Circuit breaker diagnostic procedure and associated computer program

Abstract

Method for diagnosing a circuit breaker and associated computer program. The present invention relates to a method for diagnosing (200) a circuit breaker comprising at least one pole, each pole comprising an electrical contact configured to switch between a closed position and an open position. The method comprises: for each pole of the circuit breaker, at least one image (208A, 208B) of the pole's electrical contact; for each pole of the circuit breaker, at least one determination (218A, 218B) of a health state of the pole from among at least two states by means of an artificial intelligence algorithm, the artificial intelligence algorithm taking as input at least one image of the pole and providing as output the health state of the pole; a determination (220A, 220B) of a health state of the circuit breaker from among at least two states, depending on the health state of each pole; and a rendering (222, 224) of the health state of the circuit breaker. Figure for the abstract: Figure 9

Description

Title of the invention: Method for diagnosing a circuit breaker and associated computer program

[0001] The present invention relates to a method for diagnosing a circuit breaker. It also relates to an associated computer program.

[0002] Circuit breakers are essential safety devices that interrupt the current in an electrical installation in the event of an electrical fault. It is therefore important to monitor the condition of the circuit breakers in an electrical installation to ensure their ability to protect the electrical system and to replace them if necessary.

[0003] The lifespan of a circuit breaker is determined, in particular, by the number of opening / closing cycles and the value of the current interrupted. By knowing the number of cycles and the current interrupted, it is therefore possible to estimate the remaining lifespan of the circuit breaker.

[0004] However, in real-world situations, information on the number of cycles completed and the current interrupted is only available for a very limited number of circuit breakers equipped with an advanced electronic trip unit that includes an accessory for counting the occurrences of the circuit breaker opening / closing and measuring the current interrupted. In all other cases, no method currently allows an operator to effectively determine the condition of a circuit breaker within an electrical installation, and therefore to deduce whether a replacement is necessary to continue providing safety functions.

[0005] The object of the invention is then to propose a diagnostic method for a circuit breaker allowing to determine in situ the state of health of a circuit breaker, not necessarily equipped with an advanced electronic tripping unit, without requiring prior removal of the circuit breaker from the electrical installation.

[0006] To this end, the invention relates to a method for diagnosing a circuit breaker comprising at least one pole, each pole comprising an electrical contact configured to switch between a closed position and an open position, characterized in that the method comprises: • for each pole of the circuit breaker, at least one shot of the electrical contact of the pole; • for each pole of the circuit breaker, at least one determination of a pole health state from among at least two states by means of an artificial intelligence algorithm, the artificial intelligence algorithm taking into account input at least one shot of the pole and output providing the state of health of the pole; • a determination of a circuit breaker health status from among at least two states, depending on the health status of each pole; and • a restoration of the health status of the circuit breaker.

[0007] Thanks to the invention, the condition of the circuit breaker is assessed in situ from a photograph taken directly on the electrical installation. In particular, the method, through the use of artificial intelligence, makes it possible to perform an analysis of the circuit breaker contacts that would be difficult or even impossible for a non-expert without in-depth knowledge of the circuit breaker's structure or the aging conditions of the contact pads. Thanks to the method of the invention, it is possible to determine whether or not the circuit breaker needs to be replaced to continue ensuring the safety of the electrical installation. Furthermore, taking a photograph avoids having to dismantle the circuit breaker, which represents a significant gain in practicality and time.

[0008] According to other advantageous aspects of the invention, the diagnostic method comprises one or more of the following features, taken individually or in all technically possible combinations:

[0009] - for each pole of the circuit breaker: • The electrical contact is imaged using an endoscopic probe; and • the method includes, prior to taking the image, an insertion of the endoscopic probe into the circuit breaker near a movable contact pad belonging to the electrical contact;

[0010] - for each pole of the circuit breaker: • the shot(s) are taken while the electrical contact of the pole is in the open position; and • the artificial intelligence algorithm used in the determination(s) of the health status of the pole is a classification algorithm taking as input the shot(s) taken while the electrical contact is in the open position and providing as output a state of the pole among a critical state, a non-compliant state, a compliant state and a good state;

[0011] - for each pole of the circuit breaker: • the shot(s) is taken while the electrical contact of the pole is in the closed position; and • The artificial intelligence algorithm used in the determination(s) of the health status of the pole is a detection algorithm anomaly taking as input one or more of the shots taken while the electrical contact is in the closed position and providing as output a state of the pole between a state of repulsion and a state of non-repulsion;

[0012] - the determination of the health status of the circuit breaker is carried out between a valid state and an invalid state, the invalid state being determined if, for at least one pole: • The health status of the pole predicted by the classification algorithm is either critical or non-compliant, or • the health state of the pole predicted by the anomaly detection algorithm is the state of repulsion;

[0013] the valid state being determined otherwise;

[0014] - the image(s) of the electrical contacts in the closed position and the determination(s) using the anomaly detection algorithm are carried out if and only if the health status of each pole predicted by the classification algorithm is the good state or the compliant state;

[0015] - the method further comprises, for each shot, at least one action adapting the shot includes converting the shot to greyscale, normalizing the shot's contrast, and resizing the shot;

[0016] - each artificial intelligence algorithm was trained during a phase of initialization prior to the process, each initialization phase comprising: • for each drive circuit breaker among a set of drive circuit breakers and for each pole of the drive circuit breaker, at least one shot of the pole contact; • for at least one shot obtained, at least one random transformation action of the shot; and • training the artificial intelligence algorithm on at least one set of training shots from among the shots;

[0017] - the initialization phase of the classification algorithm further includes: • for each image taken, an expert assigns an initial, actual health status of the pole from among an actual critical status, an actual non-compliant status, an actual compliant status, and an actual good status; and • a distribution of the obtained images into a first training set, a first validation set and a first test set, each first set including at least one image belonging to each state among the actual critical state, the actual non-compliant state, the actual compliant state and the actual good state, the training including training, validation and testing of the classification algorithm on the first training set, the first validation set and the first test set respectively;

[0018] - the initialization phase of the anomaly detection algorithm includes in besides : • for each shot, an expert's assignment of a second actual state of health of the pole from among a state of actual repulsion and a state of actual non-repulsion; and • a distribution of the shots obtained into a second training set, a second validation set and a second test set, the second training set comprising only shots whose second real health state is the real repulsion state and the second validation and test sets comprising at least one shot belonging to each state among the real repulsion state and the real non-repulsion state, the training comprising training, validation and testing of the anomaly detection algorithm on the second training set, the second validation set and the second test set respectively;

[0019] - the random transformation action includes at least one action among a random modification of the contrast of the shot, a random modification of the brightness of the shot, a random rotation of the shot and a random horizontal or vertical shift of the shot.

[0020] The invention also relates to a computer program comprising software instructions which, when executed by a computer, implement a diagnostic method as defined above.

[0021] The invention will become clearer upon reading the following description, given solely by way of non-limiting example, and made with reference to the drawings in which:

[0022] [Fig-1] [Fig. 1] is a perspective view of a device enabling the implementation of a process according to the invention;

[0023] [Fig.2] [Fig.2] is a cross-sectional diagram of part of a circuit breaker and part of the device of [Fig.1];

[0024] [Fig.3] [Fig.3] is a longitudinal section of the device of [Fig.1];

[0025] [Fig.4] [Fig.4] is a perspective view of a guide belonging to the device of the [Fig.1] from two distinct viewpoints;

[0026] [Fig.5] [Fig.5] is a perspective view of a spacer, a first spacer ring and a fourth spacer ring belonging to the device of [Fig.1]

[0027] [Fig.6] [Fig.6] is a perspective view and a cross-sectional view of a second spacer ring and a perspective view of a third spacer ring belonging to the device in [Fig.1];

[0028] [Fig.7] [Fig.7] is a side view of the device of [Fig.1] in position disengaged;

[0029] [Fig.8] [Fig.8] is a logic diagram of an initialization phase according to the invention;

[0030] [Fig.9] [Fig.9] is a flowchart of a diagnostic method according to the invention.

[0031] Figure 1 represents a device 1 for taking a picture of a contact of a circuit breaker 3, shown partially in [Fig.2].

[0032] Circuit breaker 3 is, for example, a molded case circuit breaker, also known as a molded case circuit breaker (MCCB). Its function is to detect an electrical fault occurring in an electrical installation (not shown) and to interrupt the current flowing in the electrical installation in the event of a fault. Circuit breaker 3 comprises at least one pole, generally three poles corresponding to three phases for a three-phase electrical installation, or four poles corresponding to three phases and a neutral for a four-pole electrical installation. The structure of a pole of circuit breaker 3 is partially visible in the cross-sectional view of [Fig. 2]. In particular, each pole comprises an electrical contact configured to switch between a closed and an open position. Each electrical contact comprises a fixed part carrying a fixed contact pad 8 and a moving part 6 carrying a movable contact pad 5.The fixed contact pad 8 and movable contact pad 5 allow the current to be established or interrupted between the fixed part and the movable part, depending on whether the movable part 6 is in an open or closed position. The cross-sectional view of [Fig.2] shows one pole, with its movable contact pad 5, the other poles being similar.

[0033] In normal operation of the circuit breaker 3, when the electrical contact of a pole is in the closed position, the circuit breaker 3 allows electric current to flow in the phase or neutral conductor corresponding to that pole. Conversely, when the electrical contact is in the open position, the circuit breaker 3 opposes the flow of current in the phase or neutral conductor corresponding to that pole.

[0034] Each pole is advantageously confined within a bulb, separated from the other poles. For each pole, the bulb defines a volume 7 communicating with the outside of the circuit breaker 3 via an escape chamber 9.

[0035] The image capture device 1, shown in perspective on [Fig.1] and in section on figures 2 and 3, is configured to take images of the electrical contacts of the circuit breaker 3. The device 1 is further connected to an external device 11, configured to implement a diagnostic method for the circuit breaker 3 detailed below.

[0036] The imaging device 1 includes an endoscopic probe 13, a guide 15 and a stop member 17. Advantageously, the imaging device 1 further includes a tube 19.

[0037] The endoscopic probe 13 comprises a probe body 21 and an optical fiber 23. The endoscopic probe 13 is configured to acquire images of surrounding elements from a distal end 25 of the optical fiber 23.

[0038] The guide 15 is configured to penetrate at least partially into the escape chamber 9 of the pole comprising the electrical contact whose image is to be captured, as shown in [Fig. 2]. To achieve this, the guide 15 is a part with a geometry complementary to a shape of the escape chamber 9. The guide 15 is shown in more detail in [Fig. 4], from two distinct viewpoints corresponding respectively to inserts A and B.

[0039] The guide 15 includes at least one straight cavity 27, configured to receive the optical fiber 23 and to guide the optical fiber 23 to a point P5 near the contact pad 5. In particular, the guide 15 is configured to guide the optical fiber 23 through the escape chamber 9, the optical fiber 23 then being guided by mechanical parts belonging to the circuit breaker 3 to arrive in the volume 7, near the contact pad 5. Thus, the guide 15 allows access to a view of the contacts of the circuit breaker 3 in situ on the electrical installation, without prior dismantling of the circuit breaker 3. The view of a contact of the circuit breaker 3 advantageously includes the movable contact pad 5, cheeks positioned on either side of the movable contact pad 5, and a spark arrestor.

[0040] The straight shape of the cavity 27 prevents the optical fiber 23 from twisting, limiting the risk of damage to the optical fiber 23. In addition, the diameter of the cavity 27 advantageously offers a clearance of approximately 0.1 mm relative to the external diameter of the optical fiber 23, which allows for easy insertion of the optical fiber 23, while providing good retention of the optical fiber 23 and relatively low optical dispersion.

[0041] As seen in Figures 3 and 4, the straight cavity 27 includes a mouth chamfer 29, facilitating the insertion of the optical fiber 23 into the cavity 27. The chamfer 29 is positioned on a face of the guide 15 opposite to the distal end 25 of the optical fiber 23 when the optical fiber 23 is inserted into the cavity 27.

[0042] In the example shown in the figures, the guide 15 comprises two cavities 27A and 27B forming a non-zero angle ai with each other, so as to guide the optical fiber 23 to the same point P5 near the movable contact pad 5, regardless of which cavity the optical fiber 23 is inserted into. In other words, axial extensions of the two cavities 27A and 27B converge at point P5. Thus, the optical fiber placing it in either of the cavities 27A and 27B allows the same contact to be visualized from two angles of view.

[0043] Advantageously, the guide 15 includes at least one flexible tab 31 allowing a width L of the guide 15 to be adapted to a width L9 of the escape chamber 9, as clearly shown in [Fig. 2]. This feature ensures that the guide 15 is held securely and accommodates tolerances on the width L9 of the escape chambers 9 from one pole to another and from one circuit breaker 3 to another.

[0044] Advantageously, the guide 15 comprises a base 16 and two lugs 18A and 18B which extend parallel to each other from the base 16. The guide 15 is monobloc.

[0045] Advantageously, the cavities 27A and 27B pass through the base 16. Each cavity 27A or 27B extends outside the leg it passes through, as visible on leg 18B of insert B) in [Fig. 4], and then inside the leg it passes through, as visible on leg 18A of insert B) in [Fig. 4]. Each of the cavities 27A and 27B advantageously opens onto the inside of the leg 18A or 18B that it passes through.

[0046] The cutting plane of [Fig.2] passes through the leg 18B and the cavity 27 visible on this figure corresponds to the cavity 27B shown in [Fig.4].

[0047] Preferably, each leg 18A or 18B is configured so as not to obstruct the passage of the optical fiber 23, nor to damage it during its progression towards point P5. Thus, the side of the leg 18A visible on insert B) of [Fig.4] has a cutout which defines a surface S18 for guiding the optical fiber 23 as it exits the cavity 27A.

[0048] The stop member 17 is configured to longitudinally hold the optical fiber 23 inserted in the guide 15 by fixing a distance d525 between the distal end 25 of the optical fiber 23 and the movable contact pad 5, as well as an angle a2 between a longitudinal axis A3 of the breaker 3 and a longitudinal axis A23 of the optical fiber 23. This feature allows greater reproducibility of the shooting by means of the device 1.

[0049] The stop member 17 is advantageously divided into a proximal stop element 33 and a distal stop element 35. The proximal stop element 33 is integral with the endoscopic probe 13 while the distal stop element 35 is integral with the guide 15, through the tube 19, as explained below.

[0050] In the illustrated example, the proximal stop element 33 comprises a spacer 37, a first spacer ring 39 and a plug 4L. These different components are shown individually in inserts A), B) and C) of [Fig.5].

[0051] The spacer 37 is configured to partially encircle the optical fiber 23 by being fixed to the optical fiber 23. To achieve this, at least one clamping screw 43, advantageously two clamping screws 43 as seen in [Fig. 3], tighten the spacer 37 so as to locally reduce a diameter of an internal volume V37 of the spacer 37 used to receive the optical fiber 23 and thus to secure the spacer 37 with the optical fiber 23. In the illustrated example, the two clamping screws 43 also serve to secure the spacer 37 with the first spacer ring 39. The spacer 37 advantageously includes at least one notch 45 to partially receive the clamping screw or screws 43.

[0052] The first spacer ring 39 at least partially encircles the spacer 37, being integral with the spacer 37. As shown in insert B) [Fig. 5], the first spacer ring 39 includes a male stop 47, as well as two holes 49 for the passage of the clamping screws 43. Advantageously, the first spacer ring 39 further includes a first external thread 51 for cooperating with the plug 4L

[0053] The plug 41, shown in insert C) of [Fig. 5], advantageously comprises a second internal thread or tapped hole 53 configured to cooperate with the first external thread 51. The plug 41 is advantageously secured to the first spacer ring 29 by means of a connecting screw 54. The plug 41 makes it easier for an operator to grasp the stop member 17 and to hold the spacer 37 in the first spacer ring 39.

[0054] As shown in Figures 1 and 3, the distal stop element 35 is connected to the guide via the pipe 19, configured to receive the optical fiber 23. In particular, the pipe 19 is fixed to the guide 15 and to the distal stop element 35 by means of mounting screws 20. The mounting screws 20 are received through respective holes on the guide 15 and on the distal stop element 35.

[0055] Thus, the optical fiber 23 passes through the proximal stop element 33, the distal stop element 35, the pipe 19 and then the guide 15.

[0056] The distal stop element 35 includes a second spacer ring 55 and, optionally, a third spacer ring 57. These elements are shown individually in [Fig.6].

[0057] In the mounted configuration of the stop member 17, the second spacer ring 55 at least partially encircles the first spacer ring 39. The second spacer ring 55 is movable in translation relative to the first spacer ring 39 along the longitudinal axis A23 of the optical fiber 23 and in rotation relative to the first spacer ring 39 around the longitudinal axis A23 of the optical fiber 23. The second spacer ring 55 includes a female stop 59, configured to cooperate with the male stop 47. Inserting the male stop 47 into the female stop 59 allows the second spacer ring 55, i.e. the distal stop element 35, to be secured with the first spacer ring 39, i.e. the proximal stop element 33, in the manner of a bayonet. In other words, inserting the male stop 47 into the female stop 59 allows the translation and rotation between the optical fiber 23 and the guide 15 relative to the longitudinal axis A23 of the optical fiber 23. Since the guide 15 is held by the breaker 3 when the guide 15 is inserted into the escape chamber 9, the insertion of the male stop 47 into the female stop 59 makes it possible to fix the position of the optical fiber 23 relative to the breaker 3, and in particular relative to the movable contact pad 5. It is therefore understood that the stop member 17 makes it possible to improve the repeatability of the imaging of the electrical contacts, by means of the endoscopic probe 13.

[0058] Advantageously, the second spacer ring 55 further includes a mouth chamfer 61 facilitating the insertion of the optical fiber 23 into the second spacer ring 55, as seen in the cross-sectional view of insert B) of [Fig.6].

[0059] In the illustrated example, the second spacer ring 55 further includes a clearance groove 63, parallel to the longitudinal axis A23 of the optical fiber 23. The clearance groove 63 is configured to cooperate with the male stop 47 and to guide, via the male stop 47, the first spacer ring 39 in translation relative to the second spacer ring 55, from the female stop 59, along the longitudinal axis A23 of the optical fiber 23, until the optical fiber 23 is entirely outside the breaker 3.

[0060] Advantageously, the second spacer ring 55 further includes a release stop 65 that prevents the first spacer ring 39 from translating relative to the second spacer ring 55 when the optical fiber 23 is completely outside the circuit breaker 3. The device 1 is then said to be in the clear position. The clear position is shown in [Fig. 7]. The clear position allows an operator to remove the guide 15 from the circuit breaker 3 without damaging the optical fiber 23, which is generally fragile and represents the majority of the cost of the device 1.

[0061] In the illustrated example, a rotation of the proximal stop element 33 relative to the distal stop element 35, while the device 1 is in the disengaged position, allows the proximal stop element 33 to be removed from the distal stop element 35. Thus, the guide 15 and the endoscopic probe 13 are again independent.

[0062] The third spacer ring 57 at least partially encircles the second spacer ring 55. The third spacer ring 57 is movable in translation along the longitudinal axis A23 of the optical fiber 23, relative to the second spacer ring 55, and in rotation about the longitudinal axis A23 of the optical fiber 23, relative to the second spacer ring 55. The third spacer ring 57 is secured to the second spacer ring 55 by means of micro-adjustment screws 67. Thus, the rotation and translation between the second spacer ring 55 and the third spacer ring 57 when the micro-adjustment screws are loosened allows to finely adjust the position of the optical fiber 23 relative to the circuit breaker 3, after inserting the male stop 47 into the female stop 59.

[0063] The various components of the imaging device 1 mentioned above, apart from the endoscopic probe 13 and the tube 19, are advantageously 3D printed from resin. More generally, the device 1 is at least partially 3D printed from resin. This manufacturing method requires that all the dimensions of these components be greater than or equal to 0.3 mm.

[0064] The external device 11 is connected to the probe body 21 by a wired or wireless link, so as to receive the images taken by the endoscopic probe 13.

[0065] The external device 11 is advantageously a smartphone or tablet on which is installed an application configured to implement the diagnostic process of the circuit breaker 3.

[0066] More generally, the external device 11 includes an electronic circuit designed to manipulate and / or transform data represented by electronic or physical quantities in registers and / or memories into other similar data corresponding to physical data in register memories or other types of display devices, transmission devices or storage devices.

[0067] As specific examples, the external device 11 is implemented in the form of a programmable logic component, such as an FPGA (Field Programmable Gate Array), or an integrated circuit, such as an ASIC (Application Specifies Integrated Circuit).

[0068] Alternatively, when the diagnostic process is performed in the form of one or more software programs, i.e., in the form of a computer program, also called a computer program product, it is further capable of being stored on a computer-readable medium, not shown. The computer-readable medium is, for example, a medium capable of storing electronic instructions and being connected to a bus of a computer system. By way of example, the readable medium is an optical disc, a magneto-optical disc, ROM, RAM, any type of non-volatile memory (e.g., FLASH or NVRAM), or a magnetic card. A computer program comprising software instructions is then stored on the readable medium.

[0069] The diagnostic method consists of determining the health status of the circuit breaker 3 from images of the electrical contacts of the circuit breaker 3 taken by the device 1 described above and by means of at least one artificial intelligence algorithm. Each artificial intelligence algorithm takes as input an image of an electrical contact of a pole and provides as output a health status of the pole. More precisely, each artificial intelligence algorithm provides a probability for Each health state of the pole is considered. Each probability ranges from 0 to a normalized maximum probability, for example, 1, 10, or 100. The sum of the probabilities obtained for each health state is then equal to the normalized maximum probability. The highest probability gives the predicted health state for the pole. The probability of the predicted health state for the pole provides a confidence level associated with that predicted health state.

[0070] Advantageously, the diagnostic process uses two distinct artificial intelligence algorithms, including a classification algorithm and an anomaly detection algorithm.

[0071] The classification algorithm is advantageously a supervised classification algorithm known to a person skilled in the art. By way of example, the classification algorithm is a neural network, for example a convolutional neural network, a recurrent neural network, or a transformer network. The classification algorithm takes as input at least one image of an electrical contact in the open position and provides as output a state of the corresponding pole among a critical state, a non-compliant state, a compliant state, and a good state, as well as the associated confidence level between 0 and the normalized maximum probability. Alternatively, the state of the pole is predicted only among a good state and a non-compliant state.

[0072] The anomaly detection algorithm is advantageously a semi-supervised learning algorithm, known to a person skilled in the art. For example, the anomaly detection algorithm is a combination of neural networks. The anomaly detection algorithm takes as input at least one image of an electrical contact in the closed position and provides as output a pole state from among a repulsion state and a non-repulsion state.

[0073] Each artificial intelligence algorithm is trained, prior to the diagnostic process, during an initialization phase 100 represented on [Fig.8] and described below.

[0074] The initialization phase 100 involves a set of drive circuit breakers. The drive circuit breakers are advantageously of the same type as the circuit breaker 3 to be diagnosed. In order to apply the diagnostic procedure to different types of circuit breakers 3, the drive phases 100 are advantageously repeated on different circuit breakers, providing an artificial intelligence algorithm trained for each type of circuit breaker.

[0075] Each initialization phase 100 includes at least a shot 108 of an electrical contact of each pole of each drive circuit breaker, a random transformation action 122 of the shot and a training 126 of the artificial intelligence algorithm on at least one set of shots including the shot.

[0076] In the example illustrated in [Fig. 8], the initialization phase 100 begins with an opening or closing 102 of the electrical contact of the drive circuit breaker to be photographed. In the initialization phase 100 of the supervised classification algorithm, this step 102 is an opening step of the electrical contact. In the initialization phase 100 of the semi-supervised anomaly detection algorithm, this step 102 is a closing step of the electrical contact.

[0077] The initialization phase 100 then includes an insertion 104 of the guide 15 into the escape chamber 9 of the pole of the drive circuit breaker comprising the electrical contact to be photographed, followed by an insertion phase 106 of the optical fiber 23 into the guide 15. The position of the optical fiber 23 relative to the circuit breaker is then fixed by inserting the male stop 47 into the female stop 59 and then by tightening the micro-adjustment screws 67. As explained previously, the distal end 25 of the optical fiber 23 is then close to the contact pad 5, at point P5.

[0078] During the image acquisition step 108, an image of the electrical contact and its environment is taken by the endoscopic probe 13 and then transmitted to the external device 11. The image advantageously includes the moving contact comprising the moving contact pad 5, the cheeks positioned on either side of the moving contact pad 5 and the spark arrestor.

[0079] The optical fiber 23 and the guide 15 are then removed from the drive circuit breaker during a removal phase 110. The removal phase advantageously includes a removal of the optical fiber 23 from the guide 15, until the device 1 is in a clear position, and then a removal of the guide 15 from the circuit breaker 3.

[0080] Then, the initialization phase includes an expert assigning 112 an actual health status of the pole, based on an analysis of the appearance of the contact pad 5 and its surroundings from the image transmitted to the external device 11, as well as the operating conditions of the circuit breaker 3. In the initialization phase 100 of the supervised classification algorithm, the actual health status of the pole is chosen by the expert from among a real critical state, a real non-compliant state, a real compliant state, and a real good state. Alternatively, the actual health status of the pole is chosen only from among a real good state and a real non-compliant state. In the initialization phase 100 of the anomaly detection algorithm, the actual health status of the pole is chosen by the expert from among a real repulsive state and a real non-repulsive state.

[0081] In an unrepresented variant, the allocation step 112 takes place before the withdrawal step 110.

[0082] The opening / closing steps 102, insertion steps 104 and 106, shooting steps 108, withdrawal steps 110 and assignment steps 112 are repeated as many times as necessary to constitute a training set of sufficient size to make the artificial intelligence algorithm reliable, for example 20 or 50 times.

[0083] In the example illustrated in [Fig. 8], a first test step 114 consists of checking whether each side of an electrical contact has been photographed, and if not, repeating from the insertion 106 of the optical fiber into the guide 15 in order to photograph the other side of the electrical contact. A second test step 116 consists of checking whether each pole of the drive circuit breaker has been photographed, and if not, repeating from the insertion 104 of the guide 15 into the escape chamber 9 in order to photograph another pole. A third test step 118 consists of checking whether each drive circuit breaker of the set of drive circuit breakers, corresponding to the required number of times, has been photographed, and if not, repeating from the opening / closing step 102 in order to photograph another drive circuit breaker. Thus, the initialization phase illustrated in [Fig.[8] allows for taking a picture of each side of each electrical contact of each drive circuit breaker.

[0084] The shots thus obtained are then distributed, during a distribution step 120, into a training set, a test set and a validation set.

[0085] In the initialization phase 100 of the supervised classification algorithm, the images are divided into a first training set, a first validation set, and a first test set. Each first set includes at least one image belonging to each of the following states: the actual critical state, the actual non-conforming state, the actual conforming state, and the actual good state. Advantageously, each state is represented in equal proportion in each of the first sets. This distribution allows for the implementation of supervised training that is familiar to a person skilled in the art.

[0086] In the initialization phase 100 of the anomaly detection algorithm, the images are divided into a second training set, a second validation set, and a second test set. The second training set comprises only images whose second actual health state is the actual repulsion state, and the second validation and test sets each comprise at least one image belonging to each of the two states: the actual repulsion state and the actual non-repulsion state. This division allows for the implementation of semi-supervised training, which is known to a person skilled in the art.

[0087] During the random transformation step 122, at least one random transformation is performed on at least one shot. The random transformation includes, for example, a random modification of a shot's contrast, a random modification of a shot's brightness, a random rotation of the shot, and / or a random horizontal or vertical shift of the shot. Advantageously, each shot is assigned a probability Random transformation. When the probability of transformation is zero, no transformation is performed. When the probability of transformation is non-zero, a combination of one or more transformations is performed. This random transformation allows for the reproduction of variability in the images captured during the diagnostic process. In other words, the random transformation makes artificial intelligence algorithms less sensitive to variations in image capture conditions.

[0088] The initialization phase 100 then includes an adaptation 124 of the images to the relevant artificial intelligence algorithm. The adaptation step 124 includes, for example, converting the images to grayscale, normalizing the image contrast, and resizing the images. The adaptation step 124 makes the images more usable by the artificial intelligence algorithm.

[0089] Finally, the training step 126 includes training, validation and testing of the artificial intelligence algorithm on the training set, validation set and test set respectively.

[0090] At the end of the initialization phase 100, the artificial intelligence algorithm concerned is trained and able to predict the state of health of a pole of the circuit breaker 3 from a photo of the pole contact.

[0091] The trained artificial intelligence algorithm is integrated into the external device 11, so that the diagnostic procedure 200, described below with reference to [Fig.9], can be implemented by the external device 11.

[0092] The diagnostic method 200 includes at least one shot 206A or 206B of the electrical contact of each pole of the circuit breaker 3, at least one determination 218A or 218B of the health status of each pole, a determination 220A and 220B of the health status of the circuit breaker 3 and a restitution 222 and 224 of the health status of the circuit breaker 3.

[0093] In the example illustrated in [Fig. 9], the diagnostic process 200 successively comprises a prediction of the state of the poles of the circuit breaker 3 by means of the supervised classification algorithm and then by means of the anomaly detection algorithm. The health status of the circuit breaker 3 is then predicted based on the predictions of the two algorithms, as described below.

[0094] In the example shown in [Fig.9], the diagnostic process 200 includes a step 202A of opening the contact of the circuit breaker 3, a step 204A of inserting the guide 15 into the escape chamber 9, a step 206A of inserting the optical fiber 23 into the guide 15, a step 208A of taking a picture of the electrical contact and a step 210A of removing the optical fiber 23 and the guide 15.

[0095] The method then includes test steps 212A and 214A similar to test steps 114 and 116 of the initialization phase 100. Test steps 212A and 214A allow the previous steps to be repeated until a shot is taken from each side of each electrical contact of circuit breaker 3.

[0096] Next, the diagnostic process 200 includes an adaptation step 216A, similar to the adaptation step 124 of the initialization phase 100, allowing the pole shots of the circuit breaker 3 to be adapted to the supervised classification algorithm.

[0097] Then, the health status of the pole is predicted by the supervised classification algorithm during the determination step 218A. As explained previously, the health status of the pole predicted by the supervised classification algorithm is a critical, non-compliant, compliant or good state.

[0098] The state of each pole is then taken into account in the 220A step of determining the health status of the circuit breaker. The health status of the circuit breaker is advantageously predicted between a valid state and an invalid state.

[0099] In particular, if the health status of at least one pole predicted by the classification algorithm is the critical or non-compliant state, then the predicted state for circuit breaker 3 is the invalid state. The step 222 for reporting the invalid state of the circuit breaker is then executed. The reporting takes, for example, the form of a display on a screen of the external device 11, allowing an operator to be informed that circuit breaker 3 is invalid and must be replaced to continue ensuring the safety of the electrical installation.

[0100] Conversely, if the classification algorithm does not predict a critical or non-compliant state for any pole, then the pole's health status is predicted by the anomaly detection algorithm to determine the presence or absence of repulsion, and thus to conclude on the health status of circuit breaker 3. Therefore, the electrical contact measurements in the closed position and the determinations using the anomaly detection algorithm are performed if and only if the health status of each pole predicted by the classification algorithm is good or compliant. If this is the case, steps 202B to 220B are executed.

[0101] Steps 202B to 220B are similar to steps 202A to 220A except for the differences mentioned below.

[0102] Unlike step 202A, which is an opening step of the electrical contacts, step 202B is a closing step of the electrical contacts of circuit breaker 3.

[0103] The adaptation step 216B adapts the shots to the anomaly detection algorithm.

[0104] Step 218B of determining the health status of each pole is performed using the anomaly detection algorithm. The predicted health status is then either the repulsion state or the non-repulsion state. If the health status of at least one pole of the circuit breaker 3 is the repulsion state, the circuit breaker is declared invalid and the step of Step 222 restores the circuit breaker to its invalid state. Conversely, if all poles are in a non-repulsive state, the circuit breaker is declared valid and step 224 restores the circuit breaker to its valid state.

[0105] In summary, the invalid state of the circuit breaker is determined if, for at least one pole, the health state of the pole predicted by the classification algorithm is the critical state or the non-compliant state, or if the health state of the pole predicted by the anomaly detection algorithm is the repulsion state; the valid state being determined otherwise.

[0106] Thus, at the end of the diagnostic process 200, the health status of the circuit breaker 3 installed in the electrical installation is known and returned by the external device 11. This knowledge makes it possible to take appropriate measures for the replacement and / or maintenance of the circuit breaker 3, in order to ensure the essential electrical safety functions of the electrical installation.

[0107] Any feature described above for one example or variant may also be implemented in the other examples or variants described above, as far as technically possible.

Claims

Demands

1. A diagnostic method (200) for a circuit breaker (3) comprising at least one pole, each pole comprising an electrical contact configured to switch between a closed position and an open position, characterized in that the method (200) comprises: • for each pole of the circuit breaker (3), at least one snapshot (208A, 208B) of the electrical contact of the pole; • for each pole of the circuit breaker (3), at least one determination (218A, 218B) of a health state of the pole from among at least two states by means of an artificial intelligence algorithm, the artificial intelligence algorithm taking as input the at least one snapshot of the pole and providing as output the health state of the pole; • a determination (220A, 220B) of a health state of the circuit breaker (3) from among at least two states, depending on the health state of each pole; and • a restitution (222, 224) of the health state of the circuit breaker (3).

2. Method (200) according to claim 1, wherein, for each pole of the circuit breaker (3): • the imaging (208A, 208B) of the electrical contact is carried out by means of an endoscopic probe (13); and • the method (200) comprises, prior to the imaging (208A, 208B), an insertion (206A, 206B) of the endoscopic probe (13) into the circuit breaker (3) in the vicinity of a movable contact pad (5) belonging to the electrical contact.

3. A method (200) according to any one of the preceding claims, wherein, for each pole of the circuit breaker (3): • the scan(s) (208A) is performed while the electrical contact of the pole is in the open position; and • the artificial intelligence algorithm used in the determination(s) (218A) of the pole's health status is a classification algorithm that takes as input the scan(s) performed while the electrical contact is in the open position and provides as output a state of the pole among a critical state, a non-compliant state, a compliant state and a good state.

4. A method (200) according to any one of the preceding claims, wherein, for each pole of the circuit breaker (3): • the shot(s) (208B) is taken while the electrical contact of the pole is in the closed position; and • the artificial intelligence algorithm used in the determination(s) (218B) of the health status of the pole is an anomaly detection algorithm taking as input the shot(s) taken while the electrical contact is in the closed position and providing as output a state of the pole from among a repulsion state and a non-repulsion state.

5. Method (200) according to claims 3 and 4, wherein the determination (220A, 220B) of the health status of the circuit breaker (3) is carried out between a valid state and an invalid state, the invalid state being determined if, for at least one pole: • the health status of the pole predicted by the classification algorithm is the critical state or the non-compliant state, or • the health status of the pole predicted by the anomaly detection algorithm is the repulsion state; the valid state being determined otherwise.

6. Method (200) according to claim 5, wherein the image(s) (208B) of the electrical contacts (5) in the closed position and the determination(s) (218B) by means of the anomaly detection algorithm are carried out if and only if the health status of each pole predicted by the classification algorithm is the good state or the compliant state.

7. A method (200) according to any one of the preceding claims, further comprising, for each shot, at least one shot adaptation action (216A, 216B) among a shot conversion to greyscale, a shot contrast normalization, and a shot resizing.

8. A method (200) according to any one of the preceding claims, wherein each artificial intelligence algorithm was trained during an initialization phase (100) prior to the process (200), each initialization phase (100) comprising: • for each drive circuit breaker among a set of drive circuit breakers and for each pole of the drive circuit breaker, at least one shot (108) of the pole contact; • for at least one shot obtained, at least one random shot transformation action (122); and • training the artificial intelligence algorithm (126) on at least one set of training shots from among the shots.

9. A method (200) according to claims 3 and 8, wherein the initialization phase (100) of the classification algorithm further comprises: • for each image capture, an assignment (112) by an expert of a first actual health status of the pole from among an actual critical status, an actual non-compliant status, an actual compliant status and an actual good status; and • a distribution (120) of the shots obtained into a first training set, a first validation set and a first test set, each first set comprising at least one shot belonging to each state among the actual critical state, the actual non-compliant state, the actual compliant state and the actual good state, the training (126) comprising training, validation and testing of the classification algorithm on the first training set, the first validation set and the first test set respectively.

10. A method according to claims 4 and 8, wherein the initialization phase (100) of the anomaly detection algorithm further comprises: for each shot, an assignment (112) by an expert of a second actual state of health of the pole from among an actual state of repulsion and an actual state of non-repulsion; and

11.

12. • a distribution (120) of the shots obtained into a second training set, a second validation set and a second test set, the second training set comprising only shots whose second real health state is the real repulsion state and the second validation and test sets comprising at least one shot belonging to each state among the real repulsion state and the real non-repulsion state, the training (126) comprising training, validation and testing of the anomaly detection algorithm on the second training set, the second validation set and the second test set respectively. Method (200) according to any one of claims 8 to 10, wherein the random transformation action (122) comprises at least one action among a random modification of a contrast of the shot, a random modification of a brightness of the shot, a random rotation of the shot and a random horizontal or vertical shift of the shot. Computer program comprising software instructions which, when executed by a computer, implement a method (200) according to any one of the preceding claims.

Images