Plant abnormality monitoring system and plant abnormality monitoring method

Abstract

To provide support information for promptly repairing a facility in a plant when an abnormality in which fluid leaks from the facility in the plant occurs.SOLUTION: A plant abnormality monitoring system comprises: an imaging device that captures an image of fluid leaking from a facility in a plant; an estimation device that includes a first learning model that has learned using learning data associating images of fluid leaking from the facility with the flow rate of the fluid, and estimates the flow rate of the fluid when a new image of the fluid leakage captured by the imaging device is input to the first learning model; and an output device that outputs an estimation value of the flow rate of the fluid estimated by the estimation device.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a plant abnormality monitoring system and a plant abnormality monitoring method.

Background Art

[0002] For example, a nuclear power plant, which is one type of plant, is provided with a reactor containment vessel that stores a reactor pressure vessel in an airtight state. The reactor containment vessel serves as a barrier to prevent the spread of radioactive substances to the surroundings of the reactor containment vessel in the event that, for example, nuclear fuel is damaged and radioactive substances are released from the reactor pressure vessel.

[0003] In order for the reactor containment vessel to fully perform its function, it is necessary to periodically or non-periodically inspect the conditions inside the reactor containment vessel, and when an abnormality occurs inside the reactor containment vessel, it is necessary to quickly identify the cause of the abnormality in order to take measures to restore the inside of the reactor containment vessel to a normal state.

[0004] For example, Patent Document 1 discloses a technique in which a guide pipe is provided between the upper and lower parts inside a reactor containment vessel, and a detector for detecting radiation dose, oxygen concentration, temperature, etc. is periodically or non-periodically moved through this guide pipe to detect whether there is an abnormality in at least any one of the radiation dose, oxygen concentration, and temperature detected inside the reactor containment vessel.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0006] Inside the reactor containment vessel, various facilities (equipment, piping, etc.) necessary for stably and continuously maintaining nuclear fission reactions in the reactor pressure vessel are provided. When an abnormality occurs where fluid (liquid, gas, etc.) leaks from the facilities inside the reactor containment vessel, with the technology of Patent Document 1, since it is impossible to grasp the location where the fluid leaks and the flow rate of the fluid at the location where the fluid leaks, it is impossible to promptly take appropriate measures for repair of the facilities inside the reactor containment vessel, and there is a problem that it may lead to a major accident.

[0007] The present invention has been made in view of the above problems, and an object is to provide a plant abnormality monitoring system and a plant abnormality monitoring method that output support information for quickly repairing facilities in a plant when an abnormality occurs where fluid leaks from the facilities in the plant.

Means for Solving the Problems

[0008] One aspect of the present invention for achieving the above object is a plant abnormality monitoring system, including an imaging device that captures an image of fluid leaking from facilities in a plant, a first learning model that has been learned using learning data in which an image of the fluid leaking from the facilities and the flow rate of the fluid are associated, and when a new image of the fluid leaking captured by the imaging device is input to the first learning model, an estimation device that estimates the flow rate of the fluid, and an output device that outputs an estimated value of the flow rate of the fluid estimated by the estimation device.

[0009] According to the plant abnormality monitoring system of the present invention, when an abnormality occurs where fluid leaks from facilities in a plant, it is possible to estimate the flow rate of the fluid at the location where the fluid leaks, and based on the estimated value of the flow rate of the fluid, provide support information for quickly repairing the facilities in the plant to the monitor.

[0010] Another aspect of the present invention for achieving the above object is a plant abnormality monitoring system, which further includes a calculation device. The estimation device includes a second learning model trained using learning data associating an image of the fluid leaking from the facility with the opening area of the location where the fluid leaks. When a new image of the fluid leaking captured by the imaging device is input into the second learning model, the opening area of the location where the fluid leaks is estimated. The calculation device calculates the flow rate of the fluid based on the opening area estimated by the estimation device. The output device outputs a difference value between an estimated value of the flow rate of the fluid estimated by the estimation device and a calculated value of the flow rate of the fluid calculated by the calculation device.

[0011] According to the plant abnormality monitoring system of the present invention, based on the difference value between the estimated value of the flow rate of the fluid estimated by the estimation device and the calculated value of the flow rate of the fluid calculated by the calculation device using the opening area estimated by the estimation device, it is possible to improve the estimation accuracy of the first learning model.

[0012] Another aspect of the present invention for achieving the above object is a plant abnormality monitoring system. The estimation device includes a third learning model trained using learning data associating an image of the fluid leaking from the facility with the facility system of the location where the fluid leaks. When a new image of the fluid leaking captured by the imaging device is input into the third learning model, the facility system of the location where the fluid leaks is estimated. The calculation device calculates the flow rate of the fluid based on the estimated value of the opening area and the estimation result of the facility system estimated by the estimation device.

[0013] According to the plant abnormality monitoring system of the present invention, based on the difference value between the estimated value of the flow rate of the fluid estimated by the estimation device and the calculated value of the flow rate of the fluid calculated by the calculation device using the estimated value of the opening area and the estimation result of the facility system estimated by the estimation device, it is possible to further improve the estimation accuracy of the first learning model.

[0014] Another one of the present invention for achieving the above object is a plant abnormality monitoring system, wherein the estimation device includes a fourth learning model that has been trained using learning data in which an image of the fluid leaking from the facility is associated with the physical properties of the fluid. When a new image of the fluid leaking, taken by the imaging device, is input into the fourth learning model, the physical properties of the fluid are estimated, and the calculation device calculates the flow rate of the fluid based on the estimated value of the opening area and the estimation result of the physical properties, estimated by the estimation device.

[0015] According to the plant abnormality monitoring system of the present invention, based on the difference value between the estimated value of the fluid flow rate estimated by the estimation device and the calculated value of the fluid flow rate calculated by the calculation device using the estimated value of the opening area estimated by the estimation device and the estimation result of the physical properties, it is possible to further improve the estimation accuracy of the first learning model.

[0016] Another one of the present invention for achieving the above object is a plant abnormality monitoring system, wherein the imaging device is a device that captures at least one of a moving image, a still image, and a thermal image.

[0017] According to the plant abnormality monitoring system of the present invention, since the images included in the learning data for which the first learning model performs learning may be any of a moving image, a still image, and a thermal image, it is possible to improve the estimation accuracy of the first learning model.

[0018] Another one of the present invention for achieving the above object is a plant abnormality monitoring system, wherein the facility in the plant can be a pipe through which the fluid flows in the reactor containment vessel of a nuclear power plant.

[0019] Another one of the present invention for achieving the above object is a plant abnormality monitoring system, wherein the fluid is a liquid or a gas.

[0020] In addition, the problems disclosed in the present application and the solutions thereto will be clarified by the section of the mode for carrying out the invention and the drawings.

Advantages of the Invention

[0021] According to the present invention, when an abnormality occurs in which a fluid leaks from equipment in a plant, the flow rate of the fluid at the location where the fluid leaks is estimated, and based on the estimated value of the flow rate of the fluid, it is possible to provide support information for quickly repairing the equipment in the plant.

Brief Description of the Drawings

[0022]

Figure 1

Figure 2A

Figure 2B

Figure 2C

Figure 2D

Figure 3A

Figure 3B

Figure 3C

Figure 4

Figure 5

Figure 6

Figure 7A

Figure 7B

Figure 7C

Figure 7D

Figure 7E

Mode for Carrying Out the Invention

[0023] From the description in this specification and the accompanying drawings, at least the following matters become clear. Hereinafter, the present invention will be described with reference to the accompanying drawings according to one embodiment thereof.

[0024] FIG. 1 is a block diagram showing a schematic configuration of the plant abnormality monitoring system 1. In this embodiment, it is assumed that the plant is a nuclear power plant, for example.

[0025] The plant abnormality monitoring system 1 is a system that monitors, within a nuclear power plant, for example, in a reactor containment vessel where a reactor pressure vessel is housed, whether fluid (liquid, gas, etc.) is leaking from various facilities (equipment, piping, etc.) necessary for stably and continuously maintaining a nuclear fission reaction within the reactor pressure vessel due to usage conditions, aging deterioration, etc. When it is determined that the facility is in an abnormal state where fluid is leaking, the system outputs support information for quickly repairing the location where the fluid is leaking.

[0026] The plant abnormality monitoring system 1 includes an information processing device 100, an imaging device 200, and an external computer 300 as means for realizing the above functions. The information processing device 100, the imaging device 200, and the external computer 300 are connected via a communication network 400 in a state where two-way communication is possible. Note that the communication network 400 is, for example, a LAN (Local Area Network), a WAN (Wide Area Network), a dedicated line, a power line communication network, various public communication networks, etc. The information processing device 100 is a computer and is installed, for example, in a control room within the nuclear power plant or a control room of an electric power company outside the nuclear power plant.

[0027] The imaging device 200 is a device having the functions of both a first camera device 210 that captures at least one of a moving image and a still image of the facilities within the reactor containment vessel and the fluid leaking from the facilities, and a second camera device (thermographic camera device) 220 that visualizes the temperature distribution of the facilities within the reactor containment vessel and the fluid leaking from the facilities as a thermal image. The imaging device 200 includes an imaging device 200A fixedly installed at a predetermined position within the reactor containment vessel and an imaging device 200B installed so as to be able to fly within the reactor containment vessel.

[0028] The imaging device 200A is installed, for example, on a base body having a panning function in the horizontal and vertical directions. The imaging device 200A has a zoom function capable of performing both wide-angle shooting and narrow-angle shooting, and is positioned within the reactor containment vessel such that it can perform fixed-point shooting of the entire facility through which fluid flows at a wide angle, and, using the panning function of the base body, can also perform shooting of the fluid leakage location in the facility through which fluid flows at a narrow angle. If it is difficult for a single imaging device 200A to perform fixed-point shooting of the entire facility through which fluid flows at a wide angle, multiple imaging devices 200A may be installed within the reactor containment vessel, and the multiple imaging devices 200A may be used to perform fixed-point shooting of the entire facility through which fluid flows in a shared manner.

[0029] Within the reactor containment vessel, for example, an unmanned aerial vehicle (drone) capable of flying near the fluid leakage location in the facility through which fluid flows by remote control is accommodated. The unmanned aerial vehicle, for example, waits at a position that serves as a base point (origin) within the reactor containment vessel, and upon receiving a flight instruction, flies from the standby position toward the vicinity of the fluid leakage location in the facility through which fluid flows according to the coordinates set within the reactor containment vessel. The imaging device 200B is integrated with this unmanned aerial vehicle, and the unmanned aerial vehicle hovers near the fluid leakage location to capture an image of the fluid leakage location. The imaging device 200B has a zoom function similar to that of the imaging device 200A, and captures an image of the fluid leakage location at an optimal angle according to the distance from the unmanned aerial vehicle to the fluid leakage location. If it is assumed that fluid leaks from multiple locations in the facility, multiple unmanned aerial vehicles and multiple imaging devices 200B integrated with the multiple unmanned aerial vehicles may be accommodated within the reactor containment vessel.

[0030] In this embodiment, it is assumed that the imaging devices 200A and 200B each include both the first camera device 210 and the second camera device 220, but there may be cases where only one of the first camera device 210 and the second camera device 220 is included.

[0031] The information processing apparatus 100 includes an input unit 110 (input device), an output unit 120 (output device), an estimation unit 130 (estimation device), a calculation unit 140 (calculation device), a storage unit 150 (storage device), and a monitoring unit 160.

[0032] The storage unit 150 stores a control program 151, learning data 152, and a learning model 153.

[0033] The control program 151 is a program executed by the information processing apparatus 100 when the plant abnormality monitoring system 1 operates, that is, when the plant abnormality monitoring system 1 monitors whether fluid is leaking from the equipment in the reactor containment vessel and determines that fluid is leaking from the equipment, and executes a series of operations to output information for quickly repairing the fluid leakage location.

[0034] The learning model 153 includes a first learning model 153A, a second learning model 153B, a third learning model 153C, and a fourth learning model 153D.

[0035] The learning data 152 includes first learning data 152A, second learning data 152B, third learning data 152C, and fourth learning data 152D. Note that the first to fourth learning data 152A to 152D are supervised data in which the feature amounts and labels used when the first to fourth learning models 153A to 153D perform learning are associated with each other.

[0036] FIG. 2A is a diagram showing an example of the feature amounts and labels constituting the first learning data 152A. FIG. 2B is a diagram showing an example of the feature amounts and labels constituting the second learning data 152B. FIG. 2C is a diagram showing an example of the feature amounts and labels constituting the third learning data 152C. FIG. 2D is a diagram showing an example of the feature amounts and labels constituting the fourth learning data 152D. In this embodiment, it is assumed that the feature amounts in the first to fourth learning data 152A to 152D are the same data.

[0037] First, the features of the first learning data 152A include data indicating (A1) a still image of the fluid, (A2) a moving image of the fluid, and (A3) a thermal image of the fluid. Here, (A1) the still image of the fluid is an image without movement indicating that the fluid is leaking from the facility. (A2) The moving image of the fluid is an image with movement over a predetermined period indicating that the fluid is leaking from the facility, and may include images before and after the still image of the fluid. The still image and the moving image of the fluid may be, for example, images taken by the imaging device 200B of the fluid leakage location in the past, or images taken in the past of the fluid leakage location in the same or similar reactor containment vessel of another nuclear power plant, or various types of image data such as images obtained in experiments in research, etc., or images in plants other than nuclear power plants. (A3) The thermal image of the fluid is an image visualizing the temperature distribution of the fluid and is an image taken together with the still image and the moving image of the fluid. On the other hand, the label of the first learning data 152A includes data indicating (A4) the flow rate of the fluid. The flow rate of the fluid is a measured value. The first learning data 152A is data associating the data indicating (A1) the still image of the fluid, (A2) the moving image of the fluid, and (A3) the thermal image of the fluid with the data indicating (A4) the flow rate of the fluid, and is stored in the storage unit 150 as data for the first learning model 153A to perform learning.

[0038] Next, the features of the second training data 152B include data indicating (B1) a still image of the fluid, (B2) a moving image of the fluid, and (B3) a thermal image of the fluid. The data indicating (B1) the still image of the fluid, (B2) the moving image of the fluid, and (B3) the thermal image of the fluid shall be the same as the data indicating (A1) the still image of the fluid, (A2) the moving image of the fluid, and (A3) the thermal image of the fluid, but may be data indicating images taken at different timings from (A1) the still image of the fluid, (A2) the moving image of the fluid, and (A3) the thermal image of the fluid. On the other hand, the label of the second training data 152B includes data indicating (B4) the opening area of the leakage location of the fluid. The opening area of the leakage location of the fluid is a measured value. The second training data 152B is data in which the data indicating these (B1) the still image of the fluid, (B2) the moving image of the fluid, and (B3) the thermal image of the fluid are associated with the data indicating (B4) the opening area of the outflow location of the fluid, and is stored in the storage unit 150 as data for the second training model 153B to perform learning.

[0039] Next, the features of the third training data 152C include data indicating (C1) a still image of the fluid, (C2) a moving image of the fluid, and (C3) a thermal image of the fluid. The data indicating (C1) a still image of the fluid, (C2) a moving image of the fluid, and (C3) a thermal image of the fluid shall be the same as the data indicating (A1) a still image of the fluid, (A2) a moving image of the fluid, and (A3) a thermal image of the fluid, but may be data indicating images taken at different timings from (A1) a still image of the fluid, (A2) a moving image of the fluid, and (A3) a thermal image of the fluid. On the other hand, the label of the third training data 152C includes data indicating (C4) the equipment system of the fluid leakage location. The equipment system of the fluid leakage location refers to the system to which the equipment location where the fluid is leaking corresponds among the entire equipment system installed in the reactor containment vessel. The equipment system is attached with colors, numbers, symbols, etc. for distinguishing each system. The storage unit 150 stores in advance, as table data 154, data related to all equipment systems. The data related to the equipment system refers to data indicating the temperature, pressure, etc. inside the equipment system in a normal state before the fluid leaks. The third training data 152C is data in which the data indicating these (C1) a still image of the fluid, (C2) a moving image of the fluid, and (C3) a thermal image of the fluid are associated with the data indicating (C4) the equipment system of the fluid leakage location, and is stored in the storage unit 150 as data for the third training model 153C to perform learning.

[0040] Next, the features of the fourth learning data 152D include data indicating (D1) a still image of the fluid, (D2) a moving image of the fluid, and (D3) a thermal image of the fluid. The data indicating (D1) a still image of the fluid, (D2) a moving image of the fluid, and (D3) a thermal image of the fluid shall be the same as the data indicating (A1) a still image of the fluid, (A2) a moving image of the fluid, and (A3) a thermal image of the fluid, but may be data indicating images taken at different timings from (A1) a still image of the fluid, (A2) a moving image of the fluid, and (A3) a thermal image of the fluid. On the other hand, the label of the fourth learning data 152D includes data indicating (D4) the physical properties of the leaking fluid. The physical properties of the leaking fluid refer to any liquid such as pure water, lubricating oil, chemicals, etc., or any gas such as water vapor, hydrogen, nitrogen, etc. The fourth learning data 152D is data associating the data indicating these (D1) a still image of the fluid, (D2) a moving image of the fluid, and (D3) a thermal image of the fluid with the data indicating (D4) the physical properties of the leaking fluid, and is stored in the storage unit 150 as data for the fourth learning model 153D to perform learning.

[0041] The leaking fluid is either a liquid or a gas. When the leaking fluid is a liquid, examples of this liquid include high-temperature and radioactive pure water injected into the reactor, normal-temperature and non-radioactive pure water injected into the reactor, low-temperature and non-radioactive pure water used to cool the equipment, lubricating oil used for the equipment, chemicals used to control the reactivity of the reactor, etc. On the other hand, when the leaking fluid is a gas, examples of this gas include water vapor, nitrogen, hydrogen, etc. Thus, there are various liquids and gases as the leaking fluid, and since the temperature of the fluid and the leaking tendency also vary, it is desirable to prepare a large number of first to fourth learning data 152A to 152D covering various situations.

[0042] FIG. 3A is a diagram showing an example of a state in which an opening is formed in a part of a pipe, which is one of the facilities inside the reactor containment vessel, and liquid leaks out as a fluid from this opening with a predetermined momentum. FIG. 3B is a diagram showing an example of a state in which an opening is formed in a part of a pipe, which is one of the facilities inside the reactor containment vessel, and liquid leaks out as a fluid from this opening with a momentum weaker than that in FIG. 3A. FIG. 3C is a diagram showing an example of a state in which an opening is formed in a part of a pipe, which is one of the facilities inside the reactor containment vessel, and liquid leaks out as a fluid from this opening with a momentum weaker than that in FIG. 3B and leaks downward along the outer peripheral surface of the pipe. When the fluid is liquid, the feature amounts of the first to fourth learning data 152A to 152D include data showing various still images, moving images, and thermal images obtained by visualizing the temperature distributions of still images and moving images, which have different states as shown in FIGS. 3A to 3C. On the other hand, when the fluid is gas, the feature amounts of the first to fourth learning data 152A to 152D will include data showing still images, moving images, and thermal images similar to FIGS. 3A to 3D. However, when the fluid cannot be visually observed, the data showing the thermal image becomes effective data as a feature amount.

[0043] The first learning model 153A is a machine learning model formed to estimate the flow rate of a fluid at the fluid leakage location when new data related to the same type of fluid as the data showing (A1) a still image of the fluid, (A2) a moving image of the fluid, and (A3) a thermal image of the fluid, which are the feature amounts of the first learning data 152A, is input when an abnormality occurs in which a fluid leaks from any of the facilities inside the reactor containment vessel by performing learning using the first learning data 152A. That is, the first learning model 153A estimates the flow rate of the fluid at the fluid leakage location according to the similarity between the new data related to the fluid and the feature amounts (A1) to (A3) related to the fluid.

[0044] [[ID=z]] The second learning model 153B is a machine learning model formed by performing learning using the second learning data 152B, and when an abnormality occurs in which fluid leaks from any facility inside the reactor containment vessel, new data related to the fluid that is of the same type as the data showing (B1) a still image of the fluid, (B2) a moving image of the fluid, and (B3) a thermal image of the fluid, which are the feature amounts of the second learning data 152B, is input, it estimates the opening area of the leakage location of the fluid in the facility. That is, the second learning model 153B estimates the opening area of the leakage location of the fluid in the facility according to the similarity between the new data related to the fluid and the above (B1) to (B3) feature amounts related to the fluid.

[0045] The third learning model 153C is a machine learning model formed by performing learning using the third learning data 152C, and when an abnormality occurs in which fluid leaks from any facility inside the reactor containment vessel, new data related to the fluid that is of the same type as the data showing (C1) a still image of the fluid, (C2) a moving image of the fluid, and (C3) a thermal image of the fluid, which are the feature amounts of the third learning data 152C, is input, it estimates the facility system to which the leakage location of the fluid corresponds. That is, the third learning model 153C estimates the facility system to which the leakage location of the fluid corresponds according to the similarity between the new data related to the fluid and the above (C1) to (C3) feature amounts related to the fluid. Incidentally, since the third learning model 153C estimates the facility system from the color, number, symbol, etc. attached to the facility from which the fluid leaks, it is possible to acquire data such as temperature and pressure within the estimated facility system from the table data 154 stored in the storage unit 150.

[0046] The fourth learning model 153D is a machine learning model formed to estimate the physical properties of a leaking fluid when an abnormality occurs in which fluid leaks from any equipment inside the reactor containment vessel by performing learning using the fourth learning data 152D. That is, when new data related to a fluid of the same type as the data indicating (D1) a still image of the fluid, (D2) a moving image of the fluid, and (D3) a thermal image of the fluid, which are the feature amounts of the fourth learning data 152D, is input, the fourth learning model 153D estimates the physical properties of the leaking fluid. In other words, the fourth learning model 153D estimates the physical properties of the leaking fluid according to the similarity between the new data related to the fluid and the feature amounts (D1) to (D3) related to the fluid.

[0047] The first to fourth learning models 153A to 153D can make more accurate predictions as they perform learning using more first to fourth learning data 152A to 152D. In the present embodiment, the first to fourth learning models 153A to 153D are, for example, deep neural networks (DNNs), but other types of models such as gradient boosting decision trees (GBDTs) may also be used.

[0048] FIG. 4 is a diagram showing the structure of a deep neural network which is an example of the learning model 153. The first to fourth learning models 153A to 153D each include three layers: an input layer 1531, an intermediate layer 1532, and an output layer 1533.

[0049] In the first learning model 153A, the input layer 1531 is a layer into which data related to a fluid of the same type as the feature amounts (A1) to (A3) of the first learning data 152A is input. The intermediate layer 1532 is a layer including one or more hidden layers composed of one or more nodes including parameters adjusted by performing learning using the first learning data 152A. The intermediate layer 1532 estimates the flow rate of the fluid at the fluid leakage location based on the data input to the input layer 1531. The output layer 1533 is a layer that outputs the estimation result of the flow rate of the fluid at the fluid leakage location by the intermediate layer 1532.

[0050] In the second learning model 153B, the input layer 1531 is a layer into which data related to a fluid of the same type as the feature quantities (B1) to (B3) of the second learning data 152B is input. The intermediate layer 1532 is a layer including one or more hidden layers composed of one or more nodes including parameters adjusted by performing learning using the second learning data 152B. The intermediate layer 1532 estimates the opening area of the leakage location of the fluid in the facility based on the data input to the input layer 1531. The output layer 1533 is a layer that outputs the estimation result of the opening area of the leakage location of the fluid by the intermediate layer 1532.

[0051] In the third learning model 153C, the input layer 1531 is a layer into which data related to a fluid of the same type as the feature quantities (C1) to (C3) of the third learning data 152C is input. The intermediate layer 1532 is a layer including one or more hidden layers composed of one or more nodes including parameters adjusted by performing learning using the third learning data 152C. The intermediate layer 1532 estimates the facility system to which the leakage location of the fluid corresponds based on the data input to the input layer 1531. The output layer 1533 is a layer that outputs the estimation result of the facility system to which the leakage location of the fluid corresponds by the intermediate layer 1532.

[0052] In the fourth learning model 153D, the input layer 1531 is a layer into which data related to a fluid of the same type as the feature quantities (D1) to (D3) of the fourth learning data 152D is input. The intermediate layer 1532 is a layer including one or more hidden layers composed of one or more nodes including parameters adjusted by performing learning using the fourth learning data 152D. The intermediate layer 1532 estimates the physical properties of the leaking fluid based on the data input to the input layer 1531. The output layer 1533 is a layer that outputs the estimation result of the physical properties of the fluid at the leakage location of the fluid by the intermediate layer 1532.

[0053] Furthermore, in the present embodiment, since the feature amounts (A1) to (A3) of the first learning data 152A, the feature amounts (B1) to (B3) of the second learning data 152B, the feature amounts (C1) to (C3) of the third learning data 152C, and the feature amounts (D1) to (D3) of the fourth learning data 152D are the same, for example, if data related to the same type of fluid as the feature amounts (A1) to (A3) of the first learning data 152A is input to the input layers 1531 of the first to fourth learning models 153A to 153D, the estimation processing by the first to fourth learning models 153A to 153D can be executed.

[0054] Returning to FIG. 1, the input unit 110 acquires data indicating still images, moving images, and thermal images inside the reactor containment vessel captured by the imaging devices 200A and 200B. For example, when the plant abnormality monitoring system 1 monitors whether fluid is leaking from the equipment inside the reactor containment vessel, the input unit 110 acquires data indicating still images, moving images, and thermal images that show the entire equipment inside the reactor containment vessel from the imaging device 200A. On the other hand, when the plant abnormality monitoring system 1 determines that fluid is leaking from a certain piece of equipment inside the reactor containment vessel based on the data indicating still images, moving images, and thermal images that show the entire equipment inside the reactor containment vessel acquired by the input unit 110 from the imaging device 200A, the input unit 110 acquires data indicating still images, moving images, and thermal images captured by the imaging device 200B moving close to the fluid leakage location of the equipment according to an instruction from the information processing device 100. Note that the still images are images captured by the imaging devices 200A and 200B at regular intervals (e.g., every 10 seconds, 30 seconds, 1 minute, etc.).

[0055] A dosimeter 510 for measuring the radiation dose inside the reactor containment vessel is provided inside the reactor containment vessel. When the plant abnormality monitoring system 1 determines that fluid is leaking from the equipment inside the reactor containment vessel, the input unit 110 acquires data indicating the radiation dose measured by the dosimeter 510 and the temperature inside the reactor containment vessel determined from the thermal image captured by the second camera device 220.

[0056] The monitoring unit 160 monitors whether fluid is leaking from the equipment inside the reactor containment vessel based on data indicating still images, moving images, and thermal images of the entire equipment inside the reactor containment vessel obtained by the input unit 110 from the imaging device 200A. When the image showing the entire equipment inside the reactor containment vessel is a still image, the monitoring unit 160 compares two sets of still images and thermal images having a certain time interval, and if there is a difference between the two still images and thermal images, it tentatively determines that fluid is leaking from some equipment inside the reactor containment vessel. On the other hand, when the image showing the entire equipment inside the reactor containment vessel is a moving image, the monitoring unit 160 compares two sets of moving images and thermal images divided at a certain time interval, and if there is a difference between the two moving images and thermal images, it tentatively determines that fluid is leaking from some equipment inside the reactor containment vessel. When the monitoring unit 160 determines that fluid is leaking from some equipment inside the reactor containment vessel, the input unit 110 stops acquiring still images, moving images, and thermal images from the imaging device 200A, and acquires still images, moving images, and thermal images from the imaging device 200B.

[0057] The estimation unit 130 inputs data indicating still images, moving images, and thermal images of the fluid leakage location inside the reactor containment vessel obtained by the input unit 110 from the imaging device 200B to the input layer 1531 of the first to fourth learning models 153A to 153D. The first learning model 153A is instructed to estimate the flow rate of the fluid at the fluid leakage location, the second learning model 153B is instructed to estimate the opening area of the equipment at the fluid leakage location, the third learning model 153C is instructed to estimate the equipment system including the fluid leakage location, and the fourth learning model 153D is instructed to estimate the physical properties of the leaking fluid. The data indicating the estimation results by the first to fourth learning models 153A to 153D is stored in the storage unit 150.

[0058] The calculation unit 140 calculates the flow rate of the fluid at the fluid leakage location and the opening area of the equipment at the fluid leakage location using the following calculation formulas for the cases where the fluid is a liquid and a gas, respectively.

[0059] In the calculation unit 140, when the second learning model 153B and the third learning model 153C perform estimation with a certain accuracy, a calculation formula using, as parameters, the estimation result of the opening area of the facility at the fluid leakage location and the estimation result of the facility system including the fluid leakage location (formula (1) when the fluid is a liquid, formula (2) when the fluid is a gas and steam, formula (3) when the fluid is a gas, and formula (4) when the fluid is a gas such as air) is used to calculate the flow rate of the fluid at the fluid leakage location.

[0060]

Number

[0061]

Number

[0062]

Number

[0063]

Number

[0064] Further, in the calculation unit 140, when the second learning model 153B and the fourth learning model 153D perform estimation with a certain accuracy, using the estimation result of the opening area of the facility at the fluid leakage location and the estimation result of the physical properties of the leaking fluid as parameters, when the fluid is a liquid, the calculation formula (1) is used, and when the fluid is a gas, the calculation formulas (2), (3), and (4) are used to calculate the fluid flow rate at the fluid leakage location.

[0065] Further, in the calculation unit 140, when the first learning model 153A and the third learning model 153C perform estimation with a certain accuracy, using the estimation result of the fluid flow rate at the fluid leakage location and the estimation result of the facility system including the fluid leakage location as parameters, when the fluid is a liquid, the calculation formula (1) is used, and when the fluid is a gas, the calculation formulas (2), (3), and (4) are used to calculate the opening area of the facility at the fluid leakage location.

[0066] Further, in the calculation unit 140, when the first learning model 153A and the fourth learning model 153D perform estimation with a certain accuracy, using the estimation result of the fluid flow rate at the fluid leakage location and the estimation result of the physical properties of the leaking fluid as parameters, when the fluid is a liquid, the calculation formula (1) is used, and when the fluid is a gas, the calculation formulas (2), (3), and (4) are used to calculate the opening area of the facility at the fluid leakage location.

[0067] Data indicating the calculation result by the calculation unit 140 is stored in the storage unit 150.

[0068] The output unit 120 outputs externally the data indicating the estimation results by the first to fourth learning models 153A to 153D stored in the storage unit 150 and the calculation result by the calculation unit 140. As a method for the output unit 120 to output the data indicating the estimation results by the first to fourth learning models 153A to 153D stored in the storage unit 150 and the calculation result by the calculation unit 140, for example, a method of displaying and outputting on a display panel, a method of outputting sound from a speaker, a method of printing and outputting on a paper medium, etc. can be considered.

[0069] The external computer 300 has a storage unit 310, stores data indicating the estimation results by the first to fourth learning models 153A to 153D stored in the storage unit 150 and the calculation results by the calculation unit 140 in the storage unit 310 in parallel with the storage unit 150, and reads out these data from the storage unit 310 as necessary and outputs them to the information processing apparatus 100.

[0070] FIG. 5 is a block diagram showing an example of the hardware of the information processing apparatus 100 used for realizing the plant abnormality monitoring system 1.

[0071] The information processing apparatus 100 includes a processor 1010, a main storage device 1020, an auxiliary storage device 1030, an input device 1040, an output device 1050, and a communication device 1060. The information processing apparatus 100 is, for example, a personal computer, an office computer, various server devices, a general-purpose machine, or the like. The information processing apparatus 100 may be realized using virtual information processing resources provided using virtualization technology, such as a virtual server provided by a cloud system, for all or part of it. The plant abnormality monitoring system 1 may be realized using a plurality of information processing apparatuses 100 connected communicably.

[0072] The processor 1010 is configured using, for example, a CPU (Central Processing Unit), an MPU (Micro Processing Unit), a GPU (Graphics Processing Unit), an FPGA (Field Programmable Gate Array), an ASIC (Application Specific Integrated Circuit), an AI (Artificial Intelligence) chip, or the like.

[0073] The main storage device 1020 is a device that stores programs and data, and is, for example, a ROM (Read Only Memory), a RAM (Random Access Memory), a non-volatile memory (NVRAM (Non Volatile RAM)), or the like.

[0074] The auxiliary storage device 1030 is, for example, an SSD (Solid State Drive), a hard disk drive, an optical storage device (such as a CD (Compact Disc), a DVD (Digital Versatile Disc), etc.), a storage system, an IC card, an SD card, a reading / writing device for a recording medium such as an optical recording medium, a storage area of a cloud server, etc. Programs and data can be read into the auxiliary storage device 1030 via a reading device for a recording medium or a communication device 2060. Programs and data stored in the auxiliary storage device 1030 are read into the main storage device 2020 at any time.

[0075] The input device 1040 is an interface for receiving external inputs, and is, for example, a keyboard, a mouse, a touch panel, a card reader, a pen-input type tablet, a voice input device, etc.

[0076] The output device 1050 is an interface for outputting various information such as the progress of processing and the results of processing. The output device 1050 is, for example, a display device (such as an LCD (Liquid Crystal Display), a graphic card, etc.) for visualizing the above various information, a device for vocalizing the above various information (a voice output device (such as a speaker)), a device for characterizing the above various information (a printing device, etc.). Note that the information processing device 100 may be configured to input and output information to and from other devices via a communication device 1060.

[0077] The input device 1040 and the output device 1050 constitute a user interface for receiving information from and presenting information to the user.

[0078] The communication device 1060 is a device that realizes communication (wired communication or wireless communication) with other devices via a communication infrastructure such as a communication network 400, and is configured using, for example, a NIC (Network Interface Card), a wireless communication module, a USB module, etc.

[0079] The information processing apparatus 100 may also be installed with, for example, an operating system, a file system, a DBMS (Data Base Management System) (relational database, NoSQL, etc.), a KVS (Key-Value Store), and the like.

[0080] The functions provided by the plant abnormality monitoring system 1 are realized by the processor 1010 of the information processing apparatus 100 executing a program (control program) read into the main storage device 1020, or are realized by the functions of the hardware (FPGA, ASIC, AI chip, etc.) itself that constitutes the plant abnormality monitoring system 1. For example, the storage unit 150 in FIG. 1 is realized by the main storage device 1020 and the auxiliary storage device 1030, the input unit 110 in FIG. 1 is realized by the input device 1040, the monitoring unit 160, the estimation unit 130, and the calculation unit 140 in FIG. 1 are realized by the processor 1010, and the output unit 120 in FIG. 1 is realized by the output device 1050 and the communication device 1060.

[0081] FIG. 6 is a flowchart showing the processing when the information processing apparatus 100 performs learning of the first to fourth learning models 153A to 153D using the first to fourth learning data 152A to 152D.

[0082] First, the information processing apparatus 100 generates the first learning data 152A in which the feature amounts (A1) to (A3) are associated with the label (A4), the second learning data 152B in which the feature amounts (B1) to (B3) are associated with the label (B4), the third learning data 152C in which the feature amounts (C1) to (C3) are associated with the label (C4), and the fourth learning data 152D in which the feature amounts (D1) to (D3) are associated with the label (D4) (S1000).

[0083] Next, the information processing apparatus 100 performs learning of the first learning model 153A using the first learning data 152A, performs learning of the second learning model 153B using the second learning data 152B, performs learning of the third learning model 153C using the third learning data 152C, and performs learning of the fourth learning model 153D using the fourth learning data 152D (S1010). Note that the information processing apparatus 100 may verify the prediction accuracy of the learned first to fourth learning models 153A to 153D. In that case, the information processing apparatus 100 prepares learning data and verification data as the first to fourth learning data 152A to 152D, performs learning of the first to fourth learning models 153A to 153D using the learning data, and performs verification of the first to fourth learning models 153A to 153D using the verification data.

[0084] FIG. 7A is an example of the processing of the information processing apparatus 100 when the plant abnormality monitoring system 1 monitors whether there is a fluid leakage from the equipment inside the reactor containment vessel, and is a flowchart showing a part of the processing. FIGS. 7B to 7E are examples of the processing of the information processing apparatus 100 when the plant abnormality monitoring system 1 monitors whether there is a fluid leakage from the equipment inside the reactor containment vessel, and are flowcharts showing other parts of the processing. Note that the estimation accuracies of the first to fourth learning models 153A to 153D differ depending on the quality and quantity of the first to fourth learning data 152A to 152D, and there may be a case where a learning model whose estimation accuracy has not reached a certain accuracy exists among the first to fourth learning models 153A to 153D. Therefore, in this flowchart, it is assumed that the estimation results of the learning models whose estimation accuracy has not reached a certain accuracy are not used, and for convenience of explanation, an explanation including both the case where the respective estimation accuracies of the first to fourth learning models 153A to 153D have reached a certain accuracy and the case where the respective estimation accuracies of the first to fourth learning models 153A to 153D have not reached a certain accuracy will be given. The monitor verifies in advance the estimation accuracies of the first to fourth learning models 153A to 153D, and does not use the estimation results of the learning models determined to have an estimation accuracy not reaching a certain accuracy, and operates the plant abnormality monitoring system 1. For example, a flag or the like for distinguishing whether the first to fourth learning models 153A to 153D have an estimation accuracy equal to or higher than a certain accuracy is attached, and the plant abnormality monitoring system 1 uses the learning models having an estimation accuracy equal to or higher than a certain accuracy by checking this flag.

[0085] First, the monitoring unit 160 continuously acquires still images, moving images, and thermal images of the entire facility inside the reactor containment vessel from the imaging device 200A via the input unit 110, detects whether there are differences between two sets of still images, moving images, and thermal images at regular time intervals, and determines whether fluid is leaking from any of the facilities inside the reactor containment vessel (S2010). If no differences are detected between two sets of still images, moving images, and thermal images at regular time intervals and it is determined that no fluid is leaking from any of the facilities inside the reactor containment vessel (S2010: NO), the monitoring unit 160 repeatedly executes the process of step S2010. Note that the monitoring unit 160 may also detect whether there are differences between the measured values and steady-state values of various plant parameters inside the reactor containment vessel, or detect whether an alarm indicating an abnormality inside the reactor containment vessel has occurred, and determine whether fluid is leaking from any of the facilities inside the reactor containment vessel.

[0086] Next, based on the information indicating the differences between two sets of still images, moving images, and thermal images at regular time intervals, the monitoring unit 160 specifies the abnormal location, which is the fluid leakage location, in coordinates or the like set inside the reactor containment vessel (S2020). Thereby, the imaging device 200B moves from the standby position to the set coordinate position and captures the abnormal location.

[0087] Next, the input unit 110 acquires data indicating the radiation dose measured by the dosimeter 510 and the temperature information obtained from the thermal image captured by the second camera device 220 of the imaging device 200A as information supplementing the state where fluid is leaking, and the storage unit 150 stores these data (S2030).

[0088] Next, the estimation unit 130 acquires data indicating still images, moving images, and thermal images from which the abnormal location can be known from the imaging device 200B via the input unit 110, and inputs these data to the input layer 1531 of the first to fourth learning models 153A to 153D (S2040).

[0089] <<Processing When It Is Determined from a Visible Image of a Still Image or a Moving Image by the Imaging Device 200B that the Fluid Is a Liquid (hereinafter Referred to as Liquid FL)>> When the estimation accuracy of the first learning model 153A is equal to or higher than a certain accuracy (S2050: YES), the first learning model 153A estimates the flow rate of the liquid FL at the leakage location of the liquid FL based on the data indicating the still image, moving image, and thermal image where the abnormal location is known (S2060).

[0090] Next, when the estimation accuracy of the second learning model 153B is equal to or higher than a certain accuracy (S2070: YES), the second learning model 153B estimates the opening area of the facility at the leakage location of the liquid FL based on the data indicating the still image, moving image, and thermal image where the abnormal location is known (S2080).

[0091] Next, when the estimation accuracy of the third learning model 153C is equal to or higher than a certain accuracy (S2090: YES), the third learning model 153C estimates the facility system including the leakage location of the liquid FL based on the data indicating the still image, moving image, and thermal image where the abnormal location is known (S2100).

[0092] Next, the calculation unit 140 uses the calculation formula (1) with the estimated value of the opening area of the facility at the leakage location of the liquid FL estimated by the second learning model 153B in step S2080 and the estimation result of the facility system including the leakage location of the liquid FL estimated by the third learning model 153C in step S2100 as parameters to calculate the flow rate of the liquid FL at the leakage location of the liquid FL (S2110).

[0093] Next, the output unit 120 outputs the difference value between the estimated value of the flow rate of the liquid FL at the leakage location of the liquid FL estimated by the first learning model 153A in step S2060, the estimated value of the opening area of the facility at the leakage location of the liquid FL estimated by the second learning model 153B in step S2080, the estimated value of the flow rate of the liquid FL, and the calculated value of the flow rate of the liquid FL calculated from the calculation formula (1) (S2120).

[0094] Also, when the estimation accuracy of the third learning model 153C is less than a certain accuracy (S2090: NO) and the estimation accuracy of the fourth learning model 153D is equal to or higher than the certain accuracy (S2130: YES), the fourth learning model 153D estimates the physical properties of the leaked liquid FL (S2140).

[0095] Next, the calculation unit 140 calculates the flow rate of the liquid FL at the leakage location of the liquid FL by using the calculation formula (1) with the estimated value of the opening area of the facility at the leakage location of the liquid FL estimated by the second learning model 153B in step S2080 and the estimation result of the physical properties of the liquid FL estimated by the fourth learning model 153D in step S2140 as parameters (S2150).

[0096] Next, the output unit 120 outputs the difference value between the estimated value of the flow rate of the liquid FL at the leakage location of the liquid FL estimated by the first learning model 153A in step S2060, the estimated value of the opening area of the facility at the leakage location of the liquid FL estimated by the second learning model 153B in step S2080, the estimated value of the flow rate of the liquid FL, and the calculated value of the flow rate of the liquid FL calculated from the calculation formula (1) (S2160).

[0097] Also, when the estimation accuracy of the fourth learning model 153D is less than a certain accuracy (S2130: NO), the output unit 120 outputs the estimated value of the flow rate of the liquid FL at the leakage location of the liquid FL estimated by the first learning model 153A in step S2060 and the estimated value of the opening area of the facility at the leakage location of the liquid FL estimated by the second learning model 153B in step S2080 (S2170).

[0098] Also, when the estimation accuracy of the second learning model 153B is less than a certain accuracy (S2070: NO) and the estimation accuracy of the third learning model 153C is equal to or higher than the certain accuracy (S2180: YES), the third learning model 153C estimates the facility system including the leakage location of the liquid FL (S2190).

[0099] Next, the calculation unit 140 calculates the opening area of the facility at the liquid FL leakage location using the calculation formula (1) with the estimated value of the flow rate of the liquid FL at the liquid FL leakage location estimated by the first learning model 153A in step S2060 and the estimation result of the facility system including the liquid FL leakage location estimated by the third learning model 153C in step S2190 as parameters (S2200).

[0100] Next, the output unit 120 outputs the estimated value of the flow rate of the liquid FL at the liquid FL leakage location estimated by the first learning model 153A in step S2060 and the calculated value of the opening area of the facility calculated from the calculation formula (1) (S2210).

[0101] Also, when the estimation accuracy of the third learning model 153C is less than a certain accuracy (S2180: NO) and the estimation accuracy of the fourth learning model 153D is equal to or higher than the certain accuracy (S2220: YES), the fourth learning model 153D estimates the physical properties of the leaking liquid FL (S2230).

[0102] Next, the calculation unit 140 calculates the opening area of the facility at the liquid FL leakage location using the calculation formula (1) with the estimated value of the flow rate of the liquid FL at the liquid FL leakage location estimated by the first learning model 153A in step S2060 and the estimation result of the physical properties of the leaking liquid FL estimated by the fourth learning model 153D in step S2230 as parameters (S2240).

[0103] Next, the output unit 120 outputs the estimated value of the flow rate of the liquid FL at the liquid FL leakage location estimated by the first learning model 153A in step S2060 and the calculated value of the opening area of the facility calculated from the calculation formula (1) (S2250).

[0104] Also, when the estimation accuracy of the fourth learning model 153D is less than the certain accuracy (S2220: NO), the output unit 120 outputs the estimated value of the flow rate of the liquid FL at the liquid FL leakage location estimated by the first learning model 153A in step S2060 (S2260).

[0105] <<Processing When It Is Determined from the Visible Image of a Still Image or Moving Image by Imaging Device 200B that the Fluid Is Gas (hereinafter Referred to as Gas FG) Instead of Liquid FL>> When the estimation accuracy of the first learning model 153A is equal to or higher than a certain accuracy (S2050: YES), the first learning model 153A estimates the flow rate of the gas FG at the gas FG leakage location based on the data indicating the change in the temperature distribution of the thermal image where the abnormal location is known (S2270).

[0106] Next, when the estimation accuracy of the second learning model 153B is equal to or higher than a certain accuracy (S2280: YES), the second learning model 153B estimates the opening area of the facility at the gas FG leakage location based on the data indicating the thermal image where the abnormal location is known (S2290).

[0107] Next, when the estimation accuracy of the third learning model 153C is equal to or higher than a certain accuracy (S2300: YES), the third learning model 153C estimates the facility system including the gas FG leakage location based on the data indicating the thermal image where the abnormal location is known (S2310).

[0108] Next, the calculation unit 140 calculates the flow rate of the gas FG at the gas FG leakage location using calculation formulas (2), (3), and (4) with the estimated value of the opening area of the facility at the gas FG leakage location estimated by the second learning model 153B in step S2290 and the estimation result of the facility system including the gas FG leakage location estimated by the third learning model 153C in step S2310 as parameters (S2320).

[0109] Next, the output unit 120 outputs the difference value between the estimated value of the flow rate of the gas FG at the gas FG leakage location estimated by the first learning model 153A in step S2270, the estimated value of the opening area of the facility at the gas FG leakage location estimated by the second learning model 153B in step S2290, the estimated value of the flow rate of the gas FG, and the calculated value of the flow rate of the gas FG calculated from the calculation formulas (2), (3), and (4) (S2330).

[0110] Also, when the estimation accuracy of the third learning model 153C is less than a certain accuracy (S2300: NO) and the estimation accuracy of the fourth learning model 153D is equal to or higher than the certain accuracy (S2340: YES), the fourth learning model 153D estimates the physical properties of the leaking gas FG (S2350).

[0111] Next, the calculation unit 140 calculates the flow rate of the gas FG at the leakage location of the gas FG by using calculation formulas (2), (3), and (4) with the estimated value of the opening area of the facility at the leakage location of the gas FG estimated by the second learning model 153B in step S2290 and the estimation result of the physical properties of the gas FG estimated by the fourth learning model 153D in step S2350 as parameters (S2360).

[0112] Next, the output unit 120 outputs the difference value between the estimated value of the flow rate of the gas FG at the leakage location of the gas FG estimated by the first learning model 153A in step S2270, the estimated value of the opening area of the facility at the leakage location of the gas FG estimated by the second learning model 153B in step S2290, the estimated value of the flow rate of the gas FG, and the calculated value of the flow rate of the gas FG calculated from the calculation formulas (2), (3), and (4) (S2370).

[0113] Also, when the estimation accuracy of the fourth learning model 153D is less than the certain accuracy (S2340: NO), the output unit 120 outputs the estimated value of the flow rate of the gas FG at the leakage location of the gas FG estimated by the first learning model 153A in step S2270 and the estimated value of the opening area of the facility at the leakage location of the gas FG estimated by the second learning model 153B in step S229 (S2380).

[0114] Also, when the estimation accuracy of the second learning model 153B is less than the certain accuracy (S2280: NO) and the estimation accuracy of the third learning model 153C is equal to or higher than the certain accuracy (S2390: YES), the third learning model 153C estimates the facility system including the leakage location of the gas FG (S2400).

[0115] Next, the calculation unit 140 calculates the opening area of the facility at the gas FG leakage location using calculation formulas (2), (3), and (4) with the estimated value of the flow rate of the gas FG at the gas FG leakage location estimated by the first learning model 153A in step S2270 and the estimation result of the facility system including the gas FG leakage location estimated by the third learning model 153C in step S2400 as parameters (S2410).

[0116] Next, the output unit 120 outputs the estimated value of the flow rate of the gas FG at the gas FG leakage location estimated by the first learning model 153A in step S2270 and the calculated value of the opening area of the facility calculated from the calculation formulas (2), (3), and (4) (S2420).

[0117] Also, when the estimation accuracy of the third learning model 153C is less than a certain accuracy (S2390: NO) and the estimation accuracy of the fourth learning model 153D is equal to or higher than the certain accuracy (S2430: YES), the fourth learning model 153D estimates the physical properties of the leaking gas FG (S2440).

[0118] Next, the calculation unit 140 calculates the opening area of the facility at the gas FG leakage location using calculation formulas (2), (3), and (4) with the estimated value of the flow rate of the gas FG at the gas FG leakage location estimated by the first learning model 153A in step S2270 and the estimation result of the physical properties of the leaking gas FG estimated by the fourth learning model 153D in step S2440 as parameters (S2450).

[0119] Next, the output unit 120 outputs the estimated value of the flow rate of the gas FG at the gas FG leakage location estimated by the first learning model 153A in step S2270 and the calculated value of the opening area of the facility calculated from the calculation formulas (2), (3), and (4) (S2460).

[0120] Also, when the estimation accuracy of the fourth learning model 153D is less than the certain accuracy (S2430: NO), the output unit 120 outputs the estimated value of the flow rate of the gas FG at the gas FG leakage location estimated by the first learning model 153A in step S2270 (S2470).

[0121] <<Other processing when it is determined from the visible image of the still image or moving image by the imaging device 200B that the fluid is a gas (hereinafter referred to as gas FG) rather than the liquid FL>> When the estimation accuracy of the first learning model 153A is less than a certain accuracy (S2050: NO) and the estimation accuracy of the second learning model 153B is equal to or higher than the certain accuracy (S2480: YES), the second learning model 153B estimates the opening area of the equipment at the leakage location of the gas FG based on the data indicating the thermal image where the abnormal location is known (S2490).

[0122] Next, when the estimation accuracy of the third learning model 153C is equal to or higher than the certain accuracy (S2500: YES), the third learning model 153C estimates the equipment system including the leakage location of the gas FG based on the data indicating the thermal image where the abnormal location is known (S2510).

[0123] Next, the calculation unit 140 calculates the flow rate of the gas FG at the leakage location of the gas FG by using the calculation formulas (2), (3), and (4) with the estimated value of the opening area of the equipment at the leakage location of the gas FG estimated by the second learning model 153B in step S2490 and the estimation result of the equipment system including the leakage location of the gas FG estimated by the third learning model 153C in step S2510 as parameters (S2520).

[0124] Next, the output unit 120 outputs the calculated value of the flow rate of the gas FG at the leakage location of the gas FG calculated from the calculation formulas (2), (3), and (4) and the estimated value of the opening area of the equipment at the leakage location of the gas FG estimated by the second learning model 153B in step S2490 (S2530).

[0125] Also, when the estimation accuracy of the third learning model 153C is less than the certain accuracy (S2500: NO) and the estimation accuracy of the fourth learning model 153D is equal to or higher than the certain accuracy (S2540: YES), the fourth learning model 153D estimates the physical properties of the leaking gas FG (S2550).

[0126] Next, the calculation unit 140 uses calculation formulas (2), (3), and (4) with the estimated value of the opening area of the facility at the gas FG leakage location estimated by the second learning model 153B in step S2490 and the estimation result of the physical properties of the gas FG estimated by the fourth learning model 153D in step S2550 as parameters to calculate the flow rate of the gas FG at the gas FG leakage location (S2560).

[0127] Next, the output unit 120 outputs the calculated value of the flow rate of the gas FG at the gas FG leakage location calculated from the calculation formulas (2), (3), and (4) and the estimated value of the opening area of the facility at the gas FG leakage location estimated by the second learning model 153B in step S2490 (S2570).

[0128] Also, when the estimation accuracy of the second learning model 153B is less than a certain accuracy (S2480: NO) and the estimation accuracy of the fourth learning model 153D is less than a certain accuracy (S2540: NO), the processing after step S2040 is executed again.

[0129] The plant abnormality monitoring system 1 can selectively output the estimated value and calculated value of the flow rate of the fluid (liquid FL, gas FG) at the leakage location of the fluid in the reactor containment vessel, and the estimated value and calculated value of the opening area of the facility at the fluid leakage location, as effective support information for repairing the fluid leakage location at an early stage while considering the radiation dose and temperature in the reactor containment vessel, according to whether the estimation accuracy of the first to fourth learning models 153A to 153D is equal to or higher than a certain accuracy.

[0130] Also, the plant abnormality monitoring system 1 outputs the difference value between the estimated value and the calculated value of the fluid flow rate as information for improving the estimation accuracy of the first learning model 153A.

[0131] As described above, the plant abnormality monitoring system 1 includes an imaging device 200 (200A, 200B) that captures an image of fluid leaking from equipment inside the reactor containment vessel, and a first learning model 153A that has been trained using first learning data 152A in which an image of fluid leaking from the equipment is associated with the flow rate of the fluid. When a new image of fluid leakage captured by the imaging device 200 is input to the first learning model 153A, an estimation unit 130 estimates the flow rate of the fluid, and an output unit 120 outputs the flow rate of the fluid estimated by the estimation unit 130.

[0132] According to the plant abnormality monitoring system 1, when an abnormality occurs in which fluid leaks from equipment inside the reactor containment vessel, the flow rate of the fluid at the leakage location of the fluid is estimated, and based on the estimated value of the flow rate of the fluid, appropriate measures for repair can be quickly taken with respect to the equipment inside the reactor containment vessel.

[0133] Further, the plant abnormality monitoring system 1 further includes a calculation unit 140. The estimation unit 130 includes a second learning model 153B that has been trained using learning data in which an image of fluid leaking from the equipment is associated with the opening area of the leakage location of the equipment. When a new image of fluid leakage captured by the imaging device 200 is input to the second learning model 153B, the opening area of the equipment at the leakage location of the fluid is estimated. The calculation unit 140 calculates the flow rate of the fluid based on the opening area estimated by the estimation unit 130, and the output unit 120 outputs a difference value between the estimated value of the flow rate of the fluid estimated by the estimation unit 130 and the calculated value of the flow rate of the fluid calculated by the calculation unit 140.

[0134] According to the plant abnormality monitoring system 1, by feeding back the difference value between the estimated value of the flow rate of the fluid estimated by the estimation unit 130 and the calculated value of the flow rate of the fluid calculated by the calculation unit 140 using the estimated value of the opening area of the equipment estimated by the estimation unit 130 to the learning of the first learning data 152A, it is possible to improve the estimation accuracy of the first learning model 153A.

[0135] In the plant abnormality monitoring system 1, the estimation unit 130 includes a third learning model 153C that has been learned using third learning data 152C in which an image of fluid leakage from equipment is associated with the equipment system including the fluid leakage location. When a new fluid leakage image captured by the imaging device 200 is input to the third learning model 153C, the equipment system including the fluid leakage location is estimated. The calculation unit 140 calculates the fluid flow rate based on the opening area of the equipment and the equipment system estimated by the estimation unit 130.

[0136] According to the plant abnormality monitoring system 1, by feeding back the difference value between the estimated value of the fluid flow rate estimated by the estimation unit 130 and the calculated value of the fluid flow rate calculated by the calculation unit 140 using the opening area of the equipment and the equipment system estimated by the estimation unit 130 to the learning of the first learning data 152A, it is possible to further improve the estimation accuracy of the first learning model 153A.

[0137] In the plant abnormality monitoring system 1, the estimation unit 130 includes a fourth learning model 153D that has been learned using fourth learning data 152D in which an image of fluid leakage from equipment is associated with the physical properties of the fluid. When a new fluid leakage image captured by the imaging device 200 is input to the fourth learning model 153D, the physical properties of the fluid are estimated. The calculation unit 140 calculates the fluid flow rate based on the opening area of the equipment and the physical properties of the fluid estimated by the estimation unit 130.

[0138] According to the plant abnormality monitoring system 1, by feeding back the difference value between the estimated value of the fluid flow rate estimated by the estimation unit 130 and the calculated value of the fluid flow rate calculated by the calculation unit 140 using the opening area of the equipment and the physical properties of the fluid estimated by the estimation unit 130 to the learning of the first learning data 152A, it is possible to further improve the estimation accuracy of the first learning model 153A.

[0139] Also, in the plant abnormality monitoring system 1, the imaging device 200 is a device that captures at least one of a moving image, a still image, and a thermal image, and can improve the estimation accuracy of the first learning model 153A.

[0140] Also, the equipment in the reactor containment vessel monitored by the plant abnormality monitoring system 1 can be a pipe through which a fluid that is liquid or gas flows in the reactor containment vessel.

[0141] The above embodiments are for facilitating the understanding of the present invention and are not for limiting and interpreting the present invention. The present invention can be changed and improved without departing from its gist, and the equivalents thereof are also included in the present invention. For example, the plant to be monitored by the plant abnormality monitoring system 1 may be an oil plant or a chemical plant other than a nuclear power plant.

Explanation of Signs

[0142] 1 Plant abnormality monitoring system 100 Information processing device 110 Input unit 120 Output unit 130 Estimation unit 140 Calculation unit 150, 310 Storage unit 151 Control program 152 Learning data 152A First learning data 152B Second learning data 152C Third learning data 152D Fourth learning data 153A First learning model 153B Second learning model 153C Third learning model 153D Fourth learning model 154 Table data 200, 200A, 200B Imaging device 210 First camera device 220 Second camera device 300 External computer 400 Communication network 510 Dosimeter

Claims

1. An imaging device that captures an image of a fluid leaking from equipment within a plant, including a first learning model trained using learning data associating an image of the fluid leaking from the equipment with the flow rate of the fluid. When a new image of the fluid leaking captured by the imaging device is input into the first learning model, an estimation device that estimates the flow rate of the fluid, an output device that outputs an estimated value of the flow rate of the fluid estimated by the estimation device, A plant abnormality monitoring system comprising.

2. The plant abnormality monitoring system according to claim 1, further including a calculation device, wherein the estimation device includes a second learning model trained using learning data associating an image of the fluid leaking from the equipment with the opening area of the location where the fluid is leaking. When a new image of the fluid leaking captured by the imaging device is input into the second learning model, the opening area of the location where the fluid is leaking is estimated, the calculation device calculates the flow rate of the fluid based on the estimated value of the opening area estimated by the estimation device, the output device outputs a difference value between the estimated value of the flow rate of the fluid estimated by the estimation device and the calculated value of the flow rate of the fluid calculated by the calculation device A plant abnormality monitoring system.

3. The plant abnormality monitoring system according to claim 2, wherein the estimation device includes a third learning model trained using learning data associating an image of the fluid leaking from the equipment with the equipment system of the location where the fluid is leaking. When a new image of the fluid leaking captured by the imaging device is input into the third learning model, the equipment system of the location where the fluid is leaking is estimated, the calculation device calculates the flow rate of the fluid based on the estimated value of the opening area and the estimated result of the equipment system estimated by the estimation device A plant abnormality monitoring system.

4. The plant abnormality monitoring system according to claim 2, wherein the estimation device includes a fourth learning model trained using learning data associating an image of the fluid leaking from the equipment with the physical properties of the fluid. When a new image of the fluid leaking captured by the imaging device is input into the fourth learning model, the physical properties of the fluid are estimated, the calculation device calculates the flow rate of the fluid based on the estimated value of the opening area and the estimated result of the physical properties estimated by the estimation device Plant abnormality monitoring system.

5. The plant abnormality monitoring system according to claim 1, wherein the imaging device is a device that captures at least one of a moving image, a still image, and a thermal image Plant abnormality monitoring system.

6. The plant abnormality monitoring system according to claim 1, wherein the facilities in the plant include pipes through which the fluid flows in the reactor containment vessel of a nuclear power plant Plant abnormality monitoring system.

7. The plant abnormality monitoring system according to any one of claims 1 to 6, wherein the fluid is a liquid or a gas Plant abnormality monitoring system.

8. A plant abnormality monitoring method using a plant abnormality monitoring system having an imaging device, an estimation device, and an output device, wherein the imaging device captures an image of fluid leakage from facilities in the plant, when a new image of fluid leakage captured by the imaging device is input to a first learning model that the estimation device has learned using learning data associating an image of fluid leakage from the facilities with the flow rate of the fluid, the flow rate of the fluid is estimated, and the output device outputs an estimated value of the flow rate of the fluid estimated by the estimation device. Plant abnormality monitoring method.

9. The plant abnormality monitoring method according to claim 8, wherein the plant abnormality monitoring system further includes a calculation device, when a new image of fluid leakage captured by the imaging device is input to a second learning model that the estimation device has learned using learning data associating an image of fluid leakage from the facilities with the opening area of the location where the fluid has leaked, the opening area of the location where the fluid has leaked is estimated, the calculation device calculates the flow rate of the fluid based on the estimated value of the opening area estimated by the estimation device, and the output device outputs a difference value between the estimated value of the flow rate of the fluid estimated by the estimation device and the calculated value of the flow rate of the fluid calculated by the calculation device. Plant abnormality monitoring method.

10. The plant abnormality monitoring method according to claim 9, when a new image of fluid leakage captured by the imaging device is input to a third learning model that the estimation device has learned using learning data associating an image of fluid leakage from the facilities with the facility system of the location where the fluid has leaked, the facility system of the location where the fluid has leaked is estimated, The calculation device calculates the flow rate of the fluid based on the estimated value of the opening area estimated by the estimation device and the estimation result of the facility system. Plant abnormality monitoring method.

11. The plant abnormality monitoring method according to claim 9, wherein When a new image of the fluid leakage is input to the fourth learning model in which the estimation device has been trained using learning data associating an image of the fluid leaking from the facility with the physical properties of the fluid, the physical properties of the fluid are estimated. The calculation device calculates the flow rate of the fluid based on the estimated value of the opening area estimated by the estimation device and the estimation result of the physical properties. Plant abnormality monitoring method.

12. The plant abnormality monitoring method according to claim 8, wherein The imaging device captures at least one of a moving image, a still image, and a thermal image. Plant abnormality monitoring method.

13. The plant abnormality monitoring method according to claim 8, wherein The facilities in the plant include pipes through which the fluid flows in the reactor containment vessel of a nuclear power plant. Plant abnormality monitoring method.

14. The plant abnormality monitoring method according to any one of claims 8 to 13, wherein The fluid is a liquid or a gas. Plant abnormality monitoring method.

Images