GB2619374A

GB2619374A - Method for managing a piece of nuclear waste

Info

Publication number: GB2619374A
Application number: GB2219229.8A
Authority: GB
Inventors: Tran Van-Xuan; Anthony Phillips Paul; Baur Julien; John Carver Matthew; Media Vazquez Gustavo
Original assignee: Electricite de France SA
Current assignee: Electricite de France SA
Priority date: 2021-12-21
Filing date: 2022-12-20
Publication date: 2023-12-06
Also published as: FR3131065B1; FR3131065A1; GB202219229D0

Abstract

The invention relates to a method and a system for managing a piece of nuclear waste 1 (figure 1) comprising the implementation by data processing means 10 of the steps of: (b1) determining at least a first characteristic of the piece of nuclear waste other than relating to the radioactivity of the piece of nuclear waste, (b2) determining a second characteristic of the piece of, nuclear waste relating to the radioactivity of the, piece of nuclear waste from data from a radioactivity sensor, (c) assigning a packaging class to the piece of nuclear waste among a plurality of packaging classes depending on the first characteristic and the second characteristic. A further step (d) may comprise packaging the waste accordingly to its class and before that the waste may be arranged in a waiting zone with other waste, spatial arrangement of all the waste determined then arranged in a container 9 (figure 1) accordingly.

Description

Method for managing a piece of nuclear waste

GENERAL TECHNICAL FIELD

The present invention relates to the field of 5 nuclear power.

More precisely, the present invention relates to the management of nuclear waste.

STATE OF THE ART

In France, the energy transition law for green growth adopted in 2015 imposes a drop in electricity from nuclear sources, resulting in the dismantling of nuclear power plants. Indeed, one objective of the law is to increase the share of nuclear power in electricity generation to 50% by 2025. 51 civil nuclear facilities are currently at a standstill in France (36 of the CEA, 9 of EDF and 6 of Orano). The first "standard" pressurised water reactor (PWR) was shut down in Fessenheim in February 2020. The French government is planning to close 14 reactors by 2035.

In the rest of the world, many nuclear power plants are being dismantled or will be dismantled, particularly in the United States, Germany, Japan, Belgium and the United Kingdom. The global dismantling market can be estimated at 300-600 billion over a century, for 450 reactors to be shut down over the next 50 years.

Furthermore, 18 nuclear power plants and 56 nuclear reactors are currently in operation in France. Worldwide, 443 nuclear reactors are in operation. All these plants require regular maintenance and modernisation work.

When a nuclear power plant is operated, maintained, modernised or dismantled, a large amount of nuclear waste is generated. These pieces of nuclear waste are different from each other. They are made of different materials, have different types and intensities of radioactivity and may contain different pollutants. These nuclear waste pieces therefore have to be separated and sorted in a manner appropriate to its specific characteristics so that they can then be stored in the long term, discarded or recycled. For this, nuclear waste pieces therefore have to be inspected, analysed, classified, moved, etc. Currently, nuclear waste sorting is implemented by humans and/or using machines which require the intervention of a human operator. Many operators are needed because they should not work for more than a certain period of time to limit their exposure to radioactivity. The sorting process is therefore costly and can involve hazards for operators. In addition, human error can occur and recyclable waste is discarded on a regular basis. This can also have serious consequences when waste that is not supposed to be packaged together is and reacts together to generate hazardous substances. In addition, nuclear waste is not optimally arranged in the storage containers and the number of storage containers is then very high. Also, additional nuclear waste is generated because the material used by the operator for sorting becomes a piece of nuclear waste.

Moreover, to sort nuclear waste, it is necessary to 30 acquire information about it. Currently, a piece of nuclear waste is manually selected from a stack of nuclear waste. Then, the piece of nuclear waste is photographed without any analysis of the photograph.

Also, information about the material of the piece of waste is acquired via specific sensors or is determined with the naked eye by an operator or by extensive laboratory analysis. These embodiments are time consuming and information recorded by operators may be inaccurate or incomplete.

Besides, waste management and sorting methods exist 10 in the traditional (non-nuclear) waste industry. However, these methods are not adapted to the features of nuclear waste.

In the automotive industry, objects travel on a conveyor belt, are identified and then caught by robot arms to be arranged in the right place (that is, usually assembled with other parts). However, these objects have predetermined shapes, predetermined destinations, and are not radioactive.

Finally, in the parcel delivery field, items are 20 arranged in boxes. However, these objects have regular shapes and are not radioactive. Furthermore, there is no problem in optimising the arrangement of multiple objects in a box since a box usually only contains one object and the storage time is very short, usually only for transport purposes.

Accordingly, nuclear waste management solutions that are cheaper, less dangerous to humans, faster and more optimal are currently lacking.

DISCLOSURE OF THE INVENTION

According to a first aspect, the present invention thus relates to a method for managing a piece of nuclear waste comprising the implementation by data processing means of the steps of: bl) determining at least a first characteristic of the piece of nuclear waste other than relating to the 5 radioactivity of the piece of nuclear waste, b2) determining a second characteristic of the piece of nuclear waste relating to the radioactivity of the piece of nuclear waste from data from a radioactivity sensor, c) assigning a packaging class to the piece of nuclear waste among a plurality of packaging classes depending on the first characteristic and the second characteristic.

According to advantageous and non-limiting characteristics: The method further comprises a step (d) of packaging the piece of nuclear waste depending on the packaging class of the piece of nuclear waste.

The method comprises a step (d0) prior to step (d) comprising arranging the piece of nuclear waste in a waiting zone in which a plurality of pieces of nuclear waste can be arranged, a step of calculating the spatial arrangement of pieces of nuclear waste in a container depending on all of the pieces of nuclear waste arranged in the waiting zone, and a step of arranging the piece of nuclear waste in the container depending on the spatial arrangement of pieces of nuclear waste.

Step (d) includes a step (dl) of grinding the piece 30 of nuclear waste with other pieces of nuclear waste of the same packaging class.

The second characteristic is a radioactivity level of the piece of nuclear waste among a plurality of radioactivity levels.

The first characteristic is a characteristic relating to the composition of the piece of nuclear waste and is determined in step (bl) from data from a composition sensor and/or a classification of an image of the piece of nuclear waste.

The first characteristic is a material.

The method comprises a step (a4) of determining the weight of the piece of nuclear waste.

The method comprises a step (a7) of determining a type of the piece of nuclear waste.

The method comprises a step (a6) of determining geometric parameters of the piece of nuclear waste from an image of the piece of nuclear waste, the geometric parameters comprising dimensions, volume and/or external surface area of the piece of nuclear waste.

The method comprises a step (a8) of cutting the piece of nuclear waste if a geometric parameter of said piece of nuclear waste is greater than a threshold of the geometric parameter of said piece of nuclear waste or if the weight of said piece of nuclear waste is greater than a weight threshold.

Geometric parameters of the piece of nuclear waste are obtained from the analysis of a 3D image of the piece of nuclear waste.

The 3D image is constructed from a plurality of 2D images obtained by rotating the piece of nuclear waste 30 relative to a camera or from a depth map acquired by a depth camera.

The method comprises a prior step (al) of identifying the piece of nuclear waste from an image of a plurality of pieces of nuclear waste, the method being implemented for at least one piece of nuclear waste of said plurality of pieces of nuclear waste.

According to a second aspect, the invention relates to a system for managing a piece of nuclear waste comprising data processing means configured to determine at least a first characteristic of the piece of nuclear waste other than relating to the radioactivity of the piece of nuclear waste, determine a second characteristic of the piece of nuclear waste relating to the radioactivity of the piece of nuclear waste from data from a radioactivity sensor and assign a packaging class to the piece of nuclear waste from a plurality of packaging classes depending on the first characteristic and the second characteristic.

According to advantageous and non-limiting characteristics: The system includes a cutting device configured to cut the piece of nuclear waste.

The system comprises at least one radioactivity 25 sensor configured to acquire radioactivity sensor data of the piece of nuclear waste and a camera configured to acquire at least one image of the piece of nuclear waste.

BRIEF DESCRIPTION OF THE FIGURES

Further characteristics and advantages of the present invention will appear upon reading the description of a preferential embodiment that follows. This description will be given with reference to the appended figures, including: - Figure 1 represents a nuclear waste management system according to one possible embodiment of the invention, - Figure 2 is a detailed flowchart of a method for managing nuclear waste according to one possible embodiment of the invention, Figure 3 is a simplified flowchart of the method for managing nuclear waste according to one possible embodiment of the invention, Figure 4 is a diagram representing the data flow while classifying a piece of nuclear waste according to one possible embodiment of the invention, Figure 5 schematically represents possible arrangements of simplified forms of pieces of nuclear waste in container modelling.

DETAILED DESCRIPTION Architecture

With reference to Figure 1, the invention provides a method for managing nuclear waste 1 by a nuclear waste management system 100. The nuclear waste management system 100 first comprises a nuclear waste characterisation system 110.

The nuclear waste characterisation system 110 comprises characterisation data processing means 114 30 configured to obtain an image of a piece of nuclear waste 1, and classify the piece of nuclear waste 1 represented by the image among a plurality of type classes of nuclear waste 1 and/or among a plurality of characteristic classes relating to the composition of nuclear waste 1 by applying to the image at least one classification model learned in a supervised manner on a learning base. The characterisation data processing means 114 are typical data processing means such as a processor of computing equipment 3.

Preferably, the characterisation data processing means 114 are configured to associate an identifier of the piece of nuclear waste 1 with the class of the piece of nuclear waste 1.

The characterisation data processing means 114 can 15 also be configured to construct a 3D image of a piece of nuclear waste 1 from a plurality of 2D images of the piece of nuclear waste 1.

Preferably, the characterisation data processing means 114 are configured to segment a piece of nuclear 20 waste 1 into an image of a plurality of pieces of nuclear waste 1 known as the segmentation image.

The nuclear waste characterisation system 110 comprises a characterisation camera 112 configured to acquire a plurality of images of a piece of nuclear waste 1. The characterisation camera 112 can be arranged on a characterisation robot arm 116. The characterisation camera 112 can therefore be moved via commands of the characterisation robot arm 116 so that the latter moves and re-orients itself so that the characterisation camera 112 moves and re-orients itself. The characterisation camera 112 can also be arranged on a fixed support. The piece of nuclear waste 1 can then be presented to the characterisation camera 112 by a characterisation robot arm 116 so that the piece of nuclear waste 1 is within the field of view of the characterisation camera 112.

Preferably, the characterisation robot arm 116 is able to catch nuclear waste 1. For example, the characterisation robot arm 116 is capable of catching pieces of nuclear waste 1 from a stack of nuclear waste 5.

The nuclear waste characterisation system 110 further comprises characterisation data storage means 118 configured to store the learning base that comprises at least one image of the piece of nuclear waste 1 associated with the class of the piece of nuclear waste 1. The characterisation data storage means 118 are typically a memory of the computing equipment 3. The learning base preferably comprises a plurality of images of pieces of nuclear waste 1 each associated with a class of nuclear waste 1. Preferably, the learning base comprises at least 10,000 images of pieces of nuclear waste 1.

The nuclear waste characterisation system 110 may also comprise an apparatus dedicated to the acquisition of a depth map representative of a piece of nuclear waste 1 such as a 3D camera or depth camera 113. For example, the depth camera 113 can be arranged on the characterisation robot arm 116.

The nuclear waste management system 100 comprises 30 management data processing means 104. The management data processing means 104 are typical data processing means such as a processor of the computing equipment 3. The management data processing means 104 are configured to determine, from the data of the radioactivity sensor 102 of a piece of nuclear waste 1, at least a first characteristic of the piece of nuclear waste 1 other than relating to the radioactivity of the piece of nuclear waste 1. The management data processing means 104 are also configured to determine a second characteristic of the piece of nuclear waste 1 relating to the radioactivity of the piece of nuclear waste 1 from data from a radioactivity sensor. Furthermore, the management data processing means 104 are configured to assign a packaging class to the piece of nuclear waste 1 among a plurality of packaging classes depending on the first characteristic and the second characteristic.

Preferably, the nuclear waste management system 100 comprises at least one radioactivity sensor 102. The radioactivity sensor 102 is configured to acquire radioactivity sensor data of the piece of nuclear waste 1. For example, the radioactivity sensor 102 is placed near a conveyor 7 on which pieces of nuclear waste 1 are deposited. In this way, the radioactivity sensor 102 can acquire radioactivity data from pieces of nuclear waste

1 passing within its field of acquisition.

The nuclear waste management system 100 may also comprise one or more management robot arms 101 configured to catch pieces of nuclear waste 1, move the pieces of nuclear waste 1 and, for example, deposit them on the conveyor V. Preferably, the nuclear waste management system 100 comprises management data storage means 107 configured to store a database of pieces of nuclear waste 1, each piece of nuclear waste 1 being identified by a unique identifier and being associated with a set of characteristics (relating to radioactivity, geometry, weight, etc.).

Still preferably, the nuclear waste management system 100 comprises grinding means 106 configured to 10 grind pieces of nuclear waste 1.

Preferably, the nuclear waste management system 100 comprises a cutting device 108 configured to cut pieces of nuclear waste 1.

The nuclear waste management system 100 may also 15 comprise a weight sensor 103 configured to acquire weight data of a piece of nuclear waste 1. For example, the weight sensor 103 is located in a management robot arm 101 so that the weight of a piece of nuclear waste 1 can be acquired when the management robot arm 101 carries a piece of nuclear waste 1.

The nuclear waste management system 100 may possibly comprise a composition sensor 105, called the material sensor 105 in the rest of the description, configured to acquire data regarding the material of a 25 piece of nuclear waste 1. The material sensor 105 is for example an X-ray fluorescence sensor. For example, the material sensor 105 is placed near a conveyor 7 on which pieces of nuclear waste 1 are deposited. In this way, the material sensor 105 can acquire material data from the pieces of nuclear waste 1 passing within its field of acquisition.

The nuclear waste management system 100 further comprises a nuclear waste packaging system 120 comprising packaging data processing means 126 configured to determine a spatial arrangement of pieces of nuclear waste 1 in a container 9, identify a piece of nuclear waste 1 to be packaged among the pieces of nuclear waste 1 depending on the spatial arrangement, introduce the piece of nuclear waste 1 into the container 9 according to the spatial arrangement, and repeat the identification and introduction steps until no piece of nuclear waste 1 can be identified for packaging in the container 9. The management data processing means 126 are typical data processing means such as a processor of the computing equipment 3. The container 9 is a container that can take various shapes. The container 9 can be in the shape of a cylindrical barrel or a parallelepiped. The container 9 is used to contain pieces of nuclear waste 1. The container 9 is therefore preferably made of a material at least partially radioactivity-tight.

Preferably, the nuclear waste packaging system 120 comprises a packaging robot arm 122 and a packaging camera 124. The packaging robot arm 122 is configured to arrange pieces of nuclear waste 1 in a container 9. The packaging robot arm 122 is therefore configured to catch and deposit pieces of nuclear waste 1. For example, the packaging robot arm 122 is configured to catch pieces of nuclear waste 1 arranged in a waiting zone or on the conveyor V. The packaging camera 124 is configured to acquire 30 images of pieces of nuclear waste 1 before and after they are introduced into a container 9. The packaging camera 124 is, for example, arranged on the packaging robot arm 122. The packaging camera 124 can therefore be moved via commands of the packaging robot arm 122 so that the latter moves and re-orients itself so that the 5 packaging camera 124 moves and re-orients itself. Preferably, the nuclear waste packaging system 120 comprises data storage means 128 configured to store data regarding the containers 9. Data storage means 128 can also store data regarding the pieces of nuclear waste 10 1 stored by the containers 9.

The elements of the nuclear waste management system 100 can be connected to each other via a network. Thus, sensor and camera data can be sent to the data processing 15 and storage means of the computing equipment 3.

Method Segmentation and acquisition With reference to Figure 3, the invention relates to a method for managing a piece of nuclear waste 1. The management method begins first with a method for characterising the piece of nuclear waste 1. With reference to Figure 2, the method for characterising the piece of nuclear waste 1 preferably starts with a step (a0) of acquiring an image of a stack of pieces of nuclear waste 1 called the segmentation image. More precisely, a characterisation camera 112 acquires the segmentation image, which is an image of a stack of pieces of nuclear waste 5. The segmentation image is typically a 2D image. The pieces of nuclear waste 5 of the stack are diverse and may be of different types, materials, radioactivity, dimensions, etc. Then, the characterisation method preferably comprises a step (al) of segmenting the piece of nuclear waste 1 in the segmentation image. From the segmentation image, the characterisation data processing means 114 segment a piece of nuclear waste 1 among the plurality of pieces of nuclear waste 1 represented in the segmentation image. By segmentation, it is meant any image segmentation operation, that is, any image processing operation which aims at gathering pixels together according to predefined criteria. The pixels are then grouped into regions, which form a paving or partition of the image. The skilled person will know how to implement this segmentation. Indeed, the skilled person knows on the one hand algorithms, such as a thresholding algorithm, and on the other hand learned models, such as a neuronal network, to implement this segmentation. Thus, a piece of nuclear waste 1 is identified among all of the pieces of nuclear waste 1 represented in the segmentation image.

From this identification, the piece of nuclear waste 1 is localised and can be caught by a characterisation robot arm 116.

Then, preferably, the method for characterising the piece of nuclear waste 1 continues with a step (a3) of acquiring at least one image of the piece of nuclear waste 1 identified called the characterisation image. A characterisation image is typically a 2D image. For this, the nuclear waste 1 is preferably caught by a characterisation robot arm 116 which presents the piece of nuclear waste to a characterisation camera 112. The piece of nuclear waste 1 can thus be rotated within the field of view of the characterisation camera 112 so that it acquires one or more characterisation images of the piece of nuclear waste 1. In another embodiment, the piece of nuclear waste 1 is caught by a characterisation robot arm 116, then deposited in a characterisation zone. In this embodiment, the characterisation camera 112 is integral with the characterisation robot arm 116. Accordingly, the characterisation robot arm 116 orients the characterisation camera 112 so that the piece of nuclear waste 1 is in its field of vision. The characterisation robot arm 116 can also turn the characterisation camera 112 so that different characterisation images of the piece of nuclear waste 1, that is, characterisation images for different orientations of the piece of nuclear waste 1, are acquired. Preferably, in a step (a2), the piece of nuclear waste 1 is illuminated when characterisation images are acquired in such a way that, regardless of the surrounding brightness, the piece of nuclear waste 1 is clearly visible on the characterisation images.

The acquired characterisation images of the piece of nuclear waste 1 are preferably stored in the characterisation data storage means 118.

The acquired characterisation images of the piece of nuclear waste 1 make it possible to obtain geometric parameters for the piece of nuclear waste 1, a type of the piece of nuclear waste 1 and characteristics relating to the composition of the piece of nuclear waste 1. The geometric parameters include, for example, dimensions, volume, external surface area, shape of the piece of nuclear waste 1. The type of the piece of nuclear waste 1 is a type of object, for example, a tube, tap, glove. For example, a characteristic relating to the composition of the piece of nuclear waste 1 is the material. The type of the piece of nuclear waste 1 and the characteristics relating to the composition of the piece of nuclear waste 1 can be obtained by classification. The geometric parameters are obtained by analysing a 3D image of the piece of nuclear waste 1.

It should also be noted that a weight value of the piece of nuclear waste 1 can be acquired in a step (a4). When the characterisation robot arm 116 holds a piece of nuclear waste 1, a weight value of the piece of nuclear waste 1 can be acquired by means of a weight sensor 103 integral with the characterisation robot arm 116. The weight of the piece of nuclear waste 1 can also be acquired in any other way, for example by depositing the piece of nuclear waste 1 on a scale.

Geometric characterisation In a preferred embodiment, the characterisation method comprises a step (a6) of determining geometric parameters of the piece of nuclear waste 1. For this, preferably, in a prior step (a5), a 3D image of the piece of nuclear waste 1 is obtained. According to a first embodiment, several characterisation images are acquired and the characterisation data processing means 114 construct a 3D image of the piece of nuclear waste 1 from the plurality of 2D characterisation images. This technique is called photogrammetry. According to a second embodiment, a depth map representative of the piece of nuclear waste 1 is acquired by a dedicated apparatus such as a 3D camera or depth camera 113 and a 3D image of the nuclear waste is constructed from this depth map. Analysis of the 3D image enables at least one geometric parameter of the piece of nuclear waste 1 to be obtained. More precisely, dimensions and the shape can be obtained from a depth map extracted from the 3D image, that is, the depth map having been acquired for constructing the 3D image according to the second embodiment or a depth map extracted from the 3D image constructed according to the first embodiment. The geometric parameters can be directly the dimensions or shape of the piece of nuclear waste 1. Also, the geometric parameters can be calculated from the dimensions or shape of the piece of nuclear waste 1 and can therefore, for example, be the volume or the external surface area of the piece of nuclear waste 1. The geometric parameters are preferably stored in the data storage means 118 and associated with the piece of nuclear waste 1, the piece of nuclear waste 1 being associated with a unique identifier.

In a certain embodiment, the unique identifier of a piece of nuclear waste 1 can be visually found by recognition of the piece of nuclear waste 1 in an image. The skilled person knows the visual recognition techniques. In another embodiment, for example, a bar code may be affixed or engraved on the piece of nuclear waste 1, which allows the unique identifier of the piece of nuclear waste 1 to be found directly. The bar code can be for example the unique identifier itself or a representation of the unique identifier. The "bar code" embodiment applies mainly to relatively large pieces of nuclear waste 1, for example of length and/or width greater than 5 cm, which therefore has an external surface area large enough for engraving or affixing a bar code thereto.

The geometric parameters allow a more precise description of the piece of nuclear waste, for example, which makes it easier to package it. Indeed, knowing the geometry of pieces of nuclear waste, it is possible to optimise the spatial arrangement of the pieces of nuclear waste 1 in a container 9. Furthermore, it also makes it possible to know whether a piece of nuclear waste is considered too large (for example, to enter a container 9) and whether it needs to be cut. Indeed, if geometric parameters of the piece of nuclear waste 1 are greater than a certain threshold of geometric parameters, a step (a8) of cutting the piece of nuclear waste 1 is implemented by the cutting device 108. It should be noted that the step (a8) of cutting the piece of nuclear waste 1 can also be implemented if the weight of the piece of nuclear waste 1 is greater than a weight threshold. Thus, if the piece of nuclear waste 1 is deemed too large, too bulky or too heavy, it is cut. The bits of the piece of nuclear waste 1 resulting from this cut will therefore be easier to be packaged or handled by robot arms. After cutting, the bits of the pieces of nuclear waste 1 are returned to the stack of pieces of nuclear waste 1. The bits of the pieces of nuclear waste 1 therefore return to the initial state of the process, that is, in a stack of pieces of nuclear waste 1. These bits of the pieces of nuclear waste 1 are now considered to be pieces of nuclear waste in their own right and identifiers are associated with each of them.

Material and type classification The management method includes a step (bl) of determining a characteristic of the piece of nuclear waste 1 other than relating to the radioactivity of the piece of nuclear waste 1. Preferably, the characteristic of the piece of nuclear waste 1 other than relating to the radioactivity of the piece of nuclear waste 1 is a characteristic relating to the composition of the piece of nuclear waste 1. Still preferably, the characteristic relating to the composition of the piece of nuclear waste 1 is a material. In a certain embodiment, this characteristic is determined by the material sensor 105. In a certain preferred embodiment, this characteristic is determined following a step (a7) of the characterisation method, of assigning a material class to the piece of nuclear waste 1. It should be noted that step (a7) also preferably comprises assigning a type class to the piece of nuclear waste 1. The characterisation method therefore comprises a step (a7) of classifying the piece of nuclear waste 1 represented by a characterisation image among a plurality of classes relating to the composition of nuclear waste 1 and/or among a plurality of type classes of nuclear waste 1 by applying to the characterisation image of at least one classification model learned in a supervised manner on a learning base. Step (a7) therefore preferably comprises a first classification (c1) implemented to determine a characteristic relating to the composition of the piece of nuclear waste and a second classification (c2) of the image implemented to determine a type of the piece of nuclear waste. Here, classes relating to the composition of nuclear waste 1 are considered to be 5 material classes. Whether it is to assign a material or type class to the piece of nuclear waste 1, it should be noted that the classification method will have the same structure. Thus, in the rest of the description, the classification of material will be referred to but it is 10 understood that the classification of type reproduces the classification of material with the difference that it is attempted to assign a type class and not a material class.

First, in a step (a70) prior to step (a7), a characterisation image of the nuclear waste 1 is obtained. This characterisation image is, for example, an image stored in the characterisation data storage means 118 following its acquisition, the acquisition having been described above (step (a3)). The obtaining step may also comprise acquiring the characterisation image. Furthermore, a material learning base is provided containing a plurality of images of pieces of nuclear waste each associated with a material. The classification model(s) that are used are learned in advance in a supervised way on this learning base. Preferably, the classification step (a7) comprises a plurality of steps of applying the classification model to the characterisation image to obtain a plurality of candidate classes. In other words, with reference to Figure 4, the classification model is independently applied several times to the characterisation image.

Preferably, as illustrated in Figure 4, it is applied at least three times. The classification model is a conventional classification model such as a tensorial algorithm, a convolutional neural network, etc. Each application of the classification model to the characterisation image allows a candidate material class to be obtained. A material class is representative of a material. Thus, a material class can be a material name or a unique material identifier associated with a particular material. As the classification model is applied several times, several candidate material classes are obtained. From these candidate classes, a material class is obtained. Preferably, the material class is associated with a final error value. This final error value is representative of the probability that the final material class obtained will correspond to the actual material of the piece of nuclear waste 1. For example, if three candidate classes are obtained and the first two candidate classes are "Steel", while the third candidate class is "Plastic", the resulting material class will be "Steel" with a probability of 66%. Therefore, the error value of the class is 33%. This classification method comprising three sub-classification steps enables a final class with a better, 25 that is, lower, final error value to be obtained and therefore the classification is more reliable.

In a preferred embodiment, the characterisation method includes repeating steps of applying the classification model to the image for obtaining candidate classes until the error value is lower than an error threshold or until the total number of repetitions reaches an application threshold. If the error value is lower than an error threshold, then it is considered that the material class probably corresponds to the actual material of the piece of nuclear waste 1. For example, if the error threshold is 10%, the "Steel" material class determined in the example set forth previously having an error value of 33% is not retained (33% being greater than 10%). The material class associated with the piece of nuclear waste 1 can then be recorded in the database storing information of the piece of nuclear waste 1 for example in the management data storage means 107. Furthermore, the characterisation image of the piece of nuclear waste 1 and its material class can be added to the material learning base. This enriches the learning base and improves classification. On the other hand, if the final error value is greater than a final error threshold, then it is considered that the material class probably does not correspond to the actual material of the piece of nuclear waste 1. Thus, a new application of the classification model to the characterisation image is implemented. In other words, a new candidate class is obtained and a new material class is obtained from all candidate classes, that is, then including the new candidate class. The material class obtained therefore has a different error value than the previous material class and this error value may be lower than the error threshold. If the error value is lower than the error threshold, the material class is retained. However, if the error value remains greater than the error threshold, a new application of the classification model to the characterisation image is implemented provided that the total number of applications remains lower than an application threshold. Indeed, to limit the duration of classification, an application threshold is determined.

For example, the number of applications can be limited to 20.

If the total number of applications exceeds 20, that is, the application threshold, a manual classification can be implemented. In other words, an operator can manually assign a material class to the piece of nuclear waste 1. The operator indicates the material of the piece of nuclear waste 1 by means of its visual expertise, laboratory analysis of the piece of nuclear waste 1 or detection of the material by the material sensor 105.

The piece of nuclear waste 1 is then assigned a material class and the characterisation image associated with the material class is recorded in the learning database to enrich the learning database.

In a preferred embodiment, several classifications are performed for a same piece of nuclear waste 1 by applying at least one classification model to different characterisation images of the piece of nuclear waste 1. In other words, the steps of obtaining a characterisation and classifying are repeated for a same piece of nuclear waste. As a result, several final material classes are obtained for a same piece of nuclear waste 1. A calculation step can be performed on these final classes to identify the final class most likely corresponding to the actual material of the piece of nuclear waste 1.

Thus, the final material associated with the piece of nuclear waste 1 is even more likely the same as the actual material of the piece of nuclear waste 1. It is therefore understood that, in this case, there is no class associated with the piece of nuclear waste 1 recorded in the database 107 each time a classification 5 is implemented for a characterisation image. The class associated with the piece of nuclear waste 1 recorded in the database 107 will be that which will be retained by analysis of all the final classes obtained for the classifications implemented for all characterisation 10 images of the same piece of nuclear waste 1.

It should be noted that, in the case of material classification, a material of the piece of nuclear waste 1 can be determined by means of the material sensor 105. In this way, the material class assigned to the piece of nuclear waste 1 and the material determined by the material sensor 105 can be compared so that the material class is confirmed with the material determined by the material sensor 105 (and vice versa, that the material determined by the material sensor 105 is confirmed with the material class). This comparison can be an additional check and, if the class assigned and the material detected by the sensor are different, a manual check can be implemented.

Thus, once the material and type classification has 25 been implemented, a material and type of the piece of nuclear waste 1 are identified. The characterisation method is considered to be completed.

Radioactivity The method for managing the piece of nuclear waste 1 also includes a step (b2) of determining a characteristic relating to the radioactivity of the piece of nuclear waste 1. The characteristic relative to the radioactivity of the piece of nuclear waste 1 may be a radioactive intensity. Preferably, the characteristic relative to the radioactivity of the piece of nuclear waste 1 is a radioactivity level of the piece of nuclear waste 1 among a plurality of radioactivity levels. Radioactivity levels include for example "Medium activity" (MA), "Low activity" (FA), "Very low activity" (TFA). Typically, the radioactivity sensor 102 measures at least one radioactivity value of the piece of nuclear waste 1. Then, the management data processing means 104 determine a radioactivity level from this value. This can simply be performed by thresholding. For example, if the radioactivity value is less than 100 Bg/g, the "Very low activity" level will be associated with the piece of nuclear waste 1. This radioactivity characteristic makes it possible to manage nuclear waste 1 appropriately. Assigning radioactivity levels to pieces of nuclear waste and not directly a radioactivity intensity value makes it easier to manage pieces of nuclear waste 1 and to be able to divide them into groups of levels.

Packaging classes The method for managing the piece of nuclear waste 1 further comprises a step (c) of assigning a packaging class to the piece of nuclear waste 1 among a plurality of packaging classes depending on a characteristic relating to the nuclear activity of the piece of nuclear waste 1 and a characteristic of the piece of nuclear waste 1 other than relating to the nuclear activity of the piece of nuclear waste 1. Preferably, the characteristic relating to the nuclear activity of the piece of nuclear waste 1 is a radioactivity level. Preferably, the characteristic of the piece of nuclear 5 waste 1 other than relating to the nuclear activity of the piece of nuclear waste 1 is a characteristic relating to the composition of the piece of nuclear waste 1. Still preferably, this characteristic of the piece of nuclear waste 1 other than relating to the nuclear activity of 10 the piece of nuclear waste 1 is a material.

It is important that the pieces of nuclear waste 1 are managed depending on their radioactivity and material. Indeed, for example, the necessary packaging will not be the same depending on the radioactivity level of the piece of nuclear waste 1. More precautions may be required for "Medium activity" waste than for "Low activity" waste. The heat generated and the reactions will not be the same between pieces of nuclear waste 1 of different radioactivity. Moreover, it is preferable that some materials are not stored together as they can react with each other. In addition, depending on the materials, the recycling possibilities are different. Accordingly, packaging classes are determined for assignment to pieces of nuclear waste 1 depending on the material and the radioactivity level of the piece of nuclear waste 1. Thus, pieces of radioactive waste 1 with different radioactivity levels will not have the same packaging class. For a same radioactivity level, pieces of nuclear waste 1 will not necessarily have the same packaging class depending on their material. For example, packaging class number cll may correspond to "Low activity" steel pieces of nuclear waste 1 and packaging class number c12 may correspond to "Medium activity" paper and plastic pieces of nuclear waste 1.

Packaging Preferably, the pieces of nuclear waste 1 are sorted depending on their packaging class so that they are arranged, in a step (d0), in a waiting zone with pieces of nuclear waste 1 of the same packaging class. Waiting zones can simply be delimited ground zones.

The method for managing the piece of nuclear waste 1 preferably comprises steps of a packaging method (d) to package the piece of nuclear waste 1 in a container 9.

According to a certain embodiment, before arranging the piece of nuclear waste 1 in a container 9, in a step (dl), pieces of nuclear waste 1 of the same packaging class are ground by the grinding means 106. Accordingly, a homogeneous cluster of nuclear waste 1 is obtained. 20 When this cluster is packaged, little space in the container 9 is unoccupied since chips of pieces of nuclear waste 1 do not leave large gaps between them. The problem of the arrangement of pieces of nuclear waste 1 in a container 9 is then solved.

According to a majority embodiment, pieces of nuclear waste 1 are not ground and pieces of nuclear waste 1 have to be arranged in a container 9 by trying to optimise the spatial arrangement of the pieces of nuclear waste in the container 9 to optimise the volume of the container 9 at the most. Indeed, it is desired to use as few containers 9 as possible to limit the costs and space taken up by a set of containers 9. For this, in a step (d2), a spatial arrangement of the pieces of nuclear waste 1 in the container 9 is determined.

First, preferably, the packaging method comprises a step of modelling pieces of nuclear waste 1 in simplified shapes and the spatial arrangement is determined depending on the simplified shapes. The simplified shapes can be parallelepipeds, cubes, pyramids or any other simple geometric shapes.

Preferably, the simplified shapes are parallelepipeds, called boxes 1' in the rest of the description. For example, the simplified shape of a piece of nuclear waste 1 can be determined from a 3D image of the piece of nuclear waste 1. The 3D image of the piece of nuclear waste 1, which is associated with the identifier of the piece of nuclear waste 1 in the management data storage means 107, can be stored in the management data storage means 107. In a certain embodiment, the 3D image can be retrieved directly for example by scanning a bar code affixed or engraved on the piece of nuclear waste 1, which allows finding the unique identifier of the piece of nuclear waste 1 and thus retrieving the associated 3D image. The scanning operation can be implemented by a scanning device integral with a packaging robot arm 122.

This box 1' is a modelling of the piece of nuclear waste 1 and is supposed to include the piece of nuclear waste 1. It is understood that modelling is implemented by means of the packaging data processing means 126. Then, an anchor point of a modelling of a container 9 is determined. The anchor point corresponds to a coordinate point (0, 0, 0) in the reference frame of the container 9 if the modelling of the container 9' is empty, that is, that no box 1' has already been arranged in the modelling of the container 9'. If the modelling of the container 9' already contains boxes 1', the coordinates of the anchor point will be different from (0, 0, 0) . Then, a first attempt to arrange the box 1' in the modelling of the container 9' is implemented at the anchor point so that an angle of the box l' is located on the anchor point. If the box 1' fits in the modelling of the container 9' according to the first arrangement attempt, then the arrangement of the box 1' in the modelling of the container 9' is retained. The fact that the box 1' fits in the modelling of the container 9' means that the box 1' does not pass through the walls of the modelling of the container 9' and therefore, in an actual container 9, an actual object identical to the box 1' could be arranged in the container 9. If the box 1' does not fit in the modelling of the container 9', the first arrangement attempt is considered as failed.

A second arrangement attempt is then implemented.

At the second arrangement attempt, the box 1' is rotated to a second orientation, an angle of the box 1' being still located on the anchor point. This angle can be the same as that located on the anchor point at the first attempt. If according to this second orientation, the box 1' fits into the modelling of the container 9', then the spatial arrangement of the box 1' in the modelling of the container 9' according to this second orientation is retained. On the other hand, if the second arrangement attempt fails, a third arrangement attempt is implemented so that the box 1' is rotated to a third orientation, different from the first and second orientations, an angle of the box 1' being still located on the anchor point. Preferably, a limited plurality of rotations can be implemented so that the number of 5 attempts is limited and the calculation of the spatial arrangement of pieces of nuclear waste 1 in a container 9 is not too time consuming. For example, in the case of boxes 1' (which are parallelepipedal), it is possible to limit the attempts to six orientation attempts as 10 illustrated in Figure 5. In Figure 5, the anchor point is illustrated with a cross.

If all arrangement attempts fail, the box 1' is not arranged in the modelling of the container 9' and another box 1' is selected to be attempted to be arranged. The box 1' that was not arranged will be arranged in another modelling of the container 9', during another implementation of the packaging method. On the other hand, if an arrangement attempt is successful, the box 1' is arranged in the modelling of the container 9' and another box 1' is selected to be arranged in the container 9. The boxes 1' that can be selected to be arranged are the modellings of the pieces of nuclear waste 1 which, for example, are present in a same waiting zone. It is therefore understood that a set of pieces of nuclear waste 1 is in a waiting zone (for example in a zone delimited on the ground) and the pieces of nuclear waste 1 are considered one by one for the development of the spatial arrangement. For this, the packaging robot arm(s) 122 move and scan each piece of nuclear waste 1.

Once the modelling of the container 9' is filled, that is, that no box 1' can be added in the modelling of the container 9', it is considered that the spatial arrangement of the pieces of nuclear waste 1 in the container 9 corresponding to the modelling of the container 9' is determined. In other words, a strategy 5 for the spatial arrangement of pieces of nuclear waste 1 in the container 9 is determined and the arrangement of the pieces of nuclear waste 1 in the container 9 can be implemented. It is therefore understood that the arrangement of the pieces of nuclear waste 1 in the 10 container 9 in the real world (and not in modelling) will be carried out.

Once the spatial arrangement has been developed, the pieces of nuclear waste 1 can be effectively packaged in a container 9. Thus, in a step (d3), the first piece of nuclear waste 1 to be packaged according to the spatial arrangement is identified and caught by the packaging robot arm 122. Then, in a step (d4), the packaging robot arm 122 introduces the piece of nuclear waste 1 into the container 9. The packaging robot arm 122 can drop the piece of nuclear waste 1 into the container 9 or carefully place it in the container 9. Then, preferably, images of the inside of the container 9 are acquired by the packaging camera 124. These images can be 3D images or 2D images from which a 3D image can be constructed. Thus, preferably, a 3D image of the inside of the container 9 is obtained. This image shows the actual arrangement of the piece of nuclear waste 1 in the container 9. Indeed, the piece of nuclear waste 1 introduced into the container 9 will not necessarily be arranged in the same way as according to the predetermined spatial arrangement. The packaging robot arm 122 does not necessarily arrange the piece of nuclear waste 1 precisely. Furthermore, in reality, the piece of nuclear waste 1 does not take a simplified shape such as a box 1'. The piece of nuclear waste 1 has a more complex shape. Therefore, it is well understood that the piece of nuclear waste 1 may not be arranged exactly as in modelling. It is therefore necessary to know the actual arrangement of the piece of nuclear waste 1 in the container 9 before adding other pieces of nuclear waste 1 into the container 9. Thus, according to a preferred embodiment, the packaging method comprises a step (d5) of updating the spatial arrangement after introduction of a piece of nuclear waste 1 into the container 9 depending on the positions of the pieces of nuclear waste 1 already introduced into the container 9 from the 3D image of the inside of the container. It is understood that the spatial arrangement may be modified and that the pieces of nuclear waste 1 which were intended to be packaged in the container 9 according to the first spatial arrangement will not necessarily be the same as the pieces of nuclear waste 1 which will be intended to be packaged in the container 9 according to the new spatial arrangement determined. Preferably, the update is performed after each introduction of a piece of nuclear waste 1 into the container 9. This update makes it possible to adapt the filling of the container 9 to the actual arrangement of the pieces of nuclear waste 1 in the container 9. In this way, the spatial arrangement is further optimised at every filling step of the container 9.

When no additional piece of nuclear waste 1 is identified for introduction into the container 9, a filling rate of the container 9 is calculated and recorded. The filling rate corresponds to the volume of the pieces of nuclear waste 1 arranged in the container 9 divided by the volume of the container 9. For example, the volume of the container 9 is stored in the packaging data storage means 128 in which the containers 9 are identified by unique container 9 identifiers. In a certain embodiment, the unique identifier of a container 9 can be visually found by recognition of the container 9 in an image. The skilled person knows the visual recognition techniques. In another embodiment, the volume of the container 9 is known, for example, by a bar code affixed to the container 9 which makes it possible to find the unique identifier of the container 9 as well as the characteristics of the container 9 such as its volume or the identifiers of nuclear waste 1 it contains. The bar code can for example be the unique identifier itself of the container 9 or a representation of the unique identifier. The volume of the pieces of nuclear waste 1 in the container 9 can be calculated by summing the volumes of each piece of nuclear waste 1, preferably with the volume of the pieces of nuclear waste 1 being recorded in the management data storage means 107.

The filling rate of the container 9 allows the calculation of an empty volume of the container 9, that is, the volume of the container 9 which remains unoccupied despite all the pieces of nuclear waste 1 it contains, and thus the calculation of a volume of chips of pieces of nuclear waste and soil which can be introduced into the container 9. Indeed, preferably, the packaging method comprises a step (d6) of introducing soil or chips of pieces of nuclear waste 1 between pieces of nuclear waste 1 previously introduced in the container 9. This increases the filling rate of the container 9 and thus prevents pieces of nuclear waste 1 from moving too much into the container 9 when the container 9 is moved. Indeed, when moving the container 9, the pieces of nuclear waste may move and strike the internal walls of the container 9 and even damage them. By increasing the filling rate of the container 9 with chips and soil, movements of pieces of nuclear waste 1 are limited.

The volume of soil and chips introduced therefore depends on the filling rate of the container 9. Preferably, soil and chips are introduced until the filling rate reaches a targeted filling rate. The filling rate is predetermined. The filling rate can be 1 (100% of the container 9 is filled) or, for example, can be set to a lower rate such as 0.9 (90% of the container 9 is filled).

During the introduction of soil and chips into the container 9, a dynamic filling rate calculation method can be implemented to follow the change in the filling rate. Thus, the introduction of soil and chips into the container 9 can be stopped when the filling rate reaches the targeted filling rate. It is therefore understood that, in this embodiment, the filling rate is calculated at regular intervals. Thus, the filling rate here corresponds to the volume of the pieces of nuclear waste 1, soil, and chips introduced into the container 9 divided by the volume of the container 9. When the calculated filling rate reaches the targeted filling rate, the introduction of soil and chips is stopped. Finally, in a step (d7), the container 9 is closed 5 and the packaging method is repeated to package pieces of nuclear waste 1 into a new container 9. In other words, a spatial arrangement is calculated for a new container 9 and pieces of nuclear waste 1 are introduced into this container 9 until no piece of nuclear waste 10 can be identified for packaging in the container 9.

The invention is not limited to the embodiment described and represented in the appended figures. Modifications remain possible, in particular with regard to the constitution of various technical characteristics or by substitution of technical equivalents, without however departing from the protection scope of the invention.

Claims

CLAIMS1. A method for managing a piece of nuclear waste (1) comprising the implementation by data processing means (104) of the steps of: b1) determining at least a first characteristic of the piece of nuclear waste (1) other than relating to the radioactivity of the piece of nuclear waste (1), b2) determining a second characteristic of the piece of nuclear waste (1) relating to the radioactivity 10 of the piece of nuclear waste (1) from data from a radioactivity sensor, c) assigning a packaging class to the piece of nuclear waste (1) among a plurality of packaging classes depending on the first characteristic and the second 15 characteristic.
2. The method according to claim 1 further comprising a step (d) of packaging the piece of nuclear waste (1) depending on the packaging class of the piece 20 of nuclear waste (1).
3. The method according to claim 2, comprising a step (d0) prior to step (d) comprising arranging the piece of nuclear waste (1) in a waiting zone in which a plurality of pieces of nuclear waste (1) can be arranged, a step of calculating the spatial arrangement of pieces nuclear waste (1) in a container (9) depending on all of the pieces of nuclear waste (1) arranged in the waiting zone, and a step of arranging the piece of nuclear waste (1) in the container (9) depending on the spatial arrangement of pieces of nuclear waste (1).
4. The method according to any one of claims 2 and 3, wherein step (d) comprises a step (dl) of grinding the piece of nuclear waste (1) with other pieces of 5 nuclear waste (1) of the same packaging class.
5. The method according to any one of claims 1 to 4, wherein the second characteristic is a radioactivity level of the piece of nuclear waste (1) among a plurality 10 of radioactivity levels.
6. The method according to any one of claims 1 to 5, wherein the first characteristic is a characteristic relating to the composition of the piece of nuclear waste (1) and is determined in step (bl) from data from a composition sensor (105) and/or a classification of an image of the piece of nuclear waste (1).
7. The method according to any one of claims 1 to 6, wherein the first characteristic is a material.
8. The method according to any one of claims 1 to 7, comprising a step (a4) of determining the weight of the piece of nuclear waste (1).
9. The method according to any one of claims 1 to 8, comprising a step (a7) of determining a type of the piece of nuclear waste (1).
10. The method according to any one of claims 1 to 9, comprising a step (a6) of determining geometric parameters of the piece of nuclear waste (1) from an image of the piece of nuclear waste (1), the geometric parameters comprising dimensions, volume and/or external surface area of the piece of nuclear waste (1).
11. The method according to claim 10, comprising a step (a8) of cutting the piece of nuclear waste (1) if a geometric parameter of said piece of nuclear waste (1) is greater than a threshold of the geometric parameter of said piece of nuclear waste (1) or if the weight of said piece of nuclear waste (1) is greater than a weight threshold.
12. The method according to any one of claims 10 15 and 11, wherein geometric parameters of the piece of nuclear waste (1) are obtained from the analysis of a 3D image of the piece of nuclear waste (1).
13. The method according to claim 12, wherein the 3D image is constructed from a plurality of 2D images obtained by rotating the piece of nuclear waste (1) relative to a camera (112) or from a depth map acquired by a depth camera (113).
14. The method according to any one of claims 1 to 13, comprising a prior step (al) of identifying the piece of nuclear waste (1) from an image of a plurality of pieces of nuclear waste (1), the method being implemented for at least one piece of nuclear waste (1) of said plurality of pieces of nuclear waste (1).
15. A system for managing a piece of nuclear waste (100) comprising data processing means (104) configured to determine at least a first characteristic of the piece of nuclear waste (1) other than relating to the 5 radioactivity of the piece of nuclear waste (1), determine a second characteristic of the piece of nuclear waste (1) relating to the radioactivity of the piece of nuclear waste (1) from data from a radioactivity sensor and assign a packaging class to the piece of nuclear 10 waste (1) from a plurality of packaging classes depending on the first characteristic and the second characteristic.
16. The system (100) according to claim 15 15 comprising a cutting device (108) configured to cut the piece of nuclear waste (1).
17. The system (100) according to any one of claims 15 and 16 comprising at least one radioactivity sensor (102) configured to acquire radioactivity sensor (102) data of the piece of nuclear waste (1) and a camera (112) configured to acquire at least one image of the piece of nuclear waste (1).