perform pipeline deterioration prediction using a mathematical model configured by a different method for each event type. Furthermore, the prediction unit 15 may perform pipeline deterioration prediction using multiple mathematical models.
[0112] A pipeline deterioration prediction using a mathematical model corresponding to corrosion deterioration will be described. In this embodiment, the prediction unit 15 performs a survival time analysis as a pipeline deterioration prediction using a mathematical model corresponding to corrosion deterioration. In this embodiment, the prediction unit 15 predicts the time until an event of corrosion deterioration occurs, and outputs a pipeline deterioration rate as a result of the pipeline deterioration prediction using the prediction result, but a predicted value of the time until corrosion deterioration occurs may be output as a result of the pipeline deterioration prediction. In this embodiment, the prediction unit 15 performs pipeline deterioration prediction regarding corrosion deterioration using a survival time analysis method such as the Kaplan-Meier method or the Cox proportional hazards model, but any method related to survival time analysis may be used to perform pipeline deterioration prediction regarding corrosion deterioration.
[0113] In this embodiment, the prediction unit 15 performs pipeline deterioration prediction using a mathematical model corresponding to an event type other than corrosion deterioration, including no leakage, electrolytic corrosion leakage, vibration leakage, and other mal-functioning leakage. In this embodiment, the prediction unit 15 performs pipeline deterioration prediction for an event type other than corrosion deterioration, using timing information including the lifetime of the pipeline as one feature.
[0114] When the pipeline attribute information of the pipeline to be predicted is data acquired in an area where an event has been investigated, such as an area where a water leakage investigation has been completed, or in a pipeline where an event has occurred, the prediction unit 15 performs pipeline deterioration prediction based on the pipeline attribute information by using a mathematical model corresponding to each event type. The prediction unit 15 performs pipeline deterioration prediction by using a multi-value classification model that has been trained by linking the feature amount related to the pipeline included in the pipeline attribute information with the teacher label related to each event type.
[0115] The classification unit 16 classifies the pipelines based on the pipeline attribute information. After performing classification, the classification unit 16 stores information on the classification result in association with the pipeline attribute information. In this embodiment, the classification unit 16 classifies the pipelines using data with low spatial autocorrelation included in the pipeline attribute information. Also, in this embodiment, the classification unit 16 performs classification using the pipeline attribute information and the inferred value of the pipeline deterioration rate of the pipeline output by the prediction unit 15 based on the pipeline attribute information, but classification may be performed using only the pipeline attribute information.
[0116] When the pipeline attribute information of the pipeline to be predicted is data of a pipeline in an uninvestigated area where no event investigation has been conducted, such as an uninvestigated area for water leakage, and for which no event information has been acquired, the prediction unit 15 performs pipeline deterioration prediction using an inference value based on the pipeline attribute information using a learned mathematical model and the classification result by the classification unit 16. Hereinafter, a method of predicting pipeline deterioration for pipelines in an uninvestigated area for which no event information has been acquired, such as an uninvestigated area for water leakage, will be described with reference to Fig. 6(b).
[0117] As shown in FIG. 6(b), the classification unit 16 classifies the pipes A, B, and C into pipes having similar characteristics based on the inferred value of the pipe deterioration rate based on the pipe attribute information and the feature amount related to the pipe included in the pipe attribute information. When predicting pipe deterioration for pipe B, which is a pipe for which event information has not been acquired, the prediction unit 15 determines a prediction result using the inferred value of the pipe deterioration rate calculated using a learned mathematical model based on the feature amount included in the pipe attribute information for pipe B and the inferred value of pipe A, which is a pipe in an area where an event has been investigated or a pipe for which an event has occurred, classified in the same classification as pipe B. When predicting pipe deterioration for pipe B, which is a pipe for which event information has not been acquired, the prediction unit 15 may output an average value of the inferred value of pipe A and the inferred value of pipe B as a prediction result.
[0118] In this embodiment, the prediction unit 15 performs pipeline deterioration prediction using statistical indicators such as the average value or median value of pipeline deterioration rates output by the classification unit 16 as inferred values of other pipelines in which events have occurred and which are classified in the same category as the pipeline for which pipeline deterioration prediction is to be performed, but the prediction unit 15 may perform pipeline deterioration prediction using information other than the inferred values regarding other pipelines included in the same category.
[0119] <1.4. Processing flow> The process flow will be described below with reference to Fig. 7 and Fig. 8. Fig. 7 shows the flow of the learning process in the base device 2. The process flows shown in Fig. 7 and Fig. 8 are examples in this embodiment, and the processes may be executed in an order different from that shown in the figures.
[0120] <1.4.1. Learning process flow> The flow of the learning process for the base model at the business base will be shown below with reference to FIG.
[0121] The acquisition unit 21 11 acquires information about pipelines, including pipeline attribute information and event information, and stores the information in a learning DB (S201). The abnormality detection unit 13 calculates an event occurrence rate for each number of laid pipelines based on the event information (S202). The abnormality detection unit 13 detects pipelines that show abnormal deterioration based on the calculated event occurrence rate, and when a pipeline shows abnormal deterioration, assigns abnormality information to the detected pipeline (S203). The combination processing unit 14 calculates the similarity of pipeline attribute information of pipelines that are closer than the acquired pipeline attribute information (S204). When the calculated similarity of adjacent pipelines is equal to or greater than a threshold, the combination processing unit 14 performs a combination process of linking pipeline attribute information of multiple pipelines whose similarity is equal to or greater than the threshold to the same pipeline ID (S205). The learning unit 12 uses the collected pipeline information, including the pipeline attribute information and event information, to learn a mathematical model and generate an integrated model (S206).
[0122] 7 is an example of a process flow for generating a base model at a business base. In this embodiment, as a learning process in the integration device 1, the acquisition unit 11 acquires information on pipelines at each base (S201), and the learning unit 12 learns the integrated model (S206).
[0123] In this embodiment, the learning unit 12 of the integration device 1 uses information about pipelines collected during the period since the previous learning time to update the integrated model by performing additional learning on last year's integrated model, but if an integrated model has not been generated, the integrated model may be generated using information about all pipelines stored in the learning DB.
[0124] In this embodiment, the integrated model is generated using machine learning methods such as fine tuning and bagging, but may be generated using other methods.
[0125] <1.4.2. Pipe deterioration prediction processing flow> Fig. 8 shows a flow of processing for predicting pipe deterioration in the base device 2. In this embodiment, the processing for predicting pipe deterioration shown in Fig. 8 is performed at a business operator's base, but may be performed at a central base.
[0126] When the acquisition unit 21 acquires information about a pipeline including pipeline attribute information (S211), the prediction unit 15 performs pipeline deterioration prediction based on the pipeline attribute information using a mathematical model, and calculates an inferred value (S212). When the acquired pipeline attribute information is pipeline attribute information acquired for a pipeline in which an event has occurred, the prediction unit 15 outputs the inferred value as a final prediction result (S213). When the acquired pipeline attribute information is pipeline attribute information for a pipeline in which event information has not been acquired, such as acquired in an area in which no event has been investigated, the classification unit 16 classifies the pipeline based on the pipeline attribute information (S214). The prediction unit 15 performs pipeline deterioration prediction based on the inferred value and the classification result (S215), and outputs a value based on the inferred value and the classification result as a final prediction result of the pipeline deterioration prediction (S216).
[0127] <1.5. Specific examples of model generation and learning> For example, each business base uses data accumulated in the current period to make a prediction using an integrated model generated from data accumulated in at least one previous period. As a specific example, when information about pipelines up to December 31, 2022 is accumulated at the business base, the integration device 1 collects information about pipelines including pipeline attribute information and event information accumulated at each base up to December 31, 2022 (referred to as pipeline data (before 2022)), and generates an integrated model (2022) using information acquired from multiple business bases as learning data. The generated integrated model (2022) or its parameters are distributed from the central base to each business base. After that, when information about pipelines at bases up to December 31, 2022 is accumulated, each business base learns this integrated model (2022) using the pipeline data (before 2022) to generate a base model (2022). If a base model is generated at the business base, the base model (2022) is used, and if there is no base model, the integrated model (2022) is used to input the pipeline attribute information (2023) and perform pipeline deterioration prediction. Next, the pipeline information management device 4 at each base collects new information about the pipeline (referred to as pipeline data (2023)) between January 1 and December 31, 2023. The integration device 1 collects information about the pipeline in 2023 (referred to as pipeline data (2023)) accumulated at each base, and generates an integrated model (2023) using information acquired from multiple business bases as learning data. In addition, the accuracy of the mathematical model (integrated model, base model) is improved by repeating the distribution of the integrated model and the generation of the base model at each base, as in the previous year.
[0128] In this embodiment, the integrated device 1 includes an acquisition unit 11 and a learning unit 12, and performs processing related to learning the integrated model. It may further include an anomaly detection unit 13, a combination processing unit 14, a prediction unit 15, and a classification unit 16, and perform processing related to adding anomaly information, combination processing, and processing related to pipeline deterioration prediction.
[0129] In this embodiment, the information on the pipelines transferred from the base device 2 to the integrating device 1 using a portable storage medium is processed data in which the process of adding anomaly information in the anomaly detection unit 13 of the base device 2 and the process of combining in the combining processing unit 14 have already been performed, but it may be data that has not been processed in the anomaly detection unit 13 and the combining processing unit 14. Furthermore, when data that has not been processed in the anomaly detection unit 13 and the combining processing unit 14 is transferred to the integrating device 1, the integrating device 1 is provided with the anomaly detection unit 13 and the combining processing unit 14 and performs the process of adding anomaly information and the combining processing. [Explanation of symbols]
[0130] 0: AI Pipeline Deterioration Prediction System 1: Integrated device 2: Base device 3: Terminal device 4: Pipeline information management device 11: Acquisition part 12: Learning Department 13: Anomaly detection section 14: Joint processing section 15: Prediction section 16: Classification section 21: Acquisition Department 10: Information processing device 101: Processing section 102: Storage section 103: Communications Department 301: Processing section 302: Storage section 303: Communications Department 304: Input section 305: Output section
Claims
1. An AI pipeline deterioration prediction system for predicting deterioration of a pipeline that supplies a fluid, comprising: The apparatus includes an acquisition unit, a prediction unit, and a classification unit, The acquisition unit acquires pipeline attribute information indicating an attribute of a pipeline, The prediction unit predicts deterioration of each pipeline based on a plurality of feature amounts included in the pipeline attribute information, and calculates an inference value; The classification unit classifies each of the pipelines based on a plurality of feature amounts included in the pipeline attribute information, the prediction unit determines a deterioration prediction result for each pipeline by using an inferred value of the pipeline for which the prediction is performed and an inferred value of another pipeline that belongs to the same classification as the pipeline for which the prediction is performed; AI pipeline deterioration prediction system.
2. The classification unit further performs classification using the inference value. The AI pipeline deterioration prediction system according to claim 1.
3. When the pipeline for which the prediction is performed is a pipeline in an area where no event has been investigated, the prediction unit determines the deterioration prediction result by utilizing an inference value of a pipeline in an area where an event has been investigated and which belongs to the same classification as the pipeline for which the prediction is performed. The AI pipeline deterioration prediction system according to claim 1.
4. A method for predicting deterioration of a pipeline for supplying a fluid, the method comprising: The method includes an acquisition step, a prediction step, and a classification step, In the acquiring step, pipeline attribute information indicating an attribute of the pipeline is acquired, In the prediction step, a deterioration prediction is performed for each of the pipelines based on a plurality of feature amounts included in the pipeline attribute information, and an inferred value is calculated; In the classification step, the pipelines are classified based on a plurality of feature amounts included in the pipeline attribute information, In the prediction step, a deterioration prediction result of each pipeline is determined by using an inferred value of the pipeline for which the prediction is performed and an inferred value of another pipeline that belongs to the same classification as the pipeline for which the prediction is performed. AI pipeline deterioration prediction method.
5. An AI pipeline deterioration prediction program for predicting deterioration of a pipeline that supplies a fluid, comprising: The AI pipeline deterioration prediction program causes a computer to function as an acquisition unit, a prediction unit, and a classification unit, The acquisition unit acquires pipeline attribute information indicating an attribute of a pipeline, The prediction unit predicts deterioration of each pipeline based on a plurality of feature amounts included in the pipeline attribute information, and calculates an inference value; The classification unit classifies each of the pipelines based on a plurality of feature amounts included in the pipeline attribute information, the prediction unit determines a deterioration prediction result for each pipeline by using an inferred value of the pipeline for which the prediction is performed and an inferred value of another pipeline that belongs to the same classification as the pipeline for which the prediction is performed; AI pipeline deterioration prediction program.
Citation Information
Patent Citations
Diagnostic method of deterioration degree of buried pipe
JP2004184344A
Plant maintenance management method and maintenance management device
JP2005180997A
Underground pipe update period prediction device, underground pipe update period prediction method, program, and computer readable recording medium
JP2020051801A
Facility maintenance management system
JP2020190994A
Pipe damage prediction
JP2021519433A
