Mapping system and method for radioactively contaminated structure using ai learning

Abstract

The present invention relates to a mapping system and method for a radioactively contaminated structure using AI learning, wherein a radiation distribution for a contamination distribution in a structure is simulated assuming a transportation means such as a virtual drone having a radiometric instrument attached thereto and the simulated radiation distribution can be used to train AI, and the AI-trained AI model estimates the contamination distribution in the structure on the basis of measurement of the radiation distribution, thereby supporting safer and more effective decontamination and dismantling tasks.

Description

Mapping System and Method for Radioactively Contaminated Structures Using Artificial Intelligence Learning

[0001] The present invention relates to a system and method for mapping radioactive contaminated structures using artificial intelligence learning, and more specifically, to simulating the radiation distribution regarding the contamination distribution within a structure. Furthermore, the present invention relates to a system and method for mapping radioactive contaminated structures using artificial intelligence learning, wherein an artificial intelligence model trained on artificial intelligence based on the measurement of radiation distribution can estimate the contamination distribution within a structure.

[0002] Radiation safety is critical when managing contaminated structures, such as nuclear facilities. Radiographic characterization, performed during the preparation or decommissioning phases of nuclear-related facilities, involves evaluating radionuclides, radioactivity, and contamination distribution of radioactive contaminated equipment or structures targeted for decommissioning. This serves as fundamental data for various preparation and operational stages, including the establishment of decommissioning plans, selection of waste treatment technologies, waste classification and disposal, and calculation of treatment costs.

[0003] To this end, since the object to be measured is divided into multiple sections based on the measurement area of ​​the radioactivity measuring device and a unique number for each section must be manually marked by the operator, there are problems such as the risk of radiation exposure for the radioactivity measuring operator and delays in the decommissioning process. In addition, since additional work is required to store and manage measurement information after measuring radioactivity for each section, there is a problem of reduced work efficiency in radioactivity measurement, and continuous efforts have been made to overcome this.

[0004] For example, Korean Published Patent Application No. 10-2024-0078500 discloses a shape recognition-based 3D radioactivity measurement and visualization method for equipment or structures subject to decommissioning of nuclear-related facilities, which enables the provision of efficient, safe, and more accurate radiological characteristic information. In this invention, a radioactivity measurement technology linked with 3D scanning technology is utilized to provide an efficient and more accurate radioactivity measurement method for radiological characteristic evaluation required when establishing decommissioning strategies and plans, such as decommissioning methods, decommissioning schedules, and decommissioning cost calculations for equipment or structures contaminated with radioactivity during the decommissioning of nuclear-related facilities. Furthermore, the invention discloses a system configuration that improves the safety and work efficiency of radiation workers through a series of automated processes involving remote control, radioactivity measurement, and analysis, and provides visualization information by radioactive waste level using a radioactivity measurement information storage and management database system based on a 3D modeling grid structure for objects subject to radioactivity measurement.

[0005] However, this method has the disadvantage that it is difficult to evaluate the contamination distribution for decontamination and demolition work because it has limited ability to model complex 3D spaces within the structure.

[0006] The objective of the present invention is to provide a system and method for mapping radioactive contaminated structures using artificial intelligence learning, which can utilize artificial intelligence learning to train artificial intelligence by simulating the radiation distribution regarding the contamination distribution within a structure assuming a mobile means such as a virtual drone equipped with a radiation detector.

[0007] Another objective of the present invention is to provide a system and method for mapping radioactive contaminated structures utilizing artificial intelligence learning, which can support safer and more effective decontamination and dismantling operations by having an AI-learned model estimate the contamination distribution within the structure based on the measurement of radiation distribution.

[0008] A radioactive contamination structure mapping system utilizing artificial intelligence learning according to the present invention may include a simulator that calculates virtual detection distribution data from virtual contamination distribution data input, a contamination detector that collects contamination concentration data, and an artificial intelligence model that takes virtual detection distribution data as input, performs learning to output virtual contamination distribution data, and then estimates contamination distribution data for the contamination concentration data.

[0009] Here, the simulator can arbitrarily generate virtual contamination distribution data for nuclear power plant structures.

[0010] In addition, virtual contamination distribution data can be distributed on the surface of at least one of the floor, walls, columns, and ceiling of the nuclear power plant structure to mimic the actual radiation contamination intensity.

[0011] Here, the virtual detection distribution data may be characterized as representing the contamination intensity in the virtual space of the nuclear power plant structure.

[0012] In addition, the virtual detection distribution data may be characterized by being calculated using the following mathematical formula.

[0013] Dxyz= Σ Cij × Rij-xyz,

[0014] Here, Dxyz is the detection amount for a voxel (x, y, z) in the detection distribution (S'),

[0015] Cij is the pollution value (concentration) for pixel coordinates (i, j) in the pollution distribution (S),

[0016] Rij-xyz is the proportionality constant.

[0017] Here, the contamination detector is placed in the space of the nuclear power plant structure to collect contamination concentration data.

[0018] In addition, the contamination detector may be characterized by being equipped with a radiation detector on the drone.

[0019] Here, the artificial intelligence model may be characterized by inputting virtual detection distribution data into the input layer.

[0020] In addition, virtual detection distribution data may include detection coordinates and contamination intensity at the detection coordinates.

[0021] Here, the artificial intelligence model can perform training so that virtual contamination distribution data is output from the output layer.

[0022] In addition, virtual contamination distribution data may include contamination coordinates and contamination intensity at the contamination coordinates.

[0023] Here, the artificial intelligence model can infer the overall contamination distribution by combining sparse virtual detection distribution data from various locations.

[0024] In addition, the artificial intelligence model can output correlations and trends between virtual contamination distribution data and virtual detection distribution data.

[0025] Here, the artificial intelligence model can provide a distribution of prediction uncertainty.

[0026] In addition, artificial intelligence models can use a 3-dimensional Convolutional Neural Network (CNN) that is effective for spatial data such as pollution distributions.

[0027] Here, the artificial intelligence model can use a recurrent neural network (RNN), which is useful for sequential data, in cases where measurements are collected over time, such as a drone moving according to a pattern.

[0028] In addition, artificial intelligence models can use an attention mechanism to improve inference accuracy by focusing on important data points.

[0029] Meanwhile, artificial intelligence models can use self-supervised learning, which performs learning by integrating noise into measurements to simulate the inaccuracy of actual sensors.

[0030] A method for mapping radioactive contaminated structures using artificial intelligence learning according to another embodiment of the present invention may include a contamination distribution generation and detection distribution calculation step for calculating virtual detection distribution data from virtual contamination distribution data virtually input in a simulator, a contamination detection step for collecting contamination concentration data from a contamination detector, and an artificial intelligence model learning step for estimating contamination distribution data with respect to contamination concentration data after performing learning in an artificial intelligence model so that virtual detection distribution data is input and virtual contamination distribution data is output.

[0031] Here, in the contamination distribution generation and detection distribution calculation steps, virtual contamination distribution data can be arbitrarily generated for the nuclear power plant structure.

[0032] In addition, virtual contamination distribution data can be distributed on the surface of any one of the floor, walls, columns, and ceiling of the nuclear power plant structure to mimic the actual intensity of radioactive contamination.

[0033] Here, the virtual detection distribution data may be characterized as representing the contamination intensity in the virtual space of the nuclear power plant structure.

[0034] In addition, the virtual detection distribution data may be characterized by being calculated using the following mathematical formula.

[0035] Dxyz= Σ Cij × Rij-xyz,

[0036] Here, Dxyz is the detection amount for a voxel (x, y, z) in the detection distribution (S'),

[0037] Cij is the pollution value (concentration) for pixel coordinates (i, j) in the pollution distribution (S),

[0038] Rij-xyz is the proportionality constant.

[0039] Here, in the contamination detection stage, contamination concentration data can be collected by placing it in the space of the nuclear power plant structure.

[0040] In addition, the contamination detection stage may be characterized by being equipped with a radiation detector in the contamination detector.

[0041] Here, the artificial intelligence model training stage can be characterized by inputting virtual detection distribution data into the input layer.

[0042] In addition, virtual detection distribution data may include detection coordinates and contamination intensity at the detection coordinates.

[0043] Here, during the artificial intelligence model training phase, training can be performed so that virtual contamination distribution data is output from the output layer.

[0044] In addition, virtual contamination distribution data may include contamination coordinates and contamination intensity at the contamination coordinates.

[0045] Here, during the AI ​​model training phase, the overall contamination distribution can be inferred by combining sparse virtual detection distribution data from various locations.

[0046] In addition, during the AI ​​model training phase, the correlation and trends between virtual contamination distribution data and virtual detection distribution data can be output.

[0047] Here, the AI ​​model training phase can provide a distribution of prediction uncertainty.

[0048] In addition, during the artificial intelligence model training phase, a 3-dimensional Convolutional Neural Network (CNN) that is effective for spatial data such as pollution distributions can be used.

[0049] Here, during the artificial intelligence model training phase, a recurrent neural network (RNN), which is useful for sequential data, can be used in cases where measurements are collected over time, such as with a drone moving according to a pattern.

[0050] In addition, during the training phase of the artificial intelligence model, an attention mechanism can be used to improve inference accuracy by focusing on important data points.

[0051] Here, in the artificial intelligence model training phase, self-supervised learning can be used to perform training by integrating noise into the measurements to simulate the inaccuracy of actual sensors.

[0052] The radioactive contamination structure mapping system and method utilizing artificial intelligence learning according to the present invention has the advantage of being able to utilize artificial intelligence for training by simulating the radiation distribution regarding the contamination distribution within the structure by assuming a mobile means of transport, such as a virtual drone equipped with a radiation detector.

[0053] In addition, the radioactive contaminated structure mapping system and method utilizing artificial intelligence learning according to the present invention has the advantage of supporting safer and more effective decontamination and dismantling operations by having an AI-learned AI model estimate the contamination distribution within the structure based on the measurement of radiation distribution.

[0054] FIG. 1 is a schematic diagram showing a radioactive contamination structure mapping system utilizing artificial intelligence learning according to one embodiment of the present invention.

[0055] FIG. 2 is a perspective view showing in detail the virtual pollution distribution data, virtual detection distribution data, pollution concentration data, and pollution distribution data of FIG. 1.

[0056] FIG. 3 is a flowchart illustrating a method for mapping radioactive contaminated structures using artificial intelligence learning according to an embodiment of the present invention.

[0057] Hereinafter, specific embodiments for carrying out the present invention will be described with reference to the attached drawings. In describing the present invention, terms such as "first," "second," etc., may be used to describe various components, but the components may not be limited by these terms. The terms are intended solely for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be named the second component, and similarly, the second component may be named the first component. Where a component is described as being connected to or coupled with another component, it may be understood that it is directly connected to or coupled with the other component, or that there may be other components in between.

[0058] The terms used in this specification are used merely to describe specific embodiments and are not intended to limit the invention. Singular expressions may include plural expressions unless the context clearly indicates otherwise. In this specification, terms such as "comprising" or "comprising" are intended to indicate the presence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof. Additionally, the shapes and sizes of elements in the drawings may be exaggerated for clearer explanation.

[0059] Hereinafter, a system and method for mapping radioactive contaminated structures using artificial intelligence learning according to the present invention will be described in detail with reference to the attached drawings.

[0060] FIG. 1 is a schematic diagram showing a radioactive contamination structure mapping system utilizing artificial intelligence learning according to one embodiment of the present invention, and FIG. 2 is a detailed perspective view for explaining FIG. 1 in detail.

[0061] Hereinafter, a radioactive contamination structure mapping system utilizing artificial intelligence learning according to an embodiment of the present invention will be described with reference to FIGS. 1 and 2.

[0062] First, referring to FIG. 1, a radioactive contamination structure mapping system utilizing artificial intelligence learning according to one embodiment of the present invention comprises a simulator (100) that calculates virtual detection distribution data (D200) from virtual input virtual contamination distribution data (D100), a contamination detector (200) that collects contamination concentration data (D300), and an artificial intelligence model (300) that takes virtual detection distribution data (D200) as input, performs learning so that virtual contamination distribution data (D100) is output, and then estimates contamination distribution data (D400) for contamination concentration data (D300).

[0063] Here, the simulator (100) simulates the measurement and mapping of radioactive contamination inside and around the nuclear power plant structure using voxelized 3D modeling.

[0064] To this end, virtual contamination distribution data (D100) is arbitrarily generated to be distributed on the surfaces of the floor, walls, columns, and ceiling of the nuclear power plant structure to mimic the actual intensity of radioactive contamination, and the contamination distribution within the structure can be defined by designing the contamination by dividing the contaminated structure into cubic voxels.

[0065] Meanwhile, the virtual contamination detector (200) can divide the space in which the nuclear power plant structure can move into cubic voxels and represent measurement points in 3D space based on virtual contamination distribution data (D100). Here, the virtual detection distribution data (D200) represents the contamination intensity displayed in the virtual space of the nuclear power plant structure and is explained in detail in FIG. 2.

[0066] At this time, the virtual contamination distribution data (D100) and the virtual detection distribution data (D200) can be used as data for training the artificial intelligence model (300).

[0067] Afterwards, the artificial intelligence model (300) performs training so that virtual detection distribution data (D200) is used as input to the input layer and virtual contamination distribution data (D100) is output from the output layer.

[0068] Here, the virtual contamination distribution data (D100) includes contamination coordinates and contamination intensity at the contamination coordinates.

[0069] Meanwhile, the artificial intelligence model (300) may infer the overall contamination distribution by combining sparse virtual detection distribution data (D200) from various locations, output the correlation and trend between the virtual contamination distribution data (D100) and the virtual detection distribution data (D200), or provide the user with the estimation reliability by providing the prediction uncertainty distribution.

[0070] At this time, the artificial intelligence model (300) may use a 3-dimensional Convolutional Neural Network (CNN) effective for spatial data such as pollution distributions, and a Recurrent Neural Network (RNN) useful for sequential data when measurements are collected over time, such as a drone moving according to a pattern. Additionally, the artificial intelligence model (300) may use an Attention Mechanism that focuses on important data points to increase inference accuracy, and self-supervised learning that performs learning by integrating noise into measurements to simulate the inaccuracy of actual sensors.

[0071] Meanwhile, once the learning is complete, the contamination detector (200) can be placed in the space of the nuclear power plant structure with a radiation detector mounted on a drone to collect contamination concentration data (D300).

[0072] In addition, pollution concentration data (D300) can be input into an artificial intelligence model (300) to estimate pollution distribution data (D400) with high accuracy.

[0073] Accordingly, the radioactive contamination structure mapping system utilizing artificial intelligence learning according to the present invention supports applications such as radioactive contamination investigation, particularly the decommissioning of contaminated facilities including nuclear reactors, and the artificial intelligence model (300) trained with this simulation data has the advantage of being able to map contamination distribution data (D400) prior to contamination removal and decommissioning with high accuracy.

[0074]

[0075] FIG. 2 is a detailed perspective view showing the virtual pollution distribution data (D100), virtual detection distribution data (D200), pollution concentration data (D300), and pollution distribution data (D400) of FIG. 1. As can be seen in FIG. 2, the virtual detection distribution data (D200), which represents pollution intensity in 3D space based on the virtual pollution distribution data (D100), can be calculated using Equation 1.

[0076] [Mathematical Formula 1]

[0077] Dxyz= Σ Cij × Rij-xyz,

[0078] Here, Dxyz is the radioactivity detected for a voxel (x, y, z) in the detection distribution (S'),

[0079] Cij is the pollution value (concentration) for pixel coordinates (i, j) in the pollution distribution (S),

[0080] Rij-xyz is the proportionality constant.

[0081]

[0082] Looking at this in detail, first, for the simulation, a contamination distribution (S) is created by assigning a contamination value (concentration) Cij to pixel coordinates (i, j) on a virtual plane.

[0083] In addition, a virtual space is created into voxels, three-dimensional coordinates (x, y, z) are assigned to the voxels, and a voxel corresponding to the location of a virtual contamination detector (200) that changes according to the movement path of an airplane, etc. is selected.

[0084] Meanwhile, the amount of radiation detected by a virtual contamination detector (200) located at an arbitrary voxel (x, y, z) starting from an arbitrary pixel (i, j) can be expressed by the formula (Cij Rij-xyz) using the proportionality constant Rij-xyz.

[0085] At this time, Rij-xyz is calculated using well-established radiation physics theories and the Monte Carlo method (probabilistic method).

[0086] Therefore, the detected amount (Dxyz) of voxel (x, y, z) must take into account radiation originating from the entire contamination distribution (S), so it can be organized as Equation 1.

[0087] Conversely, the set of radioactivity detection amounts at the locations of the virtual contamination detector (200) is expressed as S'={ Dxyz}, and the contamination distribution (S) can be estimated from the detection distribution (S') using various algorithms, and the artificial intelligence can be trained by producing multiple combinations of the contamination distribution (S) and the detection distribution (S') through simulation.

[0088] The artificial intelligence trained in this way is capable of estimating the pollution distribution data (D400) with high accuracy from the pollution concentration data (D300) measured by the pollution detector (200) in a real-world situation.

[0089] Here, virtual contamination distribution data (D100) can input contamination data for specific areas of a building or nuclear power plant structure using flexibility for various shapes (e.g., square, triangle, circle). That is, 3D voxels representing structures (e.g., rooms, walls) can be defined, and contamination can be added to specified areas and shapes and contamination intensity assigned to each voxel according to user input.

[0090] In addition, the virtual detection distribution data (D200) can define measurement locations within a building to verify contamination levels (e.g., drone paths or fixed locations), divide the structure into voxels, specify measurement points, and store the measurement locations along with the measurement values ​​in a coordinate list.

[0091] Meanwhile, the simulator (100) can simulate realistic measurements by calculating the interaction between radiation emitted from contaminated pixels and measurement equipment at specific points.

[0092] This can calculate the radiation contribution of nearby contaminated voxels for each measurement point, use distance reduction to account for lower intensity as distance increases, and sum the contributions to obtain the final reading at each measurement point to produce virtual detection distribution data (D200).

[0093]

[0094] FIG. 3 is a flowchart illustrating a method for mapping radioactive contaminated structures using artificial intelligence learning according to an embodiment of the present invention.

[0095] As can be seen in FIG. 3, the method for mapping radioactive contaminated structures using artificial intelligence learning comprises: a contamination distribution generation and detection distribution calculation step (S100) for calculating virtual detection distribution data (D200) from virtual contamination distribution data (D100) input in a simulator (100); a contamination detection step (S200) for collecting contamination concentration data (D300) from a contamination detector (200); and an artificial intelligence model learning step (S300) for estimating contamination distribution data (D400) for contamination concentration data (D300) after performing learning in an artificial intelligence model (300) so that virtual detection distribution data (D200) is input and virtual contamination distribution data (D100) is output.

[0096] Here, in the contamination distribution generation and detection distribution calculation step (S100), the measurement and mapping of radioactive contamination inside and around the nuclear power plant structure is simulated using voxelized 3D modeling.

[0097] To this end, virtual contamination distribution data (D100) is arbitrarily generated to be distributed on the surfaces of the floor, walls, columns, and ceiling of the nuclear power plant structure to mimic the actual intensity of radioactive contamination, and the contamination distribution within the structure can be defined by designing the contamination by dividing the contaminated structure into cubic voxels.

[0098] Meanwhile, the virtual contamination detector (200) can divide the space in which the nuclear power plant structure can move into cubic voxels and represent measurement points in 3D space based on virtual contamination distribution data (D100). Here, the virtual detection distribution data (D200) represents the contamination intensity displayed in the virtual space of the nuclear power plant structure, as explained in detail in FIG. 2.

[0099] At this time, the virtual contamination distribution data (D100) and the virtual detection distribution data (D200) can be used as data for training in the artificial intelligence model training stage (S300).

[0100] Subsequently, in the artificial intelligence model training stage (S300), virtual detection distribution data (D200) is used as the input to the input layer, and training is performed so that virtual contamination distribution data (D100) is output from the output layer.

[0101] Here, the virtual contamination distribution data (D100) includes contamination coordinates and contamination intensity at the contamination coordinates.

[0102] Meanwhile, in the artificial intelligence model training stage (S300), the overall contamination distribution may be inferred by combining sparse virtual detection distribution data (D200) from various locations, or the correlation and trend between the virtual contamination distribution data (D100) and the virtual detection distribution data (D200) may be output, or the prediction uncertainty distribution may be provided together to provide the user with estimation reliability.

[0103] At this time, in the artificial intelligence model training stage (S300), a 3-dimensional convolutional neural network (CNN) effective for spatial data such as pollution distributions is used, a recurrent neural network (RNN) useful for sequential data in cases where measurements are collected over time, such as drones moving according to patterns is used, an attention mechanism that focuses on important data points to increase inference accuracy is used, and self-supervised learning that performs learning by integrating noise into measurements to simulate the inaccuracy of actual sensors may also be used.

[0104] Once the learning is complete, in the contamination detection step (S200), a radiation detector is installed on the contamination detector (200) and placed in the space of the nuclear power plant structure to collect contamination concentration data (D300).

[0105] In addition, pollution concentration data (D300) can be input into an artificial intelligence model (300) to estimate pollution distribution data (D400) with high accuracy.

[0106] Accordingly, the method for mapping radioactive contaminated structures using artificial intelligence learning according to the present invention supports applications such as radioactive contamination investigation, particularly the decommissioning of contaminated facilities including nuclear reactors, and the artificial intelligence model (300) trained with this simulation data can map the contamination distribution data (D400) prior to decontamination and decommissioning with high accuracy.

[0107]

[0108] As described above, the radioactive contaminated structure mapping system and method utilizing artificial intelligence learning according to the present invention has the advantage of being able to utilize artificial intelligence for training by simulating the radiation distribution regarding the contamination distribution within the structure assuming a mobile means such as a virtual drone equipped with a radiation detector. Furthermore, by having the artificial intelligence-trained model estimate the contamination distribution within the structure based on the measurement of the radiation distribution, it has the advantage of supporting safer and more effective decontamination and dismantling operations.

[0109]

[0110] Those skilled in the art will understand that the various exemplary logic blocks, modules, processors, means, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be implemented by electronic hardware, various forms of programs or design code (referred to herein as software for convenience), or a combination of all such. To clearly illustrate this interoperability between hardware and software, various exemplary components, blocks, modules, circuits, and steps have been generally described above in relation to their functions. Whether such functions are implemented as hardware or software depends on the design constraints imposed on the specific application and the overall system. Those skilled in the art may implement the functions described in various ways for each specific application, but such implementation decisions should not be interpreted as being outside the scope of the invention.

[0111] The various embodiments presented herein may be implemented as methods, devices, or articles manufactured using standard programming and / or engineering techniques. The term "article manufactured" includes a computer program, a carrier, or a medium accessible from any computer-readable storage device. For example, computer-readable storage media include, but are not limited to, magnetic storage devices (e.g., hard disks, floppy disks, magnetic strips, etc.), optical discs (e.g., CDs, DVDs, etc.), smart cards, and flash memory devices (e.g., EEPROMs, cards, sticks, key drives, etc.). Additionally, the various storage media presented herein include one or more devices and / or other machine-readable media for storing information.

[0112] It should be understood that the specific order or hierarchy of steps in the presented processes is merely an example of exemplary approaches. It should be understood that, based on design priorities, the specific order or hierarchy of steps in the processes may be rearranged within the scope of the invention. The appended method claims provide various step elements in a sample order, but do not imply limitation to the specific order or hierarchy presented.

[0113] The description of the presented embodiments is provided so that any person skilled in the art may use or practice the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments without departing from the scope of the present invention. Thus, the present invention is not limited to the embodiments presented herein, but should be interpreted in the broadest possible scope consistent with the principles and novel features presented herein.

Claims

1. A simulator that calculates virtual detection distribution data from virtual input virtual contamination distribution data; A contamination detector that collects contamination concentration data; and A radioactive contamination structure mapping system utilizing artificial intelligence learning, comprising: an artificial intelligence model that takes the virtual detection distribution data as input, performs learning to produce the virtual contamination distribution data as output, and then estimates the contamination distribution data for the contamination concentration data.

2. In Paragraph 1, The above simulator is a radioactive contamination structure mapping system utilizing artificial intelligence learning, characterized by arbitrarily generating the above virtual contamination distribution data for nuclear power plant structures.

3. In Paragraph 2, A radioactive contamination structure mapping system utilizing artificial intelligence learning, characterized in that the above-mentioned virtual contamination distribution data is distributed on at least one surface among the floor, walls, columns, and ceiling of the above-mentioned nuclear power plant structure to mimic the actual radiation contamination intensity.

4. In Paragraph 3, A radioactive contamination structure mapping system utilizing artificial intelligence learning, characterized in that the above virtual detection distribution data represents the contamination intensity shown in the virtual space of the above nuclear power plant structure.

5. In Paragraph 4, A radioactive contamination structure mapping system utilizing artificial intelligence learning, characterized in that the above-mentioned virtual detection distribution data is calculated using the following mathematical formula. Dxyz= Σ Cij × Rij-xyz, Here, Dxyz is the detection amount for a voxel (x, y, z) in the detection distribution (S'), Cij is the pollution value (concentration) for pixel coordinates (i, j) in the pollution distribution (S), Rij-xyz is the proportionality constant.

6. In Paragraph 1, A radioactive contamination structure mapping system utilizing artificial intelligence learning, characterized in that the above-mentioned contamination detector is placed in the space of the above-mentioned nuclear power plant structure to collect the above-mentioned contamination concentration data.

7. In Paragraph 6, The above contamination detector is a radioactive contamination structure mapping system utilizing artificial intelligence learning, characterized by equipping a drone with a radiation detector.

8. In Paragraph 1, A radioactive contamination structure mapping system utilizing artificial intelligence learning, characterized by the above artificial intelligence model inputting the above virtual detection distribution data into the input layer.

9. In Paragraph 8, A radioactive contamination structure mapping system utilizing artificial intelligence learning, characterized in that the virtual detection distribution data includes detection coordinates and contamination intensity at the detection coordinates.

10. In Paragraph 9, A radioactive contamination structure mapping system utilizing artificial intelligence learning, characterized in that the above artificial intelligence model performs learning so that the above virtual contamination distribution data is output from the output layer.

11. In Paragraph 10, A radioactive contamination structure mapping system utilizing artificial intelligence learning, characterized in that the virtual contamination distribution data includes contamination coordinates and contamination intensity at the contamination coordinates.

12. In Paragraph 11, A radioactive contamination structure mapping system utilizing artificial intelligence learning, characterized by the above artificial intelligence model inferring the overall contamination distribution by combining the above-mentioned virtual detection distribution data from multiple locations.

13. In Paragraph 12, A radioactive contamination structure mapping system utilizing artificial intelligence learning, characterized in that the above artificial intelligence model outputs correlations and trends between the above virtual contamination distribution data and the above virtual detection distribution data.

14. In Paragraph 13, A radioactive contamination structure mapping system utilizing artificial intelligence learning, characterized by the above artificial intelligence model providing a prediction uncertainty distribution.

15. In Paragraph 14, The above artificial intelligence model is a radioactive contamination structure mapping system utilizing artificial intelligence learning, characterized by using a 3-dimensional Convolutional Neural Network (CNN) that is effective for spatial data such as contamination distribution.

16. In Paragraph 14, A radioactive contamination structure mapping system utilizing artificial intelligence learning, characterized by the above artificial intelligence model using a recurrent neural network (RNN) which is useful for sequential data when measurements are collected over time, such as a drone moving according to a pattern.

17. In Paragraph 14, A radioactive contamination structure mapping system utilizing artificial intelligence learning, characterized by the above-mentioned artificial intelligence model using an attention mechanism that focuses on important data points to increase inference accuracy.

18. In Paragraph 14, A radioactive contamination structure mapping system utilizing artificial intelligence learning, characterized by the above artificial intelligence model using self-supervised learning that performs learning by integrating noise into the measurement values ​​to simulate the inaccuracy of the actual sensor.

19. A contamination distribution generation and detection distribution calculation step for calculating virtual detection distribution data from virtual contamination distribution data virtually input in a simulator; A contamination detection step for collecting contamination concentration data from a contamination detector; and A method for mapping radioactive contaminated structures using artificial intelligence learning, comprising: an artificial intelligence model learning step in which the above virtual detection distribution data is used as input and the above virtual contamination distribution data is output, and then the above contamination distribution data is estimated for the above contamination concentration data.

20. In Paragraph 19, A method for mapping radioactive contaminated structures using artificial intelligence learning, characterized by generating virtual contamination distribution data for nuclear power plant structures in the contamination distribution generation and detection distribution calculation steps.

21. In Paragraph 20, A method for mapping radioactive contamination structures using artificial intelligence learning, characterized in that the above-mentioned virtual contamination distribution data is distributed on the surface of any one of the floor, wall, column, and ceiling of the nuclear power plant structure to mimic the actual radioactive contamination intensity.

22. In Paragraph 21, A method for mapping radioactive contamination structures using artificial intelligence learning, characterized in that the above virtual detection distribution data represents the contamination intensity shown in the virtual space of the above nuclear power plant structure.

23. In Paragraph 22, A method for mapping radioactive contaminated structures using artificial intelligence learning, characterized in that the above-mentioned virtual detection distribution data is calculated using the following mathematical formula. Dxyz= Σ Cij × Rij-xyz, Here, Dxyz is the detection amount for a voxel (x, y, z) in the detection distribution (S'), Cij is the pollution value (concentration) for pixel coordinates (i, j) in the pollution distribution (S), Rij-xyz is the proportionality constant.

24. In Paragraph 19, A method for mapping radioactive contaminated structures using artificial intelligence learning, characterized in that the contamination detection step above involves placing the contaminant concentration data in the space of the nuclear power plant structure.

25. In Paragraph 24, A method for mapping radioactive contaminated structures using artificial intelligence learning, characterized in that, in the contamination detection step above, a radioactivity detector is equipped on the contamination detector.

26. In Paragraph 19, A method for mapping radioactive contaminated structures using artificial intelligence learning, characterized by inputting the virtual detection distribution data into the input layer during the artificial intelligence model learning stage.

27. In Paragraph 26, A method for mapping radioactive contaminated structures using artificial intelligence learning, characterized in that the virtual detection distribution data includes detection coordinates and contamination intensity at the detection coordinates.

28. In Paragraph 27, A method for mapping radioactive contaminated structures using artificial intelligence learning, characterized in that, in the artificial intelligence model learning step, learning is performed so that the virtual contamination distribution data is output from the output layer.

29. In Paragraph 28, A method for mapping radioactive contaminated structures using artificial intelligence learning, characterized in that the virtual contamination distribution data includes contamination coordinates and contamination intensity at the contamination coordinates.

30. In Paragraph 29, A method for mapping radioactive contaminated structures using artificial intelligence learning, characterized by inferring the overall contamination distribution by combining the sparse virtual detection distribution data from multiple locations during the artificial intelligence model learning stage.

31. In Paragraph 30, A method for mapping radioactive contaminated structures using artificial intelligence learning, characterized by outputting correlations and trends between the virtual contamination distribution data and the virtual detection distribution data during the artificial intelligence model learning stage.

32. In Paragraph 31, A method for mapping radioactive contaminated structures using artificial intelligence learning, characterized by providing a prediction uncertainty distribution in the artificial intelligence model learning stage.

33. In Paragraph 32, A method for mapping radioactive contaminated structures using artificial intelligence learning, characterized by using a 3-dimensional Convolutional Neural Network (CNN) effective for spatial data such as contamination distribution in the artificial intelligence model learning stage.

34. In Paragraph 32, A method for mapping radioactive contaminated structures using artificial intelligence learning, characterized by using a recurrent neural network (RNN) which is useful for sequential data in the artificial intelligence model learning stage, such as in cases where measurements are collected over time, like a drone moving according to a pattern.

35. In Paragraph 32, A method for mapping radioactive contaminated structures using artificial intelligence learning, characterized by using an attention mechanism that focuses on important data points to increase inference accuracy during the artificial intelligence model learning stage.

36. In Paragraph 32, A method for mapping radioactive contaminated structures using artificial intelligence learning, characterized by using self-supervised learning in the artificial intelligence model learning stage to perform learning by integrating noise into the measured values ​​to simulate the inaccuracy of the actual sensor.

Images